Lines Matching +full:soft +full:- +full:compensation +full:- +full:method
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>
109 ((struct kvm_vcpu *)(ctxt)->vcpu)
112 * - enable syscall per default because its emulated by KVM
113 * - enable LME and LMA per default on 64 bit KVM
144 *(((struct kvm_x86_ops *)0)->func));
147 #include <asm/kvm-x86-ops.h>
164 /* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */
176 * Flags to manipulate forced emulation behavior (any non-zero value will
183 int __read_mostly pi_inject_timer = -1;
443 * List of MSRs that control the existence of MSR-based features, i.e. MSRs
456 (KVM_LAST_EMULATED_VMX_MSR - KVM_FIRST_EMULATED_VMX_MSR + 1)];
548 size - useroffset, NULL); in kvm_alloc_emulator_cache()
557 vcpu->arch.apf.gfns[i] = ~0; in kvm_async_pf_hash_reset()
573 if (msrs->registered) { in kvm_on_user_return()
574 msrs->registered = false; in kvm_on_user_return()
579 values = &msrs->values[slot]; in kvm_on_user_return()
580 if (values->host != values->curr) { in kvm_on_user_return()
581 wrmsrl(kvm_uret_msrs_list[slot], values->host); in kvm_on_user_return()
582 values->curr = values->host; in kvm_on_user_return()
607 return -1; in kvm_add_user_return_msr()
622 return -1; in kvm_find_user_return_msr()
634 msrs->values[i].host = value; in kvm_user_return_msr_cpu_online()
635 msrs->values[i].curr = value; in kvm_user_return_msr_cpu_online()
644 value = (value & mask) | (msrs->values[slot].host & ~mask); in kvm_set_user_return_msr()
645 if (value == msrs->values[slot].curr) in kvm_set_user_return_msr()
651 msrs->values[slot].curr = value; in kvm_set_user_return_msr()
652 if (!msrs->registered) { in kvm_set_user_return_msr()
653 msrs->urn.on_user_return = kvm_on_user_return; in kvm_set_user_return_msr()
654 user_return_notifier_register(&msrs->urn); in kvm_set_user_return_msr()
655 msrs->registered = true; in kvm_set_user_return_msr()
665 if (msrs->registered) in drop_user_return_notifiers()
666 kvm_on_user_return(&msrs->urn); in drop_user_return_notifiers()
720 * #DBs can be trap-like or fault-like, the caller must check other CPU in exception_type()
739 if (!ex->has_payload) in kvm_deliver_exception_payload()
742 switch (ex->vector) { in kvm_deliver_exception_payload()
745 * "Certain debug exceptions may clear bit 0-3. The in kvm_deliver_exception_payload()
749 vcpu->arch.dr6 &= ~DR_TRAP_BITS; in kvm_deliver_exception_payload()
758 * Active low bits should be cleared if 1-setting in payload. in kvm_deliver_exception_payload()
759 * Active high bits should be set if 1-setting in payload. in kvm_deliver_exception_payload()
766 vcpu->arch.dr6 |= DR6_ACTIVE_LOW; in kvm_deliver_exception_payload()
767 vcpu->arch.dr6 |= ex->payload; in kvm_deliver_exception_payload()
768 vcpu->arch.dr6 ^= ex->payload & DR6_ACTIVE_LOW; in kvm_deliver_exception_payload()
776 vcpu->arch.dr6 &= ~BIT(12); in kvm_deliver_exception_payload()
779 vcpu->arch.cr2 = ex->payload; in kvm_deliver_exception_payload()
783 ex->has_payload = false; in kvm_deliver_exception_payload()
784 ex->payload = 0; in kvm_deliver_exception_payload()
792 struct kvm_queued_exception *ex = &vcpu->arch.exception_vmexit; in kvm_queue_exception_vmexit()
794 ex->vector = vector; in kvm_queue_exception_vmexit()
795 ex->injected = false; in kvm_queue_exception_vmexit()
796 ex->pending = true; in kvm_queue_exception_vmexit()
797 ex->has_error_code = has_error_code; in kvm_queue_exception_vmexit()
798 ex->error_code = error_code; in kvm_queue_exception_vmexit()
799 ex->has_payload = has_payload; in kvm_queue_exception_vmexit()
800 ex->payload = payload; in kvm_queue_exception_vmexit()
813 * If the exception is destined for L2, morph it to a VM-Exit if L1 in kvm_multiple_exception()
817 kvm_x86_ops.nested_ops->is_exception_vmexit(vcpu, nr, error_code)) { in kvm_multiple_exception()
823 if (!vcpu->arch.exception.pending && !vcpu->arch.exception.injected) { in kvm_multiple_exception()
825 vcpu->arch.exception.pending = true; in kvm_multiple_exception()
826 vcpu->arch.exception.injected = false; in kvm_multiple_exception()
828 vcpu->arch.exception.has_error_code = has_error; in kvm_multiple_exception()
829 vcpu->arch.exception.vector = nr; in kvm_multiple_exception()
830 vcpu->arch.exception.error_code = error_code; in kvm_multiple_exception()
831 vcpu->arch.exception.has_payload = has_payload; in kvm_multiple_exception()
832 vcpu->arch.exception.payload = payload; in kvm_multiple_exception()
835 &vcpu->arch.exception); in kvm_multiple_exception()
840 prev_nr = vcpu->arch.exception.vector; in kvm_multiple_exception()
842 /* triple fault -> shutdown */ in kvm_multiple_exception()
854 vcpu->arch.exception.injected = false; in kvm_multiple_exception()
855 vcpu->arch.exception.pending = false; in kvm_multiple_exception()
860 that instruction re-execution will regenerate lost in kvm_multiple_exception()
891 * On VM-Entry, an exception can be pending if and only if event in kvm_requeue_exception()
901 * re-checking is incorrect if _L1_ injected the exception, in which in kvm_requeue_exception()
906 vcpu->arch.exception.injected = true; in kvm_requeue_exception()
907 vcpu->arch.exception.has_error_code = has_error_code; in kvm_requeue_exception()
908 vcpu->arch.exception.vector = nr; in kvm_requeue_exception()
909 vcpu->arch.exception.error_code = error_code; in kvm_requeue_exception()
910 vcpu->arch.exception.has_payload = false; in kvm_requeue_exception()
911 vcpu->arch.exception.payload = 0; in kvm_requeue_exception()
939 ++vcpu->stat.pf_guest; in kvm_inject_page_fault()
942 * Async #PF in L2 is always forwarded to L1 as a VM-Exit regardless of in kvm_inject_page_fault()
945 if (is_guest_mode(vcpu) && fault->async_page_fault) in kvm_inject_page_fault()
947 true, fault->error_code, in kvm_inject_page_fault()
948 true, fault->address); in kvm_inject_page_fault()
950 kvm_queue_exception_e_p(vcpu, PF_VECTOR, fault->error_code, in kvm_inject_page_fault()
951 fault->address); in kvm_inject_page_fault()
958 WARN_ON_ONCE(fault->vector != PF_VECTOR); in kvm_inject_emulated_page_fault()
960 fault_mmu = fault->nested_page_fault ? vcpu->arch.mmu : in kvm_inject_emulated_page_fault()
961 vcpu->arch.walk_mmu; in kvm_inject_emulated_page_fault()
967 if ((fault->error_code & PFERR_PRESENT_MASK) && in kvm_inject_emulated_page_fault()
968 !(fault->error_code & PFERR_RSVD_MASK)) in kvm_inject_emulated_page_fault()
969 kvm_mmu_invalidate_addr(vcpu, fault_mmu, fault->address, in kvm_inject_emulated_page_fault()
972 fault_mmu->inject_page_fault(vcpu, fault); in kvm_inject_emulated_page_fault()
978 atomic_inc(&vcpu->arch.nmi_queued); in kvm_inject_nmi()
1012 return vcpu->arch.reserved_gpa_bits | rsvd_bits(5, 8) | rsvd_bits(1, 2); in pdptr_rsvd_bits()
1020 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in load_pdptrs()
1025 u64 pdpte[ARRAY_SIZE(mmu->pdptrs)]; in load_pdptrs()
1053 if (!tdp_enabled && memcmp(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs))) in load_pdptrs()
1054 kvm_mmu_free_roots(vcpu->kvm, mmu, KVM_MMU_ROOT_CURRENT); in load_pdptrs()
1056 memcpy(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs)); in load_pdptrs()
1059 vcpu->arch.pdptrs_from_userspace = false; in load_pdptrs()
1084 * CR0.WP is incorporated into the MMU role, but only for non-nested, in kvm_post_set_cr0()
1131 if ((vcpu->arch.efer & EFER_LME) && !is_paging(vcpu) && in kvm_set_cr0()
1142 if (!(vcpu->arch.efer & EFER_LME) && (cr0 & X86_CR0_PG) && in kvm_set_cr0()
1167 if (vcpu->arch.guest_state_protected) in kvm_load_guest_xsave_state()
1172 if (vcpu->arch.xcr0 != kvm_host.xcr0) in kvm_load_guest_xsave_state()
1173 xsetbv(XCR_XFEATURE_ENABLED_MASK, vcpu->arch.xcr0); in kvm_load_guest_xsave_state()
1176 vcpu->arch.ia32_xss != kvm_host.xss) in kvm_load_guest_xsave_state()
1177 wrmsrl(MSR_IA32_XSS, vcpu->arch.ia32_xss); in kvm_load_guest_xsave_state()
1181 vcpu->arch.pkru != vcpu->arch.host_pkru && in kvm_load_guest_xsave_state()
1182 ((vcpu->arch.xcr0 & XFEATURE_MASK_PKRU) || in kvm_load_guest_xsave_state()
1184 wrpkru(vcpu->arch.pkru); in kvm_load_guest_xsave_state()
1190 if (vcpu->arch.guest_state_protected) in kvm_load_host_xsave_state()
1194 ((vcpu->arch.xcr0 & XFEATURE_MASK_PKRU) || in kvm_load_host_xsave_state()
1196 vcpu->arch.pkru = rdpkru(); in kvm_load_host_xsave_state()
1197 if (vcpu->arch.pkru != vcpu->arch.host_pkru) in kvm_load_host_xsave_state()
1198 wrpkru(vcpu->arch.host_pkru); in kvm_load_host_xsave_state()
1203 if (vcpu->arch.xcr0 != kvm_host.xcr0) in kvm_load_host_xsave_state()
1207 vcpu->arch.ia32_xss != kvm_host.xss) in kvm_load_host_xsave_state()
1217 return vcpu->arch.guest_supported_xcr0 & XFEATURE_MASK_USER_DYNAMIC; in kvm_guest_supported_xfd()
1224 u64 old_xcr0 = vcpu->arch.xcr0; in __kvm_set_xcr()
1240 valid_bits = vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FP; in __kvm_set_xcr()
1259 vcpu->arch.xcr0 = xcr0; in __kvm_set_xcr()
1262 vcpu->arch.cpuid_dynamic_bits_dirty = true; in __kvm_set_xcr()
1291 * If CR4.PCIDE is changed 0 -> 1, there is no need to flush the TLB in kvm_post_set_cr4()
1305 * - CR4.PCIDE is changed from 1 to 0 in kvm_post_set_cr4()
1306 * - CR4.PGE is toggled in kvm_post_set_cr4()
1317 * - CR4.SMEP is changed from 0 to 1 in kvm_post_set_cr4()
1318 * - CR4.PAE is toggled in kvm_post_set_cr4()
1360 struct kvm_mmu *mmu = vcpu->arch.mmu; in kvm_invalidate_pcid()
1395 if (kvm_get_pcid(vcpu, mmu->prev_roots[i].pgd) == pcid) in kvm_invalidate_pcid()
1398 kvm_mmu_free_roots(vcpu->kvm, mmu, roots_to_free); in kvm_invalidate_pcid()
1431 vcpu->arch.cr3 = cr3; in kvm_set_cr3()
1440 * and it's impossible to use a non-zero PCID when PCID is disabled, in kvm_set_cr3()
1457 vcpu->arch.cr8 = cr8; in kvm_set_cr8()
1467 return vcpu->arch.cr8; in kvm_get_cr8()
1475 if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) { in kvm_update_dr0123()
1477 vcpu->arch.eff_db[i] = vcpu->arch.db[i]; in kvm_update_dr0123()
1485 if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) in kvm_update_dr7()
1486 dr7 = vcpu->arch.guest_debug_dr7; in kvm_update_dr7()
1488 dr7 = vcpu->arch.dr7; in kvm_update_dr7()
1490 vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_BP_ENABLED; in kvm_update_dr7()
1492 vcpu->arch.switch_db_regs |= KVM_DEBUGREG_BP_ENABLED; in kvm_update_dr7()
1510 size_t size = ARRAY_SIZE(vcpu->arch.db); in kvm_set_dr()
1514 vcpu->arch.db[array_index_nospec(dr, size)] = val; in kvm_set_dr()
1515 if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) in kvm_set_dr()
1516 vcpu->arch.eff_db[dr] = val; in kvm_set_dr()
1522 vcpu->arch.dr6 = (val & DR6_VOLATILE) | kvm_dr6_fixed(vcpu); in kvm_set_dr()
1528 vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1; in kvm_set_dr()
1539 size_t size = ARRAY_SIZE(vcpu->arch.db); in kvm_get_dr()
1543 return vcpu->arch.db[array_index_nospec(dr, size)]; in kvm_get_dr()
1546 return vcpu->arch.dr6; in kvm_get_dr()
1549 return vcpu->arch.dr7; in kvm_get_dr()
1573 * 10 - MISC_PACKAGE_CTRLS
1574 * 11 - ENERGY_FILTERING_CTL
1575 * 12 - DOITM
1576 * 18 - FB_CLEAR_CTRL
1577 * 21 - XAPIC_DISABLE_STATUS
1578 * 23 - OVERCLOCKING_STATUS
1709 u64 old_efer = vcpu->arch.efer; in set_efer()
1710 u64 efer = msr_info->data; in set_efer()
1716 if (!msr_info->host_initiated) { in set_efer()
1721 (vcpu->arch.efer & EFER_LME) != (efer & EFER_LME)) in set_efer()
1726 efer |= vcpu->arch.efer & EFER_LMA; in set_efer()
1754 struct kvm *kvm = vcpu->kvm; in kvm_msr_allowed()
1763 idx = srcu_read_lock(&kvm->srcu); in kvm_msr_allowed()
1765 msr_filter = srcu_dereference(kvm->arch.msr_filter, &kvm->srcu); in kvm_msr_allowed()
1771 allowed = msr_filter->default_allow; in kvm_msr_allowed()
1772 ranges = msr_filter->ranges; in kvm_msr_allowed()
1774 for (i = 0; i < msr_filter->count; i++) { in kvm_msr_allowed()
1781 allowed = test_bit(index - start, bitmap); in kvm_msr_allowed()
1787 srcu_read_unlock(&kvm->srcu, idx); in kvm_msr_allowed()
1796 * Returns 0 on success, non-0 otherwise.
1817 * non-canonical address is written on Intel but not on in __kvm_set_msr()
1818 * AMD (which ignores the top 32-bits, because it does in __kvm_set_msr()
1819 * not implement 64-bit SYSENTER). in __kvm_set_msr()
1821 * 64-bit code should hence be able to write a non-canonical in __kvm_set_msr()
1823 * vmentry does not fail on Intel after writing a non-canonical in __kvm_set_msr()
1825 * invokes 64-bit SYSENTER. in __kvm_set_msr()
1844 * clear the bits. This ensures cross-vendor migration will in __kvm_set_msr()
1877 * Returns 0 on success, non-0 otherwise.
1944 if (!vcpu->run->msr.error) { in complete_userspace_rdmsr()
1945 kvm_rax_write(vcpu, (u32)vcpu->run->msr.data); in complete_userspace_rdmsr()
1946 kvm_rdx_write(vcpu, vcpu->run->msr.data >> 32); in complete_userspace_rdmsr()
1952 return complete_emulated_insn_gp(vcpu, vcpu->run->msr.error); in complete_emulated_msr_access()
1963 return kvm_x86_call(complete_emulated_msr)(vcpu, vcpu->run->msr.error); in complete_fast_msr_access()
1992 if (!(vcpu->kvm->arch.user_space_msr_mask & msr_reason)) in kvm_msr_user_space()
1995 vcpu->run->exit_reason = exit_reason; in kvm_msr_user_space()
1996 vcpu->run->msr.error = 0; in kvm_msr_user_space()
1997 memset(vcpu->run->msr.pad, 0, sizeof(vcpu->run->msr.pad)); in kvm_msr_user_space()
1998 vcpu->run->msr.reason = msr_reason; in kvm_msr_user_space()
1999 vcpu->run->msr.index = index; in kvm_msr_user_space()
2000 vcpu->run->msr.data = data; in kvm_msr_user_space()
2001 vcpu->arch.complete_userspace_io = completion; in kvm_msr_user_space()
2017 kvm_rax_write(vcpu, data & -1u); in kvm_emulate_rdmsr()
2018 kvm_rdx_write(vcpu, (data >> 32) & -1u); in kvm_emulate_rdmsr()
2080 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS)) in kvm_emulate_monitor_mwait()
2083 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT)) in kvm_emulate_monitor_mwait()
2086 enabled = vcpu->arch.ia32_misc_enable_msr & MSR_IA32_MISC_ENABLE_MWAIT; in kvm_emulate_monitor_mwait()
2111 return READ_ONCE(vcpu->mode) == EXITING_GUEST_MODE || in kvm_vcpu_exit_request()
2117 * i.e. the sending of IPI, sending IPI early in the VM-Exit flow reduces
2124 if (!lapic_in_kernel(vcpu) || !apic_x2apic_mode(vcpu->arch.apic)) in handle_fastpath_set_x2apic_icr_irqoff()
2131 return kvm_x2apic_icr_write(vcpu->arch.apic, data); in handle_fastpath_set_x2apic_icr_irqoff()
2205 return -EINVAL; in do_set_msr()
2237 write_seqcount_begin(&vdata->seq); in update_pvclock_gtod()
2240 vdata->clock.vclock_mode = tk->tkr_mono.clock->vdso_clock_mode; in update_pvclock_gtod()
2241 vdata->clock.cycle_last = tk->tkr_mono.cycle_last; in update_pvclock_gtod()
2242 vdata->clock.mask = tk->tkr_mono.mask; in update_pvclock_gtod()
2243 vdata->clock.mult = tk->tkr_mono.mult; in update_pvclock_gtod()
2244 vdata->clock.shift = tk->tkr_mono.shift; in update_pvclock_gtod()
2245 vdata->clock.base_cycles = tk->tkr_mono.xtime_nsec; in update_pvclock_gtod()
2246 vdata->clock.offset = tk->tkr_mono.base; in update_pvclock_gtod()
2248 vdata->raw_clock.vclock_mode = tk->tkr_raw.clock->vdso_clock_mode; in update_pvclock_gtod()
2249 vdata->raw_clock.cycle_last = tk->tkr_raw.cycle_last; in update_pvclock_gtod()
2250 vdata->raw_clock.mask = tk->tkr_raw.mask; in update_pvclock_gtod()
2251 vdata->raw_clock.mult = tk->tkr_raw.mult; in update_pvclock_gtod()
2252 vdata->raw_clock.shift = tk->tkr_raw.shift; in update_pvclock_gtod()
2253 vdata->raw_clock.base_cycles = tk->tkr_raw.xtime_nsec; in update_pvclock_gtod()
2254 vdata->raw_clock.offset = tk->tkr_raw.base; in update_pvclock_gtod()
2256 vdata->wall_time_sec = tk->xtime_sec; in update_pvclock_gtod()
2258 vdata->offs_boot = tk->offs_boot; in update_pvclock_gtod()
2260 write_seqcount_end(&vdata->seq); in update_pvclock_gtod()
2320 struct kvm_arch *ka = &vcpu->kvm->arch; in kvm_write_system_time()
2322 if (vcpu->vcpu_id == 0 && !host_initiated) { in kvm_write_system_time()
2323 if (ka->boot_vcpu_runs_old_kvmclock != old_msr) in kvm_write_system_time()
2326 ka->boot_vcpu_runs_old_kvmclock = old_msr; in kvm_write_system_time()
2329 vcpu->arch.time = system_time; in kvm_write_system_time()
2334 kvm_gpc_activate(&vcpu->arch.pv_time, system_time & ~1ULL, in kvm_write_system_time()
2337 kvm_gpc_deactivate(&vcpu->arch.pv_time); in kvm_write_system_time()
2360 shift--; in kvm_get_time_scale()
2405 vcpu->arch.tsc_catchup = 1; in set_tsc_khz()
2406 vcpu->arch.tsc_always_catchup = 1; in set_tsc_khz()
2410 return -1; in set_tsc_khz()
2414 /* TSC scaling required - calculate ratio */ in set_tsc_khz()
2419 pr_warn_ratelimited("Invalid TSC scaling ratio - virtual-tsc-khz=%u\n", in set_tsc_khz()
2421 return -1; in set_tsc_khz()
2437 return -1; in kvm_set_tsc_khz()
2442 &vcpu->arch.virtual_tsc_shift, in kvm_set_tsc_khz()
2443 &vcpu->arch.virtual_tsc_mult); in kvm_set_tsc_khz()
2444 vcpu->arch.virtual_tsc_khz = user_tsc_khz; in kvm_set_tsc_khz()
2450 * rate. If so, no scaling or compensation need be done. in kvm_set_tsc_khz()
2452 thresh_lo = adjust_tsc_khz(tsc_khz, -tsc_tolerance_ppm); in kvm_set_tsc_khz()
2464 u64 tsc = pvclock_scale_delta(kernel_ns-vcpu->arch.this_tsc_nsec, in compute_guest_tsc()
2465 vcpu->arch.virtual_tsc_mult, in compute_guest_tsc()
2466 vcpu->arch.virtual_tsc_shift); in compute_guest_tsc()
2467 tsc += vcpu->arch.this_tsc_write; in compute_guest_tsc()
2481 struct kvm_arch *ka = &vcpu->kvm->arch; in kvm_track_tsc_matching()
2489 bool use_master_clock = (ka->nr_vcpus_matched_tsc + 1 == in kvm_track_tsc_matching()
2490 atomic_read(&vcpu->kvm->online_vcpus)) && in kvm_track_tsc_matching()
2491 gtod_is_based_on_tsc(gtod->clock.vclock_mode); in kvm_track_tsc_matching()
2499 if ((ka->use_master_clock && new_generation) || in kvm_track_tsc_matching()
2500 (ka->use_master_clock != use_master_clock)) in kvm_track_tsc_matching()
2503 trace_kvm_track_tsc(vcpu->vcpu_id, ka->nr_vcpus_matched_tsc, in kvm_track_tsc_matching()
2504 atomic_read(&vcpu->kvm->online_vcpus), in kvm_track_tsc_matching()
2505 ka->use_master_clock, gtod->clock.vclock_mode); in kvm_track_tsc_matching()
2512 * The most significant 64-N bits (mult) of ratio represent the
2515 * point number (mult + frac * 2^(-N)).
2538 tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio); in kvm_compute_l1_tsc_offset()
2540 return target_tsc - tsc; in kvm_compute_l1_tsc_offset()
2545 return vcpu->arch.l1_tsc_offset + in kvm_read_l1_tsc()
2546 kvm_scale_tsc(host_tsc, vcpu->arch.l1_tsc_scaling_ratio); in kvm_read_l1_tsc()
2577 if (vcpu->arch.guest_tsc_protected) in kvm_vcpu_write_tsc_offset()
2580 trace_kvm_write_tsc_offset(vcpu->vcpu_id, in kvm_vcpu_write_tsc_offset()
2581 vcpu->arch.l1_tsc_offset, in kvm_vcpu_write_tsc_offset()
2584 vcpu->arch.l1_tsc_offset = l1_offset; in kvm_vcpu_write_tsc_offset()
2592 vcpu->arch.tsc_offset = kvm_calc_nested_tsc_offset( in kvm_vcpu_write_tsc_offset()
2597 vcpu->arch.tsc_offset = l1_offset; in kvm_vcpu_write_tsc_offset()
2604 vcpu->arch.l1_tsc_scaling_ratio = l1_multiplier; in kvm_vcpu_write_tsc_multiplier()
2608 vcpu->arch.tsc_scaling_ratio = kvm_calc_nested_tsc_multiplier( in kvm_vcpu_write_tsc_multiplier()
2612 vcpu->arch.tsc_scaling_ratio = l1_multiplier; in kvm_vcpu_write_tsc_multiplier()
2622 * TSC is marked unstable when we're running on Hyper-V, in kvm_check_tsc_unstable()
2639 struct kvm *kvm = vcpu->kvm; in __kvm_synchronize_tsc()
2641 lockdep_assert_held(&kvm->arch.tsc_write_lock); in __kvm_synchronize_tsc()
2643 if (vcpu->arch.guest_tsc_protected) in __kvm_synchronize_tsc()
2647 vcpu->kvm->arch.user_set_tsc = true; in __kvm_synchronize_tsc()
2653 kvm->arch.last_tsc_nsec = ns; in __kvm_synchronize_tsc()
2654 kvm->arch.last_tsc_write = tsc; in __kvm_synchronize_tsc()
2655 kvm->arch.last_tsc_khz = vcpu->arch.virtual_tsc_khz; in __kvm_synchronize_tsc()
2656 kvm->arch.last_tsc_offset = offset; in __kvm_synchronize_tsc()
2658 vcpu->arch.last_guest_tsc = tsc; in __kvm_synchronize_tsc()
2670 * These values are tracked in kvm->arch.cur_xxx variables. in __kvm_synchronize_tsc()
2672 kvm->arch.cur_tsc_generation++; in __kvm_synchronize_tsc()
2673 kvm->arch.cur_tsc_nsec = ns; in __kvm_synchronize_tsc()
2674 kvm->arch.cur_tsc_write = tsc; in __kvm_synchronize_tsc()
2675 kvm->arch.cur_tsc_offset = offset; in __kvm_synchronize_tsc()
2676 kvm->arch.nr_vcpus_matched_tsc = 0; in __kvm_synchronize_tsc()
2677 } else if (vcpu->arch.this_tsc_generation != kvm->arch.cur_tsc_generation) { in __kvm_synchronize_tsc()
2678 kvm->arch.nr_vcpus_matched_tsc++; in __kvm_synchronize_tsc()
2682 vcpu->arch.this_tsc_generation = kvm->arch.cur_tsc_generation; in __kvm_synchronize_tsc()
2683 vcpu->arch.this_tsc_nsec = kvm->arch.cur_tsc_nsec; in __kvm_synchronize_tsc()
2684 vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write; in __kvm_synchronize_tsc()
2692 struct kvm *kvm = vcpu->kvm; in kvm_synchronize_tsc()
2698 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_synchronize_tsc()
2701 elapsed = ns - kvm->arch.last_tsc_nsec; in kvm_synchronize_tsc()
2703 if (vcpu->arch.virtual_tsc_khz) { in kvm_synchronize_tsc()
2710 } else if (kvm->arch.user_set_tsc) { in kvm_synchronize_tsc()
2711 u64 tsc_exp = kvm->arch.last_tsc_write + in kvm_synchronize_tsc()
2713 u64 tsc_hz = vcpu->arch.virtual_tsc_khz * 1000LL; in kvm_synchronize_tsc()
2715 * Here lies UAPI baggage: when a user-initiated TSC write has in kvm_synchronize_tsc()
2726 * come from the kernel's default vCPU creation. Make the 1-second in kvm_synchronize_tsc()
2738 * compensation code attempt to catch up if we fall behind, but in kvm_synchronize_tsc()
2742 vcpu->arch.virtual_tsc_khz == kvm->arch.last_tsc_khz) { in kvm_synchronize_tsc()
2744 offset = kvm->arch.cur_tsc_offset; in kvm_synchronize_tsc()
2754 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_synchronize_tsc()
2760 u64 tsc_offset = vcpu->arch.l1_tsc_offset; in adjust_tsc_offset_guest()
2766 if (vcpu->arch.l1_tsc_scaling_ratio != kvm_caps.default_tsc_scaling_ratio) in adjust_tsc_offset_host()
2769 vcpu->arch.l1_tsc_scaling_ratio); in adjust_tsc_offset_host()
2801 switch (clock->vclock_mode) { in vgettsc()
2807 v = (tsc_pg_val - clock->cycle_last) & in vgettsc()
2808 clock->mask; in vgettsc()
2817 v = (*tsc_timestamp - clock->cycle_last) & in vgettsc()
2818 clock->mask; in vgettsc()
2827 return v * clock->mult; in vgettsc()
2832 * frequency of CLOCK_MONOTONIC_RAW (hence adding gtos->offs_boot).
2842 seq = read_seqcount_begin(>od->seq); in do_kvmclock_base()
2843 ns = gtod->raw_clock.base_cycles; in do_kvmclock_base()
2844 ns += vgettsc(>od->raw_clock, tsc_timestamp, &mode); in do_kvmclock_base()
2845 ns >>= gtod->raw_clock.shift; in do_kvmclock_base()
2846 ns += ktime_to_ns(ktime_add(gtod->raw_clock.offset, gtod->offs_boot)); in do_kvmclock_base()
2847 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_kvmclock_base()
2865 seq = read_seqcount_begin(>od->seq); in do_monotonic()
2866 ns = gtod->clock.base_cycles; in do_monotonic()
2867 ns += vgettsc(>od->clock, tsc_timestamp, &mode); in do_monotonic()
2868 ns >>= gtod->clock.shift; in do_monotonic()
2869 ns += ktime_to_ns(gtod->clock.offset); in do_monotonic()
2870 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_monotonic()
2884 seq = read_seqcount_begin(>od->seq); in do_realtime()
2885 ts->tv_sec = gtod->wall_time_sec; in do_realtime()
2886 ns = gtod->clock.base_cycles; in do_realtime()
2887 ns += vgettsc(>od->clock, tsc_timestamp, &mode); in do_realtime()
2888 ns >>= gtod->clock.shift; in do_realtime()
2889 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_realtime()
2891 ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns); in do_realtime()
2892 ts->tv_nsec = ns; in do_realtime()
2960 * 4. ret0 = timespec0 + (rdtsc - tsc0) |
2961 * 5. | ret1 = timespec1 + (rdtsc - tsc1)
2962 * | ret1 = timespec0 + N + (rdtsc - (tsc0 + M))
2966 * - ret0 < ret1
2967 * - timespec0 + (rdtsc - tsc0) < timespec0 + N + (rdtsc - (tsc0 + M))
2969 * - 0 < N - M => M < N
2988 struct kvm_arch *ka = &kvm->arch; in pvclock_update_vm_gtod_copy()
2992 lockdep_assert_held(&kvm->arch.tsc_write_lock); in pvclock_update_vm_gtod_copy()
2993 vcpus_matched = (ka->nr_vcpus_matched_tsc + 1 == in pvclock_update_vm_gtod_copy()
2994 atomic_read(&kvm->online_vcpus)); in pvclock_update_vm_gtod_copy()
3001 &ka->master_kernel_ns, in pvclock_update_vm_gtod_copy()
3002 &ka->master_cycle_now); in pvclock_update_vm_gtod_copy()
3004 ka->use_master_clock = host_tsc_clocksource && vcpus_matched in pvclock_update_vm_gtod_copy()
3005 && !ka->backwards_tsc_observed in pvclock_update_vm_gtod_copy()
3006 && !ka->boot_vcpu_runs_old_kvmclock; in pvclock_update_vm_gtod_copy()
3008 if (ka->use_master_clock) in pvclock_update_vm_gtod_copy()
3012 trace_kvm_update_master_clock(ka->use_master_clock, vclock_mode, in pvclock_update_vm_gtod_copy()
3024 raw_spin_lock_irq(&kvm->arch.tsc_write_lock); in __kvm_start_pvclock_update()
3025 write_seqcount_begin(&kvm->arch.pvclock_sc); in __kvm_start_pvclock_update()
3038 struct kvm_arch *ka = &kvm->arch; in kvm_end_pvclock_update()
3042 write_seqcount_end(&ka->pvclock_sc); in kvm_end_pvclock_update()
3043 raw_spin_unlock_irq(&ka->tsc_write_lock); in kvm_end_pvclock_update()
3062 * per-CPU value (which may be zero if a CPU is going offline). Note, tsc_khz
3076 /* Called within read_seqcount_begin/retry for kvm->pvclock_sc. */
3079 struct kvm_arch *ka = &kvm->arch; in __get_kvmclock()
3085 data->flags = 0; in __get_kvmclock()
3086 if (ka->use_master_clock && in __get_kvmclock()
3091 if (kvm_get_walltime_and_clockread(&ts, &data->host_tsc)) { in __get_kvmclock()
3092 data->realtime = ts.tv_nsec + NSEC_PER_SEC * ts.tv_sec; in __get_kvmclock()
3093 data->flags |= KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC; in __get_kvmclock()
3096 data->host_tsc = rdtsc(); in __get_kvmclock()
3098 data->flags |= KVM_CLOCK_TSC_STABLE; in __get_kvmclock()
3099 hv_clock.tsc_timestamp = ka->master_cycle_now; in __get_kvmclock()
3100 hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; in __get_kvmclock()
3104 data->clock = __pvclock_read_cycles(&hv_clock, data->host_tsc); in __get_kvmclock()
3106 data->clock = get_kvmclock_base_ns() + ka->kvmclock_offset; in __get_kvmclock()
3114 struct kvm_arch *ka = &kvm->arch; in get_kvmclock()
3118 seq = read_seqcount_begin(&ka->pvclock_sc); in get_kvmclock()
3120 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in get_kvmclock()
3142 read_lock_irqsave(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3144 read_unlock_irqrestore(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3149 read_lock_irqsave(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3152 guest_hv_clock = (void *)(gpc->khva + offset); in kvm_setup_guest_pvclock()
3161 guest_hv_clock->version = hv_clock.version = (guest_hv_clock->version + 1) | 1; in kvm_setup_guest_pvclock()
3165 hv_clock.flags |= (guest_hv_clock->flags & PVCLOCK_GUEST_STOPPED); in kvm_setup_guest_pvclock()
3171 guest_hv_clock->version = ++hv_clock.version; in kvm_setup_guest_pvclock()
3174 read_unlock_irqrestore(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3176 trace_kvm_pvclock_update(vcpu->vcpu_id, &hv_clock); in kvm_setup_guest_pvclock()
3184 struct kvm_vcpu_arch *vcpu = &v->arch; in kvm_guest_time_update()
3185 struct kvm_arch *ka = &v->kvm->arch; in kvm_guest_time_update()
3198 seq = read_seqcount_begin(&ka->pvclock_sc); in kvm_guest_time_update()
3199 use_master_clock = ka->use_master_clock; in kvm_guest_time_update()
3201 host_tsc = ka->master_cycle_now; in kvm_guest_time_update()
3202 kernel_ns = ka->master_kernel_ns; in kvm_guest_time_update()
3204 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in kvm_guest_time_update()
3225 * 2) Broken TSC compensation resets the base at each VCPU in kvm_guest_time_update()
3231 if (vcpu->tsc_catchup) { in kvm_guest_time_update()
3234 adjust_tsc_offset_guest(v, tsc - tsc_timestamp); in kvm_guest_time_update()
3245 v->arch.l1_tsc_scaling_ratio); in kvm_guest_time_update()
3247 if (unlikely(vcpu->hw_tsc_khz != tgt_tsc_khz)) { in kvm_guest_time_update()
3249 &vcpu->pvclock_tsc_shift, in kvm_guest_time_update()
3250 &vcpu->pvclock_tsc_mul); in kvm_guest_time_update()
3251 vcpu->hw_tsc_khz = tgt_tsc_khz; in kvm_guest_time_update()
3254 hv_clock.tsc_shift = vcpu->pvclock_tsc_shift; in kvm_guest_time_update()
3255 hv_clock.tsc_to_system_mul = vcpu->pvclock_tsc_mul; in kvm_guest_time_update()
3257 hv_clock.system_time = kernel_ns + v->kvm->arch.kvmclock_offset; in kvm_guest_time_update()
3258 vcpu->last_guest_tsc = tsc_timestamp; in kvm_guest_time_update()
3265 if (vcpu->pv_time.active) { in kvm_guest_time_update()
3271 if (vcpu->pvclock_set_guest_stopped_request) { in kvm_guest_time_update()
3273 vcpu->pvclock_set_guest_stopped_request = false; in kvm_guest_time_update()
3275 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->pv_time, 0); in kvm_guest_time_update()
3280 kvm_hv_setup_tsc_page(v->kvm, &hv_clock); in kvm_guest_time_update()
3291 if (ka->xen.hvm_config.flags & KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE) in kvm_guest_time_update()
3294 if (vcpu->xen.vcpu_info_cache.active) in kvm_guest_time_update()
3295 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->xen.vcpu_info_cache, in kvm_guest_time_update()
3297 if (vcpu->xen.vcpu_time_info_cache.active) in kvm_guest_time_update()
3298 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->xen.vcpu_time_info_cache, 0); in kvm_guest_time_update()
3326 struct kvm_arch *ka = &kvm->arch; in kvm_get_wall_clock_epoch()
3332 seq = read_seqcount_begin(&ka->pvclock_sc); in kvm_get_wall_clock_epoch()
3335 if (!ka->use_master_clock) in kvm_get_wall_clock_epoch()
3348 local_tsc_khz = 0; /* Fall back to old method */ in kvm_get_wall_clock_epoch()
3357 hv_clock.tsc_timestamp = ka->master_cycle_now; in kvm_get_wall_clock_epoch()
3358 hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; in kvm_get_wall_clock_epoch()
3360 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in kvm_get_wall_clock_epoch()
3366 * since 1970-01-01. in kvm_get_wall_clock_epoch()
3372 return ts.tv_nsec + NSEC_PER_SEC * ts.tv_sec - in kvm_get_wall_clock_epoch()
3376 return ktime_get_real_ns() - get_kvmclock_ns(kvm); in kvm_get_wall_clock_epoch()
3381 * vcpu->cpu migration, should not allow system_timestamp from
3387 * We need to rate-limit these requests though, as they can
3390 * by the delay we use to rate-limit the updates.
3412 struct kvm *kvm = v->kvm; in kvm_gen_kvmclock_update()
3415 schedule_delayed_work(&kvm->arch.kvmclock_update_work, in kvm_gen_kvmclock_update()
3428 schedule_delayed_work(&kvm->arch.kvmclock_update_work, 0); in kvmclock_sync_fn()
3429 schedule_delayed_work(&kvm->arch.kvmclock_sync_work, in kvmclock_sync_fn()
3450 return !!(vcpu->arch.msr_hwcr & BIT_ULL(18)); in can_set_mci_status()
3457 u64 mcg_cap = vcpu->arch.mcg_cap; in set_msr_mce()
3459 u32 msr = msr_info->index; in set_msr_mce()
3460 u64 data = msr_info->data; in set_msr_mce()
3465 vcpu->arch.mcg_status = data; in set_msr_mce()
3469 (data || !msr_info->host_initiated)) in set_msr_mce()
3473 vcpu->arch.mcg_ctl = data; in set_msr_mce()
3475 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in set_msr_mce()
3476 last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1; in set_msr_mce()
3480 if (!(mcg_cap & MCG_CMCI_P) && (data || !msr_info->host_initiated)) in set_msr_mce()
3485 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2, in set_msr_mce()
3486 last_msr + 1 - MSR_IA32_MC0_CTL2); in set_msr_mce()
3487 vcpu->arch.mci_ctl2_banks[offset] = data; in set_msr_mce()
3489 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in set_msr_mce()
3490 last_msr = MSR_IA32_MCx_CTL(bank_num) - 1; in set_msr_mce()
3502 * single-bit ECC data errors. in set_msr_mce()
3510 * AMD-based CPUs allow non-zero values, but if and only if in set_msr_mce()
3513 if (!msr_info->host_initiated && is_mci_status_msr(msr) && in set_msr_mce()
3517 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL, in set_msr_mce()
3518 last_msr + 1 - MSR_IA32_MC0_CTL); in set_msr_mce()
3519 vcpu->arch.mce_banks[offset] = data; in set_msr_mce()
3531 return (vcpu->arch.apf.msr_en_val & mask) == mask; in kvm_pv_async_pf_enabled()
3553 vcpu->arch.apf.msr_en_val = data; in kvm_pv_enable_async_pf()
3561 if (kvm_gfn_to_hva_cache_init(vcpu->kvm, &vcpu->arch.apf.data, gpa, in kvm_pv_enable_async_pf()
3565 vcpu->arch.apf.send_always = (data & KVM_ASYNC_PF_SEND_ALWAYS); in kvm_pv_enable_async_pf()
3566 vcpu->arch.apf.delivery_as_pf_vmexit = data & KVM_ASYNC_PF_DELIVERY_AS_PF_VMEXIT; in kvm_pv_enable_async_pf()
3575 /* Bits 8-63 are reserved */ in kvm_pv_enable_async_pf_int()
3582 vcpu->arch.apf.msr_int_val = data; in kvm_pv_enable_async_pf_int()
3584 vcpu->arch.apf.vec = data & KVM_ASYNC_PF_VEC_MASK; in kvm_pv_enable_async_pf_int()
3591 kvm_gpc_deactivate(&vcpu->arch.pv_time); in kvmclock_reset()
3592 vcpu->arch.time = 0; in kvmclock_reset()
3597 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_all()
3606 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_guest()
3622 * Flushing all "guest" TLB is always a superset of Hyper-V's fine in kvm_vcpu_flush_tlb_guest()
3631 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_current()
3639 * prior before nested VM-Enter/VM-Exit.
3653 struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache; in record_steal_time()
3656 gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS; in record_steal_time()
3660 if (kvm_xen_msr_enabled(vcpu->kvm)) { in record_steal_time()
3665 if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED)) in record_steal_time()
3668 if (WARN_ON_ONCE(current->mm != vcpu->kvm->mm)) in record_steal_time()
3671 slots = kvm_memslots(vcpu->kvm); in record_steal_time()
3673 if (unlikely(slots->generation != ghc->generation || in record_steal_time()
3674 gpa != ghc->gpa || in record_steal_time()
3675 kvm_is_error_hva(ghc->hva) || !ghc->memslot)) { in record_steal_time()
3677 BUILD_BUG_ON((sizeof(*st) - 1) & KVM_STEAL_VALID_BITS); in record_steal_time()
3679 if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gpa, sizeof(*st)) || in record_steal_time()
3680 kvm_is_error_hva(ghc->hva) || !ghc->memslot) in record_steal_time()
3684 st = (struct kvm_steal_time __user *)ghc->hva; in record_steal_time()
3691 int err = -EFAULT; in record_steal_time()
3702 "+m" (st->preempted)); in record_steal_time()
3708 vcpu->arch.st.preempted = 0; in record_steal_time()
3710 trace_kvm_pv_tlb_flush(vcpu->vcpu_id, in record_steal_time()
3721 unsafe_put_user(0, &st->preempted, out); in record_steal_time()
3722 vcpu->arch.st.preempted = 0; in record_steal_time()
3725 unsafe_get_user(version, &st->version, out); in record_steal_time()
3730 unsafe_put_user(version, &st->version, out); in record_steal_time()
3734 unsafe_get_user(steal, &st->steal, out); in record_steal_time()
3735 steal += current->sched_info.run_delay - in record_steal_time()
3736 vcpu->arch.st.last_steal; in record_steal_time()
3737 vcpu->arch.st.last_steal = current->sched_info.run_delay; in record_steal_time()
3738 unsafe_put_user(steal, &st->steal, out); in record_steal_time()
3741 unsafe_put_user(version, &st->version, out); in record_steal_time()
3746 mark_page_dirty_in_slot(vcpu->kvm, ghc->memslot, gpa_to_gfn(ghc->gpa)); in record_steal_time()
3751 u32 msr = msr_info->index; in kvm_set_msr_common()
3752 u64 data = msr_info->data; in kvm_set_msr_common()
3755 * Do not allow host-initiated writes to trigger the Xen hypercall in kvm_set_msr_common()
3759 if (kvm_xen_is_hypercall_page_msr(vcpu->kvm, msr) && in kvm_set_msr_common()
3760 !msr_info->host_initiated) in kvm_set_msr_common()
3775 if (msr_info->host_initiated) in kvm_set_msr_common()
3776 vcpu->arch.microcode_version = data; in kvm_set_msr_common()
3779 if (!msr_info->host_initiated || in kvm_set_msr_common()
3782 vcpu->arch.arch_capabilities = data; in kvm_set_msr_common()
3785 if (!msr_info->host_initiated || in kvm_set_msr_common()
3797 if (vcpu->arch.perf_capabilities == data) in kvm_set_msr_common()
3800 vcpu->arch.perf_capabilities = data; in kvm_set_msr_common()
3806 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3834 if (!msr_info->host_initiated && in kvm_set_msr_common()
3861 vcpu->arch.msr_hwcr = data; in kvm_set_msr_common()
3873 vcpu->arch.pat = data; in kvm_set_msr_common()
3879 return kvm_apic_set_base(vcpu, data, msr_info->host_initiated); in kvm_set_msr_common()
3887 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3888 s64 adj = data - vcpu->arch.ia32_tsc_adjust_msr; in kvm_set_msr_common()
3895 vcpu->arch.ia32_tsc_adjust_msr = data; in kvm_set_msr_common()
3899 u64 old_val = vcpu->arch.ia32_misc_enable_msr; in kvm_set_msr_common()
3901 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3911 if (!kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT) && in kvm_set_msr_common()
3915 vcpu->arch.ia32_misc_enable_msr = data; in kvm_set_msr_common()
3916 vcpu->arch.cpuid_dynamic_bits_dirty = true; in kvm_set_msr_common()
3918 vcpu->arch.ia32_misc_enable_msr = data; in kvm_set_msr_common()
3923 if (!IS_ENABLED(CONFIG_KVM_SMM) || !msr_info->host_initiated) in kvm_set_msr_common()
3925 vcpu->arch.smbase = data; in kvm_set_msr_common()
3928 vcpu->arch.msr_ia32_power_ctl = data; in kvm_set_msr_common()
3931 if (msr_info->host_initiated) { in kvm_set_msr_common()
3933 } else if (!vcpu->arch.guest_tsc_protected) { in kvm_set_msr_common()
3934 u64 adj = kvm_compute_l1_tsc_offset(vcpu, data) - vcpu->arch.l1_tsc_offset; in kvm_set_msr_common()
3936 vcpu->arch.ia32_tsc_adjust_msr += adj; in kvm_set_msr_common()
3940 if (!msr_info->host_initiated && in kvm_set_msr_common()
3950 vcpu->arch.ia32_xss = data; in kvm_set_msr_common()
3951 vcpu->arch.cpuid_dynamic_bits_dirty = true; in kvm_set_msr_common()
3954 if (!msr_info->host_initiated) in kvm_set_msr_common()
3956 vcpu->arch.smi_count = data; in kvm_set_msr_common()
3962 vcpu->kvm->arch.wall_clock = data; in kvm_set_msr_common()
3963 kvm_write_wall_clock(vcpu->kvm, data, 0); in kvm_set_msr_common()
3969 vcpu->kvm->arch.wall_clock = data; in kvm_set_msr_common()
3970 kvm_write_wall_clock(vcpu->kvm, data, 0); in kvm_set_msr_common()
3976 kvm_write_system_time(vcpu, data, false, msr_info->host_initiated); in kvm_set_msr_common()
3982 kvm_write_system_time(vcpu, data, true, msr_info->host_initiated); in kvm_set_msr_common()
4002 vcpu->arch.apf.pageready_pending = false; in kvm_set_msr_common()
4016 vcpu->arch.st.msr_val = data; in kvm_set_msr_common()
4037 if (data & (-1ULL << 1)) in kvm_set_msr_common()
4040 vcpu->arch.msr_kvm_poll_control = data; in kvm_set_msr_common()
4045 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in kvm_set_msr_common()
4046 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in kvm_set_msr_common()
4063 * all pre-dating SVM, but a recommended workaround from in kvm_set_msr_common()
4081 msr_info->host_initiated); in kvm_set_msr_common()
4084 /* Drop writes to this legacy MSR -- see rdmsr in kvm_set_msr_common()
4092 vcpu->arch.osvw.length = data; in kvm_set_msr_common()
4097 vcpu->arch.osvw.status = data; in kvm_set_msr_common()
4100 if (!msr_info->host_initiated) in kvm_set_msr_common()
4102 vcpu->arch.msr_platform_info = data; in kvm_set_msr_common()
4109 vcpu->arch.msr_misc_features_enables = data; in kvm_set_msr_common()
4113 if (!msr_info->host_initiated && in kvm_set_msr_common()
4120 fpu_update_guest_xfd(&vcpu->arch.guest_fpu, data); in kvm_set_msr_common()
4123 if (!msr_info->host_initiated && in kvm_set_msr_common()
4130 vcpu->arch.guest_fpu.xfd_err = data; in kvm_set_msr_common()
4146 u64 mcg_cap = vcpu->arch.mcg_cap; in get_msr_mce()
4156 data = vcpu->arch.mcg_cap; in get_msr_mce()
4161 data = vcpu->arch.mcg_ctl; in get_msr_mce()
4164 data = vcpu->arch.mcg_status; in get_msr_mce()
4166 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in get_msr_mce()
4167 last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1; in get_msr_mce()
4173 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2, in get_msr_mce()
4174 last_msr + 1 - MSR_IA32_MC0_CTL2); in get_msr_mce()
4175 data = vcpu->arch.mci_ctl2_banks[offset]; in get_msr_mce()
4177 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in get_msr_mce()
4178 last_msr = MSR_IA32_MCx_CTL(bank_num) - 1; in get_msr_mce()
4182 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL, in get_msr_mce()
4183 last_msr + 1 - MSR_IA32_MC0_CTL); in get_msr_mce()
4184 data = vcpu->arch.mce_banks[offset]; in get_msr_mce()
4195 switch (msr_info->index) { in kvm_get_msr_common()
4218 * so for existing CPU-specific MSRs. in kvm_get_msr_common()
4225 msr_info->data = 0; in kvm_get_msr_common()
4231 if (kvm_pmu_is_valid_msr(vcpu, msr_info->index)) in kvm_get_msr_common()
4233 msr_info->data = 0; in kvm_get_msr_common()
4236 msr_info->data = vcpu->arch.microcode_version; in kvm_get_msr_common()
4241 msr_info->data = vcpu->arch.arch_capabilities; in kvm_get_msr_common()
4246 msr_info->data = vcpu->arch.perf_capabilities; in kvm_get_msr_common()
4249 msr_info->data = vcpu->arch.msr_ia32_power_ctl; in kvm_get_msr_common()
4258 * return L1's TSC value to ensure backwards-compatible in kvm_get_msr_common()
4263 if (msr_info->host_initiated) { in kvm_get_msr_common()
4264 offset = vcpu->arch.l1_tsc_offset; in kvm_get_msr_common()
4265 ratio = vcpu->arch.l1_tsc_scaling_ratio; in kvm_get_msr_common()
4267 offset = vcpu->arch.tsc_offset; in kvm_get_msr_common()
4268 ratio = vcpu->arch.tsc_scaling_ratio; in kvm_get_msr_common()
4271 msr_info->data = kvm_scale_tsc(rdtsc(), ratio) + offset; in kvm_get_msr_common()
4275 msr_info->data = vcpu->arch.pat; in kvm_get_msr_common()
4280 return kvm_mtrr_get_msr(vcpu, msr_info->index, &msr_info->data); in kvm_get_msr_common()
4282 msr_info->data = 3; in kvm_get_msr_common()
4296 msr_info->data = 1 << 24; in kvm_get_msr_common()
4299 msr_info->data = vcpu->arch.apic_base; in kvm_get_msr_common()
4302 return kvm_x2apic_msr_read(vcpu, msr_info->index, &msr_info->data); in kvm_get_msr_common()
4304 msr_info->data = kvm_get_lapic_tscdeadline_msr(vcpu); in kvm_get_msr_common()
4307 msr_info->data = (u64)vcpu->arch.ia32_tsc_adjust_msr; in kvm_get_msr_common()
4310 msr_info->data = vcpu->arch.ia32_misc_enable_msr; in kvm_get_msr_common()
4313 if (!IS_ENABLED(CONFIG_KVM_SMM) || !msr_info->host_initiated) in kvm_get_msr_common()
4315 msr_info->data = vcpu->arch.smbase; in kvm_get_msr_common()
4318 msr_info->data = vcpu->arch.smi_count; in kvm_get_msr_common()
4322 msr_info->data = 1000ULL; in kvm_get_msr_common()
4324 msr_info->data |= (((uint64_t)4ULL) << 40); in kvm_get_msr_common()
4327 msr_info->data = vcpu->arch.efer; in kvm_get_msr_common()
4333 msr_info->data = vcpu->kvm->arch.wall_clock; in kvm_get_msr_common()
4339 msr_info->data = vcpu->kvm->arch.wall_clock; in kvm_get_msr_common()
4345 msr_info->data = vcpu->arch.time; in kvm_get_msr_common()
4351 msr_info->data = vcpu->arch.time; in kvm_get_msr_common()
4357 msr_info->data = vcpu->arch.apf.msr_en_val; in kvm_get_msr_common()
4363 msr_info->data = vcpu->arch.apf.msr_int_val; in kvm_get_msr_common()
4369 msr_info->data = 0; in kvm_get_msr_common()
4375 msr_info->data = vcpu->arch.st.msr_val; in kvm_get_msr_common()
4381 msr_info->data = vcpu->arch.pv_eoi.msr_val; in kvm_get_msr_common()
4387 msr_info->data = vcpu->arch.msr_kvm_poll_control; in kvm_get_msr_common()
4394 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in kvm_get_msr_common()
4395 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in kvm_get_msr_common()
4396 return get_msr_mce(vcpu, msr_info->index, &msr_info->data, in kvm_get_msr_common()
4397 msr_info->host_initiated); in kvm_get_msr_common()
4399 if (!msr_info->host_initiated && in kvm_get_msr_common()
4402 msr_info->data = vcpu->arch.ia32_xss; in kvm_get_msr_common()
4406 * Provide expected ramp-up count for K7. All other in kvm_get_msr_common()
4414 msr_info->data = 0x20000000; in kvm_get_msr_common()
4428 msr_info->index, &msr_info->data, in kvm_get_msr_common()
4429 msr_info->host_initiated); in kvm_get_msr_common()
4442 msr_info->data = 0xbe702111; in kvm_get_msr_common()
4447 msr_info->data = vcpu->arch.osvw.length; in kvm_get_msr_common()
4452 msr_info->data = vcpu->arch.osvw.status; in kvm_get_msr_common()
4455 if (!msr_info->host_initiated && in kvm_get_msr_common()
4456 !vcpu->kvm->arch.guest_can_read_msr_platform_info) in kvm_get_msr_common()
4458 msr_info->data = vcpu->arch.msr_platform_info; in kvm_get_msr_common()
4461 msr_info->data = vcpu->arch.msr_misc_features_enables; in kvm_get_msr_common()
4464 msr_info->data = vcpu->arch.msr_hwcr; in kvm_get_msr_common()
4468 if (!msr_info->host_initiated && in kvm_get_msr_common()
4472 msr_info->data = vcpu->arch.guest_fpu.fpstate->xfd; in kvm_get_msr_common()
4475 if (!msr_info->host_initiated && in kvm_get_msr_common()
4479 msr_info->data = vcpu->arch.guest_fpu.xfd_err; in kvm_get_msr_common()
4483 if (kvm_pmu_is_valid_msr(vcpu, msr_info->index)) in kvm_get_msr_common()
4504 for (i = 0; i < msrs->nmsrs; ++i) in __msr_io()
4526 r = -EFAULT; in msr_io()
4530 r = -E2BIG; in msr_io()
4535 entries = memdup_user(user_msrs->entries, size); in msr_io()
4543 if (writeback && copy_to_user(user_msrs->entries, entries, size)) in msr_io()
4544 r = -EFAULT; in msr_io()
4579 r = -EFAULT; in kvm_ioctl_get_supported_hv_cpuid()
4583 r = kvm_get_hv_cpuid(vcpu, &cpuid, cpuid_arg->entries); in kvm_ioctl_get_supported_hv_cpuid()
4587 r = -EFAULT; in kvm_ioctl_get_supported_hv_cpuid()
4602 return kvm && kvm->arch.has_protected_state ? 0 : KVM_SYNC_X86_VALID_FIELDS; in kvm_sync_valid_fields()
4761 r = kvm_x86_ops.nested_ops->get_state ? in kvm_vm_ioctl_check_extension()
4762 kvm_x86_ops.nested_ops->get_state(NULL, NULL, 0) : 0; in kvm_vm_ioctl_check_extension()
4769 r = kvm_x86_ops.nested_ops->enable_evmcs != NULL; in kvm_vm_ioctl_check_extension()
4814 if (attr->group) { in __kvm_x86_dev_get_attr()
4816 return kvm_x86_call(dev_get_attr)(attr->group, attr->attr, val); in __kvm_x86_dev_get_attr()
4817 return -ENXIO; in __kvm_x86_dev_get_attr()
4820 switch (attr->attr) { in __kvm_x86_dev_get_attr()
4825 return -ENXIO; in __kvm_x86_dev_get_attr()
4831 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_x86_dev_get_attr()
4840 return -EFAULT; in kvm_x86_dev_get_attr()
4864 r = -EFAULT; in kvm_arch_dev_ioctl()
4871 r = -E2BIG; in kvm_arch_dev_ioctl()
4874 r = -EFAULT; in kvm_arch_dev_ioctl()
4875 if (copy_to_user(user_msr_list->indices, &msrs_to_save, in kvm_arch_dev_ioctl()
4878 if (copy_to_user(user_msr_list->indices + num_msrs_to_save, in kvm_arch_dev_ioctl()
4890 r = -EFAULT; in kvm_arch_dev_ioctl()
4894 r = kvm_dev_ioctl_get_cpuid(&cpuid, cpuid_arg->entries, in kvm_arch_dev_ioctl()
4899 r = -EFAULT; in kvm_arch_dev_ioctl()
4906 r = -EFAULT; in kvm_arch_dev_ioctl()
4917 r = -EFAULT; in kvm_arch_dev_ioctl()
4924 r = -E2BIG; in kvm_arch_dev_ioctl()
4927 r = -EFAULT; in kvm_arch_dev_ioctl()
4928 if (copy_to_user(user_msr_list->indices, &msr_based_features, in kvm_arch_dev_ioctl()
4944 r = -EFAULT; in kvm_arch_dev_ioctl()
4952 r = -EFAULT; in kvm_arch_dev_ioctl()
4959 r = -EINVAL; in kvm_arch_dev_ioctl()
4973 return kvm_arch_has_noncoherent_dma(vcpu->kvm); in need_emulate_wbinvd()
4980 vcpu->arch.l1tf_flush_l1d = true; in kvm_arch_vcpu_load()
4982 if (vcpu->scheduled_out && pmu->version && pmu->event_count) { in kvm_arch_vcpu_load()
4983 pmu->need_cleanup = true; in kvm_arch_vcpu_load()
4990 cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_load()
4991 else if (vcpu->cpu != -1 && vcpu->cpu != cpu) in kvm_arch_vcpu_load()
4992 smp_call_function_single(vcpu->cpu, in kvm_arch_vcpu_load()
4999 vcpu->arch.host_pkru = read_pkru(); in kvm_arch_vcpu_load()
5002 if (unlikely(vcpu->arch.tsc_offset_adjustment)) { in kvm_arch_vcpu_load()
5003 adjust_tsc_offset_host(vcpu, vcpu->arch.tsc_offset_adjustment); in kvm_arch_vcpu_load()
5004 vcpu->arch.tsc_offset_adjustment = 0; in kvm_arch_vcpu_load()
5008 if (unlikely(vcpu->cpu != cpu) || kvm_check_tsc_unstable()) { in kvm_arch_vcpu_load()
5009 s64 tsc_delta = !vcpu->arch.last_host_tsc ? 0 : in kvm_arch_vcpu_load()
5010 rdtsc() - vcpu->arch.last_host_tsc; in kvm_arch_vcpu_load()
5016 vcpu->arch.last_guest_tsc); in kvm_arch_vcpu_load()
5018 if (!vcpu->arch.guest_tsc_protected) in kvm_arch_vcpu_load()
5019 vcpu->arch.tsc_catchup = 1; in kvm_arch_vcpu_load()
5027 * kvmclock on vcpu->cpu migration in kvm_arch_vcpu_load()
5029 if (!vcpu->kvm->arch.use_master_clock || vcpu->cpu == -1) in kvm_arch_vcpu_load()
5031 if (vcpu->cpu != cpu) in kvm_arch_vcpu_load()
5033 vcpu->cpu = cpu; in kvm_arch_vcpu_load()
5041 struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache; in kvm_steal_time_set_preempted()
5045 gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS; in kvm_steal_time_set_preempted()
5048 * The vCPU can be marked preempted if and only if the VM-Exit was on in kvm_steal_time_set_preempted()
5052 * preempted if and only if the VM-Exit was due to a host interrupt. in kvm_steal_time_set_preempted()
5054 if (!vcpu->arch.at_instruction_boundary) { in kvm_steal_time_set_preempted()
5055 vcpu->stat.preemption_other++; in kvm_steal_time_set_preempted()
5059 vcpu->stat.preemption_reported++; in kvm_steal_time_set_preempted()
5060 if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED)) in kvm_steal_time_set_preempted()
5063 if (vcpu->arch.st.preempted) in kvm_steal_time_set_preempted()
5067 if (unlikely(current->mm != vcpu->kvm->mm)) in kvm_steal_time_set_preempted()
5070 slots = kvm_memslots(vcpu->kvm); in kvm_steal_time_set_preempted()
5072 if (unlikely(slots->generation != ghc->generation || in kvm_steal_time_set_preempted()
5073 gpa != ghc->gpa || in kvm_steal_time_set_preempted()
5074 kvm_is_error_hva(ghc->hva) || !ghc->memslot)) in kvm_steal_time_set_preempted()
5077 st = (struct kvm_steal_time __user *)ghc->hva; in kvm_steal_time_set_preempted()
5078 BUILD_BUG_ON(sizeof(st->preempted) != sizeof(preempted)); in kvm_steal_time_set_preempted()
5080 if (!copy_to_user_nofault(&st->preempted, &preempted, sizeof(preempted))) in kvm_steal_time_set_preempted()
5081 vcpu->arch.st.preempted = KVM_VCPU_PREEMPTED; in kvm_steal_time_set_preempted()
5083 mark_page_dirty_in_slot(vcpu->kvm, ghc->memslot, gpa_to_gfn(ghc->gpa)); in kvm_steal_time_set_preempted()
5090 if (vcpu->preempted) { in kvm_arch_vcpu_put()
5092 * Assume protected guests are in-kernel. Inefficient yielding in kvm_arch_vcpu_put()
5096 vcpu->arch.preempted_in_kernel = vcpu->arch.guest_state_protected || in kvm_arch_vcpu_put()
5103 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_put()
5104 if (kvm_xen_msr_enabled(vcpu->kvm)) in kvm_arch_vcpu_put()
5108 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_put()
5112 vcpu->arch.last_host_tsc = rdtsc(); in kvm_arch_vcpu_put()
5159 * instruction boundary and with no events half-injected. in kvm_vcpu_ready_for_interrupt_injection()
5170 if (irq->irq >= KVM_NR_INTERRUPTS) in kvm_vcpu_ioctl_interrupt()
5171 return -EINVAL; in kvm_vcpu_ioctl_interrupt()
5173 if (!irqchip_in_kernel(vcpu->kvm)) { in kvm_vcpu_ioctl_interrupt()
5174 kvm_queue_interrupt(vcpu, irq->irq, false); in kvm_vcpu_ioctl_interrupt()
5180 * With in-kernel LAPIC, we only use this to inject EXTINT, so in kvm_vcpu_ioctl_interrupt()
5181 * fail for in-kernel 8259. in kvm_vcpu_ioctl_interrupt()
5183 if (pic_in_kernel(vcpu->kvm)) in kvm_vcpu_ioctl_interrupt()
5184 return -ENXIO; in kvm_vcpu_ioctl_interrupt()
5186 if (vcpu->arch.pending_external_vector != -1) in kvm_vcpu_ioctl_interrupt()
5187 return -EEXIST; in kvm_vcpu_ioctl_interrupt()
5189 vcpu->arch.pending_external_vector = irq->irq; in kvm_vcpu_ioctl_interrupt()
5204 if (tac->flags) in vcpu_ioctl_tpr_access_reporting()
5205 return -EINVAL; in vcpu_ioctl_tpr_access_reporting()
5206 vcpu->arch.tpr_access_reporting = !!tac->enabled; in vcpu_ioctl_tpr_access_reporting()
5216 r = -EINVAL; in kvm_vcpu_ioctl_x86_setup_mce()
5222 vcpu->arch.mcg_cap = mcg_cap; in kvm_vcpu_ioctl_x86_setup_mce()
5225 vcpu->arch.mcg_ctl = ~(u64)0; in kvm_vcpu_ioctl_x86_setup_mce()
5228 vcpu->arch.mce_banks[bank*4] = ~(u64)0; in kvm_vcpu_ioctl_x86_setup_mce()
5230 vcpu->arch.mci_ctl2_banks[bank] = 0; in kvm_vcpu_ioctl_x86_setup_mce()
5243 * - none of the bits for Machine Check Exceptions are set
5244 * - both the VAL (valid) and UC (uncorrectable) bits are set
5245 * MCI_STATUS_PCC - Processor Context Corrupted
5246 * MCI_STATUS_S - Signaled as a Machine Check Exception
5247 * MCI_STATUS_AR - Software recoverable Action Required
5251 return !mce->mcg_status && in is_ucna()
5252 !(mce->status & (MCI_STATUS_PCC | MCI_STATUS_S | MCI_STATUS_AR)) && in is_ucna()
5253 (mce->status & MCI_STATUS_VAL) && in is_ucna()
5254 (mce->status & MCI_STATUS_UC); in is_ucna()
5259 u64 mcg_cap = vcpu->arch.mcg_cap; in kvm_vcpu_x86_set_ucna()
5261 banks[1] = mce->status; in kvm_vcpu_x86_set_ucna()
5262 banks[2] = mce->addr; in kvm_vcpu_x86_set_ucna()
5263 banks[3] = mce->misc; in kvm_vcpu_x86_set_ucna()
5264 vcpu->arch.mcg_status = mce->mcg_status; in kvm_vcpu_x86_set_ucna()
5267 !(vcpu->arch.mci_ctl2_banks[mce->bank] & MCI_CTL2_CMCI_EN)) in kvm_vcpu_x86_set_ucna()
5271 kvm_apic_local_deliver(vcpu->arch.apic, APIC_LVTCMCI); in kvm_vcpu_x86_set_ucna()
5279 u64 mcg_cap = vcpu->arch.mcg_cap; in kvm_vcpu_ioctl_x86_set_mce()
5281 u64 *banks = vcpu->arch.mce_banks; in kvm_vcpu_ioctl_x86_set_mce()
5283 if (mce->bank >= bank_num || !(mce->status & MCI_STATUS_VAL)) in kvm_vcpu_ioctl_x86_set_mce()
5284 return -EINVAL; in kvm_vcpu_ioctl_x86_set_mce()
5286 banks += array_index_nospec(4 * mce->bank, 4 * bank_num); in kvm_vcpu_ioctl_x86_set_mce()
5295 if ((mce->status & MCI_STATUS_UC) && (mcg_cap & MCG_CTL_P) && in kvm_vcpu_ioctl_x86_set_mce()
5296 vcpu->arch.mcg_ctl != ~(u64)0) in kvm_vcpu_ioctl_x86_set_mce()
5302 if ((mce->status & MCI_STATUS_UC) && banks[0] != ~(u64)0) in kvm_vcpu_ioctl_x86_set_mce()
5304 if (mce->status & MCI_STATUS_UC) { in kvm_vcpu_ioctl_x86_set_mce()
5305 if ((vcpu->arch.mcg_status & MCG_STATUS_MCIP) || in kvm_vcpu_ioctl_x86_set_mce()
5311 mce->status |= MCI_STATUS_OVER; in kvm_vcpu_ioctl_x86_set_mce()
5312 banks[2] = mce->addr; in kvm_vcpu_ioctl_x86_set_mce()
5313 banks[3] = mce->misc; in kvm_vcpu_ioctl_x86_set_mce()
5314 vcpu->arch.mcg_status = mce->mcg_status; in kvm_vcpu_ioctl_x86_set_mce()
5315 banks[1] = mce->status; in kvm_vcpu_ioctl_x86_set_mce()
5320 mce->status |= MCI_STATUS_OVER; in kvm_vcpu_ioctl_x86_set_mce()
5321 banks[2] = mce->addr; in kvm_vcpu_ioctl_x86_set_mce()
5322 banks[3] = mce->misc; in kvm_vcpu_ioctl_x86_set_mce()
5323 banks[1] = mce->status; in kvm_vcpu_ioctl_x86_set_mce()
5344 * non-exiting _injected_ exception, and a pending exiting exception. in kvm_vcpu_ioctl_x86_get_vcpu_events()
5345 * In that case, ignore the VM-Exiting exception as it's an extension in kvm_vcpu_ioctl_x86_get_vcpu_events()
5348 if (vcpu->arch.exception_vmexit.pending && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5349 !vcpu->arch.exception.pending && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5350 !vcpu->arch.exception.injected) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5351 ex = &vcpu->arch.exception_vmexit; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5353 ex = &vcpu->arch.exception; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5358 * intercepts #PF, ditto for DR6 and #DBs. If the per-VM capability, in kvm_vcpu_ioctl_x86_get_vcpu_events()
5363 if (!vcpu->kvm->arch.exception_payload_enabled && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5364 ex->pending && ex->has_payload) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5375 if (!kvm_exception_is_soft(ex->vector)) { in kvm_vcpu_ioctl_x86_get_vcpu_events()
5376 events->exception.injected = ex->injected; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5377 events->exception.pending = ex->pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5383 if (!vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5384 events->exception.injected |= ex->pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5386 events->exception.nr = ex->vector; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5387 events->exception.has_error_code = ex->has_error_code; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5388 events->exception.error_code = ex->error_code; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5389 events->exception_has_payload = ex->has_payload; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5390 events->exception_payload = ex->payload; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5392 events->interrupt.injected = in kvm_vcpu_ioctl_x86_get_vcpu_events()
5393 vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5394 events->interrupt.nr = vcpu->arch.interrupt.nr; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5395 events->interrupt.shadow = kvm_x86_call(get_interrupt_shadow)(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5397 events->nmi.injected = vcpu->arch.nmi_injected; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5398 events->nmi.pending = kvm_get_nr_pending_nmis(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5399 events->nmi.masked = kvm_x86_call(get_nmi_mask)(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5401 /* events->sipi_vector is never valid when reporting to user space */ in kvm_vcpu_ioctl_x86_get_vcpu_events()
5404 events->smi.smm = is_smm(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5405 events->smi.pending = vcpu->arch.smi_pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5406 events->smi.smm_inside_nmi = in kvm_vcpu_ioctl_x86_get_vcpu_events()
5407 !!(vcpu->arch.hflags & HF_SMM_INSIDE_NMI_MASK); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5409 events->smi.latched_init = kvm_lapic_latched_init(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5411 events->flags = (KVM_VCPUEVENT_VALID_NMI_PENDING in kvm_vcpu_ioctl_x86_get_vcpu_events()
5414 if (vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5415 events->flags |= KVM_VCPUEVENT_VALID_PAYLOAD; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5416 if (vcpu->kvm->arch.triple_fault_event) { in kvm_vcpu_ioctl_x86_get_vcpu_events()
5417 events->triple_fault.pending = kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5418 events->flags |= KVM_VCPUEVENT_VALID_TRIPLE_FAULT; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5425 if (events->flags & ~(KVM_VCPUEVENT_VALID_NMI_PENDING in kvm_vcpu_ioctl_x86_set_vcpu_events()
5431 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5433 if (events->flags & KVM_VCPUEVENT_VALID_PAYLOAD) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5434 if (!vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5435 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5436 if (events->exception.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5437 events->exception.injected = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5439 events->exception_has_payload = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5441 events->exception.pending = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5442 events->exception_has_payload = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5445 if ((events->exception.injected || events->exception.pending) && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5446 (events->exception.nr > 31 || events->exception.nr == NMI_VECTOR)) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5447 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5450 if (events->flags & KVM_VCPUEVENT_VALID_SMM && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5451 (events->smi.smm || events->smi.pending) && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5452 vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5453 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5459 * morph the exception to a VM-Exit if appropriate. Do this only for in kvm_vcpu_ioctl_x86_set_vcpu_events()
5460 * pending exceptions, already-injected exceptions are not subject to in kvm_vcpu_ioctl_x86_set_vcpu_events()
5463 * pending exception, which in turn may cause a spurious VM-Exit. in kvm_vcpu_ioctl_x86_set_vcpu_events()
5465 vcpu->arch.exception_from_userspace = events->exception.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5467 vcpu->arch.exception_vmexit.pending = false; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5469 vcpu->arch.exception.injected = events->exception.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5470 vcpu->arch.exception.pending = events->exception.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5471 vcpu->arch.exception.vector = events->exception.nr; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5472 vcpu->arch.exception.has_error_code = events->exception.has_error_code; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5473 vcpu->arch.exception.error_code = events->exception.error_code; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5474 vcpu->arch.exception.has_payload = events->exception_has_payload; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5475 vcpu->arch.exception.payload = events->exception_payload; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5477 vcpu->arch.interrupt.injected = events->interrupt.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5478 vcpu->arch.interrupt.nr = events->interrupt.nr; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5479 vcpu->arch.interrupt.soft = events->interrupt.soft; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5480 if (events->flags & KVM_VCPUEVENT_VALID_SHADOW) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5482 events->interrupt.shadow); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5484 vcpu->arch.nmi_injected = events->nmi.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5485 if (events->flags & KVM_VCPUEVENT_VALID_NMI_PENDING) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5486 vcpu->arch.nmi_pending = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5487 atomic_set(&vcpu->arch.nmi_queued, events->nmi.pending); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5488 if (events->nmi.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5491 kvm_x86_call(set_nmi_mask)(vcpu, events->nmi.masked); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5493 if (events->flags & KVM_VCPUEVENT_VALID_SIPI_VECTOR && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5495 vcpu->arch.apic->sipi_vector = events->sipi_vector; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5497 if (events->flags & KVM_VCPUEVENT_VALID_SMM) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5499 if (!!(vcpu->arch.hflags & HF_SMM_MASK) != events->smi.smm) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5501 kvm_smm_changed(vcpu, events->smi.smm); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5504 vcpu->arch.smi_pending = events->smi.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5506 if (events->smi.smm) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5507 if (events->smi.smm_inside_nmi) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5508 vcpu->arch.hflags |= HF_SMM_INSIDE_NMI_MASK; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5510 vcpu->arch.hflags &= ~HF_SMM_INSIDE_NMI_MASK; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5514 if (events->smi.smm || events->smi.pending || in kvm_vcpu_ioctl_x86_set_vcpu_events()
5515 events->smi.smm_inside_nmi) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5516 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5520 if (events->smi.latched_init) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5521 set_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5523 clear_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5527 if (events->flags & KVM_VCPUEVENT_VALID_TRIPLE_FAULT) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5528 if (!vcpu->kvm->arch.triple_fault_event) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5529 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5530 if (events->triple_fault.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5546 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_get_debugregs()
5547 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_get_debugregs()
5548 return -EINVAL; in kvm_vcpu_ioctl_x86_get_debugregs()
5552 BUILD_BUG_ON(ARRAY_SIZE(vcpu->arch.db) != ARRAY_SIZE(dbgregs->db)); in kvm_vcpu_ioctl_x86_get_debugregs()
5553 for (i = 0; i < ARRAY_SIZE(vcpu->arch.db); i++) in kvm_vcpu_ioctl_x86_get_debugregs()
5554 dbgregs->db[i] = vcpu->arch.db[i]; in kvm_vcpu_ioctl_x86_get_debugregs()
5556 dbgregs->dr6 = vcpu->arch.dr6; in kvm_vcpu_ioctl_x86_get_debugregs()
5557 dbgregs->dr7 = vcpu->arch.dr7; in kvm_vcpu_ioctl_x86_get_debugregs()
5566 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_set_debugregs()
5567 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_set_debugregs()
5568 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5570 if (dbgregs->flags) in kvm_vcpu_ioctl_x86_set_debugregs()
5571 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5573 if (!kvm_dr6_valid(dbgregs->dr6)) in kvm_vcpu_ioctl_x86_set_debugregs()
5574 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5575 if (!kvm_dr7_valid(dbgregs->dr7)) in kvm_vcpu_ioctl_x86_set_debugregs()
5576 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5578 for (i = 0; i < ARRAY_SIZE(vcpu->arch.db); i++) in kvm_vcpu_ioctl_x86_set_debugregs()
5579 vcpu->arch.db[i] = dbgregs->db[i]; in kvm_vcpu_ioctl_x86_set_debugregs()
5582 vcpu->arch.dr6 = dbgregs->dr6; in kvm_vcpu_ioctl_x86_set_debugregs()
5583 vcpu->arch.dr7 = dbgregs->dr7; in kvm_vcpu_ioctl_x86_set_debugregs()
5605 u64 supported_xcr0 = vcpu->arch.guest_supported_xcr0 | in kvm_vcpu_ioctl_x86_get_xsave2()
5608 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_vcpu_ioctl_x86_get_xsave2()
5609 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_vcpu_ioctl_x86_get_xsave2()
5611 fpu_copy_guest_fpstate_to_uabi(&vcpu->arch.guest_fpu, state, size, in kvm_vcpu_ioctl_x86_get_xsave2()
5612 supported_xcr0, vcpu->arch.pkru); in kvm_vcpu_ioctl_x86_get_xsave2()
5619 return kvm_vcpu_ioctl_x86_get_xsave2(vcpu, (void *)guest_xsave->region, in kvm_vcpu_ioctl_x86_get_xsave()
5620 sizeof(guest_xsave->region)); in kvm_vcpu_ioctl_x86_get_xsave()
5626 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_vcpu_ioctl_x86_set_xsave()
5627 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_vcpu_ioctl_x86_set_xsave()
5629 return fpu_copy_uabi_to_guest_fpstate(&vcpu->arch.guest_fpu, in kvm_vcpu_ioctl_x86_set_xsave()
5630 guest_xsave->region, in kvm_vcpu_ioctl_x86_set_xsave()
5632 &vcpu->arch.pkru); in kvm_vcpu_ioctl_x86_set_xsave()
5638 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_get_xcrs()
5639 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_get_xcrs()
5640 return -EINVAL; in kvm_vcpu_ioctl_x86_get_xcrs()
5643 guest_xcrs->nr_xcrs = 0; in kvm_vcpu_ioctl_x86_get_xcrs()
5647 guest_xcrs->nr_xcrs = 1; in kvm_vcpu_ioctl_x86_get_xcrs()
5648 guest_xcrs->flags = 0; in kvm_vcpu_ioctl_x86_get_xcrs()
5649 guest_xcrs->xcrs[0].xcr = XCR_XFEATURE_ENABLED_MASK; in kvm_vcpu_ioctl_x86_get_xcrs()
5650 guest_xcrs->xcrs[0].value = vcpu->arch.xcr0; in kvm_vcpu_ioctl_x86_get_xcrs()
5659 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_set_xcrs()
5660 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_set_xcrs()
5661 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5664 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5666 if (guest_xcrs->nr_xcrs > KVM_MAX_XCRS || guest_xcrs->flags) in kvm_vcpu_ioctl_x86_set_xcrs()
5667 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5669 for (i = 0; i < guest_xcrs->nr_xcrs; i++) in kvm_vcpu_ioctl_x86_set_xcrs()
5671 if (guest_xcrs->xcrs[i].xcr == XCR_XFEATURE_ENABLED_MASK) { in kvm_vcpu_ioctl_x86_set_xcrs()
5673 guest_xcrs->xcrs[i].value); in kvm_vcpu_ioctl_x86_set_xcrs()
5677 r = -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5689 if (!vcpu->arch.pv_time.active) in kvm_set_guest_paused()
5690 return -EINVAL; in kvm_set_guest_paused()
5691 vcpu->arch.pvclock_set_guest_stopped_request = true; in kvm_set_guest_paused()
5701 switch (attr->attr) { in kvm_arch_tsc_has_attr()
5706 r = -ENXIO; in kvm_arch_tsc_has_attr()
5715 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_arch_tsc_get_attr()
5718 switch (attr->attr) { in kvm_arch_tsc_get_attr()
5720 r = -EFAULT; in kvm_arch_tsc_get_attr()
5721 if (put_user(vcpu->arch.l1_tsc_offset, uaddr)) in kvm_arch_tsc_get_attr()
5726 r = -ENXIO; in kvm_arch_tsc_get_attr()
5735 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_arch_tsc_set_attr()
5736 struct kvm *kvm = vcpu->kvm; in kvm_arch_tsc_set_attr()
5739 switch (attr->attr) { in kvm_arch_tsc_set_attr()
5745 r = -EFAULT; in kvm_arch_tsc_set_attr()
5749 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_arch_tsc_set_attr()
5751 matched = (vcpu->arch.virtual_tsc_khz && in kvm_arch_tsc_set_attr()
5752 kvm->arch.last_tsc_khz == vcpu->arch.virtual_tsc_khz && in kvm_arch_tsc_set_attr()
5753 kvm->arch.last_tsc_offset == offset); in kvm_arch_tsc_set_attr()
5755 tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio) + offset; in kvm_arch_tsc_set_attr()
5759 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_arch_tsc_set_attr()
5765 r = -ENXIO; in kvm_arch_tsc_set_attr()
5779 return -EFAULT; in kvm_vcpu_ioctl_device_attr()
5782 return -ENXIO; in kvm_vcpu_ioctl_device_attr()
5802 if (cap->flags) in kvm_vcpu_ioctl_enable_cap()
5803 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5805 switch (cap->cap) { in kvm_vcpu_ioctl_enable_cap()
5808 if (cap->args[0]) in kvm_vcpu_ioctl_enable_cap()
5809 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5813 if (!irqchip_in_kernel(vcpu->kvm)) in kvm_vcpu_ioctl_enable_cap()
5814 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5815 return kvm_hv_activate_synic(vcpu, cap->cap == in kvm_vcpu_ioctl_enable_cap()
5823 if (!kvm_x86_ops.nested_ops->enable_evmcs) in kvm_vcpu_ioctl_enable_cap()
5824 return -ENOTTY; in kvm_vcpu_ioctl_enable_cap()
5825 r = kvm_x86_ops.nested_ops->enable_evmcs(vcpu, &vmcs_version); in kvm_vcpu_ioctl_enable_cap()
5827 user_ptr = (void __user *)(uintptr_t)cap->args[0]; in kvm_vcpu_ioctl_enable_cap()
5830 r = -EFAULT; in kvm_vcpu_ioctl_enable_cap()
5836 return -ENOTTY; in kvm_vcpu_ioctl_enable_cap()
5841 return kvm_hv_set_enforce_cpuid(vcpu, cap->args[0]); in kvm_vcpu_ioctl_enable_cap()
5845 vcpu->arch.pv_cpuid.enforce = cap->args[0]; in kvm_vcpu_ioctl_enable_cap()
5848 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5855 struct kvm_vcpu *vcpu = filp->private_data; in kvm_arch_vcpu_ioctl()
5871 r = -EINVAL; in kvm_arch_vcpu_ioctl()
5876 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
5882 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5889 r = -EINVAL; in kvm_arch_vcpu_ioctl()
5904 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5922 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5925 r = kvm_vcpu_ioctl_set_cpuid(vcpu, &cpuid, cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5932 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5936 cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5943 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5947 cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5950 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5957 int idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
5959 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
5963 int idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
5965 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
5971 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5977 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5987 r = -EINVAL; in kvm_arch_vcpu_ioctl()
5990 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5993 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
5995 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6001 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6010 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6021 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6030 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6046 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6056 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6065 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6066 if (vcpu->arch.guest_fpu.uabi_size > sizeof(struct kvm_xsave)) in kvm_arch_vcpu_ioctl()
6070 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6078 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6085 int size = vcpu->arch.guest_fpu.uabi_size; in kvm_arch_vcpu_ioctl()
6098 int size = vcpu->arch.guest_fpu.uabi_size; in kvm_arch_vcpu_ioctl()
6101 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6109 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6119 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6127 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6147 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6163 r = vcpu->arch.virtual_tsc_khz; in kvm_arch_vcpu_ioctl()
6173 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6183 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6184 if (!kvm_x86_ops.nested_ops->get_state) in kvm_arch_vcpu_ioctl()
6187 BUILD_BUG_ON(sizeof(user_data_size) != sizeof(user_kvm_nested_state->size)); in kvm_arch_vcpu_ioctl()
6188 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6189 if (get_user(user_data_size, &user_kvm_nested_state->size)) in kvm_arch_vcpu_ioctl()
6192 r = kvm_x86_ops.nested_ops->get_state(vcpu, user_kvm_nested_state, in kvm_arch_vcpu_ioctl()
6198 if (put_user(r, &user_kvm_nested_state->size)) in kvm_arch_vcpu_ioctl()
6199 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6201 r = -E2BIG; in kvm_arch_vcpu_ioctl()
6213 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6214 if (!kvm_x86_ops.nested_ops->set_state) in kvm_arch_vcpu_ioctl()
6217 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6221 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6236 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
6237 r = kvm_x86_ops.nested_ops->set_state(vcpu, user_kvm_nested_state, &kvm_state); in kvm_arch_vcpu_ioctl()
6238 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6250 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6255 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6261 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6269 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6270 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl()
6271 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl()
6275 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6279 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6286 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6287 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl()
6288 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl()
6306 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6324 if (addr > (unsigned int)(-3 * PAGE_SIZE)) in kvm_vm_ioctl_set_tss_addr()
6325 return -EINVAL; in kvm_vm_ioctl_set_tss_addr()
6340 return -EINVAL; in kvm_vm_ioctl_set_nr_mmu_pages()
6342 mutex_lock(&kvm->slots_lock); in kvm_vm_ioctl_set_nr_mmu_pages()
6345 kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages; in kvm_vm_ioctl_set_nr_mmu_pages()
6347 mutex_unlock(&kvm->slots_lock); in kvm_vm_ioctl_set_nr_mmu_pages()
6353 struct kvm_pic *pic = kvm->arch.vpic; in kvm_vm_ioctl_get_irqchip()
6357 switch (chip->chip_id) { in kvm_vm_ioctl_get_irqchip()
6359 memcpy(&chip->chip.pic, &pic->pics[0], in kvm_vm_ioctl_get_irqchip()
6363 memcpy(&chip->chip.pic, &pic->pics[1], in kvm_vm_ioctl_get_irqchip()
6367 kvm_get_ioapic(kvm, &chip->chip.ioapic); in kvm_vm_ioctl_get_irqchip()
6370 r = -EINVAL; in kvm_vm_ioctl_get_irqchip()
6378 struct kvm_pic *pic = kvm->arch.vpic; in kvm_vm_ioctl_set_irqchip()
6382 switch (chip->chip_id) { in kvm_vm_ioctl_set_irqchip()
6384 spin_lock(&pic->lock); in kvm_vm_ioctl_set_irqchip()
6385 memcpy(&pic->pics[0], &chip->chip.pic, in kvm_vm_ioctl_set_irqchip()
6387 spin_unlock(&pic->lock); in kvm_vm_ioctl_set_irqchip()
6390 spin_lock(&pic->lock); in kvm_vm_ioctl_set_irqchip()
6391 memcpy(&pic->pics[1], &chip->chip.pic, in kvm_vm_ioctl_set_irqchip()
6393 spin_unlock(&pic->lock); in kvm_vm_ioctl_set_irqchip()
6396 kvm_set_ioapic(kvm, &chip->chip.ioapic); in kvm_vm_ioctl_set_irqchip()
6399 r = -EINVAL; in kvm_vm_ioctl_set_irqchip()
6408 struct kvm_kpit_state *kps = &kvm->arch.vpit->pit_state; in kvm_vm_ioctl_get_pit()
6410 BUILD_BUG_ON(sizeof(*ps) != sizeof(kps->channels)); in kvm_vm_ioctl_get_pit()
6412 mutex_lock(&kps->lock); in kvm_vm_ioctl_get_pit()
6413 memcpy(ps, &kps->channels, sizeof(*ps)); in kvm_vm_ioctl_get_pit()
6414 mutex_unlock(&kps->lock); in kvm_vm_ioctl_get_pit()
6421 struct kvm_pit *pit = kvm->arch.vpit; in kvm_vm_ioctl_set_pit()
6423 mutex_lock(&pit->pit_state.lock); in kvm_vm_ioctl_set_pit()
6424 memcpy(&pit->pit_state.channels, ps, sizeof(*ps)); in kvm_vm_ioctl_set_pit()
6426 kvm_pit_load_count(pit, i, ps->channels[i].count, 0); in kvm_vm_ioctl_set_pit()
6427 mutex_unlock(&pit->pit_state.lock); in kvm_vm_ioctl_set_pit()
6433 mutex_lock(&kvm->arch.vpit->pit_state.lock); in kvm_vm_ioctl_get_pit2()
6434 memcpy(ps->channels, &kvm->arch.vpit->pit_state.channels, in kvm_vm_ioctl_get_pit2()
6435 sizeof(ps->channels)); in kvm_vm_ioctl_get_pit2()
6436 ps->flags = kvm->arch.vpit->pit_state.flags; in kvm_vm_ioctl_get_pit2()
6437 mutex_unlock(&kvm->arch.vpit->pit_state.lock); in kvm_vm_ioctl_get_pit2()
6438 memset(&ps->reserved, 0, sizeof(ps->reserved)); in kvm_vm_ioctl_get_pit2()
6447 struct kvm_pit *pit = kvm->arch.vpit; in kvm_vm_ioctl_set_pit2()
6449 mutex_lock(&pit->pit_state.lock); in kvm_vm_ioctl_set_pit2()
6450 prev_legacy = pit->pit_state.flags & KVM_PIT_FLAGS_HPET_LEGACY; in kvm_vm_ioctl_set_pit2()
6451 cur_legacy = ps->flags & KVM_PIT_FLAGS_HPET_LEGACY; in kvm_vm_ioctl_set_pit2()
6454 memcpy(&pit->pit_state.channels, &ps->channels, in kvm_vm_ioctl_set_pit2()
6455 sizeof(pit->pit_state.channels)); in kvm_vm_ioctl_set_pit2()
6456 pit->pit_state.flags = ps->flags; in kvm_vm_ioctl_set_pit2()
6458 kvm_pit_load_count(pit, i, pit->pit_state.channels[i].count, in kvm_vm_ioctl_set_pit2()
6460 mutex_unlock(&pit->pit_state.lock); in kvm_vm_ioctl_set_pit2()
6467 struct kvm_pit *pit = kvm->arch.vpit; in kvm_vm_ioctl_reinject()
6469 /* pit->pit_state.lock was overloaded to prevent userspace from getting in kvm_vm_ioctl_reinject()
6473 mutex_lock(&pit->pit_state.lock); in kvm_vm_ioctl_reinject()
6474 kvm_pit_set_reinject(pit, control->pit_reinject); in kvm_vm_ioctl_reinject()
6475 mutex_unlock(&pit->pit_state.lock); in kvm_vm_ioctl_reinject()
6486 * on all VM-Exits, thus we only need to kick running vCPUs to force a in kvm_arch_sync_dirty_log()
6487 * VM-Exit. in kvm_arch_sync_dirty_log()
6503 return -ENXIO; in kvm_vm_ioctl_irq_line()
6505 irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, in kvm_vm_ioctl_irq_line()
6506 irq_event->irq, irq_event->level, in kvm_vm_ioctl_irq_line()
6516 if (cap->flags) in kvm_vm_ioctl_enable_cap()
6517 return -EINVAL; in kvm_vm_ioctl_enable_cap()
6519 switch (cap->cap) { in kvm_vm_ioctl_enable_cap()
6521 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6522 if (cap->args[0] & ~KVM_X86_VALID_QUIRKS) in kvm_vm_ioctl_enable_cap()
6526 kvm->arch.disabled_quirks = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6530 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6531 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6532 if (cap->args[0] > MAX_NR_RESERVED_IOAPIC_PINS) in kvm_vm_ioctl_enable_cap()
6534 r = -EEXIST; in kvm_vm_ioctl_enable_cap()
6537 if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6541 kvm->arch.irqchip_mode = KVM_IRQCHIP_SPLIT; in kvm_vm_ioctl_enable_cap()
6542 kvm->arch.nr_reserved_ioapic_pins = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6546 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6550 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6551 if (cap->args[0] & ~KVM_X2APIC_API_VALID_FLAGS) in kvm_vm_ioctl_enable_cap()
6554 if (cap->args[0] & KVM_X2APIC_API_USE_32BIT_IDS) in kvm_vm_ioctl_enable_cap()
6555 kvm->arch.x2apic_format = true; in kvm_vm_ioctl_enable_cap()
6556 if (cap->args[0] & KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK) in kvm_vm_ioctl_enable_cap()
6557 kvm->arch.x2apic_broadcast_quirk_disabled = true; in kvm_vm_ioctl_enable_cap()
6562 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6563 if (cap->args[0] & ~kvm_get_allowed_disable_exits()) in kvm_vm_ioctl_enable_cap()
6566 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6567 if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6570 #define SMT_RSB_MSG "This processor is affected by the Cross-Thread Return Predictions vulnerabilit… in kvm_vm_ioctl_enable_cap()
6575 (cap->args[0] & ~KVM_X86_DISABLE_EXITS_PAUSE)) in kvm_vm_ioctl_enable_cap()
6578 if (cap->args[0] & KVM_X86_DISABLE_EXITS_PAUSE) in kvm_vm_ioctl_enable_cap()
6579 kvm->arch.pause_in_guest = true; in kvm_vm_ioctl_enable_cap()
6580 if (cap->args[0] & KVM_X86_DISABLE_EXITS_MWAIT) in kvm_vm_ioctl_enable_cap()
6581 kvm->arch.mwait_in_guest = true; in kvm_vm_ioctl_enable_cap()
6582 if (cap->args[0] & KVM_X86_DISABLE_EXITS_HLT) in kvm_vm_ioctl_enable_cap()
6583 kvm->arch.hlt_in_guest = true; in kvm_vm_ioctl_enable_cap()
6584 if (cap->args[0] & KVM_X86_DISABLE_EXITS_CSTATE) in kvm_vm_ioctl_enable_cap()
6585 kvm->arch.cstate_in_guest = true; in kvm_vm_ioctl_enable_cap()
6588 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6591 kvm->arch.guest_can_read_msr_platform_info = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6595 kvm->arch.exception_payload_enabled = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6599 kvm->arch.triple_fault_event = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6603 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6604 if (cap->args[0] & ~KVM_MSR_EXIT_REASON_VALID_MASK) in kvm_vm_ioctl_enable_cap()
6606 kvm->arch.user_space_msr_mask = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6610 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6611 if (cap->args[0] & ~KVM_BUS_LOCK_DETECTION_VALID_MODE) in kvm_vm_ioctl_enable_cap()
6614 if ((cap->args[0] & KVM_BUS_LOCK_DETECTION_OFF) && in kvm_vm_ioctl_enable_cap()
6615 (cap->args[0] & KVM_BUS_LOCK_DETECTION_EXIT)) in kvm_vm_ioctl_enable_cap()
6619 cap->args[0] & KVM_BUS_LOCK_DETECTION_EXIT) in kvm_vm_ioctl_enable_cap()
6620 kvm->arch.bus_lock_detection_enabled = true; in kvm_vm_ioctl_enable_cap()
6627 r = sgx_set_attribute(&allowed_attributes, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6634 kvm->arch.sgx_provisioning_allowed = true; in kvm_vm_ioctl_enable_cap()
6636 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6641 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6645 r = kvm_x86_call(vm_copy_enc_context_from)(kvm, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6648 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6652 r = kvm_x86_call(vm_move_enc_context_from)(kvm, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6655 if (cap->args[0] & ~KVM_EXIT_HYPERCALL_VALID_MASK) { in kvm_vm_ioctl_enable_cap()
6656 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6659 kvm->arch.hypercall_exit_enabled = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6663 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6664 if (cap->args[0] & ~1) in kvm_vm_ioctl_enable_cap()
6666 kvm->arch.exit_on_emulation_error = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6670 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6671 if (!enable_pmu || (cap->args[0] & ~KVM_CAP_PMU_VALID_MASK)) in kvm_vm_ioctl_enable_cap()
6674 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6675 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6676 kvm->arch.enable_pmu = !(cap->args[0] & KVM_PMU_CAP_DISABLE); in kvm_vm_ioctl_enable_cap()
6679 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6682 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6683 if (cap->args[0] > KVM_MAX_VCPU_IDS) in kvm_vm_ioctl_enable_cap()
6686 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6687 if (kvm->arch.bsp_vcpu_id > cap->args[0]) { in kvm_vm_ioctl_enable_cap()
6689 } else if (kvm->arch.max_vcpu_ids == cap->args[0]) { in kvm_vm_ioctl_enable_cap()
6691 } else if (!kvm->arch.max_vcpu_ids) { in kvm_vm_ioctl_enable_cap()
6692 kvm->arch.max_vcpu_ids = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6695 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6698 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6699 if ((u32)cap->args[0] & ~KVM_X86_NOTIFY_VMEXIT_VALID_BITS) in kvm_vm_ioctl_enable_cap()
6703 if (!((u32)cap->args[0] & KVM_X86_NOTIFY_VMEXIT_ENABLED)) in kvm_vm_ioctl_enable_cap()
6705 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6706 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6707 kvm->arch.notify_window = cap->args[0] >> 32; in kvm_vm_ioctl_enable_cap()
6708 kvm->arch.notify_vmexit_flags = (u32)cap->args[0]; in kvm_vm_ioctl_enable_cap()
6711 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6714 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6728 r = -EPERM; in kvm_vm_ioctl_enable_cap()
6732 if (cap->args[0]) in kvm_vm_ioctl_enable_cap()
6735 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6736 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6737 kvm->arch.disable_nx_huge_pages = true; in kvm_vm_ioctl_enable_cap()
6740 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6743 u64 bus_cycle_ns = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6750 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6756 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6758 r = -ENXIO; in kvm_vm_ioctl_enable_cap()
6759 else if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6760 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6762 kvm->arch.apic_bus_cycle_ns = bus_cycle_ns; in kvm_vm_ioctl_enable_cap()
6763 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6767 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6781 msr_filter->default_allow = default_allow; in kvm_alloc_msr_filter()
6792 for (i = 0; i < msr_filter->count; i++) in kvm_free_msr_filter()
6793 kfree(msr_filter->ranges[i].bitmap); in kvm_free_msr_filter()
6804 if (!user_range->nmsrs) in kvm_add_msr_filter()
6807 if (user_range->flags & ~KVM_MSR_FILTER_RANGE_VALID_MASK) in kvm_add_msr_filter()
6808 return -EINVAL; in kvm_add_msr_filter()
6810 if (!user_range->flags) in kvm_add_msr_filter()
6811 return -EINVAL; in kvm_add_msr_filter()
6813 bitmap_size = BITS_TO_LONGS(user_range->nmsrs) * sizeof(long); in kvm_add_msr_filter()
6815 return -EINVAL; in kvm_add_msr_filter()
6817 bitmap = memdup_user((__user u8*)user_range->bitmap, bitmap_size); in kvm_add_msr_filter()
6821 msr_filter->ranges[msr_filter->count] = (struct msr_bitmap_range) { in kvm_add_msr_filter()
6822 .flags = user_range->flags, in kvm_add_msr_filter()
6823 .base = user_range->base, in kvm_add_msr_filter()
6824 .nmsrs = user_range->nmsrs, in kvm_add_msr_filter()
6828 msr_filter->count++; in kvm_add_msr_filter()
6841 if (filter->flags & ~KVM_MSR_FILTER_VALID_MASK) in kvm_vm_ioctl_set_msr_filter()
6842 return -EINVAL; in kvm_vm_ioctl_set_msr_filter()
6844 for (i = 0; i < ARRAY_SIZE(filter->ranges); i++) in kvm_vm_ioctl_set_msr_filter()
6845 empty &= !filter->ranges[i].nmsrs; in kvm_vm_ioctl_set_msr_filter()
6847 default_allow = !(filter->flags & KVM_MSR_FILTER_DEFAULT_DENY); in kvm_vm_ioctl_set_msr_filter()
6849 return -EINVAL; in kvm_vm_ioctl_set_msr_filter()
6853 return -ENOMEM; in kvm_vm_ioctl_set_msr_filter()
6855 for (i = 0; i < ARRAY_SIZE(filter->ranges); i++) { in kvm_vm_ioctl_set_msr_filter()
6856 r = kvm_add_msr_filter(new_filter, &filter->ranges[i]); in kvm_vm_ioctl_set_msr_filter()
6863 mutex_lock(&kvm->lock); in kvm_vm_ioctl_set_msr_filter()
6864 old_filter = rcu_replace_pointer(kvm->arch.msr_filter, new_filter, in kvm_vm_ioctl_set_msr_filter()
6865 mutex_is_locked(&kvm->lock)); in kvm_vm_ioctl_set_msr_filter()
6866 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_set_msr_filter()
6867 synchronize_srcu(&kvm->srcu); in kvm_vm_ioctl_set_msr_filter()
6896 struct kvm *kvm = filp->private_data; in kvm_arch_vm_compat_ioctl()
6897 long r = -ENOTTY; in kvm_arch_vm_compat_ioctl()
6908 return -EFAULT; in kvm_arch_vm_compat_ioctl()
6916 .flags = cr->flags, in kvm_arch_vm_compat_ioctl()
6917 .nmsrs = cr->nmsrs, in kvm_arch_vm_compat_ioctl()
6918 .base = cr->base, in kvm_arch_vm_compat_ioctl()
6919 .bitmap = (__u8 *)(ulong)cr->bitmap, in kvm_arch_vm_compat_ioctl()
6966 return -EFAULT; in kvm_vm_ioctl_get_clock()
6973 struct kvm_arch *ka = &kvm->arch; in kvm_vm_ioctl_set_clock()
6978 return -EFAULT; in kvm_vm_ioctl_set_clock()
6985 return -EINVAL; in kvm_vm_ioctl_set_clock()
7005 data.clock += now_real_ns - data.realtime; in kvm_vm_ioctl_set_clock()
7008 if (ka->use_master_clock) in kvm_vm_ioctl_set_clock()
7009 now_raw_ns = ka->master_kernel_ns; in kvm_vm_ioctl_set_clock()
7012 ka->kvmclock_offset = data.clock - now_raw_ns; in kvm_vm_ioctl_set_clock()
7019 struct kvm *kvm = filp->private_data; in kvm_arch_vm_ioctl()
7021 int r = -ENOTTY; in kvm_arch_vm_ioctl()
7023 * This union makes it completely explicit to gcc-3.x in kvm_arch_vm_ioctl()
7040 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7041 r = -EINVAL; in kvm_arch_vm_ioctl()
7042 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
7044 r = -EFAULT; in kvm_arch_vm_ioctl()
7049 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7056 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7058 r = -EEXIST; in kvm_arch_vm_ioctl()
7062 r = -EINVAL; in kvm_arch_vm_ioctl()
7063 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
7082 /* Write kvm->irq_routing before enabling irqchip_in_kernel. */ in kvm_arch_vm_ioctl()
7084 kvm->arch.irqchip_mode = KVM_IRQCHIP_KERNEL; in kvm_arch_vm_ioctl()
7087 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7094 r = -EFAULT; in kvm_arch_vm_ioctl()
7099 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7100 r = -EEXIST; in kvm_arch_vm_ioctl()
7101 if (kvm->arch.vpit) in kvm_arch_vm_ioctl()
7103 r = -ENOENT; in kvm_arch_vm_ioctl()
7106 r = -ENOMEM; in kvm_arch_vm_ioctl()
7107 kvm->arch.vpit = kvm_create_pit(kvm, u.pit_config.flags); in kvm_arch_vm_ioctl()
7108 if (kvm->arch.vpit) in kvm_arch_vm_ioctl()
7111 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7123 r = -ENXIO; in kvm_arch_vm_ioctl()
7129 r = -EFAULT; in kvm_arch_vm_ioctl()
7147 r = -ENXIO; in kvm_arch_vm_ioctl()
7156 r = -EFAULT; in kvm_arch_vm_ioctl()
7159 r = -ENXIO; in kvm_arch_vm_ioctl()
7160 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7165 r = -EFAULT; in kvm_arch_vm_ioctl()
7172 r = -EFAULT; in kvm_arch_vm_ioctl()
7175 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7176 r = -ENXIO; in kvm_arch_vm_ioctl()
7177 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7181 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7185 r = -ENXIO; in kvm_arch_vm_ioctl()
7186 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7191 r = -EFAULT; in kvm_arch_vm_ioctl()
7198 r = -EFAULT; in kvm_arch_vm_ioctl()
7201 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7202 r = -ENXIO; in kvm_arch_vm_ioctl()
7203 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7207 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7212 r = -EFAULT; in kvm_arch_vm_ioctl()
7215 r = -ENXIO; in kvm_arch_vm_ioctl()
7216 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7223 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7224 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
7225 r = -EBUSY; in kvm_arch_vm_ioctl()
7227 (kvm->arch.max_vcpu_ids && arg > kvm->arch.max_vcpu_ids)) in kvm_arch_vm_ioctl()
7228 r = -EINVAL; in kvm_arch_vm_ioctl()
7230 kvm->arch.bsp_vcpu_id = arg; in kvm_arch_vm_ioctl()
7231 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7236 r = -EFAULT; in kvm_arch_vm_ioctl()
7245 r = -EFAULT; in kvm_arch_vm_ioctl()
7250 r = -EFAULT; in kvm_arch_vm_ioctl()
7256 r = -EFAULT; in kvm_arch_vm_ioctl()
7265 r = -EFAULT; in kvm_arch_vm_ioctl()
7281 r = -EINVAL; in kvm_arch_vm_ioctl()
7291 WRITE_ONCE(kvm->arch.default_tsc_khz, user_tsc_khz); in kvm_arch_vm_ioctl()
7297 r = READ_ONCE(kvm->arch.default_tsc_khz); in kvm_arch_vm_ioctl()
7301 r = -ENOTTY; in kvm_arch_vm_ioctl()
7311 r = -EFAULT; in kvm_arch_vm_ioctl()
7315 r = -ENOTTY; in kvm_arch_vm_ioctl()
7325 r = -EFAULT; in kvm_arch_vm_ioctl()
7329 r = -ENOTTY; in kvm_arch_vm_ioctl()
7340 r = -EFAULT; in kvm_arch_vm_ioctl()
7355 return -EFAULT; in kvm_arch_vm_ioctl()
7361 r = -ENOTTY; in kvm_arch_vm_ioctl()
7421 (msr_index - MSR_IA32_RTIT_ADDR0_A >= in kvm_probe_msr_to_save()
7426 MSR_ARCH_PERFMON_PERFCTR0 + KVM_MAX_NR_GP_COUNTERS - 1: in kvm_probe_msr_to_save()
7427 if (msr_index - MSR_ARCH_PERFMON_PERFCTR0 >= in kvm_probe_msr_to_save()
7432 MSR_ARCH_PERFMON_EVENTSEL0 + KVM_MAX_NR_GP_COUNTERS - 1: in kvm_probe_msr_to_save()
7433 if (msr_index - MSR_ARCH_PERFMON_EVENTSEL0 >= in kvm_probe_msr_to_save()
7438 MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_MAX_NR_FIXED_COUNTERS - 1: in kvm_probe_msr_to_save()
7439 if (msr_index - MSR_ARCH_PERFMON_FIXED_CTR0 >= in kvm_probe_msr_to_save()
7508 !kvm_iodevice_write(vcpu, &vcpu->arch.apic->dev, addr, n, v)) in vcpu_mmio_write()
7513 len -= n; in vcpu_mmio_write()
7528 !kvm_iodevice_read(vcpu, &vcpu->arch.apic->dev, in vcpu_mmio_read()
7535 len -= n; in vcpu_mmio_read()
7557 struct kvm_mmu *mmu = vcpu->arch.mmu; in translate_nested_gpa()
7562 /* NPT walks are always user-walks */ in translate_nested_gpa()
7564 t_gpa = mmu->gva_to_gpa(vcpu, mmu, gpa, access, exception); in translate_nested_gpa()
7572 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_read()
7575 return mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in kvm_mmu_gva_to_gpa_read()
7582 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_write()
7586 return mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in kvm_mmu_gva_to_gpa_write()
7594 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_system()
7596 return mmu->gva_to_gpa(vcpu, mmu, gva, 0, exception); in kvm_mmu_gva_to_gpa_system()
7603 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_read_guest_virt_helper()
7608 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access, exception); in kvm_read_guest_virt_helper()
7609 unsigned offset = addr & (PAGE_SIZE-1); in kvm_read_guest_virt_helper()
7610 unsigned toread = min(bytes, (unsigned)PAGE_SIZE - offset); in kvm_read_guest_virt_helper()
7622 bytes -= toread; in kvm_read_guest_virt_helper()
7636 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_fetch_guest_virt()
7642 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access|PFERR_FETCH_MASK, in kvm_fetch_guest_virt()
7647 offset = addr & (PAGE_SIZE-1); in kvm_fetch_guest_virt()
7649 bytes = (unsigned)PAGE_SIZE - offset; in kvm_fetch_guest_virt()
7695 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_write_guest_virt_helper()
7700 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access, exception); in kvm_write_guest_virt_helper()
7701 unsigned offset = addr & (PAGE_SIZE-1); in kvm_write_guest_virt_helper()
7702 unsigned towrite = min(bytes, (unsigned)PAGE_SIZE - offset); in kvm_write_guest_virt_helper()
7713 bytes -= towrite; in kvm_write_guest_virt_helper()
7741 vcpu->arch.l1tf_flush_l1d = true; in kvm_write_guest_virt_system()
7801 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in vcpu_mmio_gva_to_gpa()
7811 !permission_fault(vcpu, vcpu->arch.walk_mmu, in vcpu_mmio_gva_to_gpa()
7812 vcpu->arch.mmio_access, 0, access))) { in vcpu_mmio_gva_to_gpa()
7813 *gpa = vcpu->arch.mmio_gfn << PAGE_SHIFT | in vcpu_mmio_gva_to_gpa()
7814 (gva & (PAGE_SIZE - 1)); in vcpu_mmio_gva_to_gpa()
7819 *gpa = mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in vcpu_mmio_gva_to_gpa()
7822 return -1; in vcpu_mmio_gva_to_gpa()
7853 if (vcpu->mmio_read_completed) { in read_prepare()
7855 vcpu->mmio_fragments[0].gpa, val); in read_prepare()
7856 vcpu->mmio_read_completed = 0; in read_prepare()
7891 struct kvm_mmio_fragment *frag = &vcpu->mmio_fragments[0]; in write_exit_mmio()
7893 memcpy(vcpu->run->mmio.data, frag->data, min(8u, frag->len)); in write_exit_mmio()
7919 bool write = ops->write; in emulator_read_write_onepage()
7921 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in emulator_read_write_onepage()
7930 if (ctxt->gpa_available && emulator_can_use_gpa(ctxt) && in emulator_read_write_onepage()
7931 (addr & ~PAGE_MASK) == (ctxt->gpa_val & ~PAGE_MASK)) { in emulator_read_write_onepage()
7932 gpa = ctxt->gpa_val; in emulator_read_write_onepage()
7940 if (!ret && ops->read_write_emulate(vcpu, gpa, val, bytes)) in emulator_read_write_onepage()
7946 handled = ops->read_write_mmio(vcpu, gpa, bytes, val); in emulator_read_write_onepage()
7951 bytes -= handled; in emulator_read_write_onepage()
7954 WARN_ON(vcpu->mmio_nr_fragments >= KVM_MAX_MMIO_FRAGMENTS); in emulator_read_write_onepage()
7955 frag = &vcpu->mmio_fragments[vcpu->mmio_nr_fragments++]; in emulator_read_write_onepage()
7956 frag->gpa = gpa; in emulator_read_write_onepage()
7957 frag->data = val; in emulator_read_write_onepage()
7958 frag->len = bytes; in emulator_read_write_onepage()
7972 if (ops->read_write_prepare && in emulator_read_write()
7973 ops->read_write_prepare(vcpu, val, bytes)) in emulator_read_write()
7976 vcpu->mmio_nr_fragments = 0; in emulator_read_write()
7979 if (((addr + bytes - 1) ^ addr) & PAGE_MASK) { in emulator_read_write()
7982 now = -addr & ~PAGE_MASK; in emulator_read_write()
7989 if (ctxt->mode != X86EMUL_MODE_PROT64) in emulator_read_write()
7992 bytes -= now; in emulator_read_write()
8000 if (!vcpu->mmio_nr_fragments) in emulator_read_write()
8003 gpa = vcpu->mmio_fragments[0].gpa; in emulator_read_write()
8005 vcpu->mmio_needed = 1; in emulator_read_write()
8006 vcpu->mmio_cur_fragment = 0; in emulator_read_write()
8008 vcpu->run->mmio.len = min(8u, vcpu->mmio_fragments[0].len); in emulator_read_write()
8009 vcpu->run->mmio.is_write = vcpu->mmio_is_write = ops->write; in emulator_read_write()
8010 vcpu->run->exit_reason = KVM_EXIT_MMIO; in emulator_read_write()
8011 vcpu->run->mmio.phys_addr = gpa; in emulator_read_write()
8013 return ops->read_write_exit_mmio(vcpu, gpa, val, bytes); in emulator_read_write()
8053 if (bytes > 8 || (bytes & (bytes - 1))) in emulator_cmpxchg_emulated()
8067 page_line_mask = ~(cache_line_size() - 1); in emulator_cmpxchg_emulated()
8071 if (((gpa + bytes - 1) & page_line_mask) != (gpa & page_line_mask)) in emulator_cmpxchg_emulated()
8129 WARN_ON_ONCE(vcpu->arch.pio.count); in emulator_pio_in_out()
8145 memset(data, 0, size * (count - i)); in emulator_pio_in_out()
8154 vcpu->arch.pio.port = port; in emulator_pio_in_out()
8155 vcpu->arch.pio.in = in; in emulator_pio_in_out()
8156 vcpu->arch.pio.count = count; in emulator_pio_in_out()
8157 vcpu->arch.pio.size = size; in emulator_pio_in_out()
8160 memset(vcpu->arch.pio_data, 0, size * count); in emulator_pio_in_out()
8162 memcpy(vcpu->arch.pio_data, data, size * count); in emulator_pio_in_out()
8164 vcpu->run->exit_reason = KVM_EXIT_IO; in emulator_pio_in_out()
8165 vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; in emulator_pio_in_out()
8166 vcpu->run->io.size = size; in emulator_pio_in_out()
8167 vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; in emulator_pio_in_out()
8168 vcpu->run->io.count = count; in emulator_pio_in_out()
8169 vcpu->run->io.port = port; in emulator_pio_in_out()
8185 int size = vcpu->arch.pio.size; in complete_emulator_pio_in()
8186 unsigned int count = vcpu->arch.pio.count; in complete_emulator_pio_in()
8187 memcpy(val, vcpu->arch.pio_data, size * count); in complete_emulator_pio_in()
8188 trace_kvm_pio(KVM_PIO_IN, vcpu->arch.pio.port, size, count, vcpu->arch.pio_data); in complete_emulator_pio_in()
8189 vcpu->arch.pio.count = 0; in complete_emulator_pio_in()
8197 if (vcpu->arch.pio.count) { in emulator_pio_in_emulated()
8245 cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); in kvm_emulate_wbinvd_noskip()
8246 on_each_cpu_mask(vcpu->arch.wbinvd_dirty_mask, in kvm_emulate_wbinvd_noskip()
8249 cpumask_clear(vcpu->arch.wbinvd_dirty_mask); in kvm_emulate_wbinvd_noskip()
8283 return (curr_cr & ~((1ULL << 32) - 1)) | new_val; in mk_cr_64()
8296 value = vcpu->arch.cr2; in emulator_get_cr()
8325 vcpu->arch.cr2 = val; in emulator_set_cr()
8338 res = -1; in emulator_set_cr()
8399 desc->type = var.type; in emulator_get_segment()
8400 desc->s = var.s; in emulator_get_segment()
8401 desc->dpl = var.dpl; in emulator_get_segment()
8402 desc->p = var.present; in emulator_get_segment()
8403 desc->avl = var.avl; in emulator_get_segment()
8404 desc->l = var.l; in emulator_get_segment()
8405 desc->d = var.db; in emulator_get_segment()
8406 desc->g = var.g; in emulator_get_segment()
8424 if (desc->g) in emulator_set_segment()
8426 var.type = desc->type; in emulator_set_segment()
8427 var.dpl = desc->dpl; in emulator_set_segment()
8428 var.db = desc->d; in emulator_set_segment()
8429 var.s = desc->s; in emulator_set_segment()
8430 var.l = desc->l; in emulator_set_segment()
8431 var.g = desc->g; in emulator_set_segment()
8432 var.avl = desc->avl; in emulator_set_segment()
8433 var.present = desc->p; in emulator_set_segment()
8506 emul_to_vcpu(ctxt)->arch.halt_request = 1; in emulator_halt()
8514 &ctxt->exception); in emulator_intercept()
8589 struct kvm *kvm = emul_to_vcpu(ctxt)->kvm; in emulator_vm_bugged()
8591 if (!kvm->vm_bugged) in emulator_vm_bugged()
8681 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in inject_emulated_exception()
8683 if (ctxt->exception.vector == PF_VECTOR) in inject_emulated_exception()
8684 kvm_inject_emulated_page_fault(vcpu, &ctxt->exception); in inject_emulated_exception()
8685 else if (ctxt->exception.error_code_valid) in inject_emulated_exception()
8686 kvm_queue_exception_e(vcpu, ctxt->exception.vector, in inject_emulated_exception()
8687 ctxt->exception.error_code); in inject_emulated_exception()
8689 kvm_queue_exception(vcpu, ctxt->exception.vector); in inject_emulated_exception()
8702 ctxt->vcpu = vcpu; in alloc_emulate_ctxt()
8703 ctxt->ops = &emulate_ops; in alloc_emulate_ctxt()
8704 vcpu->arch.emulate_ctxt = ctxt; in alloc_emulate_ctxt()
8711 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in init_emulate_ctxt()
8716 ctxt->gpa_available = false; in init_emulate_ctxt()
8717 ctxt->eflags = kvm_get_rflags(vcpu); in init_emulate_ctxt()
8718 ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0; in init_emulate_ctxt()
8720 ctxt->eip = kvm_rip_read(vcpu); in init_emulate_ctxt()
8721 ctxt->mode = (!is_protmode(vcpu)) ? X86EMUL_MODE_REAL : in init_emulate_ctxt()
8722 (ctxt->eflags & X86_EFLAGS_VM) ? X86EMUL_MODE_VM86 : in init_emulate_ctxt()
8726 ctxt->interruptibility = 0; in init_emulate_ctxt()
8727 ctxt->have_exception = false; in init_emulate_ctxt()
8728 ctxt->exception.vector = -1; in init_emulate_ctxt()
8729 ctxt->perm_ok = false; in init_emulate_ctxt()
8732 vcpu->arch.emulate_regs_need_sync_from_vcpu = false; in init_emulate_ctxt()
8737 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in kvm_inject_realmode_interrupt()
8742 ctxt->op_bytes = 2; in kvm_inject_realmode_interrupt()
8743 ctxt->ad_bytes = 2; in kvm_inject_realmode_interrupt()
8744 ctxt->_eip = ctxt->eip + inc_eip; in kvm_inject_realmode_interrupt()
8750 ctxt->eip = ctxt->_eip; in kvm_inject_realmode_interrupt()
8751 kvm_rip_write(vcpu, ctxt->eip); in kvm_inject_realmode_interrupt()
8752 kvm_set_rflags(vcpu, ctxt->eflags); in kvm_inject_realmode_interrupt()
8760 struct kvm_run *run = vcpu->run; in prepare_emulation_failure_exit()
8773 run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in prepare_emulation_failure_exit()
8774 run->emulation_failure.suberror = KVM_INTERNAL_ERROR_EMULATION; in prepare_emulation_failure_exit()
8786 run->emulation_failure.flags = 0; in prepare_emulation_failure_exit()
8789 BUILD_BUG_ON((sizeof(run->emulation_failure.insn_size) + in prepare_emulation_failure_exit()
8790 sizeof(run->emulation_failure.insn_bytes) != 16)); in prepare_emulation_failure_exit()
8792 run->emulation_failure.flags |= in prepare_emulation_failure_exit()
8794 run->emulation_failure.insn_size = insn_size; in prepare_emulation_failure_exit()
8795 memset(run->emulation_failure.insn_bytes, 0x90, in prepare_emulation_failure_exit()
8796 sizeof(run->emulation_failure.insn_bytes)); in prepare_emulation_failure_exit()
8797 memcpy(run->emulation_failure.insn_bytes, insn_bytes, insn_size); in prepare_emulation_failure_exit()
8800 memcpy(&run->internal.data[info_start], info, sizeof(info)); in prepare_emulation_failure_exit()
8801 memcpy(&run->internal.data[info_start + ARRAY_SIZE(info)], data, in prepare_emulation_failure_exit()
8804 run->emulation_failure.ndata = info_start + ARRAY_SIZE(info) + ndata; in prepare_emulation_failure_exit()
8809 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in prepare_emulation_ctxt_failure_exit()
8811 prepare_emulation_failure_exit(vcpu, NULL, 0, ctxt->fetch.data, in prepare_emulation_ctxt_failure_exit()
8812 ctxt->fetch.end - ctxt->fetch.data); in prepare_emulation_ctxt_failure_exit()
8831 struct kvm_run *run = vcpu->run; in kvm_prepare_event_vectoring_exit()
8838 run->internal.data[ndata++] = info2; in kvm_prepare_event_vectoring_exit()
8839 run->internal.data[ndata++] = reason; in kvm_prepare_event_vectoring_exit()
8840 run->internal.data[ndata++] = info1; in kvm_prepare_event_vectoring_exit()
8841 run->internal.data[ndata++] = gpa; in kvm_prepare_event_vectoring_exit()
8842 run->internal.data[ndata++] = vcpu->arch.last_vmentry_cpu; in kvm_prepare_event_vectoring_exit()
8844 run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in kvm_prepare_event_vectoring_exit()
8845 run->internal.suberror = KVM_INTERNAL_ERROR_DELIVERY_EV; in kvm_prepare_event_vectoring_exit()
8846 run->internal.ndata = ndata; in kvm_prepare_event_vectoring_exit()
8852 struct kvm *kvm = vcpu->kvm; in handle_emulation_failure()
8854 ++vcpu->stat.insn_emulation_fail; in handle_emulation_failure()
8862 if (kvm->arch.exit_on_emulation_error || in handle_emulation_failure()
8891 * a SPTE and write-protect the gfn to resolve the !PRESENT fault, and in kvm_unprotect_and_retry_on_failure()
8900 * table, unprotect the gfn (zap any relevant SPTEs) and re-enter the in kvm_unprotect_and_retry_on_failure()
8901 * guest to let the CPU re-execute the instruction in the hope that the in kvm_unprotect_and_retry_on_failure()
8936 struct kvm_run *kvm_run = vcpu->run; in kvm_vcpu_do_singlestep()
8938 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) { in kvm_vcpu_do_singlestep()
8939 kvm_run->debug.arch.dr6 = DR6_BS | DR6_ACTIVE_LOW; in kvm_vcpu_do_singlestep()
8940 kvm_run->debug.arch.pc = kvm_get_linear_rip(vcpu); in kvm_vcpu_do_singlestep()
8941 kvm_run->debug.arch.exception = DB_VECTOR; in kvm_vcpu_do_singlestep()
8942 kvm_run->exit_reason = KVM_EXIT_DEBUG; in kvm_vcpu_do_singlestep()
9012 if (unlikely(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) && in kvm_vcpu_check_code_breakpoint()
9013 (vcpu->arch.guest_debug_dr7 & DR7_BP_EN_MASK)) { in kvm_vcpu_check_code_breakpoint()
9014 struct kvm_run *kvm_run = vcpu->run; in kvm_vcpu_check_code_breakpoint()
9017 vcpu->arch.guest_debug_dr7, in kvm_vcpu_check_code_breakpoint()
9018 vcpu->arch.eff_db); in kvm_vcpu_check_code_breakpoint()
9021 kvm_run->debug.arch.dr6 = dr6 | DR6_ACTIVE_LOW; in kvm_vcpu_check_code_breakpoint()
9022 kvm_run->debug.arch.pc = eip; in kvm_vcpu_check_code_breakpoint()
9023 kvm_run->debug.arch.exception = DB_VECTOR; in kvm_vcpu_check_code_breakpoint()
9024 kvm_run->exit_reason = KVM_EXIT_DEBUG; in kvm_vcpu_check_code_breakpoint()
9030 if (unlikely(vcpu->arch.dr7 & DR7_BP_EN_MASK) && in kvm_vcpu_check_code_breakpoint()
9034 vcpu->arch.dr7, in kvm_vcpu_check_code_breakpoint()
9035 vcpu->arch.db); in kvm_vcpu_check_code_breakpoint()
9049 switch (ctxt->opcode_len) { in is_vmware_backdoor_opcode()
9051 switch (ctxt->b) { in is_vmware_backdoor_opcode()
9068 switch (ctxt->b) { in is_vmware_backdoor_opcode()
9081 * (and wrong) when emulating on an intercepted fault-like exception[*], as
9091 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in x86_decode_emulated_instruction()
9099 ++vcpu->stat.insn_emulation; in x86_decode_emulated_instruction()
9109 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in x86_emulate_instruction()
9135 vcpu->arch.l1tf_flush_l1d = true; in x86_emulate_instruction()
9142 * are fault-like and are higher priority than any faults on in x86_emulate_instruction()
9160 if (ctxt->have_exception && in x86_emulate_instruction()
9163 * #UD should result in just EMULATION_FAILED, and trap-like in x86_emulate_instruction()
9166 WARN_ON_ONCE(ctxt->exception.vector == UD_VECTOR || in x86_emulate_instruction()
9167 exception_type(ctxt->exception.vector) == EXCPT_TRAP); in x86_emulate_instruction()
9185 * injecting single-step #DBs. in x86_emulate_instruction()
9188 if (ctxt->mode != X86EMUL_MODE_PROT64) in x86_emulate_instruction()
9189 ctxt->eip = (u32)ctxt->_eip; in x86_emulate_instruction()
9191 ctxt->eip = ctxt->_eip; in x86_emulate_instruction()
9198 kvm_rip_write(vcpu, ctxt->eip); in x86_emulate_instruction()
9199 if (ctxt->eflags & X86_EFLAGS_RF) in x86_emulate_instruction()
9200 kvm_set_rflags(vcpu, ctxt->eflags & ~X86_EFLAGS_RF); in x86_emulate_instruction()
9205 * If emulation was caused by a write-protection #PF on a non-page_table in x86_emulate_instruction()
9217 if (vcpu->arch.emulate_regs_need_sync_from_vcpu) { in x86_emulate_instruction()
9218 vcpu->arch.emulate_regs_need_sync_from_vcpu = false; in x86_emulate_instruction()
9225 ctxt->exception.address = cr2_or_gpa; in x86_emulate_instruction()
9228 if (vcpu->arch.mmu->root_role.direct) { in x86_emulate_instruction()
9229 ctxt->gpa_available = true; in x86_emulate_instruction()
9230 ctxt->gpa_val = cr2_or_gpa; in x86_emulate_instruction()
9234 ctxt->exception.address = 0; in x86_emulate_instruction()
9250 if (ctxt->have_exception) { in x86_emulate_instruction()
9251 WARN_ON_ONCE(vcpu->mmio_needed && !vcpu->mmio_is_write); in x86_emulate_instruction()
9252 vcpu->mmio_needed = false; in x86_emulate_instruction()
9255 } else if (vcpu->arch.pio.count) { in x86_emulate_instruction()
9256 if (!vcpu->arch.pio.in) { in x86_emulate_instruction()
9257 /* FIXME: return into emulator if single-stepping. */ in x86_emulate_instruction()
9258 vcpu->arch.pio.count = 0; in x86_emulate_instruction()
9261 vcpu->arch.complete_userspace_io = complete_emulated_pio; in x86_emulate_instruction()
9264 } else if (vcpu->mmio_needed) { in x86_emulate_instruction()
9265 ++vcpu->stat.mmio_exits; in x86_emulate_instruction()
9267 if (!vcpu->mmio_is_write) in x86_emulate_instruction()
9270 vcpu->arch.complete_userspace_io = complete_emulated_mmio; in x86_emulate_instruction()
9271 } else if (vcpu->arch.complete_userspace_io) { in x86_emulate_instruction()
9282 toggle_interruptibility(vcpu, ctxt->interruptibility); in x86_emulate_instruction()
9283 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in x86_emulate_instruction()
9286 * Note, EXCPT_DB is assumed to be fault-like as the emulator in x86_emulate_instruction()
9288 * of which are fault-like. in x86_emulate_instruction()
9290 if (!ctxt->have_exception || in x86_emulate_instruction()
9291 exception_type(ctxt->exception.vector) == EXCPT_TRAP) { in x86_emulate_instruction()
9293 if (ctxt->is_branch) in x86_emulate_instruction()
9295 kvm_rip_write(vcpu, ctxt->eip); in x86_emulate_instruction()
9296 if (r && (ctxt->tf || (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP))) in x86_emulate_instruction()
9299 __kvm_set_rflags(vcpu, ctxt->eflags); in x86_emulate_instruction()
9308 if (unlikely((ctxt->eflags & ~rflags) & X86_EFLAGS_IF)) in x86_emulate_instruction()
9311 vcpu->arch.emulate_regs_need_sync_to_vcpu = true; in x86_emulate_instruction()
9331 vcpu->arch.pio.count = 0; in complete_fast_pio_out_port_0x7e()
9337 vcpu->arch.pio.count = 0; in complete_fast_pio_out()
9339 if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.pio.linear_rip))) in complete_fast_pio_out()
9359 kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_OUT_7E_INC_RIP)) { in kvm_fast_pio_out()
9360 vcpu->arch.complete_userspace_io = in kvm_fast_pio_out()
9364 vcpu->arch.pio.linear_rip = kvm_get_linear_rip(vcpu); in kvm_fast_pio_out()
9365 vcpu->arch.complete_userspace_io = complete_fast_pio_out; in kvm_fast_pio_out()
9375 BUG_ON(vcpu->arch.pio.count != 1); in complete_fast_pio_in()
9377 if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.pio.linear_rip))) { in complete_fast_pio_in()
9378 vcpu->arch.pio.count = 0; in complete_fast_pio_in()
9383 val = (vcpu->arch.pio.size < 4) ? kvm_rax_read(vcpu) : 0; in complete_fast_pio_in()
9406 vcpu->arch.pio.linear_rip = kvm_get_linear_rip(vcpu); in kvm_fast_pio_in()
9407 vcpu->arch.complete_userspace_io = complete_fast_pio_in; in kvm_fast_pio_in()
9438 khz = freq->new; in tsc_khz_changed()
9459 /* TSC frequency always matches when on Hyper-V */ in kvm_hyperv_tsc_notifier()
9527 if (vcpu->cpu != cpu) in __kvmclock_cpufreq_notifier()
9530 if (vcpu->cpu != raw_smp_processor_id()) in __kvmclock_cpufreq_notifier()
9536 if (freq->old < freq->new && send_ipi) { in __kvmclock_cpufreq_notifier()
9559 if (val == CPUFREQ_PRECHANGE && freq->old > freq->new) in kvmclock_cpufreq_notifier()
9561 if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new) in kvmclock_cpufreq_notifier()
9564 for_each_cpu(cpu, freq->policy->cpus) in kvmclock_cpufreq_notifier()
9592 if (policy->cpuinfo.max_freq) in kvm_timer_init()
9593 max_tsc_khz = policy->cpuinfo.max_freq; in kvm_timer_init()
9651 if (!gtod_is_based_on_tsc(gtod->clock.vclock_mode) && in pvclock_gtod_notify()
9664 memcpy(&kvm_x86_ops, ops->runtime_ops, sizeof(kvm_x86_ops)); in kvm_ops_update()
9674 #include <asm/kvm-x86-ops.h> in kvm_ops_update()
9677 kvm_pmu_ops_update(ops->pmu_ops); in kvm_ops_update()
9695 return -EIO; in kvm_x86_check_processor_compatibility()
9714 return -EEXIST; in kvm_x86_vendor_init()
9724 return -EOPNOTSUPP; in kvm_x86_vendor_init()
9729 return -EOPNOTSUPP; in kvm_x86_vendor_init()
9742 return -EIO; in kvm_x86_vendor_init()
9750 return -ENOMEM; in kvm_x86_vendor_init()
9756 r = -ENOMEM; in kvm_x86_vendor_init()
9778 kvm_init_pmu_capability(ops->pmu_ops); in kvm_x86_vendor_init()
9783 r = ops->hardware_setup(); in kvm_x86_vendor_init()
9802 if (pi_inject_timer == -1) in kvm_x86_vendor_init()
9811 kvm_register_perf_callbacks(ops->handle_intel_pt_intr); in kvm_x86_vendor_init()
9891 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9897 if (vcpu->arch.tsc_always_catchup) in kvm_pv_clock_pairing()
9898 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9901 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9910 if (kvm_write_guest(vcpu->kvm, paddr, &clock_pairing, in kvm_pv_clock_pairing()
9912 ret = -KVM_EFAULT; in kvm_pv_clock_pairing()
9921 * @apicid - apicid of vcpu to be kicked.
9941 return (READ_ONCE(kvm->arch.apicv_inhibit_reasons) == 0); in kvm_apicv_activated()
9947 ulong vm_reasons = READ_ONCE(vcpu->kvm->arch.apicv_inhibit_reasons); in kvm_vcpu_apicv_activated()
9975 set_or_clear_apicv_inhibit(&kvm->arch.apicv_inhibit_reasons, reason, true); in kvm_apicv_init()
9977 init_rwsem(&kvm->arch.apicv_update_lock); in kvm_apicv_init()
9985 vcpu->stat.directed_yield_attempted++; in kvm_sched_yield()
9991 map = rcu_dereference(vcpu->kvm->arch.apic_map); in kvm_sched_yield()
9993 if (likely(map) && dest_id <= map->max_apic_id && map->phys_map[dest_id]) in kvm_sched_yield()
9994 target = map->phys_map[dest_id]->vcpu; in kvm_sched_yield()
9998 if (!target || !READ_ONCE(target->ready)) in kvm_sched_yield()
10008 vcpu->stat.directed_yield_successful++; in kvm_sched_yield()
10016 u64 ret = vcpu->run->hypercall.ret; in complete_hypercall_exit()
10032 ++vcpu->stat.hypercalls; in ____kvm_emulate_hypercall()
10045 ret = -KVM_EPERM; in ____kvm_emulate_hypercall()
10049 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
10059 kvm_pv_kick_cpu_op(vcpu->kvm, a1); in ____kvm_emulate_hypercall()
10072 ret = kvm_pv_send_ipi(vcpu->kvm, a0, a1, a2, a3, op_64_bit); in ____kvm_emulate_hypercall()
10084 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
10085 if (!user_exit_on_hypercall(vcpu->kvm, KVM_HC_MAP_GPA_RANGE)) in ____kvm_emulate_hypercall()
10090 ret = -KVM_EINVAL; in ____kvm_emulate_hypercall()
10094 vcpu->run->exit_reason = KVM_EXIT_HYPERCALL; in ____kvm_emulate_hypercall()
10095 vcpu->run->hypercall.nr = KVM_HC_MAP_GPA_RANGE; in ____kvm_emulate_hypercall()
10097 * In principle this should have been -KVM_ENOSYS, but userspace (QEMU <=9.2) in ____kvm_emulate_hypercall()
10098 * assumed that vcpu->run->hypercall.ret is never changed by KVM and thus that in ____kvm_emulate_hypercall()
10100 * vcpu->run->hypercall.ret, ensuring that it is zero to not break QEMU. in ____kvm_emulate_hypercall()
10102 vcpu->run->hypercall.ret = 0; in ____kvm_emulate_hypercall()
10103 vcpu->run->hypercall.args[0] = gpa; in ____kvm_emulate_hypercall()
10104 vcpu->run->hypercall.args[1] = npages; in ____kvm_emulate_hypercall()
10105 vcpu->run->hypercall.args[2] = attrs; in ____kvm_emulate_hypercall()
10106 vcpu->run->hypercall.flags = 0; in ____kvm_emulate_hypercall()
10108 vcpu->run->hypercall.flags |= KVM_EXIT_HYPERCALL_LONG_MODE; in ____kvm_emulate_hypercall()
10110 WARN_ON_ONCE(vcpu->run->hypercall.flags & KVM_EXIT_HYPERCALL_MBZ); in ____kvm_emulate_hypercall()
10111 vcpu->arch.complete_userspace_io = complete_hypercall; in ____kvm_emulate_hypercall()
10115 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
10120 vcpu->run->hypercall.ret = ret; in ____kvm_emulate_hypercall()
10127 if (kvm_xen_hypercall_enabled(vcpu->kvm)) in kvm_emulate_hypercall()
10150 if (!kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_FIX_HYPERCALL_INSN)) { in emulator_fix_hypercall()
10151 ctxt->exception.error_code_valid = false; in emulator_fix_hypercall()
10152 ctxt->exception.vector = UD_VECTOR; in emulator_fix_hypercall()
10153 ctxt->have_exception = true; in emulator_fix_hypercall()
10160 &ctxt->exception); in emulator_fix_hypercall()
10165 return vcpu->run->request_interrupt_window && in dm_request_for_irq_injection()
10166 likely(!pic_in_kernel(vcpu->kvm)); in dm_request_for_irq_injection()
10169 /* Called within kvm->srcu read side. */
10172 struct kvm_run *kvm_run = vcpu->run; in post_kvm_run_save()
10174 kvm_run->if_flag = kvm_x86_call(get_if_flag)(vcpu); in post_kvm_run_save()
10175 kvm_run->cr8 = kvm_get_cr8(vcpu); in post_kvm_run_save()
10176 kvm_run->apic_base = vcpu->arch.apic_base; in post_kvm_run_save()
10178 kvm_run->ready_for_interrupt_injection = in post_kvm_run_save()
10179 pic_in_kernel(vcpu->kvm) || in post_kvm_run_save()
10183 kvm_run->flags |= KVM_RUN_X86_SMM; in post_kvm_run_save()
10185 kvm_run->flags |= KVM_RUN_X86_GUEST_MODE; in post_kvm_run_save()
10198 if (vcpu->arch.apic->apicv_active) in update_cr8_intercept()
10201 if (!vcpu->arch.apic->vapic_addr) in update_cr8_intercept()
10204 max_irr = -1; in update_cr8_intercept()
10206 if (max_irr != -1) in update_cr8_intercept()
10218 kvm_x86_ops.nested_ops->triple_fault(vcpu); in kvm_check_nested_events()
10222 return kvm_x86_ops.nested_ops->check_events(vcpu); in kvm_check_nested_events()
10231 * is injected as intercepted #PF VM-Exits for AMD's Paged Real Mode do in kvm_inject_exception()
10234 vcpu->arch.exception.has_error_code &= is_protmode(vcpu); in kvm_inject_exception()
10236 trace_kvm_inj_exception(vcpu->arch.exception.vector, in kvm_inject_exception()
10237 vcpu->arch.exception.has_error_code, in kvm_inject_exception()
10238 vcpu->arch.exception.error_code, in kvm_inject_exception()
10239 vcpu->arch.exception.injected); in kvm_inject_exception()
10249 * injected as part of a previous VM-Enter, but weren't successfully delivered
10250 * and need to be re-injected.
10255 * also be able to re-inject NMIs and IRQs in the middle of an instruction.
10256 * I.e. for exceptions and re-injected events, NOT invoking this on instruction
10261 * instruction boundaries for asynchronous events. However, because VM-Exits
10267 * But, if a VM-Exit occurs during instruction execution, and KVM does NOT skip
10290 * Process nested events first, as nested VM-Exit supersedes event in kvm_check_and_inject_events()
10291 * re-injection. If there's an event queued for re-injection, it will in kvm_check_and_inject_events()
10292 * be saved into the appropriate vmc{b,s}12 fields on nested VM-Exit. in kvm_check_and_inject_events()
10300 * Re-inject exceptions and events *especially* if immediate entry+exit in kvm_check_and_inject_events()
10304 * Don't re-inject an NMI or interrupt if there is a pending exception. in kvm_check_and_inject_events()
10313 * as the exception "occurred" before the exit to userspace. Trap-like in kvm_check_and_inject_events()
10315 * And while fault-like exceptions, e.g. #GP and #PF, are the lowest in kvm_check_and_inject_events()
10318 * Thus a pending fault-like exception means the fault occurred on the in kvm_check_and_inject_events()
10322 if (vcpu->arch.exception.injected) in kvm_check_and_inject_events()
10326 else if (vcpu->arch.nmi_injected) in kvm_check_and_inject_events()
10328 else if (vcpu->arch.interrupt.injected) in kvm_check_and_inject_events()
10332 * Exceptions that morph to VM-Exits are handled above, and pending in kvm_check_and_inject_events()
10333 * exceptions on top of injected exceptions that do not VM-Exit should in kvm_check_and_inject_events()
10336 WARN_ON_ONCE(vcpu->arch.exception.injected && in kvm_check_and_inject_events()
10337 vcpu->arch.exception.pending); in kvm_check_and_inject_events()
10341 * nested VM-Enter or event re-injection so that a different pending in kvm_check_and_inject_events()
10344 * Otherwise, continue processing events even if VM-Exit occurred. The in kvm_check_and_inject_events()
10345 * VM-Exit will have cleared exceptions that were meant for L2, but in kvm_check_and_inject_events()
10352 * A pending exception VM-Exit should either result in nested VM-Exit in kvm_check_and_inject_events()
10353 * or force an immediate re-entry and exit to/from L2, and exception in kvm_check_and_inject_events()
10354 * VM-Exits cannot be injected (flag should _never_ be set). in kvm_check_and_inject_events()
10356 WARN_ON_ONCE(vcpu->arch.exception_vmexit.injected || in kvm_check_and_inject_events()
10357 vcpu->arch.exception_vmexit.pending); in kvm_check_and_inject_events()
10361 * to re-inject a previous event. See above comments on re-injecting in kvm_check_and_inject_events()
10366 if (vcpu->arch.exception.pending) { in kvm_check_and_inject_events()
10368 * Fault-class exceptions, except #DBs, set RF=1 in the RFLAGS in kvm_check_and_inject_events()
10369 * value pushed on the stack. Trap-like exception and all #DBs in kvm_check_and_inject_events()
10370 * leave RF as-is (KVM follows Intel's behavior in this regard; in kvm_check_and_inject_events()
10375 * fault-like. They do _not_ set RF, a la code breakpoints. in kvm_check_and_inject_events()
10377 if (exception_type(vcpu->arch.exception.vector) == EXCPT_FAULT) in kvm_check_and_inject_events()
10381 if (vcpu->arch.exception.vector == DB_VECTOR) { in kvm_check_and_inject_events()
10382 kvm_deliver_exception_payload(vcpu, &vcpu->arch.exception); in kvm_check_and_inject_events()
10383 if (vcpu->arch.dr7 & DR7_GD) { in kvm_check_and_inject_events()
10384 vcpu->arch.dr7 &= ~DR7_GD; in kvm_check_and_inject_events()
10391 vcpu->arch.exception.pending = false; in kvm_check_and_inject_events()
10392 vcpu->arch.exception.injected = true; in kvm_check_and_inject_events()
10398 if (vcpu->guest_debug & KVM_GUESTDBG_BLOCKIRQ) in kvm_check_and_inject_events()
10403 * due to architectural conditions (e.g. IF=0) a window-open exit in kvm_check_and_inject_events()
10404 * will re-request KVM_REQ_EVENT. Sometimes however an event is pending in kvm_check_and_inject_events()
10410 * The kvm_x86_ops hooks communicate this by returning -EBUSY. in kvm_check_and_inject_events()
10413 if (vcpu->arch.smi_pending) { in kvm_check_and_inject_events()
10415 -EBUSY; in kvm_check_and_inject_events()
10419 vcpu->arch.smi_pending = false; in kvm_check_and_inject_events()
10420 ++vcpu->arch.smi_count; in kvm_check_and_inject_events()
10428 if (vcpu->arch.nmi_pending) { in kvm_check_and_inject_events()
10430 -EBUSY; in kvm_check_and_inject_events()
10434 --vcpu->arch.nmi_pending; in kvm_check_and_inject_events()
10435 vcpu->arch.nmi_injected = true; in kvm_check_and_inject_events()
10440 if (vcpu->arch.nmi_pending) in kvm_check_and_inject_events()
10446 -EBUSY; in kvm_check_and_inject_events()
10452 if (!WARN_ON_ONCE(irq == -1)) { in kvm_check_and_inject_events()
10463 kvm_x86_ops.nested_ops->has_events && in kvm_check_and_inject_events()
10464 kvm_x86_ops.nested_ops->has_events(vcpu, true)) in kvm_check_and_inject_events()
10469 * is done emulating and should only propagate the to-be-injected event in kvm_check_and_inject_events()
10471 * infinite loop as KVM will bail from VM-Enter to inject the pending in kvm_check_and_inject_events()
10479 WARN_ON_ONCE(vcpu->arch.exception.pending || in kvm_check_and_inject_events()
10480 vcpu->arch.exception_vmexit.pending); in kvm_check_and_inject_events()
10484 if (r == -EBUSY) { in kvm_check_and_inject_events()
10505 if (kvm_x86_call(get_nmi_mask)(vcpu) || vcpu->arch.nmi_injected) in process_nmi()
10512 * tracked in vcpu->arch.nmi_pending. in process_nmi()
10515 limit--; in process_nmi()
10517 vcpu->arch.nmi_pending += atomic_xchg(&vcpu->arch.nmi_queued, 0); in process_nmi()
10518 vcpu->arch.nmi_pending = min(vcpu->arch.nmi_pending, limit); in process_nmi()
10520 if (vcpu->arch.nmi_pending && in process_nmi()
10522 vcpu->arch.nmi_pending--; in process_nmi()
10524 if (vcpu->arch.nmi_pending) in process_nmi()
10531 return vcpu->arch.nmi_pending + in kvm_get_nr_pending_nmis()
10548 struct kvm_lapic *apic = vcpu->arch.apic; in __kvm_vcpu_update_apicv()
10554 down_read(&vcpu->kvm->arch.apicv_update_lock); in __kvm_vcpu_update_apicv()
10561 if (apic->apicv_active == activate) in __kvm_vcpu_update_apicv()
10564 apic->apicv_active = activate; in __kvm_vcpu_update_apicv()
10574 if (!apic->apicv_active) in __kvm_vcpu_update_apicv()
10579 up_read(&vcpu->kvm->arch.apicv_update_lock); in __kvm_vcpu_update_apicv()
10599 if (apic_x2apic_mode(vcpu->arch.apic) && in kvm_vcpu_update_apicv()
10611 lockdep_assert_held_write(&kvm->arch.apicv_update_lock); in __kvm_set_or_clear_apicv_inhibit()
10616 old = new = kvm->arch.apicv_inhibit_reasons; in __kvm_set_or_clear_apicv_inhibit()
10634 kvm->arch.apicv_inhibit_reasons = new; in __kvm_set_or_clear_apicv_inhibit()
10637 int idx = srcu_read_lock(&kvm->srcu); in __kvm_set_or_clear_apicv_inhibit()
10640 srcu_read_unlock(&kvm->srcu, idx); in __kvm_set_or_clear_apicv_inhibit()
10643 kvm->arch.apicv_inhibit_reasons = new; in __kvm_set_or_clear_apicv_inhibit()
10653 down_write(&kvm->arch.apicv_update_lock); in kvm_set_or_clear_apicv_inhibit()
10655 up_write(&kvm->arch.apicv_update_lock); in kvm_set_or_clear_apicv_inhibit()
10664 bitmap_zero(vcpu->arch.ioapic_handled_vectors, 256); in vcpu_scan_ioapic()
10668 if (irqchip_split(vcpu->kvm)) in vcpu_scan_ioapic()
10669 kvm_scan_ioapic_routes(vcpu, vcpu->arch.ioapic_handled_vectors); in vcpu_scan_ioapic()
10670 else if (ioapic_in_kernel(vcpu->kvm)) in vcpu_scan_ioapic()
10671 kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors); in vcpu_scan_ioapic()
10674 vcpu->arch.load_eoi_exitmap_pending = true; in vcpu_scan_ioapic()
10681 if (!kvm_apic_hw_enabled(vcpu->arch.apic)) in vcpu_load_eoi_exitmap()
10689 vcpu->arch.ioapic_handled_vectors, in vcpu_load_eoi_exitmap()
10690 to_hv_synic(vcpu)->vec_bitmap, 256); in vcpu_load_eoi_exitmap()
10696 vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors); in vcpu_load_eoi_exitmap()
10713 * Called within kvm->srcu read side.
10730 r = -EIO; in vcpu_enter_guest()
10740 if (unlikely(!kvm_x86_ops.nested_ops->get_nested_state_pages(vcpu))) { in vcpu_enter_guest()
10750 kvm_update_masterclock(vcpu->kvm); in vcpu_enter_guest()
10774 * Fall back to a "full" guest flush if Hyper-V's precise in vcpu_enter_guest()
10775 * flushing fails. Note, Hyper-V's flushing is per-vCPU, but in vcpu_enter_guest()
10786 vcpu->run->exit_reason = KVM_EXIT_TPR_ACCESS; in vcpu_enter_guest()
10792 kvm_x86_ops.nested_ops->triple_fault(vcpu); in vcpu_enter_guest()
10795 vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; in vcpu_enter_guest()
10796 vcpu->mmio_needed = 0; in vcpu_enter_guest()
10803 vcpu->arch.apf.halted = true; in vcpu_enter_guest()
10820 BUG_ON(vcpu->arch.pending_ioapic_eoi > 255); in vcpu_enter_guest()
10821 if (test_bit(vcpu->arch.pending_ioapic_eoi, in vcpu_enter_guest()
10822 vcpu->arch.ioapic_handled_vectors)) { in vcpu_enter_guest()
10823 vcpu->run->exit_reason = KVM_EXIT_IOAPIC_EOI; in vcpu_enter_guest()
10824 vcpu->run->eoi.vector = in vcpu_enter_guest()
10825 vcpu->arch.pending_ioapic_eoi; in vcpu_enter_guest()
10838 vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT; in vcpu_enter_guest()
10839 vcpu->run->system_event.type = KVM_SYSTEM_EVENT_CRASH; in vcpu_enter_guest()
10840 vcpu->run->system_event.ndata = 0; in vcpu_enter_guest()
10845 vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT; in vcpu_enter_guest()
10846 vcpu->run->system_event.type = KVM_SYSTEM_EVENT_RESET; in vcpu_enter_guest()
10847 vcpu->run->system_event.ndata = 0; in vcpu_enter_guest()
10854 vcpu->run->exit_reason = KVM_EXIT_HYPERV; in vcpu_enter_guest()
10855 vcpu->run->hyperv = hv_vcpu->exit; in vcpu_enter_guest()
10862 * KVM_REQ_CLOCK_UPDATE, because Hyper-V SynIC timers in vcpu_enter_guest()
10863 * depend on the guest clock being up-to-date in vcpu_enter_guest()
10880 if (vcpu->arch.mp_state != KVM_MP_STATE_RUNNABLE) { in vcpu_enter_guest()
10889 ++vcpu->stat.req_event; in vcpu_enter_guest()
10895 if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) { in vcpu_enter_guest()
10930 /* Store vcpu->apicv_active before vcpu->mode. */ in vcpu_enter_guest()
10931 smp_store_release(&vcpu->mode, IN_GUEST_MODE); in vcpu_enter_guest()
10936 * 1) We should set ->mode before checking ->requests. Please see in vcpu_enter_guest()
10939 * 2) For APICv, we should set ->mode before checking PID.ON. This in vcpu_enter_guest()
10960 vcpu->mode = OUTSIDE_GUEST_MODE; in vcpu_enter_guest()
10976 if (vcpu->arch.guest_fpu.xfd_err) in vcpu_enter_guest()
10977 wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); in vcpu_enter_guest()
10979 if (unlikely(vcpu->arch.switch_db_regs)) { in vcpu_enter_guest()
10981 set_debugreg(vcpu->arch.eff_db[0], 0); in vcpu_enter_guest()
10982 set_debugreg(vcpu->arch.eff_db[1], 1); in vcpu_enter_guest()
10983 set_debugreg(vcpu->arch.eff_db[2], 2); in vcpu_enter_guest()
10984 set_debugreg(vcpu->arch.eff_db[3], 3); in vcpu_enter_guest()
10986 if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) in vcpu_enter_guest()
10987 kvm_x86_call(set_dr6)(vcpu, vcpu->arch.dr6); in vcpu_enter_guest()
10992 vcpu->arch.host_debugctl = get_debugctlmsr(); in vcpu_enter_guest()
11000 * per-VM state, and responding vCPUs must wait for the update in vcpu_enter_guest()
11019 /* Note, VM-Exits that go down the "slow" path are accounted below. */ in vcpu_enter_guest()
11020 ++vcpu->stat.exits; in vcpu_enter_guest()
11029 if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { in vcpu_enter_guest()
11030 WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); in vcpu_enter_guest()
11046 vcpu->arch.last_vmentry_cpu = vcpu->cpu; in vcpu_enter_guest()
11047 vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc()); in vcpu_enter_guest()
11049 vcpu->mode = OUTSIDE_GUEST_MODE; in vcpu_enter_guest()
11054 * rely on the fact that guest_fpu::xfd is up-to-date (e.g. in vcpu_enter_guest()
11057 if (vcpu->arch.xfd_no_write_intercept) in vcpu_enter_guest()
11062 if (vcpu->arch.guest_fpu.xfd_err) in vcpu_enter_guest()
11067 * VM-Exit on SVM and any ticks that occur between VM-Exit and now. in vcpu_enter_guest()
11074 ++vcpu->stat.exits; in vcpu_enter_guest()
11106 if (unlikely(vcpu->arch.tsc_always_catchup)) in vcpu_enter_guest()
11109 if (vcpu->arch.apic_attention) in vcpu_enter_guest()
11122 if (unlikely(vcpu->arch.apic_attention)) in vcpu_enter_guest()
11130 return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE && in kvm_vcpu_running()
11131 !vcpu->arch.apf.halted); in kvm_vcpu_running()
11136 if (!list_empty_careful(&vcpu->async_pf.done)) in kvm_vcpu_has_events()
11143 if (vcpu->arch.pv.pv_unhalted) in kvm_vcpu_has_events()
11150 (vcpu->arch.nmi_pending && in kvm_vcpu_has_events()
11156 (vcpu->arch.smi_pending && in kvm_vcpu_has_events()
11174 kvm_x86_ops.nested_ops->has_events && in kvm_vcpu_has_events()
11175 kvm_x86_ops.nested_ops->has_events(vcpu, false)) in kvm_vcpu_has_events()
11189 /* Called within kvm->srcu read side. */
11196 * Switch to the software timer before halt-polling/blocking as in vcpu_block()
11199 * Switch before halt-polling so that KVM recognizes an expired in vcpu_block()
11207 if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED) in vcpu_block()
11228 * state field (AMD does not have a similar field and a VM-Exit always in vcpu_block()
11234 WARN_ON_ONCE(r == -EBUSY); in vcpu_block()
11241 switch(vcpu->arch.mp_state) { in vcpu_block()
11247 vcpu->arch.apf.halted = false; in vcpu_block()
11258 /* Called within kvm->srcu read side. */
11263 vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; in vcpu_run()
11272 vcpu->arch.at_instruction_boundary = false; in vcpu_run()
11292 vcpu->run->exit_reason = KVM_EXIT_IRQ_WINDOW_OPEN; in vcpu_run()
11293 ++vcpu->stat.request_irq_exits; in vcpu_run()
11313 * local APIC is in-kernel, the run loop will detect the non-runnable in __kvm_emulate_halt()
11318 ++vcpu->stat.halt_exits; in __kvm_emulate_halt()
11325 vcpu->run->exit_reason = reason; in __kvm_emulate_halt()
11340 * TODO: we might be squashing a GUESTDBG_SINGLESTEP-triggered in kvm_emulate_halt()
11382 return vcpu->arch.preempted_in_kernel; in kvm_arch_vcpu_preempted_in_kernel()
11387 if (READ_ONCE(vcpu->arch.pv.pv_unhalted)) in kvm_arch_dy_runnable()
11407 BUG_ON(!vcpu->arch.pio.count); in complete_emulated_pio()
11432 struct kvm_run *run = vcpu->run; in complete_emulated_mmio()
11436 BUG_ON(!vcpu->mmio_needed); in complete_emulated_mmio()
11439 frag = &vcpu->mmio_fragments[vcpu->mmio_cur_fragment]; in complete_emulated_mmio()
11440 len = min(8u, frag->len); in complete_emulated_mmio()
11441 if (!vcpu->mmio_is_write) in complete_emulated_mmio()
11442 memcpy(frag->data, run->mmio.data, len); in complete_emulated_mmio()
11444 if (frag->len <= 8) { in complete_emulated_mmio()
11447 vcpu->mmio_cur_fragment++; in complete_emulated_mmio()
11450 frag->data += len; in complete_emulated_mmio()
11451 frag->gpa += len; in complete_emulated_mmio()
11452 frag->len -= len; in complete_emulated_mmio()
11455 if (vcpu->mmio_cur_fragment >= vcpu->mmio_nr_fragments) { in complete_emulated_mmio()
11456 vcpu->mmio_needed = 0; in complete_emulated_mmio()
11458 /* FIXME: return into emulator if single-stepping. */ in complete_emulated_mmio()
11459 if (vcpu->mmio_is_write) in complete_emulated_mmio()
11461 vcpu->mmio_read_completed = 1; in complete_emulated_mmio()
11465 run->exit_reason = KVM_EXIT_MMIO; in complete_emulated_mmio()
11466 run->mmio.phys_addr = frag->gpa; in complete_emulated_mmio()
11467 if (vcpu->mmio_is_write) in complete_emulated_mmio()
11468 memcpy(run->mmio.data, frag->data, min(8u, frag->len)); in complete_emulated_mmio()
11469 run->mmio.len = min(8u, frag->len); in complete_emulated_mmio()
11470 run->mmio.is_write = vcpu->mmio_is_write; in complete_emulated_mmio()
11471 vcpu->arch.complete_userspace_io = complete_emulated_mmio; in complete_emulated_mmio()
11478 /* Exclude PKRU, it's restored separately immediately after VM-Exit. */ in kvm_load_guest_fpu()
11479 fpu_swap_kvm_fpstate(&vcpu->arch.guest_fpu, true); in kvm_load_guest_fpu()
11486 fpu_swap_kvm_fpstate(&vcpu->arch.guest_fpu, false); in kvm_put_guest_fpu()
11487 ++vcpu->stat.fpu_reload; in kvm_put_guest_fpu()
11493 struct kvm_queued_exception *ex = &vcpu->arch.exception; in kvm_arch_vcpu_ioctl_run()
11494 struct kvm_run *kvm_run = vcpu->run; in kvm_arch_vcpu_ioctl_run()
11498 r = kvm_mmu_post_init_vm(vcpu->kvm); in kvm_arch_vcpu_ioctl_run()
11504 kvm_run->flags = 0; in kvm_arch_vcpu_ioctl_run()
11508 if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) { in kvm_arch_vcpu_ioctl_run()
11509 if (!vcpu->wants_to_run) { in kvm_arch_vcpu_ioctl_run()
11510 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11532 r = -EAGAIN; in kvm_arch_vcpu_ioctl_run()
11534 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11535 kvm_run->exit_reason = KVM_EXIT_INTR; in kvm_arch_vcpu_ioctl_run()
11536 ++vcpu->stat.signal_exits; in kvm_arch_vcpu_ioctl_run()
11541 sync_valid_fields = kvm_sync_valid_fields(vcpu->kvm); in kvm_arch_vcpu_ioctl_run()
11542 if ((kvm_run->kvm_valid_regs & ~sync_valid_fields) || in kvm_arch_vcpu_ioctl_run()
11543 (kvm_run->kvm_dirty_regs & ~sync_valid_fields)) { in kvm_arch_vcpu_ioctl_run()
11544 r = -EINVAL; in kvm_arch_vcpu_ioctl_run()
11548 if (kvm_run->kvm_dirty_regs) { in kvm_arch_vcpu_ioctl_run()
11554 /* re-sync apic's tpr */ in kvm_arch_vcpu_ioctl_run()
11556 if (kvm_set_cr8(vcpu, kvm_run->cr8) != 0) { in kvm_arch_vcpu_ioctl_run()
11557 r = -EINVAL; in kvm_arch_vcpu_ioctl_run()
11564 * a pending VM-Exit if L1 wants to intercept the exception. in kvm_arch_vcpu_ioctl_run()
11566 if (vcpu->arch.exception_from_userspace && is_guest_mode(vcpu) && in kvm_arch_vcpu_ioctl_run()
11567 kvm_x86_ops.nested_ops->is_exception_vmexit(vcpu, ex->vector, in kvm_arch_vcpu_ioctl_run()
11568 ex->error_code)) { in kvm_arch_vcpu_ioctl_run()
11569 kvm_queue_exception_vmexit(vcpu, ex->vector, in kvm_arch_vcpu_ioctl_run()
11570 ex->has_error_code, ex->error_code, in kvm_arch_vcpu_ioctl_run()
11571 ex->has_payload, ex->payload); in kvm_arch_vcpu_ioctl_run()
11572 ex->injected = false; in kvm_arch_vcpu_ioctl_run()
11573 ex->pending = false; in kvm_arch_vcpu_ioctl_run()
11575 vcpu->arch.exception_from_userspace = false; in kvm_arch_vcpu_ioctl_run()
11577 if (unlikely(vcpu->arch.complete_userspace_io)) { in kvm_arch_vcpu_ioctl_run()
11578 int (*cui)(struct kvm_vcpu *) = vcpu->arch.complete_userspace_io; in kvm_arch_vcpu_ioctl_run()
11579 vcpu->arch.complete_userspace_io = NULL; in kvm_arch_vcpu_ioctl_run()
11584 WARN_ON_ONCE(vcpu->arch.pio.count); in kvm_arch_vcpu_ioctl_run()
11585 WARN_ON_ONCE(vcpu->mmio_needed); in kvm_arch_vcpu_ioctl_run()
11588 if (!vcpu->wants_to_run) { in kvm_arch_vcpu_ioctl_run()
11589 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11601 if (kvm_run->kvm_valid_regs && likely(!vcpu->arch.guest_state_protected)) in kvm_arch_vcpu_ioctl_run()
11613 if (vcpu->arch.emulate_regs_need_sync_to_vcpu) { in __get_regs()
11621 emulator_writeback_register_cache(vcpu->arch.emulate_ctxt); in __get_regs()
11622 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in __get_regs()
11624 regs->rax = kvm_rax_read(vcpu); in __get_regs()
11625 regs->rbx = kvm_rbx_read(vcpu); in __get_regs()
11626 regs->rcx = kvm_rcx_read(vcpu); in __get_regs()
11627 regs->rdx = kvm_rdx_read(vcpu); in __get_regs()
11628 regs->rsi = kvm_rsi_read(vcpu); in __get_regs()
11629 regs->rdi = kvm_rdi_read(vcpu); in __get_regs()
11630 regs->rsp = kvm_rsp_read(vcpu); in __get_regs()
11631 regs->rbp = kvm_rbp_read(vcpu); in __get_regs()
11633 regs->r8 = kvm_r8_read(vcpu); in __get_regs()
11634 regs->r9 = kvm_r9_read(vcpu); in __get_regs()
11635 regs->r10 = kvm_r10_read(vcpu); in __get_regs()
11636 regs->r11 = kvm_r11_read(vcpu); in __get_regs()
11637 regs->r12 = kvm_r12_read(vcpu); in __get_regs()
11638 regs->r13 = kvm_r13_read(vcpu); in __get_regs()
11639 regs->r14 = kvm_r14_read(vcpu); in __get_regs()
11640 regs->r15 = kvm_r15_read(vcpu); in __get_regs()
11643 regs->rip = kvm_rip_read(vcpu); in __get_regs()
11644 regs->rflags = kvm_get_rflags(vcpu); in __get_regs()
11649 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_get_regs()
11650 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_get_regs()
11651 return -EINVAL; in kvm_arch_vcpu_ioctl_get_regs()
11661 vcpu->arch.emulate_regs_need_sync_from_vcpu = true; in __set_regs()
11662 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in __set_regs()
11664 kvm_rax_write(vcpu, regs->rax); in __set_regs()
11665 kvm_rbx_write(vcpu, regs->rbx); in __set_regs()
11666 kvm_rcx_write(vcpu, regs->rcx); in __set_regs()
11667 kvm_rdx_write(vcpu, regs->rdx); in __set_regs()
11668 kvm_rsi_write(vcpu, regs->rsi); in __set_regs()
11669 kvm_rdi_write(vcpu, regs->rdi); in __set_regs()
11670 kvm_rsp_write(vcpu, regs->rsp); in __set_regs()
11671 kvm_rbp_write(vcpu, regs->rbp); in __set_regs()
11673 kvm_r8_write(vcpu, regs->r8); in __set_regs()
11674 kvm_r9_write(vcpu, regs->r9); in __set_regs()
11675 kvm_r10_write(vcpu, regs->r10); in __set_regs()
11676 kvm_r11_write(vcpu, regs->r11); in __set_regs()
11677 kvm_r12_write(vcpu, regs->r12); in __set_regs()
11678 kvm_r13_write(vcpu, regs->r13); in __set_regs()
11679 kvm_r14_write(vcpu, regs->r14); in __set_regs()
11680 kvm_r15_write(vcpu, regs->r15); in __set_regs()
11683 kvm_rip_write(vcpu, regs->rip); in __set_regs()
11684 kvm_set_rflags(vcpu, regs->rflags | X86_EFLAGS_FIXED); in __set_regs()
11686 vcpu->arch.exception.pending = false; in __set_regs()
11687 vcpu->arch.exception_vmexit.pending = false; in __set_regs()
11694 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_set_regs()
11695 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_regs()
11696 return -EINVAL; in kvm_arch_vcpu_ioctl_set_regs()
11708 if (vcpu->arch.guest_state_protected) in __get_sregs_common()
11711 kvm_get_segment(vcpu, &sregs->cs, VCPU_SREG_CS); in __get_sregs_common()
11712 kvm_get_segment(vcpu, &sregs->ds, VCPU_SREG_DS); in __get_sregs_common()
11713 kvm_get_segment(vcpu, &sregs->es, VCPU_SREG_ES); in __get_sregs_common()
11714 kvm_get_segment(vcpu, &sregs->fs, VCPU_SREG_FS); in __get_sregs_common()
11715 kvm_get_segment(vcpu, &sregs->gs, VCPU_SREG_GS); in __get_sregs_common()
11716 kvm_get_segment(vcpu, &sregs->ss, VCPU_SREG_SS); in __get_sregs_common()
11718 kvm_get_segment(vcpu, &sregs->tr, VCPU_SREG_TR); in __get_sregs_common()
11719 kvm_get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); in __get_sregs_common()
11722 sregs->idt.limit = dt.size; in __get_sregs_common()
11723 sregs->idt.base = dt.address; in __get_sregs_common()
11725 sregs->gdt.limit = dt.size; in __get_sregs_common()
11726 sregs->gdt.base = dt.address; in __get_sregs_common()
11728 sregs->cr2 = vcpu->arch.cr2; in __get_sregs_common()
11729 sregs->cr3 = kvm_read_cr3(vcpu); in __get_sregs_common()
11732 sregs->cr0 = kvm_read_cr0(vcpu); in __get_sregs_common()
11733 sregs->cr4 = kvm_read_cr4(vcpu); in __get_sregs_common()
11734 sregs->cr8 = kvm_get_cr8(vcpu); in __get_sregs_common()
11735 sregs->efer = vcpu->arch.efer; in __get_sregs_common()
11736 sregs->apic_base = vcpu->arch.apic_base; in __get_sregs_common()
11743 if (vcpu->arch.guest_state_protected) in __get_sregs()
11746 if (vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft) in __get_sregs()
11747 set_bit(vcpu->arch.interrupt.nr, in __get_sregs()
11748 (unsigned long *)sregs->interrupt_bitmap); in __get_sregs()
11757 if (vcpu->arch.guest_state_protected) in __get_sregs2()
11762 sregs2->pdptrs[i] = kvm_pdptr_read(vcpu, i); in __get_sregs2()
11763 sregs2->flags |= KVM_SREGS2_FLAGS_PDPTRS_VALID; in __get_sregs2()
11770 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_get_sregs()
11771 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_get_sregs()
11772 return -EINVAL; in kvm_arch_vcpu_ioctl_get_sregs()
11796 if ((vcpu->arch.mp_state == KVM_MP_STATE_HALTED || in kvm_arch_vcpu_ioctl_get_mpstate()
11797 vcpu->arch.mp_state == KVM_MP_STATE_AP_RESET_HOLD) && in kvm_arch_vcpu_ioctl_get_mpstate()
11798 vcpu->arch.pv.pv_unhalted) in kvm_arch_vcpu_ioctl_get_mpstate()
11799 mp_state->mp_state = KVM_MP_STATE_RUNNABLE; in kvm_arch_vcpu_ioctl_get_mpstate()
11801 mp_state->mp_state = vcpu->arch.mp_state; in kvm_arch_vcpu_ioctl_get_mpstate()
11815 int ret = -EINVAL; in kvm_arch_vcpu_ioctl_set_mpstate()
11819 switch (mp_state->mp_state) { in kvm_arch_vcpu_ioctl_set_mpstate()
11842 if ((!kvm_apic_init_sipi_allowed(vcpu) || vcpu->arch.smi_pending) && in kvm_arch_vcpu_ioctl_set_mpstate()
11843 (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED || in kvm_arch_vcpu_ioctl_set_mpstate()
11844 mp_state->mp_state == KVM_MP_STATE_INIT_RECEIVED)) in kvm_arch_vcpu_ioctl_set_mpstate()
11847 if (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED) { in kvm_arch_vcpu_ioctl_set_mpstate()
11849 set_bit(KVM_APIC_SIPI, &vcpu->arch.apic->pending_events); in kvm_arch_vcpu_ioctl_set_mpstate()
11851 kvm_set_mp_state(vcpu, mp_state->mp_state); in kvm_arch_vcpu_ioctl_set_mpstate()
11863 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in kvm_task_switch()
11875 if (ret || vcpu->mmio_needed) { in kvm_task_switch()
11876 vcpu->mmio_needed = false; in kvm_task_switch()
11877 vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in kvm_task_switch()
11878 vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; in kvm_task_switch()
11879 vcpu->run->internal.ndata = 0; in kvm_task_switch()
11883 kvm_rip_write(vcpu, ctxt->eip); in kvm_task_switch()
11884 kvm_set_rflags(vcpu, ctxt->eflags); in kvm_task_switch()
11891 if ((sregs->efer & EFER_LME) && (sregs->cr0 & X86_CR0_PG)) { in kvm_is_valid_sregs()
11894 * 64-bit mode (though maybe in a 32-bit code segment). in kvm_is_valid_sregs()
11897 if (!(sregs->cr4 & X86_CR4_PAE) || !(sregs->efer & EFER_LMA)) in kvm_is_valid_sregs()
11899 if (!kvm_vcpu_is_legal_cr3(vcpu, sregs->cr3)) in kvm_is_valid_sregs()
11903 * Not in 64-bit mode: EFER.LMA is clear and the code in kvm_is_valid_sregs()
11904 * segment cannot be 64-bit. in kvm_is_valid_sregs()
11906 if (sregs->efer & EFER_LMA || sregs->cs.l) in kvm_is_valid_sregs()
11910 return kvm_is_valid_cr4(vcpu, sregs->cr4) && in kvm_is_valid_sregs()
11911 kvm_is_valid_cr0(vcpu, sregs->cr0); in kvm_is_valid_sregs()
11921 return -EINVAL; in __set_sregs_common()
11923 if (kvm_apic_set_base(vcpu, sregs->apic_base, true)) in __set_sregs_common()
11924 return -EINVAL; in __set_sregs_common()
11926 if (vcpu->arch.guest_state_protected) in __set_sregs_common()
11929 dt.size = sregs->idt.limit; in __set_sregs_common()
11930 dt.address = sregs->idt.base; in __set_sregs_common()
11932 dt.size = sregs->gdt.limit; in __set_sregs_common()
11933 dt.address = sregs->gdt.base; in __set_sregs_common()
11936 vcpu->arch.cr2 = sregs->cr2; in __set_sregs_common()
11937 *mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3; in __set_sregs_common()
11938 vcpu->arch.cr3 = sregs->cr3; in __set_sregs_common()
11940 kvm_x86_call(post_set_cr3)(vcpu, sregs->cr3); in __set_sregs_common()
11942 kvm_set_cr8(vcpu, sregs->cr8); in __set_sregs_common()
11944 *mmu_reset_needed |= vcpu->arch.efer != sregs->efer; in __set_sregs_common()
11945 kvm_x86_call(set_efer)(vcpu, sregs->efer); in __set_sregs_common()
11947 *mmu_reset_needed |= kvm_read_cr0(vcpu) != sregs->cr0; in __set_sregs_common()
11948 kvm_x86_call(set_cr0)(vcpu, sregs->cr0); in __set_sregs_common()
11950 *mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4; in __set_sregs_common()
11951 kvm_x86_call(set_cr4)(vcpu, sregs->cr4); in __set_sregs_common()
11954 idx = srcu_read_lock(&vcpu->kvm->srcu); in __set_sregs_common()
11959 srcu_read_unlock(&vcpu->kvm->srcu, idx); in __set_sregs_common()
11962 kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); in __set_sregs_common()
11963 kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS); in __set_sregs_common()
11964 kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES); in __set_sregs_common()
11965 kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS); in __set_sregs_common()
11966 kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS); in __set_sregs_common()
11967 kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS); in __set_sregs_common()
11969 kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); in __set_sregs_common()
11970 kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); in __set_sregs_common()
11976 sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 && in __set_sregs_common()
11999 (const unsigned long *)sregs->interrupt_bitmap, max_bits); in __set_sregs()
12012 bool valid_pdptrs = sregs2->flags & KVM_SREGS2_FLAGS_PDPTRS_VALID; in __set_sregs2()
12013 bool pae = (sregs2->cr0 & X86_CR0_PG) && (sregs2->cr4 & X86_CR4_PAE) && in __set_sregs2()
12014 !(sregs2->efer & EFER_LMA); in __set_sregs2()
12017 if (sregs2->flags & ~KVM_SREGS2_FLAGS_PDPTRS_VALID) in __set_sregs2()
12018 return -EINVAL; in __set_sregs2()
12020 if (valid_pdptrs && (!pae || vcpu->arch.guest_state_protected)) in __set_sregs2()
12021 return -EINVAL; in __set_sregs2()
12030 kvm_pdptr_write(vcpu, i, sregs2->pdptrs[i]); in __set_sregs2()
12034 vcpu->arch.pdptrs_from_userspace = true; in __set_sregs2()
12048 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_set_sregs()
12049 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_sregs()
12050 return -EINVAL; in kvm_arch_vcpu_ioctl_set_sregs()
12067 down_write(&kvm->arch.apicv_update_lock); in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12070 if (vcpu->guest_debug & KVM_GUESTDBG_BLOCKIRQ) { in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12076 up_write(&kvm->arch.apicv_update_lock); in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12085 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_guest_debug()
12086 return -EINVAL; in kvm_arch_vcpu_ioctl_set_guest_debug()
12090 if (dbg->control & (KVM_GUESTDBG_INJECT_DB | KVM_GUESTDBG_INJECT_BP)) { in kvm_arch_vcpu_ioctl_set_guest_debug()
12091 r = -EBUSY; in kvm_arch_vcpu_ioctl_set_guest_debug()
12094 if (dbg->control & KVM_GUESTDBG_INJECT_DB) in kvm_arch_vcpu_ioctl_set_guest_debug()
12106 vcpu->guest_debug = dbg->control; in kvm_arch_vcpu_ioctl_set_guest_debug()
12107 if (!(vcpu->guest_debug & KVM_GUESTDBG_ENABLE)) in kvm_arch_vcpu_ioctl_set_guest_debug()
12108 vcpu->guest_debug = 0; in kvm_arch_vcpu_ioctl_set_guest_debug()
12110 if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { in kvm_arch_vcpu_ioctl_set_guest_debug()
12112 vcpu->arch.eff_db[i] = dbg->arch.debugreg[i]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12113 vcpu->arch.guest_debug_dr7 = dbg->arch.debugreg[7]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12116 vcpu->arch.eff_db[i] = vcpu->arch.db[i]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12120 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) in kvm_arch_vcpu_ioctl_set_guest_debug()
12121 vcpu->arch.singlestep_rip = kvm_get_linear_rip(vcpu); in kvm_arch_vcpu_ioctl_set_guest_debug()
12131 kvm_arch_vcpu_guestdbg_update_apicv_inhibit(vcpu->kvm); in kvm_arch_vcpu_ioctl_set_guest_debug()
12146 unsigned long vaddr = tr->linear_address; in kvm_arch_vcpu_ioctl_translate()
12152 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl_translate()
12154 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl_translate()
12155 tr->physical_address = gpa; in kvm_arch_vcpu_ioctl_translate()
12156 tr->valid = gpa != INVALID_GPA; in kvm_arch_vcpu_ioctl_translate()
12157 tr->writeable = 1; in kvm_arch_vcpu_ioctl_translate()
12158 tr->usermode = 0; in kvm_arch_vcpu_ioctl_translate()
12168 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_arch_vcpu_ioctl_get_fpu()
12169 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_arch_vcpu_ioctl_get_fpu()
12173 fxsave = &vcpu->arch.guest_fpu.fpstate->regs.fxsave; in kvm_arch_vcpu_ioctl_get_fpu()
12174 memcpy(fpu->fpr, fxsave->st_space, 128); in kvm_arch_vcpu_ioctl_get_fpu()
12175 fpu->fcw = fxsave->cwd; in kvm_arch_vcpu_ioctl_get_fpu()
12176 fpu->fsw = fxsave->swd; in kvm_arch_vcpu_ioctl_get_fpu()
12177 fpu->ftwx = fxsave->twd; in kvm_arch_vcpu_ioctl_get_fpu()
12178 fpu->last_opcode = fxsave->fop; in kvm_arch_vcpu_ioctl_get_fpu()
12179 fpu->last_ip = fxsave->rip; in kvm_arch_vcpu_ioctl_get_fpu()
12180 fpu->last_dp = fxsave->rdp; in kvm_arch_vcpu_ioctl_get_fpu()
12181 memcpy(fpu->xmm, fxsave->xmm_space, sizeof(fxsave->xmm_space)); in kvm_arch_vcpu_ioctl_get_fpu()
12191 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_arch_vcpu_ioctl_set_fpu()
12192 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_arch_vcpu_ioctl_set_fpu()
12196 fxsave = &vcpu->arch.guest_fpu.fpstate->regs.fxsave; in kvm_arch_vcpu_ioctl_set_fpu()
12198 memcpy(fxsave->st_space, fpu->fpr, 128); in kvm_arch_vcpu_ioctl_set_fpu()
12199 fxsave->cwd = fpu->fcw; in kvm_arch_vcpu_ioctl_set_fpu()
12200 fxsave->swd = fpu->fsw; in kvm_arch_vcpu_ioctl_set_fpu()
12201 fxsave->twd = fpu->ftwx; in kvm_arch_vcpu_ioctl_set_fpu()
12202 fxsave->fop = fpu->last_opcode; in kvm_arch_vcpu_ioctl_set_fpu()
12203 fxsave->rip = fpu->last_ip; in kvm_arch_vcpu_ioctl_set_fpu()
12204 fxsave->rdp = fpu->last_dp; in kvm_arch_vcpu_ioctl_set_fpu()
12205 memcpy(fxsave->xmm_space, fpu->xmm, sizeof(fxsave->xmm_space)); in kvm_arch_vcpu_ioctl_set_fpu()
12215 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_REGS) in store_regs()
12216 __get_regs(vcpu, &vcpu->run->s.regs.regs); in store_regs()
12218 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_SREGS) in store_regs()
12219 __get_sregs(vcpu, &vcpu->run->s.regs.sregs); in store_regs()
12221 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_EVENTS) in store_regs()
12223 vcpu, &vcpu->run->s.regs.events); in store_regs()
12228 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_REGS) { in sync_regs()
12229 __set_regs(vcpu, &vcpu->run->s.regs.regs); in sync_regs()
12230 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_REGS; in sync_regs()
12233 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_SREGS) { in sync_regs()
12234 struct kvm_sregs sregs = vcpu->run->s.regs.sregs; in sync_regs()
12237 return -EINVAL; in sync_regs()
12239 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_SREGS; in sync_regs()
12242 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_EVENTS) { in sync_regs()
12243 struct kvm_vcpu_events events = vcpu->run->s.regs.events; in sync_regs()
12246 return -EINVAL; in sync_regs()
12248 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_EVENTS; in sync_regs()
12256 if (kvm_check_tsc_unstable() && kvm->created_vcpus) in kvm_arch_vcpu_precreate()
12260 if (!kvm->arch.max_vcpu_ids) in kvm_arch_vcpu_precreate()
12261 kvm->arch.max_vcpu_ids = KVM_MAX_VCPU_IDS; in kvm_arch_vcpu_precreate()
12263 if (id >= kvm->arch.max_vcpu_ids) in kvm_arch_vcpu_precreate()
12264 return -EINVAL; in kvm_arch_vcpu_precreate()
12274 vcpu->arch.last_vmentry_cpu = -1; in kvm_arch_vcpu_create()
12275 vcpu->arch.regs_avail = ~0; in kvm_arch_vcpu_create()
12276 vcpu->arch.regs_dirty = ~0; in kvm_arch_vcpu_create()
12278 kvm_gpc_init(&vcpu->arch.pv_time, vcpu->kvm); in kvm_arch_vcpu_create()
12280 if (!irqchip_in_kernel(vcpu->kvm) || kvm_vcpu_is_reset_bsp(vcpu)) in kvm_arch_vcpu_create()
12293 r = -ENOMEM; in kvm_arch_vcpu_create()
12298 vcpu->arch.pio_data = page_address(page); in kvm_arch_vcpu_create()
12300 vcpu->arch.mce_banks = kcalloc(KVM_MAX_MCE_BANKS * 4, sizeof(u64), in kvm_arch_vcpu_create()
12302 vcpu->arch.mci_ctl2_banks = kcalloc(KVM_MAX_MCE_BANKS, sizeof(u64), in kvm_arch_vcpu_create()
12304 if (!vcpu->arch.mce_banks || !vcpu->arch.mci_ctl2_banks) in kvm_arch_vcpu_create()
12306 vcpu->arch.mcg_cap = KVM_MAX_MCE_BANKS; in kvm_arch_vcpu_create()
12308 if (!zalloc_cpumask_var(&vcpu->arch.wbinvd_dirty_mask, in kvm_arch_vcpu_create()
12315 if (!fpu_alloc_guest_fpstate(&vcpu->arch.guest_fpu)) { in kvm_arch_vcpu_create()
12322 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_STUFF_FEATURE_MSRS)) { in kvm_arch_vcpu_create()
12323 vcpu->arch.arch_capabilities = kvm_get_arch_capabilities(); in kvm_arch_vcpu_create()
12324 vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT; in kvm_arch_vcpu_create()
12325 vcpu->arch.perf_capabilities = kvm_caps.supported_perf_cap; in kvm_arch_vcpu_create()
12329 vcpu->arch.pending_external_vector = -1; in kvm_arch_vcpu_create()
12330 vcpu->arch.preempted_in_kernel = false; in kvm_arch_vcpu_create()
12333 vcpu->arch.hv_root_tdp = INVALID_PAGE; in kvm_arch_vcpu_create()
12343 kvm_set_tsc_khz(vcpu, vcpu->kvm->arch.default_tsc_khz); in kvm_arch_vcpu_create()
12350 fpu_free_guest_fpstate(&vcpu->arch.guest_fpu); in kvm_arch_vcpu_create()
12352 kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt); in kvm_arch_vcpu_create()
12354 free_cpumask_var(vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_create()
12356 kfree(vcpu->arch.mce_banks); in kvm_arch_vcpu_create()
12357 kfree(vcpu->arch.mci_ctl2_banks); in kvm_arch_vcpu_create()
12358 free_page((unsigned long)vcpu->arch.pio_data); in kvm_arch_vcpu_create()
12368 struct kvm *kvm = vcpu->kvm; in kvm_arch_vcpu_postcreate()
12370 if (mutex_lock_killable(&vcpu->mutex)) in kvm_arch_vcpu_postcreate()
12377 vcpu->arch.msr_kvm_poll_control = 1; in kvm_arch_vcpu_postcreate()
12379 mutex_unlock(&vcpu->mutex); in kvm_arch_vcpu_postcreate()
12381 if (kvmclock_periodic_sync && vcpu->vcpu_idx == 0) in kvm_arch_vcpu_postcreate()
12382 schedule_delayed_work(&kvm->arch.kvmclock_sync_work, in kvm_arch_vcpu_postcreate()
12397 kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt); in kvm_arch_vcpu_destroy()
12398 free_cpumask_var(vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_destroy()
12399 fpu_free_guest_fpstate(&vcpu->arch.guest_fpu); in kvm_arch_vcpu_destroy()
12404 kfree(vcpu->arch.mce_banks); in kvm_arch_vcpu_destroy()
12405 kfree(vcpu->arch.mci_ctl2_banks); in kvm_arch_vcpu_destroy()
12407 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_destroy()
12409 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_destroy()
12410 free_page((unsigned long)vcpu->arch.pio_data); in kvm_arch_vcpu_destroy()
12411 kvfree(vcpu->arch.cpuid_entries); in kvm_arch_vcpu_destroy()
12421 * Several of the "set" flows, e.g. ->set_cr0(), read other registers in kvm_vcpu_reset()
12431 * SVM doesn't unconditionally VM-Exit on INIT and SHUTDOWN, thus it's in kvm_vcpu_reset()
12442 vcpu->arch.hflags = 0; in kvm_vcpu_reset()
12444 vcpu->arch.smi_pending = 0; in kvm_vcpu_reset()
12445 vcpu->arch.smi_count = 0; in kvm_vcpu_reset()
12446 atomic_set(&vcpu->arch.nmi_queued, 0); in kvm_vcpu_reset()
12447 vcpu->arch.nmi_pending = 0; in kvm_vcpu_reset()
12448 vcpu->arch.nmi_injected = false; in kvm_vcpu_reset()
12452 memset(vcpu->arch.db, 0, sizeof(vcpu->arch.db)); in kvm_vcpu_reset()
12454 vcpu->arch.dr6 = DR6_ACTIVE_LOW; in kvm_vcpu_reset()
12455 vcpu->arch.dr7 = DR7_FIXED_1; in kvm_vcpu_reset()
12458 vcpu->arch.cr2 = 0; in kvm_vcpu_reset()
12461 vcpu->arch.apf.msr_en_val = 0; in kvm_vcpu_reset()
12462 vcpu->arch.apf.msr_int_val = 0; in kvm_vcpu_reset()
12463 vcpu->arch.st.msr_val = 0; in kvm_vcpu_reset()
12469 vcpu->arch.apf.halted = false; in kvm_vcpu_reset()
12471 if (vcpu->arch.guest_fpu.fpstate && kvm_mpx_supported()) { in kvm_vcpu_reset()
12472 struct fpstate *fpstate = vcpu->arch.guest_fpu.fpstate; in kvm_vcpu_reset()
12489 vcpu->arch.smbase = 0x30000; in kvm_vcpu_reset()
12491 vcpu->arch.pat = MSR_IA32_CR_PAT_DEFAULT; in kvm_vcpu_reset()
12493 vcpu->arch.msr_misc_features_enables = 0; in kvm_vcpu_reset()
12494 vcpu->arch.ia32_misc_enable_msr = MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL | in kvm_vcpu_reset()
12502 memset(vcpu->arch.regs, 0, sizeof(vcpu->arch.regs)); in kvm_vcpu_reset()
12513 kvm_rdx_write(vcpu, cpuid_0x1 ? cpuid_0x1->eax : 0x600); in kvm_vcpu_reset()
12520 vcpu->arch.cr3 = 0; in kvm_vcpu_reset()
12542 * which PCIDs have to be flushed. However, CR0.WP and the paging-related in kvm_vcpu_reset()
12612 if (!stable && vcpu->cpu == smp_processor_id()) in kvm_arch_enable_virtualization_cpu()
12614 if (stable && vcpu->arch.last_host_tsc > local_tsc) { in kvm_arch_enable_virtualization_cpu()
12616 if (vcpu->arch.last_host_tsc > max_tsc) in kvm_arch_enable_virtualization_cpu()
12617 max_tsc = vcpu->arch.last_host_tsc; in kvm_arch_enable_virtualization_cpu()
12647 * N.B. - this code below runs only on platforms with reliable TSC, in kvm_arch_enable_virtualization_cpu()
12661 u64 delta_cyc = max_tsc - local_tsc; in kvm_arch_enable_virtualization_cpu()
12663 kvm->arch.backwards_tsc_observed = true; in kvm_arch_enable_virtualization_cpu()
12665 vcpu->arch.tsc_offset_adjustment += delta_cyc; in kvm_arch_enable_virtualization_cpu()
12666 vcpu->arch.last_host_tsc = local_tsc; in kvm_arch_enable_virtualization_cpu()
12676 kvm->arch.last_tsc_nsec = 0; in kvm_arch_enable_virtualization_cpu()
12677 kvm->arch.last_tsc_write = 0; in kvm_arch_enable_virtualization_cpu()
12692 return vcpu->kvm->arch.bsp_vcpu_id == vcpu->vcpu_id; in kvm_vcpu_is_reset_bsp()
12697 return (vcpu->arch.apic_base & MSR_IA32_APICBASE_BSP) != 0; in kvm_vcpu_is_bsp()
12703 kfree(kvm->arch.hv_pa_pg); in kvm_arch_free_vm()
12715 return -EINVAL; in kvm_arch_init_vm()
12717 kvm->arch.vm_type = type; in kvm_arch_init_vm()
12718 kvm->arch.has_private_mem = in kvm_arch_init_vm()
12721 kvm->arch.pre_fault_allowed = in kvm_arch_init_vm()
12734 INIT_HLIST_HEAD(&kvm->arch.mask_notifier_list); in kvm_arch_init_vm()
12735 atomic_set(&kvm->arch.noncoherent_dma_count, 0); in kvm_arch_init_vm()
12738 set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap); in kvm_arch_init_vm()
12739 /* Reserve bit 1 of irq_sources_bitmap for irqfd-resampler */ in kvm_arch_init_vm()
12741 &kvm->arch.irq_sources_bitmap); in kvm_arch_init_vm()
12743 raw_spin_lock_init(&kvm->arch.tsc_write_lock); in kvm_arch_init_vm()
12744 mutex_init(&kvm->arch.apic_map_lock); in kvm_arch_init_vm()
12745 seqcount_raw_spinlock_init(&kvm->arch.pvclock_sc, &kvm->arch.tsc_write_lock); in kvm_arch_init_vm()
12746 kvm->arch.kvmclock_offset = -get_kvmclock_base_ns(); in kvm_arch_init_vm()
12748 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_arch_init_vm()
12750 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_arch_init_vm()
12752 kvm->arch.default_tsc_khz = max_tsc_khz ? : tsc_khz; in kvm_arch_init_vm()
12753 kvm->arch.apic_bus_cycle_ns = APIC_BUS_CYCLE_NS_DEFAULT; in kvm_arch_init_vm()
12754 kvm->arch.guest_can_read_msr_platform_info = true; in kvm_arch_init_vm()
12755 kvm->arch.enable_pmu = enable_pmu; in kvm_arch_init_vm()
12758 spin_lock_init(&kvm->arch.hv_root_tdp_lock); in kvm_arch_init_vm()
12759 kvm->arch.hv_root_tdp = INVALID_PAGE; in kvm_arch_init_vm()
12762 INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn); in kvm_arch_init_vm()
12763 INIT_DELAYED_WORK(&kvm->arch.kvmclock_sync_work, kvmclock_sync_fn); in kvm_arch_init_vm()
12776 once_init(&kvm->arch.nx_once); in kvm_arch_init_vm()
12799 * -errno: on error
12804 * GPA->HVA translation will not change. However, the HVA is a user
12816 lockdep_assert_held(&kvm->slots_lock); in __x86_set_memory_region()
12819 return ERR_PTR_USR(-EINVAL); in __x86_set_memory_region()
12823 if (slot && slot->npages) in __x86_set_memory_region()
12824 return ERR_PTR_USR(-EEXIST); in __x86_set_memory_region()
12835 if (!slot || !slot->npages) in __x86_set_memory_region()
12838 old_npages = slot->npages; in __x86_set_memory_region()
12839 hva = slot->userspace_addr; in __x86_set_memory_region()
12867 * is unsafe, i.e. will lead to use-after-free. The PIT also needs to in kvm_arch_pre_destroy_vm()
12870 cancel_delayed_work_sync(&kvm->arch.kvmclock_sync_work); in kvm_arch_pre_destroy_vm()
12871 cancel_delayed_work_sync(&kvm->arch.kvmclock_update_work); in kvm_arch_pre_destroy_vm()
12880 if (current->mm == kvm->mm) { in kvm_arch_destroy_vm()
12886 mutex_lock(&kvm->slots_lock); in kvm_arch_destroy_vm()
12892 mutex_unlock(&kvm->slots_lock); in kvm_arch_destroy_vm()
12895 kvm_free_msr_filter(srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1)); in kvm_arch_destroy_vm()
12898 kvfree(rcu_dereference_check(kvm->arch.apic_map, 1)); in kvm_arch_destroy_vm()
12899 kfree(srcu_dereference_check(kvm->arch.pmu_event_filter, &kvm->srcu, 1)); in kvm_arch_destroy_vm()
12912 vfree(slot->arch.rmap[i]); in memslot_rmap_free()
12913 slot->arch.rmap[i] = NULL; in memslot_rmap_free()
12924 vfree(slot->arch.lpage_info[i - 1]); in kvm_arch_free_memslot()
12925 slot->arch.lpage_info[i - 1] = NULL; in kvm_arch_free_memslot()
12933 const int sz = sizeof(*slot->arch.rmap[0]); in memslot_rmap_alloc()
12940 if (slot->arch.rmap[i]) in memslot_rmap_alloc()
12943 slot->arch.rmap[i] = __vcalloc(lpages, sz, GFP_KERNEL_ACCOUNT); in memslot_rmap_alloc()
12944 if (!slot->arch.rmap[i]) { in memslot_rmap_alloc()
12946 return -ENOMEM; in memslot_rmap_alloc()
12956 unsigned long npages = slot->npages; in kvm_alloc_memslot_metadata()
12964 memset(&slot->arch, 0, sizeof(slot->arch)); in kvm_alloc_memslot_metadata()
12984 slot->arch.lpage_info[i - 1] = linfo; in kvm_alloc_memslot_metadata()
12986 if (slot->base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) in kvm_alloc_memslot_metadata()
12988 if ((slot->base_gfn + npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) in kvm_alloc_memslot_metadata()
12989 linfo[lpages - 1].disallow_lpage = 1; in kvm_alloc_memslot_metadata()
12990 ugfn = slot->userspace_addr >> PAGE_SHIFT; in kvm_alloc_memslot_metadata()
12995 if ((slot->base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1)) { in kvm_alloc_memslot_metadata()
13016 vfree(slot->arch.lpage_info[i - 1]); in kvm_alloc_memslot_metadata()
13017 slot->arch.lpage_info[i - 1] = NULL; in kvm_alloc_memslot_metadata()
13019 return -ENOMEM; in kvm_alloc_memslot_metadata()
13028 * memslots->generation has been incremented. in kvm_arch_memslots_updated()
13033 /* Force re-initialization of steal_time cache */ in kvm_arch_memslots_updated()
13048 return -EINVAL; in kvm_arch_prepare_memory_region()
13051 if ((new->base_gfn + new->npages - 1) > kvm_mmu_max_gfn()) in kvm_arch_prepare_memory_region()
13052 return -EINVAL; in kvm_arch_prepare_memory_region()
13054 if (kvm_is_gfn_alias(kvm, new->base_gfn + new->npages - 1)) in kvm_arch_prepare_memory_region()
13055 return -EINVAL; in kvm_arch_prepare_memory_region()
13061 memcpy(&new->arch, &old->arch, sizeof(old->arch)); in kvm_arch_prepare_memory_region()
13063 return -EIO; in kvm_arch_prepare_memory_region()
13076 nr_slots = atomic_read(&kvm->nr_memslots_dirty_logging); in kvm_mmu_update_cpu_dirty_logging()
13086 u32 old_flags = old ? old->flags : 0; in kvm_mmu_slot_apply_flags()
13087 u32 new_flags = new ? new->flags : 0; in kvm_mmu_slot_apply_flags()
13107 * CREATE: No shadow pages exist, thus nothing to write-protect in kvm_mmu_slot_apply_flags()
13116 * READONLY and non-flags changes were filtered out above, and the only in kvm_mmu_slot_apply_flags()
13136 * Initially-all-set does not require write protecting any page, in kvm_mmu_slot_apply_flags()
13161 * write-protected before returning to userspace, i.e. before in kvm_mmu_slot_apply_flags()
13168 * Specifically, KVM also write-protects guest page tables to in kvm_mmu_slot_apply_flags()
13177 * To handle these scenarios, KVM uses a separate software-only in kvm_mmu_slot_apply_flags()
13178 * bit (MMU-writable) to track if a SPTE is !writable due to in kvm_mmu_slot_apply_flags()
13179 * a guest page table being write-protected (KVM clears the in kvm_mmu_slot_apply_flags()
13180 * MMU-writable flag when write-protecting for shadow paging). in kvm_mmu_slot_apply_flags()
13182 * The use of MMU-writable is also the primary motivation for in kvm_mmu_slot_apply_flags()
13185 * !MMU-writable SPTE, KVM must flush if it encounters any in kvm_mmu_slot_apply_flags()
13186 * MMU-writable SPTE regardless of whether the actual hardware in kvm_mmu_slot_apply_flags()
13189 * write access" helpers to ignore MMU-writable entirely. in kvm_mmu_slot_apply_flags()
13192 * access-tracked SPTEs is particularly relevant). in kvm_mmu_slot_apply_flags()
13206 if (!kvm->arch.n_requested_mmu_pages && in kvm_arch_commit_memory_region()
13210 nr_mmu_pages = kvm->nr_memslot_pages / KVM_MEMSLOT_PAGES_TO_MMU_PAGES_RATIO; in kvm_arch_commit_memory_region()
13226 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_in_kernel()
13236 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_get_ip()
13255 if (vcpu->arch.guest_state_protected) in kvm_get_linear_rip()
13276 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) in kvm_get_rflags()
13284 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP && in __kvm_set_rflags()
13285 kvm_is_linear_rip(vcpu, vcpu->arch.singlestep_rip)) in __kvm_set_rflags()
13306 return (key + 1) & (ASYNC_PF_PER_VCPU - 1); in kvm_async_pf_next_probe()
13313 while (vcpu->arch.apf.gfns[key] != ~0) in kvm_add_async_pf_gfn()
13316 vcpu->arch.apf.gfns[key] = gfn; in kvm_add_async_pf_gfn()
13325 (vcpu->arch.apf.gfns[key] != gfn && in kvm_async_pf_gfn_slot()
13326 vcpu->arch.apf.gfns[key] != ~0); i++) in kvm_async_pf_gfn_slot()
13334 return vcpu->arch.apf.gfns[kvm_async_pf_gfn_slot(vcpu, gfn)] == gfn; in kvm_find_async_pf_gfn()
13343 if (WARN_ON_ONCE(vcpu->arch.apf.gfns[i] != gfn)) in kvm_del_async_pf_gfn()
13347 vcpu->arch.apf.gfns[i] = ~0; in kvm_del_async_pf_gfn()
13350 if (vcpu->arch.apf.gfns[j] == ~0) in kvm_del_async_pf_gfn()
13352 k = kvm_async_pf_hash_fn(vcpu->arch.apf.gfns[j]); in kvm_del_async_pf_gfn()
13359 vcpu->arch.apf.gfns[i] = vcpu->arch.apf.gfns[j]; in kvm_del_async_pf_gfn()
13368 return kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.apf.data, &reason, in apf_put_user_notpresent()
13376 return kvm_write_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data, in apf_put_user_ready()
13385 if (kvm_read_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data, in apf_pageready_slot_free()
13398 if (!vcpu->arch.apf.send_always && in kvm_can_deliver_async_pf()
13399 (vcpu->arch.guest_state_protected || !kvm_x86_call(get_cpl)(vcpu))) in kvm_can_deliver_async_pf()
13407 return vcpu->arch.apf.delivery_as_pf_vmexit; in kvm_can_deliver_async_pf()
13425 if (kvm_hlt_in_guest(vcpu->kvm) && !kvm_can_deliver_async_pf(vcpu)) in kvm_can_do_async_pf()
13440 trace_kvm_async_pf_not_present(work->arch.token, work->cr2_or_gpa); in kvm_arch_async_page_not_present()
13441 kvm_add_async_pf_gfn(vcpu, work->arch.gfn); in kvm_arch_async_page_not_present()
13449 fault.address = work->arch.token; in kvm_arch_async_page_not_present()
13472 .vector = vcpu->arch.apf.vec in kvm_arch_async_page_present()
13475 if (work->wakeup_all) in kvm_arch_async_page_present()
13476 work->arch.token = ~0; /* broadcast wakeup */ in kvm_arch_async_page_present()
13478 kvm_del_async_pf_gfn(vcpu, work->arch.gfn); in kvm_arch_async_page_present()
13479 trace_kvm_async_pf_ready(work->arch.token, work->cr2_or_gpa); in kvm_arch_async_page_present()
13481 if ((work->wakeup_all || work->notpresent_injected) && in kvm_arch_async_page_present()
13483 !apf_put_user_ready(vcpu, work->arch.token)) { in kvm_arch_async_page_present()
13484 vcpu->arch.apf.pageready_pending = true; in kvm_arch_async_page_present()
13488 vcpu->arch.apf.halted = false; in kvm_arch_async_page_present()
13495 if (!vcpu->arch.apf.pageready_pending) in kvm_arch_async_page_present_queued()
13509 if (atomic_inc_return(&kvm->arch.assigned_device_count) == 1) in kvm_arch_start_assignment()
13516 atomic_dec(&kvm->arch.assigned_device_count); in kvm_arch_end_assignment()
13522 return raw_atomic_read(&kvm->arch.assigned_device_count); in kvm_arch_has_assigned_device()
13529 * Non-coherent DMA assignment and de-assignment may affect whether or in kvm_noncoherent_dma_assignment_start_or_stop()
13532 * (or last) non-coherent device is (un)registered to so that new SPTEs in kvm_noncoherent_dma_assignment_start_or_stop()
13541 if (atomic_inc_return(&kvm->arch.noncoherent_dma_count) == 1) in kvm_arch_register_noncoherent_dma()
13548 if (!atomic_dec_return(&kvm->arch.noncoherent_dma_count)) in kvm_arch_unregister_noncoherent_dma()
13555 return atomic_read(&kvm->arch.noncoherent_dma_count); in kvm_arch_has_noncoherent_dma()
13571 irqfd->producer = prod; in kvm_arch_irq_bypass_add_producer()
13572 kvm_arch_start_assignment(irqfd->kvm); in kvm_arch_irq_bypass_add_producer()
13573 ret = kvm_x86_call(pi_update_irte)(irqfd->kvm, in kvm_arch_irq_bypass_add_producer()
13574 prod->irq, irqfd->gsi, 1); in kvm_arch_irq_bypass_add_producer()
13576 kvm_arch_end_assignment(irqfd->kvm); in kvm_arch_irq_bypass_add_producer()
13588 WARN_ON(irqfd->producer != prod); in kvm_arch_irq_bypass_del_producer()
13589 irqfd->producer = NULL; in kvm_arch_irq_bypass_del_producer()
13593 * remapped mode, so we can re-use the current implementation in kvm_arch_irq_bypass_del_producer()
13597 ret = kvm_x86_call(pi_update_irte)(irqfd->kvm, in kvm_arch_irq_bypass_del_producer()
13598 prod->irq, irqfd->gsi, 0); in kvm_arch_irq_bypass_del_producer()
13601 " fails: %d\n", irqfd->consumer.token, ret); in kvm_arch_irq_bypass_del_producer()
13603 kvm_arch_end_assignment(irqfd->kvm); in kvm_arch_irq_bypass_del_producer()
13615 if (new->type != KVM_IRQ_ROUTING_MSI) in kvm_arch_irqfd_route_changed()
13618 return !!memcmp(&old->msi, &new->msi, sizeof(new->msi)); in kvm_arch_irqfd_route_changed()
13628 return (vcpu->arch.msr_kvm_poll_control & 1) == 0; in kvm_arch_no_poll()
13674 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_fixup_and_inject_pf_error()
13680 mmu->gva_to_gpa(vcpu, mmu, gva, access, &fault) != INVALID_GPA) { in kvm_fixup_and_inject_pf_error()
13682 * If vcpu->arch.walk_mmu->gva_to_gpa succeeded, the page in kvm_fixup_and_inject_pf_error()
13693 vcpu->arch.walk_mmu->inject_page_fault(vcpu, &fault); in kvm_fixup_and_inject_pf_error()
13706 if (KVM_BUG_ON(!e, vcpu->kvm)) in kvm_handle_memory_failure()
13707 return -EIO; in kvm_handle_memory_failure()
13717 * doesn't seem to be a real use-case behind such requests, just return in kvm_handle_memory_failure()
13773 * page tables, so a non-global flush just degenerates to a in kvm_handle_invpcid()
13792 struct kvm_run *run = vcpu->run; in complete_sev_es_emulated_mmio()
13796 BUG_ON(!vcpu->mmio_needed); in complete_sev_es_emulated_mmio()
13799 frag = &vcpu->mmio_fragments[vcpu->mmio_cur_fragment]; in complete_sev_es_emulated_mmio()
13800 len = min(8u, frag->len); in complete_sev_es_emulated_mmio()
13801 if (!vcpu->mmio_is_write) in complete_sev_es_emulated_mmio()
13802 memcpy(frag->data, run->mmio.data, len); in complete_sev_es_emulated_mmio()
13804 if (frag->len <= 8) { in complete_sev_es_emulated_mmio()
13807 vcpu->mmio_cur_fragment++; in complete_sev_es_emulated_mmio()
13810 frag->data += len; in complete_sev_es_emulated_mmio()
13811 frag->gpa += len; in complete_sev_es_emulated_mmio()
13812 frag->len -= len; in complete_sev_es_emulated_mmio()
13815 if (vcpu->mmio_cur_fragment >= vcpu->mmio_nr_fragments) { in complete_sev_es_emulated_mmio()
13816 vcpu->mmio_needed = 0; in complete_sev_es_emulated_mmio()
13824 run->mmio.phys_addr = frag->gpa; in complete_sev_es_emulated_mmio()
13825 run->mmio.len = min(8u, frag->len); in complete_sev_es_emulated_mmio()
13826 run->mmio.is_write = vcpu->mmio_is_write; in complete_sev_es_emulated_mmio()
13827 if (run->mmio.is_write) in complete_sev_es_emulated_mmio()
13828 memcpy(run->mmio.data, frag->data, min(8u, frag->len)); in complete_sev_es_emulated_mmio()
13829 run->exit_reason = KVM_EXIT_MMIO; in complete_sev_es_emulated_mmio()
13831 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in complete_sev_es_emulated_mmio()
13843 return -EINVAL; in kvm_sev_es_mmio_write()
13849 bytes -= handled; in kvm_sev_es_mmio_write()
13854 frag = vcpu->mmio_fragments; in kvm_sev_es_mmio_write()
13855 vcpu->mmio_nr_fragments = 1; in kvm_sev_es_mmio_write()
13856 frag->len = bytes; in kvm_sev_es_mmio_write()
13857 frag->gpa = gpa; in kvm_sev_es_mmio_write()
13858 frag->data = data; in kvm_sev_es_mmio_write()
13860 vcpu->mmio_needed = 1; in kvm_sev_es_mmio_write()
13861 vcpu->mmio_cur_fragment = 0; in kvm_sev_es_mmio_write()
13863 vcpu->run->mmio.phys_addr = gpa; in kvm_sev_es_mmio_write()
13864 vcpu->run->mmio.len = min(8u, frag->len); in kvm_sev_es_mmio_write()
13865 vcpu->run->mmio.is_write = 1; in kvm_sev_es_mmio_write()
13866 memcpy(vcpu->run->mmio.data, frag->data, min(8u, frag->len)); in kvm_sev_es_mmio_write()
13867 vcpu->run->exit_reason = KVM_EXIT_MMIO; in kvm_sev_es_mmio_write()
13869 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in kvm_sev_es_mmio_write()
13882 return -EINVAL; in kvm_sev_es_mmio_read()
13888 bytes -= handled; in kvm_sev_es_mmio_read()
13893 frag = vcpu->mmio_fragments; in kvm_sev_es_mmio_read()
13894 vcpu->mmio_nr_fragments = 1; in kvm_sev_es_mmio_read()
13895 frag->len = bytes; in kvm_sev_es_mmio_read()
13896 frag->gpa = gpa; in kvm_sev_es_mmio_read()
13897 frag->data = data; in kvm_sev_es_mmio_read()
13899 vcpu->mmio_needed = 1; in kvm_sev_es_mmio_read()
13900 vcpu->mmio_cur_fragment = 0; in kvm_sev_es_mmio_read()
13902 vcpu->run->mmio.phys_addr = gpa; in kvm_sev_es_mmio_read()
13903 vcpu->run->mmio.len = min(8u, frag->len); in kvm_sev_es_mmio_read()
13904 vcpu->run->mmio.is_write = 0; in kvm_sev_es_mmio_read()
13905 vcpu->run->exit_reason = KVM_EXIT_MMIO; in kvm_sev_es_mmio_read()
13907 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in kvm_sev_es_mmio_read()
13915 vcpu->arch.sev_pio_count -= count; in advance_sev_es_emulated_pio()
13916 vcpu->arch.sev_pio_data += count * size; in advance_sev_es_emulated_pio()
13924 int size = vcpu->arch.pio.size; in complete_sev_es_emulated_outs()
13925 int port = vcpu->arch.pio.port; in complete_sev_es_emulated_outs()
13927 vcpu->arch.pio.count = 0; in complete_sev_es_emulated_outs()
13928 if (vcpu->arch.sev_pio_count) in complete_sev_es_emulated_outs()
13938 min_t(unsigned int, PAGE_SIZE / size, vcpu->arch.sev_pio_count); in kvm_sev_es_outs()
13939 int ret = emulator_pio_out(vcpu, size, port, vcpu->arch.sev_pio_data, count); in kvm_sev_es_outs()
13947 if (!vcpu->arch.sev_pio_count) in kvm_sev_es_outs()
13951 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_outs; in kvm_sev_es_outs()
13960 unsigned count = vcpu->arch.pio.count; in complete_sev_es_emulated_ins()
13961 int size = vcpu->arch.pio.size; in complete_sev_es_emulated_ins()
13962 int port = vcpu->arch.pio.port; in complete_sev_es_emulated_ins()
13964 complete_emulator_pio_in(vcpu, vcpu->arch.sev_pio_data); in complete_sev_es_emulated_ins()
13966 if (vcpu->arch.sev_pio_count) in complete_sev_es_emulated_ins()
13976 min_t(unsigned int, PAGE_SIZE / size, vcpu->arch.sev_pio_count); in kvm_sev_es_ins()
13977 if (!emulator_pio_in(vcpu, size, port, vcpu->arch.sev_pio_data, count)) in kvm_sev_es_ins()
13982 if (!vcpu->arch.sev_pio_count) in kvm_sev_es_ins()
13986 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_ins; in kvm_sev_es_ins()
13994 vcpu->arch.sev_pio_data = data; in kvm_sev_es_string_io()
13995 vcpu->arch.sev_pio_count = count; in kvm_sev_es_string_io()