Index: releng/12.0/sys/amd64/vmm/intel/vmx.c =================================================================== --- releng/12.0/sys/amd64/vmm/intel/vmx.c (revision 340626) +++ releng/12.0/sys/amd64/vmm/intel/vmx.c (revision 340627) @@ -1,3753 +1,3754 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2011 NetApp, Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "vmm_lapic.h" #include "vmm_host.h" #include "vmm_ioport.h" #include "vmm_ktr.h" #include "vmm_stat.h" #include "vatpic.h" #include "vlapic.h" #include "vlapic_priv.h" #include "ept.h" #include "vmx_cpufunc.h" #include "vmx.h" #include "vmx_msr.h" #include "x86.h" #include "vmx_controls.h" #define PINBASED_CTLS_ONE_SETTING \ (PINBASED_EXTINT_EXITING | \ PINBASED_NMI_EXITING | \ PINBASED_VIRTUAL_NMI) #define PINBASED_CTLS_ZERO_SETTING 0 #define PROCBASED_CTLS_WINDOW_SETTING \ (PROCBASED_INT_WINDOW_EXITING | \ PROCBASED_NMI_WINDOW_EXITING) #define PROCBASED_CTLS_ONE_SETTING \ (PROCBASED_SECONDARY_CONTROLS | \ PROCBASED_MWAIT_EXITING | \ PROCBASED_MONITOR_EXITING | \ PROCBASED_IO_EXITING | \ PROCBASED_MSR_BITMAPS | \ PROCBASED_CTLS_WINDOW_SETTING | \ PROCBASED_CR8_LOAD_EXITING | \ PROCBASED_CR8_STORE_EXITING) #define PROCBASED_CTLS_ZERO_SETTING \ (PROCBASED_CR3_LOAD_EXITING | \ PROCBASED_CR3_STORE_EXITING | \ PROCBASED_IO_BITMAPS) #define PROCBASED_CTLS2_ONE_SETTING PROCBASED2_ENABLE_EPT #define PROCBASED_CTLS2_ZERO_SETTING 0 #define VM_EXIT_CTLS_ONE_SETTING \ (VM_EXIT_SAVE_DEBUG_CONTROLS | \ VM_EXIT_HOST_LMA | \ VM_EXIT_SAVE_EFER | \ VM_EXIT_LOAD_EFER | \ VM_EXIT_ACKNOWLEDGE_INTERRUPT) #define VM_EXIT_CTLS_ZERO_SETTING 0 #define VM_ENTRY_CTLS_ONE_SETTING \ (VM_ENTRY_LOAD_DEBUG_CONTROLS | \ VM_ENTRY_LOAD_EFER) #define VM_ENTRY_CTLS_ZERO_SETTING \ (VM_ENTRY_INTO_SMM | \ VM_ENTRY_DEACTIVATE_DUAL_MONITOR) #define HANDLED 1 #define UNHANDLED 0 static MALLOC_DEFINE(M_VMX, "vmx", "vmx"); static MALLOC_DEFINE(M_VLAPIC, "vlapic", "vlapic"); SYSCTL_DECL(_hw_vmm); SYSCTL_NODE(_hw_vmm, OID_AUTO, vmx, CTLFLAG_RW, NULL, NULL); int vmxon_enabled[MAXCPU]; static char vmxon_region[MAXCPU][PAGE_SIZE] __aligned(PAGE_SIZE); static uint32_t pinbased_ctls, procbased_ctls, procbased_ctls2; static uint32_t exit_ctls, entry_ctls; static uint64_t cr0_ones_mask, cr0_zeros_mask; SYSCTL_ULONG(_hw_vmm_vmx, OID_AUTO, cr0_ones_mask, CTLFLAG_RD, &cr0_ones_mask, 0, NULL); SYSCTL_ULONG(_hw_vmm_vmx, OID_AUTO, cr0_zeros_mask, CTLFLAG_RD, &cr0_zeros_mask, 0, NULL); static uint64_t cr4_ones_mask, cr4_zeros_mask; SYSCTL_ULONG(_hw_vmm_vmx, OID_AUTO, cr4_ones_mask, CTLFLAG_RD, &cr4_ones_mask, 0, NULL); SYSCTL_ULONG(_hw_vmm_vmx, OID_AUTO, cr4_zeros_mask, CTLFLAG_RD, &cr4_zeros_mask, 0, NULL); static int vmx_initialized; SYSCTL_INT(_hw_vmm_vmx, OID_AUTO, initialized, CTLFLAG_RD, &vmx_initialized, 0, "Intel VMX initialized"); /* * Optional capabilities */ static SYSCTL_NODE(_hw_vmm_vmx, OID_AUTO, cap, CTLFLAG_RW, NULL, NULL); static int cap_halt_exit; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, halt_exit, CTLFLAG_RD, &cap_halt_exit, 0, "HLT triggers a VM-exit"); static int cap_pause_exit; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, pause_exit, CTLFLAG_RD, &cap_pause_exit, 0, "PAUSE triggers a VM-exit"); static int cap_unrestricted_guest; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, unrestricted_guest, CTLFLAG_RD, &cap_unrestricted_guest, 0, "Unrestricted guests"); static int cap_monitor_trap; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, monitor_trap, CTLFLAG_RD, &cap_monitor_trap, 0, "Monitor trap flag"); static int cap_invpcid; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, invpcid, CTLFLAG_RD, &cap_invpcid, 0, "Guests are allowed to use INVPCID"); static int virtual_interrupt_delivery; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, virtual_interrupt_delivery, CTLFLAG_RD, &virtual_interrupt_delivery, 0, "APICv virtual interrupt delivery support"); static int posted_interrupts; SYSCTL_INT(_hw_vmm_vmx_cap, OID_AUTO, posted_interrupts, CTLFLAG_RD, &posted_interrupts, 0, "APICv posted interrupt support"); static int pirvec = -1; SYSCTL_INT(_hw_vmm_vmx, OID_AUTO, posted_interrupt_vector, CTLFLAG_RD, &pirvec, 0, "APICv posted interrupt vector"); static struct unrhdr *vpid_unr; static u_int vpid_alloc_failed; SYSCTL_UINT(_hw_vmm_vmx, OID_AUTO, vpid_alloc_failed, CTLFLAG_RD, &vpid_alloc_failed, 0, NULL); static int guest_l1d_flush; SYSCTL_INT(_hw_vmm_vmx, OID_AUTO, l1d_flush, CTLFLAG_RD, &guest_l1d_flush, 0, NULL); static int guest_l1d_flush_sw; SYSCTL_INT(_hw_vmm_vmx, OID_AUTO, l1d_flush_sw, CTLFLAG_RD, &guest_l1d_flush_sw, 0, NULL); static struct msr_entry msr_load_list[1] __aligned(16); /* * The definitions of SDT probes for VMX. */ SDT_PROBE_DEFINE3(vmm, vmx, exit, entry, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE4(vmm, vmx, exit, taskswitch, "struct vmx *", "int", "struct vm_exit *", "struct vm_task_switch *"); SDT_PROBE_DEFINE4(vmm, vmx, exit, craccess, "struct vmx *", "int", "struct vm_exit *", "uint64_t"); SDT_PROBE_DEFINE4(vmm, vmx, exit, rdmsr, "struct vmx *", "int", "struct vm_exit *", "uint32_t"); SDT_PROBE_DEFINE5(vmm, vmx, exit, wrmsr, "struct vmx *", "int", "struct vm_exit *", "uint32_t", "uint64_t"); SDT_PROBE_DEFINE3(vmm, vmx, exit, halt, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, mtrap, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, pause, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, intrwindow, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE4(vmm, vmx, exit, interrupt, "struct vmx *", "int", "struct vm_exit *", "uint32_t"); SDT_PROBE_DEFINE3(vmm, vmx, exit, nmiwindow, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, inout, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, cpuid, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE5(vmm, vmx, exit, exception, "struct vmx *", "int", "struct vm_exit *", "uint32_t", "int"); SDT_PROBE_DEFINE5(vmm, vmx, exit, nestedfault, "struct vmx *", "int", "struct vm_exit *", "uint64_t", "uint64_t"); SDT_PROBE_DEFINE4(vmm, vmx, exit, mmiofault, "struct vmx *", "int", "struct vm_exit *", "uint64_t"); SDT_PROBE_DEFINE3(vmm, vmx, exit, eoi, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, apicaccess, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE4(vmm, vmx, exit, apicwrite, "struct vmx *", "int", "struct vm_exit *", "struct vlapic *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, xsetbv, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, monitor, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, mwait, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE3(vmm, vmx, exit, vminsn, "struct vmx *", "int", "struct vm_exit *"); SDT_PROBE_DEFINE4(vmm, vmx, exit, unknown, "struct vmx *", "int", "struct vm_exit *", "uint32_t"); SDT_PROBE_DEFINE4(vmm, vmx, exit, return, "struct vmx *", "int", "struct vm_exit *", "int"); /* * Use the last page below 4GB as the APIC access address. This address is * occupied by the boot firmware so it is guaranteed that it will not conflict * with a page in system memory. */ #define APIC_ACCESS_ADDRESS 0xFFFFF000 static int vmx_getdesc(void *arg, int vcpu, int reg, struct seg_desc *desc); static int vmx_getreg(void *arg, int vcpu, int reg, uint64_t *retval); static int vmxctx_setreg(struct vmxctx *vmxctx, int reg, uint64_t val); static void vmx_inject_pir(struct vlapic *vlapic); #ifdef KTR static const char * exit_reason_to_str(int reason) { static char reasonbuf[32]; switch (reason) { case EXIT_REASON_EXCEPTION: return "exception"; case EXIT_REASON_EXT_INTR: return "extint"; case EXIT_REASON_TRIPLE_FAULT: return "triplefault"; case EXIT_REASON_INIT: return "init"; case EXIT_REASON_SIPI: return "sipi"; case EXIT_REASON_IO_SMI: return "iosmi"; case EXIT_REASON_SMI: return "smi"; case EXIT_REASON_INTR_WINDOW: return "intrwindow"; case EXIT_REASON_NMI_WINDOW: return "nmiwindow"; case EXIT_REASON_TASK_SWITCH: return "taskswitch"; case EXIT_REASON_CPUID: return "cpuid"; case EXIT_REASON_GETSEC: return "getsec"; case EXIT_REASON_HLT: return "hlt"; case EXIT_REASON_INVD: return "invd"; case EXIT_REASON_INVLPG: return "invlpg"; case EXIT_REASON_RDPMC: return "rdpmc"; case EXIT_REASON_RDTSC: return "rdtsc"; case EXIT_REASON_RSM: return "rsm"; case EXIT_REASON_VMCALL: return "vmcall"; case EXIT_REASON_VMCLEAR: return "vmclear"; case EXIT_REASON_VMLAUNCH: return "vmlaunch"; case EXIT_REASON_VMPTRLD: return "vmptrld"; case EXIT_REASON_VMPTRST: return "vmptrst"; case EXIT_REASON_VMREAD: return "vmread"; case EXIT_REASON_VMRESUME: return "vmresume"; case EXIT_REASON_VMWRITE: return "vmwrite"; case EXIT_REASON_VMXOFF: return "vmxoff"; case EXIT_REASON_VMXON: return "vmxon"; case EXIT_REASON_CR_ACCESS: return "craccess"; case EXIT_REASON_DR_ACCESS: return "draccess"; case EXIT_REASON_INOUT: return "inout"; case EXIT_REASON_RDMSR: return "rdmsr"; case EXIT_REASON_WRMSR: return "wrmsr"; case EXIT_REASON_INVAL_VMCS: return "invalvmcs"; case EXIT_REASON_INVAL_MSR: return "invalmsr"; case EXIT_REASON_MWAIT: return "mwait"; case EXIT_REASON_MTF: return "mtf"; case EXIT_REASON_MONITOR: return "monitor"; case EXIT_REASON_PAUSE: return "pause"; case EXIT_REASON_MCE_DURING_ENTRY: return "mce-during-entry"; case EXIT_REASON_TPR: return "tpr"; case EXIT_REASON_APIC_ACCESS: return "apic-access"; case EXIT_REASON_GDTR_IDTR: return "gdtridtr"; case EXIT_REASON_LDTR_TR: return "ldtrtr"; case EXIT_REASON_EPT_FAULT: return "eptfault"; case EXIT_REASON_EPT_MISCONFIG: return "eptmisconfig"; case EXIT_REASON_INVEPT: return "invept"; case EXIT_REASON_RDTSCP: return "rdtscp"; case EXIT_REASON_VMX_PREEMPT: return "vmxpreempt"; case EXIT_REASON_INVVPID: return "invvpid"; case EXIT_REASON_WBINVD: return "wbinvd"; case EXIT_REASON_XSETBV: return "xsetbv"; case EXIT_REASON_APIC_WRITE: return "apic-write"; default: snprintf(reasonbuf, sizeof(reasonbuf), "%d", reason); return (reasonbuf); } } #endif /* KTR */ static int vmx_allow_x2apic_msrs(struct vmx *vmx) { int i, error; error = 0; /* * Allow readonly access to the following x2APIC MSRs from the guest. */ error += guest_msr_ro(vmx, MSR_APIC_ID); error += guest_msr_ro(vmx, MSR_APIC_VERSION); error += guest_msr_ro(vmx, MSR_APIC_LDR); error += guest_msr_ro(vmx, MSR_APIC_SVR); for (i = 0; i < 8; i++) error += guest_msr_ro(vmx, MSR_APIC_ISR0 + i); for (i = 0; i < 8; i++) error += guest_msr_ro(vmx, MSR_APIC_TMR0 + i); for (i = 0; i < 8; i++) error += guest_msr_ro(vmx, MSR_APIC_IRR0 + i); error += guest_msr_ro(vmx, MSR_APIC_ESR); error += guest_msr_ro(vmx, MSR_APIC_LVT_TIMER); error += guest_msr_ro(vmx, MSR_APIC_LVT_THERMAL); error += guest_msr_ro(vmx, MSR_APIC_LVT_PCINT); error += guest_msr_ro(vmx, MSR_APIC_LVT_LINT0); error += guest_msr_ro(vmx, MSR_APIC_LVT_LINT1); error += guest_msr_ro(vmx, MSR_APIC_LVT_ERROR); error += guest_msr_ro(vmx, MSR_APIC_ICR_TIMER); error += guest_msr_ro(vmx, MSR_APIC_DCR_TIMER); error += guest_msr_ro(vmx, MSR_APIC_ICR); /* * Allow TPR, EOI and SELF_IPI MSRs to be read and written by the guest. * * These registers get special treatment described in the section * "Virtualizing MSR-Based APIC Accesses". */ error += guest_msr_rw(vmx, MSR_APIC_TPR); error += guest_msr_rw(vmx, MSR_APIC_EOI); error += guest_msr_rw(vmx, MSR_APIC_SELF_IPI); return (error); } u_long vmx_fix_cr0(u_long cr0) { return ((cr0 | cr0_ones_mask) & ~cr0_zeros_mask); } u_long vmx_fix_cr4(u_long cr4) { return ((cr4 | cr4_ones_mask) & ~cr4_zeros_mask); } static void vpid_free(int vpid) { if (vpid < 0 || vpid > 0xffff) panic("vpid_free: invalid vpid %d", vpid); /* * VPIDs [0,VM_MAXCPU] are special and are not allocated from * the unit number allocator. */ if (vpid > VM_MAXCPU) free_unr(vpid_unr, vpid); } static void vpid_alloc(uint16_t *vpid, int num) { int i, x; if (num <= 0 || num > VM_MAXCPU) panic("invalid number of vpids requested: %d", num); /* * If the "enable vpid" execution control is not enabled then the * VPID is required to be 0 for all vcpus. */ if ((procbased_ctls2 & PROCBASED2_ENABLE_VPID) == 0) { for (i = 0; i < num; i++) vpid[i] = 0; return; } /* * Allocate a unique VPID for each vcpu from the unit number allocator. */ for (i = 0; i < num; i++) { x = alloc_unr(vpid_unr); if (x == -1) break; else vpid[i] = x; } if (i < num) { atomic_add_int(&vpid_alloc_failed, 1); /* * If the unit number allocator does not have enough unique * VPIDs then we need to allocate from the [1,VM_MAXCPU] range. * * These VPIDs are not be unique across VMs but this does not * affect correctness because the combined mappings are also * tagged with the EP4TA which is unique for each VM. * * It is still sub-optimal because the invvpid will invalidate * combined mappings for a particular VPID across all EP4TAs. */ while (i-- > 0) vpid_free(vpid[i]); for (i = 0; i < num; i++) vpid[i] = i + 1; } } static void vpid_init(void) { /* * VPID 0 is required when the "enable VPID" execution control is * disabled. * * VPIDs [1,VM_MAXCPU] are used as the "overflow namespace" when the * unit number allocator does not have sufficient unique VPIDs to * satisfy the allocation. * * The remaining VPIDs are managed by the unit number allocator. */ vpid_unr = new_unrhdr(VM_MAXCPU + 1, 0xffff, NULL); } static void vmx_disable(void *arg __unused) { struct invvpid_desc invvpid_desc = { 0 }; struct invept_desc invept_desc = { 0 }; if (vmxon_enabled[curcpu]) { /* * See sections 25.3.3.3 and 25.3.3.4 in Intel Vol 3b. * * VMXON or VMXOFF are not required to invalidate any TLB * caching structures. This prevents potential retention of * cached information in the TLB between distinct VMX episodes. */ invvpid(INVVPID_TYPE_ALL_CONTEXTS, invvpid_desc); invept(INVEPT_TYPE_ALL_CONTEXTS, invept_desc); vmxoff(); } load_cr4(rcr4() & ~CR4_VMXE); } static int vmx_cleanup(void) { if (pirvec >= 0) lapic_ipi_free(pirvec); if (vpid_unr != NULL) { delete_unrhdr(vpid_unr); vpid_unr = NULL; } if (nmi_flush_l1d_sw == 1) nmi_flush_l1d_sw = 0; smp_rendezvous(NULL, vmx_disable, NULL, NULL); return (0); } static void vmx_enable(void *arg __unused) { int error; uint64_t feature_control; feature_control = rdmsr(MSR_IA32_FEATURE_CONTROL); if ((feature_control & IA32_FEATURE_CONTROL_LOCK) == 0 || (feature_control & IA32_FEATURE_CONTROL_VMX_EN) == 0) { wrmsr(MSR_IA32_FEATURE_CONTROL, feature_control | IA32_FEATURE_CONTROL_VMX_EN | IA32_FEATURE_CONTROL_LOCK); } load_cr4(rcr4() | CR4_VMXE); *(uint32_t *)vmxon_region[curcpu] = vmx_revision(); error = vmxon(vmxon_region[curcpu]); if (error == 0) vmxon_enabled[curcpu] = 1; } static void vmx_restore(void) { if (vmxon_enabled[curcpu]) vmxon(vmxon_region[curcpu]); } static int vmx_init(int ipinum) { int error, use_tpr_shadow; uint64_t basic, fixed0, fixed1, feature_control; uint32_t tmp, procbased2_vid_bits; /* CPUID.1:ECX[bit 5] must be 1 for processor to support VMX */ if (!(cpu_feature2 & CPUID2_VMX)) { printf("vmx_init: processor does not support VMX operation\n"); return (ENXIO); } /* * Verify that MSR_IA32_FEATURE_CONTROL lock and VMXON enable bits * are set (bits 0 and 2 respectively). */ feature_control = rdmsr(MSR_IA32_FEATURE_CONTROL); if ((feature_control & IA32_FEATURE_CONTROL_LOCK) == 1 && (feature_control & IA32_FEATURE_CONTROL_VMX_EN) == 0) { printf("vmx_init: VMX operation disabled by BIOS\n"); return (ENXIO); } /* * Verify capabilities MSR_VMX_BASIC: * - bit 54 indicates support for INS/OUTS decoding */ basic = rdmsr(MSR_VMX_BASIC); if ((basic & (1UL << 54)) == 0) { printf("vmx_init: processor does not support desired basic " "capabilities\n"); return (EINVAL); } /* Check support for primary processor-based VM-execution controls */ error = vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS, MSR_VMX_TRUE_PROCBASED_CTLS, PROCBASED_CTLS_ONE_SETTING, PROCBASED_CTLS_ZERO_SETTING, &procbased_ctls); if (error) { printf("vmx_init: processor does not support desired primary " "processor-based controls\n"); return (error); } /* Clear the processor-based ctl bits that are set on demand */ procbased_ctls &= ~PROCBASED_CTLS_WINDOW_SETTING; /* Check support for secondary processor-based VM-execution controls */ error = vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2, PROCBASED_CTLS2_ONE_SETTING, PROCBASED_CTLS2_ZERO_SETTING, &procbased_ctls2); if (error) { printf("vmx_init: processor does not support desired secondary " "processor-based controls\n"); return (error); } /* Check support for VPID */ error = vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2, PROCBASED2_ENABLE_VPID, 0, &tmp); if (error == 0) procbased_ctls2 |= PROCBASED2_ENABLE_VPID; /* Check support for pin-based VM-execution controls */ error = vmx_set_ctlreg(MSR_VMX_PINBASED_CTLS, MSR_VMX_TRUE_PINBASED_CTLS, PINBASED_CTLS_ONE_SETTING, PINBASED_CTLS_ZERO_SETTING, &pinbased_ctls); if (error) { printf("vmx_init: processor does not support desired " "pin-based controls\n"); return (error); } /* Check support for VM-exit controls */ error = vmx_set_ctlreg(MSR_VMX_EXIT_CTLS, MSR_VMX_TRUE_EXIT_CTLS, VM_EXIT_CTLS_ONE_SETTING, VM_EXIT_CTLS_ZERO_SETTING, &exit_ctls); if (error) { printf("vmx_init: processor does not support desired " "exit controls\n"); return (error); } /* Check support for VM-entry controls */ error = vmx_set_ctlreg(MSR_VMX_ENTRY_CTLS, MSR_VMX_TRUE_ENTRY_CTLS, VM_ENTRY_CTLS_ONE_SETTING, VM_ENTRY_CTLS_ZERO_SETTING, &entry_ctls); if (error) { printf("vmx_init: processor does not support desired " "entry controls\n"); return (error); } /* * Check support for optional features by testing them * as individual bits */ cap_halt_exit = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS, MSR_VMX_TRUE_PROCBASED_CTLS, PROCBASED_HLT_EXITING, 0, &tmp) == 0); cap_monitor_trap = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS, MSR_VMX_PROCBASED_CTLS, PROCBASED_MTF, 0, &tmp) == 0); cap_pause_exit = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS, MSR_VMX_TRUE_PROCBASED_CTLS, PROCBASED_PAUSE_EXITING, 0, &tmp) == 0); cap_unrestricted_guest = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2, PROCBASED2_UNRESTRICTED_GUEST, 0, &tmp) == 0); cap_invpcid = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2, PROCBASED2_ENABLE_INVPCID, 0, &tmp) == 0); /* * Check support for virtual interrupt delivery. */ procbased2_vid_bits = (PROCBASED2_VIRTUALIZE_APIC_ACCESSES | PROCBASED2_VIRTUALIZE_X2APIC_MODE | PROCBASED2_APIC_REGISTER_VIRTUALIZATION | PROCBASED2_VIRTUAL_INTERRUPT_DELIVERY); use_tpr_shadow = (vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS, MSR_VMX_TRUE_PROCBASED_CTLS, PROCBASED_USE_TPR_SHADOW, 0, &tmp) == 0); error = vmx_set_ctlreg(MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2, procbased2_vid_bits, 0, &tmp); if (error == 0 && use_tpr_shadow) { virtual_interrupt_delivery = 1; TUNABLE_INT_FETCH("hw.vmm.vmx.use_apic_vid", &virtual_interrupt_delivery); } if (virtual_interrupt_delivery) { procbased_ctls |= PROCBASED_USE_TPR_SHADOW; procbased_ctls2 |= procbased2_vid_bits; procbased_ctls2 &= ~PROCBASED2_VIRTUALIZE_X2APIC_MODE; /* * No need to emulate accesses to %CR8 if virtual * interrupt delivery is enabled. */ procbased_ctls &= ~PROCBASED_CR8_LOAD_EXITING; procbased_ctls &= ~PROCBASED_CR8_STORE_EXITING; /* * Check for Posted Interrupts only if Virtual Interrupt * Delivery is enabled. */ error = vmx_set_ctlreg(MSR_VMX_PINBASED_CTLS, MSR_VMX_TRUE_PINBASED_CTLS, PINBASED_POSTED_INTERRUPT, 0, &tmp); if (error == 0) { pirvec = lapic_ipi_alloc(pti ? &IDTVEC(justreturn1_pti) : &IDTVEC(justreturn)); if (pirvec < 0) { if (bootverbose) { printf("vmx_init: unable to allocate " "posted interrupt vector\n"); } } else { posted_interrupts = 1; TUNABLE_INT_FETCH("hw.vmm.vmx.use_apic_pir", &posted_interrupts); } } } if (posted_interrupts) pinbased_ctls |= PINBASED_POSTED_INTERRUPT; /* Initialize EPT */ error = ept_init(ipinum); if (error) { printf("vmx_init: ept initialization failed (%d)\n", error); return (error); } - guest_l1d_flush = (cpu_ia32_arch_caps & IA32_ARCH_CAP_RDCL_NO) == 0; + guest_l1d_flush = (cpu_ia32_arch_caps & + IA32_ARCH_CAP_SKIP_L1DFL_VMENTRY) == 0; TUNABLE_INT_FETCH("hw.vmm.l1d_flush", &guest_l1d_flush); /* * L1D cache flush is enabled. Use IA32_FLUSH_CMD MSR when * available. Otherwise fall back to the software flush * method which loads enough data from the kernel text to * flush existing L1D content, both on VMX entry and on NMI * return. */ if (guest_l1d_flush) { if ((cpu_stdext_feature3 & CPUID_STDEXT3_L1D_FLUSH) == 0) { guest_l1d_flush_sw = 1; TUNABLE_INT_FETCH("hw.vmm.l1d_flush_sw", &guest_l1d_flush_sw); } if (guest_l1d_flush_sw) { if (nmi_flush_l1d_sw <= 1) nmi_flush_l1d_sw = 1; } else { msr_load_list[0].index = MSR_IA32_FLUSH_CMD; msr_load_list[0].val = IA32_FLUSH_CMD_L1D; } } /* * Stash the cr0 and cr4 bits that must be fixed to 0 or 1 */ fixed0 = rdmsr(MSR_VMX_CR0_FIXED0); fixed1 = rdmsr(MSR_VMX_CR0_FIXED1); cr0_ones_mask = fixed0 & fixed1; cr0_zeros_mask = ~fixed0 & ~fixed1; /* * CR0_PE and CR0_PG can be set to zero in VMX non-root operation * if unrestricted guest execution is allowed. */ if (cap_unrestricted_guest) cr0_ones_mask &= ~(CR0_PG | CR0_PE); /* * Do not allow the guest to set CR0_NW or CR0_CD. */ cr0_zeros_mask |= (CR0_NW | CR0_CD); fixed0 = rdmsr(MSR_VMX_CR4_FIXED0); fixed1 = rdmsr(MSR_VMX_CR4_FIXED1); cr4_ones_mask = fixed0 & fixed1; cr4_zeros_mask = ~fixed0 & ~fixed1; vpid_init(); vmx_msr_init(); /* enable VMX operation */ smp_rendezvous(NULL, vmx_enable, NULL, NULL); vmx_initialized = 1; return (0); } static void vmx_trigger_hostintr(int vector) { uintptr_t func; struct gate_descriptor *gd; gd = &idt[vector]; KASSERT(vector >= 32 && vector <= 255, ("vmx_trigger_hostintr: " "invalid vector %d", vector)); KASSERT(gd->gd_p == 1, ("gate descriptor for vector %d not present", vector)); KASSERT(gd->gd_type == SDT_SYSIGT, ("gate descriptor for vector %d " "has invalid type %d", vector, gd->gd_type)); KASSERT(gd->gd_dpl == SEL_KPL, ("gate descriptor for vector %d " "has invalid dpl %d", vector, gd->gd_dpl)); KASSERT(gd->gd_selector == GSEL(GCODE_SEL, SEL_KPL), ("gate descriptor " "for vector %d has invalid selector %d", vector, gd->gd_selector)); KASSERT(gd->gd_ist == 0, ("gate descriptor for vector %d has invalid " "IST %d", vector, gd->gd_ist)); func = ((long)gd->gd_hioffset << 16 | gd->gd_looffset); vmx_call_isr(func); } static int vmx_setup_cr_shadow(int which, struct vmcs *vmcs, uint32_t initial) { int error, mask_ident, shadow_ident; uint64_t mask_value; if (which != 0 && which != 4) panic("vmx_setup_cr_shadow: unknown cr%d", which); if (which == 0) { mask_ident = VMCS_CR0_MASK; mask_value = cr0_ones_mask | cr0_zeros_mask; shadow_ident = VMCS_CR0_SHADOW; } else { mask_ident = VMCS_CR4_MASK; mask_value = cr4_ones_mask | cr4_zeros_mask; shadow_ident = VMCS_CR4_SHADOW; } error = vmcs_setreg(vmcs, 0, VMCS_IDENT(mask_ident), mask_value); if (error) return (error); error = vmcs_setreg(vmcs, 0, VMCS_IDENT(shadow_ident), initial); if (error) return (error); return (0); } #define vmx_setup_cr0_shadow(vmcs,init) vmx_setup_cr_shadow(0, (vmcs), (init)) #define vmx_setup_cr4_shadow(vmcs,init) vmx_setup_cr_shadow(4, (vmcs), (init)) static void * vmx_vminit(struct vm *vm, pmap_t pmap) { uint16_t vpid[VM_MAXCPU]; int i, error; struct vmx *vmx; struct vmcs *vmcs; uint32_t exc_bitmap; vmx = malloc(sizeof(struct vmx), M_VMX, M_WAITOK | M_ZERO); if ((uintptr_t)vmx & PAGE_MASK) { panic("malloc of struct vmx not aligned on %d byte boundary", PAGE_SIZE); } vmx->vm = vm; vmx->eptp = eptp(vtophys((vm_offset_t)pmap->pm_pml4)); /* * Clean up EPTP-tagged guest physical and combined mappings * * VMX transitions are not required to invalidate any guest physical * mappings. So, it may be possible for stale guest physical mappings * to be present in the processor TLBs. * * Combined mappings for this EP4TA are also invalidated for all VPIDs. */ ept_invalidate_mappings(vmx->eptp); msr_bitmap_initialize(vmx->msr_bitmap); /* * It is safe to allow direct access to MSR_GSBASE and MSR_FSBASE. * The guest FSBASE and GSBASE are saved and restored during * vm-exit and vm-entry respectively. The host FSBASE and GSBASE are * always restored from the vmcs host state area on vm-exit. * * The SYSENTER_CS/ESP/EIP MSRs are identical to FS/GSBASE in * how they are saved/restored so can be directly accessed by the * guest. * * MSR_EFER is saved and restored in the guest VMCS area on a * VM exit and entry respectively. It is also restored from the * host VMCS area on a VM exit. * * The TSC MSR is exposed read-only. Writes are disallowed as * that will impact the host TSC. If the guest does a write * the "use TSC offsetting" execution control is enabled and the * difference between the host TSC and the guest TSC is written * into the TSC offset in the VMCS. */ if (guest_msr_rw(vmx, MSR_GSBASE) || guest_msr_rw(vmx, MSR_FSBASE) || guest_msr_rw(vmx, MSR_SYSENTER_CS_MSR) || guest_msr_rw(vmx, MSR_SYSENTER_ESP_MSR) || guest_msr_rw(vmx, MSR_SYSENTER_EIP_MSR) || guest_msr_rw(vmx, MSR_EFER) || guest_msr_ro(vmx, MSR_TSC)) panic("vmx_vminit: error setting guest msr access"); vpid_alloc(vpid, VM_MAXCPU); if (virtual_interrupt_delivery) { error = vm_map_mmio(vm, DEFAULT_APIC_BASE, PAGE_SIZE, APIC_ACCESS_ADDRESS); /* XXX this should really return an error to the caller */ KASSERT(error == 0, ("vm_map_mmio(apicbase) error %d", error)); } for (i = 0; i < VM_MAXCPU; i++) { vmcs = &vmx->vmcs[i]; vmcs->identifier = vmx_revision(); error = vmclear(vmcs); if (error != 0) { panic("vmx_vminit: vmclear error %d on vcpu %d\n", error, i); } vmx_msr_guest_init(vmx, i); error = vmcs_init(vmcs); KASSERT(error == 0, ("vmcs_init error %d", error)); VMPTRLD(vmcs); error = 0; error += vmwrite(VMCS_HOST_RSP, (u_long)&vmx->ctx[i]); error += vmwrite(VMCS_EPTP, vmx->eptp); error += vmwrite(VMCS_PIN_BASED_CTLS, pinbased_ctls); error += vmwrite(VMCS_PRI_PROC_BASED_CTLS, procbased_ctls); error += vmwrite(VMCS_SEC_PROC_BASED_CTLS, procbased_ctls2); error += vmwrite(VMCS_EXIT_CTLS, exit_ctls); error += vmwrite(VMCS_ENTRY_CTLS, entry_ctls); error += vmwrite(VMCS_MSR_BITMAP, vtophys(vmx->msr_bitmap)); error += vmwrite(VMCS_VPID, vpid[i]); if (guest_l1d_flush && !guest_l1d_flush_sw) { vmcs_write(VMCS_ENTRY_MSR_LOAD, pmap_kextract( (vm_offset_t)&msr_load_list[0])); vmcs_write(VMCS_ENTRY_MSR_LOAD_COUNT, nitems(msr_load_list)); vmcs_write(VMCS_EXIT_MSR_STORE, 0); vmcs_write(VMCS_EXIT_MSR_STORE_COUNT, 0); } /* exception bitmap */ if (vcpu_trace_exceptions(vm, i)) exc_bitmap = 0xffffffff; else exc_bitmap = 1 << IDT_MC; error += vmwrite(VMCS_EXCEPTION_BITMAP, exc_bitmap); vmx->ctx[i].guest_dr6 = DBREG_DR6_RESERVED1; error += vmwrite(VMCS_GUEST_DR7, DBREG_DR7_RESERVED1); if (virtual_interrupt_delivery) { error += vmwrite(VMCS_APIC_ACCESS, APIC_ACCESS_ADDRESS); error += vmwrite(VMCS_VIRTUAL_APIC, vtophys(&vmx->apic_page[i])); error += vmwrite(VMCS_EOI_EXIT0, 0); error += vmwrite(VMCS_EOI_EXIT1, 0); error += vmwrite(VMCS_EOI_EXIT2, 0); error += vmwrite(VMCS_EOI_EXIT3, 0); } if (posted_interrupts) { error += vmwrite(VMCS_PIR_VECTOR, pirvec); error += vmwrite(VMCS_PIR_DESC, vtophys(&vmx->pir_desc[i])); } VMCLEAR(vmcs); KASSERT(error == 0, ("vmx_vminit: error customizing the vmcs")); vmx->cap[i].set = 0; vmx->cap[i].proc_ctls = procbased_ctls; vmx->cap[i].proc_ctls2 = procbased_ctls2; vmx->state[i].nextrip = ~0; vmx->state[i].lastcpu = NOCPU; vmx->state[i].vpid = vpid[i]; /* * Set up the CR0/4 shadows, and init the read shadow * to the power-on register value from the Intel Sys Arch. * CR0 - 0x60000010 * CR4 - 0 */ error = vmx_setup_cr0_shadow(vmcs, 0x60000010); if (error != 0) panic("vmx_setup_cr0_shadow %d", error); error = vmx_setup_cr4_shadow(vmcs, 0); if (error != 0) panic("vmx_setup_cr4_shadow %d", error); vmx->ctx[i].pmap = pmap; } return (vmx); } static int vmx_handle_cpuid(struct vm *vm, int vcpu, struct vmxctx *vmxctx) { int handled, func; func = vmxctx->guest_rax; handled = x86_emulate_cpuid(vm, vcpu, (uint32_t*)(&vmxctx->guest_rax), (uint32_t*)(&vmxctx->guest_rbx), (uint32_t*)(&vmxctx->guest_rcx), (uint32_t*)(&vmxctx->guest_rdx)); return (handled); } static __inline void vmx_run_trace(struct vmx *vmx, int vcpu) { #ifdef KTR VCPU_CTR1(vmx->vm, vcpu, "Resume execution at %#lx", vmcs_guest_rip()); #endif } static __inline void vmx_exit_trace(struct vmx *vmx, int vcpu, uint64_t rip, uint32_t exit_reason, int handled) { #ifdef KTR VCPU_CTR3(vmx->vm, vcpu, "%s %s vmexit at 0x%0lx", handled ? "handled" : "unhandled", exit_reason_to_str(exit_reason), rip); #endif } static __inline void vmx_astpending_trace(struct vmx *vmx, int vcpu, uint64_t rip) { #ifdef KTR VCPU_CTR1(vmx->vm, vcpu, "astpending vmexit at 0x%0lx", rip); #endif } static VMM_STAT_INTEL(VCPU_INVVPID_SAVED, "Number of vpid invalidations saved"); static VMM_STAT_INTEL(VCPU_INVVPID_DONE, "Number of vpid invalidations done"); /* * Invalidate guest mappings identified by its vpid from the TLB. */ static __inline void vmx_invvpid(struct vmx *vmx, int vcpu, pmap_t pmap, int running) { struct vmxstate *vmxstate; struct invvpid_desc invvpid_desc; vmxstate = &vmx->state[vcpu]; if (vmxstate->vpid == 0) return; if (!running) { /* * Set the 'lastcpu' to an invalid host cpu. * * This will invalidate TLB entries tagged with the vcpu's * vpid the next time it runs via vmx_set_pcpu_defaults(). */ vmxstate->lastcpu = NOCPU; return; } KASSERT(curthread->td_critnest > 0, ("%s: vcpu %d running outside " "critical section", __func__, vcpu)); /* * Invalidate all mappings tagged with 'vpid' * * We do this because this vcpu was executing on a different host * cpu when it last ran. We do not track whether it invalidated * mappings associated with its 'vpid' during that run. So we must * assume that the mappings associated with 'vpid' on 'curcpu' are * stale and invalidate them. * * Note that we incur this penalty only when the scheduler chooses to * move the thread associated with this vcpu between host cpus. * * Note also that this will invalidate mappings tagged with 'vpid' * for "all" EP4TAs. */ if (pmap->pm_eptgen == vmx->eptgen[curcpu]) { invvpid_desc._res1 = 0; invvpid_desc._res2 = 0; invvpid_desc.vpid = vmxstate->vpid; invvpid_desc.linear_addr = 0; invvpid(INVVPID_TYPE_SINGLE_CONTEXT, invvpid_desc); vmm_stat_incr(vmx->vm, vcpu, VCPU_INVVPID_DONE, 1); } else { /* * The invvpid can be skipped if an invept is going to * be performed before entering the guest. The invept * will invalidate combined mappings tagged with * 'vmx->eptp' for all vpids. */ vmm_stat_incr(vmx->vm, vcpu, VCPU_INVVPID_SAVED, 1); } } static void vmx_set_pcpu_defaults(struct vmx *vmx, int vcpu, pmap_t pmap) { struct vmxstate *vmxstate; vmxstate = &vmx->state[vcpu]; if (vmxstate->lastcpu == curcpu) return; vmxstate->lastcpu = curcpu; vmm_stat_incr(vmx->vm, vcpu, VCPU_MIGRATIONS, 1); vmcs_write(VMCS_HOST_TR_BASE, vmm_get_host_trbase()); vmcs_write(VMCS_HOST_GDTR_BASE, vmm_get_host_gdtrbase()); vmcs_write(VMCS_HOST_GS_BASE, vmm_get_host_gsbase()); vmx_invvpid(vmx, vcpu, pmap, 1); } /* * We depend on 'procbased_ctls' to have the Interrupt Window Exiting bit set. */ CTASSERT((PROCBASED_CTLS_ONE_SETTING & PROCBASED_INT_WINDOW_EXITING) != 0); static void __inline vmx_set_int_window_exiting(struct vmx *vmx, int vcpu) { if ((vmx->cap[vcpu].proc_ctls & PROCBASED_INT_WINDOW_EXITING) == 0) { vmx->cap[vcpu].proc_ctls |= PROCBASED_INT_WINDOW_EXITING; vmcs_write(VMCS_PRI_PROC_BASED_CTLS, vmx->cap[vcpu].proc_ctls); VCPU_CTR0(vmx->vm, vcpu, "Enabling interrupt window exiting"); } } static void __inline vmx_clear_int_window_exiting(struct vmx *vmx, int vcpu) { KASSERT((vmx->cap[vcpu].proc_ctls & PROCBASED_INT_WINDOW_EXITING) != 0, ("intr_window_exiting not set: %#x", vmx->cap[vcpu].proc_ctls)); vmx->cap[vcpu].proc_ctls &= ~PROCBASED_INT_WINDOW_EXITING; vmcs_write(VMCS_PRI_PROC_BASED_CTLS, vmx->cap[vcpu].proc_ctls); VCPU_CTR0(vmx->vm, vcpu, "Disabling interrupt window exiting"); } static void __inline vmx_set_nmi_window_exiting(struct vmx *vmx, int vcpu) { if ((vmx->cap[vcpu].proc_ctls & PROCBASED_NMI_WINDOW_EXITING) == 0) { vmx->cap[vcpu].proc_ctls |= PROCBASED_NMI_WINDOW_EXITING; vmcs_write(VMCS_PRI_PROC_BASED_CTLS, vmx->cap[vcpu].proc_ctls); VCPU_CTR0(vmx->vm, vcpu, "Enabling NMI window exiting"); } } static void __inline vmx_clear_nmi_window_exiting(struct vmx *vmx, int vcpu) { KASSERT((vmx->cap[vcpu].proc_ctls & PROCBASED_NMI_WINDOW_EXITING) != 0, ("nmi_window_exiting not set %#x", vmx->cap[vcpu].proc_ctls)); vmx->cap[vcpu].proc_ctls &= ~PROCBASED_NMI_WINDOW_EXITING; vmcs_write(VMCS_PRI_PROC_BASED_CTLS, vmx->cap[vcpu].proc_ctls); VCPU_CTR0(vmx->vm, vcpu, "Disabling NMI window exiting"); } int vmx_set_tsc_offset(struct vmx *vmx, int vcpu, uint64_t offset) { int error; if ((vmx->cap[vcpu].proc_ctls & PROCBASED_TSC_OFFSET) == 0) { vmx->cap[vcpu].proc_ctls |= PROCBASED_TSC_OFFSET; vmcs_write(VMCS_PRI_PROC_BASED_CTLS, vmx->cap[vcpu].proc_ctls); VCPU_CTR0(vmx->vm, vcpu, "Enabling TSC offsetting"); } error = vmwrite(VMCS_TSC_OFFSET, offset); return (error); } #define NMI_BLOCKING (VMCS_INTERRUPTIBILITY_NMI_BLOCKING | \ VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING) #define HWINTR_BLOCKING (VMCS_INTERRUPTIBILITY_STI_BLOCKING | \ VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING) static void vmx_inject_nmi(struct vmx *vmx, int vcpu) { uint32_t gi, info; gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); KASSERT((gi & NMI_BLOCKING) == 0, ("vmx_inject_nmi: invalid guest " "interruptibility-state %#x", gi)); info = vmcs_read(VMCS_ENTRY_INTR_INFO); KASSERT((info & VMCS_INTR_VALID) == 0, ("vmx_inject_nmi: invalid " "VM-entry interruption information %#x", info)); /* * Inject the virtual NMI. The vector must be the NMI IDT entry * or the VMCS entry check will fail. */ info = IDT_NMI | VMCS_INTR_T_NMI | VMCS_INTR_VALID; vmcs_write(VMCS_ENTRY_INTR_INFO, info); VCPU_CTR0(vmx->vm, vcpu, "Injecting vNMI"); /* Clear the request */ vm_nmi_clear(vmx->vm, vcpu); } static void vmx_inject_interrupts(struct vmx *vmx, int vcpu, struct vlapic *vlapic, uint64_t guestrip) { int vector, need_nmi_exiting, extint_pending; uint64_t rflags, entryinfo; uint32_t gi, info; if (vmx->state[vcpu].nextrip != guestrip) { gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); if (gi & HWINTR_BLOCKING) { VCPU_CTR2(vmx->vm, vcpu, "Guest interrupt blocking " "cleared due to rip change: %#lx/%#lx", vmx->state[vcpu].nextrip, guestrip); gi &= ~HWINTR_BLOCKING; vmcs_write(VMCS_GUEST_INTERRUPTIBILITY, gi); } } if (vm_entry_intinfo(vmx->vm, vcpu, &entryinfo)) { KASSERT((entryinfo & VMCS_INTR_VALID) != 0, ("%s: entry " "intinfo is not valid: %#lx", __func__, entryinfo)); info = vmcs_read(VMCS_ENTRY_INTR_INFO); KASSERT((info & VMCS_INTR_VALID) == 0, ("%s: cannot inject " "pending exception: %#lx/%#x", __func__, entryinfo, info)); info = entryinfo; vector = info & 0xff; if (vector == IDT_BP || vector == IDT_OF) { /* * VT-x requires #BP and #OF to be injected as software * exceptions. */ info &= ~VMCS_INTR_T_MASK; info |= VMCS_INTR_T_SWEXCEPTION; } if (info & VMCS_INTR_DEL_ERRCODE) vmcs_write(VMCS_ENTRY_EXCEPTION_ERROR, entryinfo >> 32); vmcs_write(VMCS_ENTRY_INTR_INFO, info); } if (vm_nmi_pending(vmx->vm, vcpu)) { /* * If there are no conditions blocking NMI injection then * inject it directly here otherwise enable "NMI window * exiting" to inject it as soon as we can. * * We also check for STI_BLOCKING because some implementations * don't allow NMI injection in this case. If we are running * on a processor that doesn't have this restriction it will * immediately exit and the NMI will be injected in the * "NMI window exiting" handler. */ need_nmi_exiting = 1; gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); if ((gi & (HWINTR_BLOCKING | NMI_BLOCKING)) == 0) { info = vmcs_read(VMCS_ENTRY_INTR_INFO); if ((info & VMCS_INTR_VALID) == 0) { vmx_inject_nmi(vmx, vcpu); need_nmi_exiting = 0; } else { VCPU_CTR1(vmx->vm, vcpu, "Cannot inject NMI " "due to VM-entry intr info %#x", info); } } else { VCPU_CTR1(vmx->vm, vcpu, "Cannot inject NMI due to " "Guest Interruptibility-state %#x", gi); } if (need_nmi_exiting) vmx_set_nmi_window_exiting(vmx, vcpu); } extint_pending = vm_extint_pending(vmx->vm, vcpu); if (!extint_pending && virtual_interrupt_delivery) { vmx_inject_pir(vlapic); return; } /* * If interrupt-window exiting is already in effect then don't bother * checking for pending interrupts. This is just an optimization and * not needed for correctness. */ if ((vmx->cap[vcpu].proc_ctls & PROCBASED_INT_WINDOW_EXITING) != 0) { VCPU_CTR0(vmx->vm, vcpu, "Skip interrupt injection due to " "pending int_window_exiting"); return; } if (!extint_pending) { /* Ask the local apic for a vector to inject */ if (!vlapic_pending_intr(vlapic, &vector)) return; /* * From the Intel SDM, Volume 3, Section "Maskable * Hardware Interrupts": * - maskable interrupt vectors [16,255] can be delivered * through the local APIC. */ KASSERT(vector >= 16 && vector <= 255, ("invalid vector %d from local APIC", vector)); } else { /* Ask the legacy pic for a vector to inject */ vatpic_pending_intr(vmx->vm, &vector); /* * From the Intel SDM, Volume 3, Section "Maskable * Hardware Interrupts": * - maskable interrupt vectors [0,255] can be delivered * through the INTR pin. */ KASSERT(vector >= 0 && vector <= 255, ("invalid vector %d from INTR", vector)); } /* Check RFLAGS.IF and the interruptibility state of the guest */ rflags = vmcs_read(VMCS_GUEST_RFLAGS); if ((rflags & PSL_I) == 0) { VCPU_CTR2(vmx->vm, vcpu, "Cannot inject vector %d due to " "rflags %#lx", vector, rflags); goto cantinject; } gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); if (gi & HWINTR_BLOCKING) { VCPU_CTR2(vmx->vm, vcpu, "Cannot inject vector %d due to " "Guest Interruptibility-state %#x", vector, gi); goto cantinject; } info = vmcs_read(VMCS_ENTRY_INTR_INFO); if (info & VMCS_INTR_VALID) { /* * This is expected and could happen for multiple reasons: * - A vectoring VM-entry was aborted due to astpending * - A VM-exit happened during event injection. * - An exception was injected above. * - An NMI was injected above or after "NMI window exiting" */ VCPU_CTR2(vmx->vm, vcpu, "Cannot inject vector %d due to " "VM-entry intr info %#x", vector, info); goto cantinject; } /* Inject the interrupt */ info = VMCS_INTR_T_HWINTR | VMCS_INTR_VALID; info |= vector; vmcs_write(VMCS_ENTRY_INTR_INFO, info); if (!extint_pending) { /* Update the Local APIC ISR */ vlapic_intr_accepted(vlapic, vector); } else { vm_extint_clear(vmx->vm, vcpu); vatpic_intr_accepted(vmx->vm, vector); /* * After we accepted the current ExtINT the PIC may * have posted another one. If that is the case, set * the Interrupt Window Exiting execution control so * we can inject that one too. * * Also, interrupt window exiting allows us to inject any * pending APIC vector that was preempted by the ExtINT * as soon as possible. This applies both for the software * emulated vlapic and the hardware assisted virtual APIC. */ vmx_set_int_window_exiting(vmx, vcpu); } VCPU_CTR1(vmx->vm, vcpu, "Injecting hwintr at vector %d", vector); return; cantinject: /* * Set the Interrupt Window Exiting execution control so we can inject * the interrupt as soon as blocking condition goes away. */ vmx_set_int_window_exiting(vmx, vcpu); } /* * If the Virtual NMIs execution control is '1' then the logical processor * tracks virtual-NMI blocking in the Guest Interruptibility-state field of * the VMCS. An IRET instruction in VMX non-root operation will remove any * virtual-NMI blocking. * * This unblocking occurs even if the IRET causes a fault. In this case the * hypervisor needs to restore virtual-NMI blocking before resuming the guest. */ static void vmx_restore_nmi_blocking(struct vmx *vmx, int vcpuid) { uint32_t gi; VCPU_CTR0(vmx->vm, vcpuid, "Restore Virtual-NMI blocking"); gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); gi |= VMCS_INTERRUPTIBILITY_NMI_BLOCKING; vmcs_write(VMCS_GUEST_INTERRUPTIBILITY, gi); } static void vmx_clear_nmi_blocking(struct vmx *vmx, int vcpuid) { uint32_t gi; VCPU_CTR0(vmx->vm, vcpuid, "Clear Virtual-NMI blocking"); gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); gi &= ~VMCS_INTERRUPTIBILITY_NMI_BLOCKING; vmcs_write(VMCS_GUEST_INTERRUPTIBILITY, gi); } static void vmx_assert_nmi_blocking(struct vmx *vmx, int vcpuid) { uint32_t gi; gi = vmcs_read(VMCS_GUEST_INTERRUPTIBILITY); KASSERT(gi & VMCS_INTERRUPTIBILITY_NMI_BLOCKING, ("NMI blocking is not in effect %#x", gi)); } static int vmx_emulate_xsetbv(struct vmx *vmx, int vcpu, struct vm_exit *vmexit) { struct vmxctx *vmxctx; uint64_t xcrval; const struct xsave_limits *limits; vmxctx = &vmx->ctx[vcpu]; limits = vmm_get_xsave_limits(); /* * Note that the processor raises a GP# fault on its own if * xsetbv is executed for CPL != 0, so we do not have to * emulate that fault here. */ /* Only xcr0 is supported. */ if (vmxctx->guest_rcx != 0) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } /* We only handle xcr0 if both the host and guest have XSAVE enabled. */ if (!limits->xsave_enabled || !(vmcs_read(VMCS_GUEST_CR4) & CR4_XSAVE)) { vm_inject_ud(vmx->vm, vcpu); return (HANDLED); } xcrval = vmxctx->guest_rdx << 32 | (vmxctx->guest_rax & 0xffffffff); if ((xcrval & ~limits->xcr0_allowed) != 0) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } if (!(xcrval & XFEATURE_ENABLED_X87)) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } /* AVX (YMM_Hi128) requires SSE. */ if (xcrval & XFEATURE_ENABLED_AVX && (xcrval & XFEATURE_AVX) != XFEATURE_AVX) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } /* * AVX512 requires base AVX (YMM_Hi128) as well as OpMask, * ZMM_Hi256, and Hi16_ZMM. */ if (xcrval & XFEATURE_AVX512 && (xcrval & (XFEATURE_AVX512 | XFEATURE_AVX)) != (XFEATURE_AVX512 | XFEATURE_AVX)) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } /* * Intel MPX requires both bound register state flags to be * set. */ if (((xcrval & XFEATURE_ENABLED_BNDREGS) != 0) != ((xcrval & XFEATURE_ENABLED_BNDCSR) != 0)) { vm_inject_gp(vmx->vm, vcpu); return (HANDLED); } /* * This runs "inside" vmrun() with the guest's FPU state, so * modifying xcr0 directly modifies the guest's xcr0, not the * host's. */ load_xcr(0, xcrval); return (HANDLED); } static uint64_t vmx_get_guest_reg(struct vmx *vmx, int vcpu, int ident) { const struct vmxctx *vmxctx; vmxctx = &vmx->ctx[vcpu]; switch (ident) { case 0: return (vmxctx->guest_rax); case 1: return (vmxctx->guest_rcx); case 2: return (vmxctx->guest_rdx); case 3: return (vmxctx->guest_rbx); case 4: return (vmcs_read(VMCS_GUEST_RSP)); case 5: return (vmxctx->guest_rbp); case 6: return (vmxctx->guest_rsi); case 7: return (vmxctx->guest_rdi); case 8: return (vmxctx->guest_r8); case 9: return (vmxctx->guest_r9); case 10: return (vmxctx->guest_r10); case 11: return (vmxctx->guest_r11); case 12: return (vmxctx->guest_r12); case 13: return (vmxctx->guest_r13); case 14: return (vmxctx->guest_r14); case 15: return (vmxctx->guest_r15); default: panic("invalid vmx register %d", ident); } } static void vmx_set_guest_reg(struct vmx *vmx, int vcpu, int ident, uint64_t regval) { struct vmxctx *vmxctx; vmxctx = &vmx->ctx[vcpu]; switch (ident) { case 0: vmxctx->guest_rax = regval; break; case 1: vmxctx->guest_rcx = regval; break; case 2: vmxctx->guest_rdx = regval; break; case 3: vmxctx->guest_rbx = regval; break; case 4: vmcs_write(VMCS_GUEST_RSP, regval); break; case 5: vmxctx->guest_rbp = regval; break; case 6: vmxctx->guest_rsi = regval; break; case 7: vmxctx->guest_rdi = regval; break; case 8: vmxctx->guest_r8 = regval; break; case 9: vmxctx->guest_r9 = regval; break; case 10: vmxctx->guest_r10 = regval; break; case 11: vmxctx->guest_r11 = regval; break; case 12: vmxctx->guest_r12 = regval; break; case 13: vmxctx->guest_r13 = regval; break; case 14: vmxctx->guest_r14 = regval; break; case 15: vmxctx->guest_r15 = regval; break; default: panic("invalid vmx register %d", ident); } } static int vmx_emulate_cr0_access(struct vmx *vmx, int vcpu, uint64_t exitqual) { uint64_t crval, regval; /* We only handle mov to %cr0 at this time */ if ((exitqual & 0xf0) != 0x00) return (UNHANDLED); regval = vmx_get_guest_reg(vmx, vcpu, (exitqual >> 8) & 0xf); vmcs_write(VMCS_CR0_SHADOW, regval); crval = regval | cr0_ones_mask; crval &= ~cr0_zeros_mask; vmcs_write(VMCS_GUEST_CR0, crval); if (regval & CR0_PG) { uint64_t efer, entry_ctls; /* * If CR0.PG is 1 and EFER.LME is 1 then EFER.LMA and * the "IA-32e mode guest" bit in VM-entry control must be * equal. */ efer = vmcs_read(VMCS_GUEST_IA32_EFER); if (efer & EFER_LME) { efer |= EFER_LMA; vmcs_write(VMCS_GUEST_IA32_EFER, efer); entry_ctls = vmcs_read(VMCS_ENTRY_CTLS); entry_ctls |= VM_ENTRY_GUEST_LMA; vmcs_write(VMCS_ENTRY_CTLS, entry_ctls); } } return (HANDLED); } static int vmx_emulate_cr4_access(struct vmx *vmx, int vcpu, uint64_t exitqual) { uint64_t crval, regval; /* We only handle mov to %cr4 at this time */ if ((exitqual & 0xf0) != 0x00) return (UNHANDLED); regval = vmx_get_guest_reg(vmx, vcpu, (exitqual >> 8) & 0xf); vmcs_write(VMCS_CR4_SHADOW, regval); crval = regval | cr4_ones_mask; crval &= ~cr4_zeros_mask; vmcs_write(VMCS_GUEST_CR4, crval); return (HANDLED); } static int vmx_emulate_cr8_access(struct vmx *vmx, int vcpu, uint64_t exitqual) { struct vlapic *vlapic; uint64_t cr8; int regnum; /* We only handle mov %cr8 to/from a register at this time. */ if ((exitqual & 0xe0) != 0x00) { return (UNHANDLED); } vlapic = vm_lapic(vmx->vm, vcpu); regnum = (exitqual >> 8) & 0xf; if (exitqual & 0x10) { cr8 = vlapic_get_cr8(vlapic); vmx_set_guest_reg(vmx, vcpu, regnum, cr8); } else { cr8 = vmx_get_guest_reg(vmx, vcpu, regnum); vlapic_set_cr8(vlapic, cr8); } return (HANDLED); } /* * From section "Guest Register State" in the Intel SDM: CPL = SS.DPL */ static int vmx_cpl(void) { uint32_t ssar; ssar = vmcs_read(VMCS_GUEST_SS_ACCESS_RIGHTS); return ((ssar >> 5) & 0x3); } static enum vm_cpu_mode vmx_cpu_mode(void) { uint32_t csar; if (vmcs_read(VMCS_GUEST_IA32_EFER) & EFER_LMA) { csar = vmcs_read(VMCS_GUEST_CS_ACCESS_RIGHTS); if (csar & 0x2000) return (CPU_MODE_64BIT); /* CS.L = 1 */ else return (CPU_MODE_COMPATIBILITY); } else if (vmcs_read(VMCS_GUEST_CR0) & CR0_PE) { return (CPU_MODE_PROTECTED); } else { return (CPU_MODE_REAL); } } static enum vm_paging_mode vmx_paging_mode(void) { if (!(vmcs_read(VMCS_GUEST_CR0) & CR0_PG)) return (PAGING_MODE_FLAT); if (!(vmcs_read(VMCS_GUEST_CR4) & CR4_PAE)) return (PAGING_MODE_32); if (vmcs_read(VMCS_GUEST_IA32_EFER) & EFER_LME) return (PAGING_MODE_64); else return (PAGING_MODE_PAE); } static uint64_t inout_str_index(struct vmx *vmx, int vcpuid, int in) { uint64_t val; int error; enum vm_reg_name reg; reg = in ? VM_REG_GUEST_RDI : VM_REG_GUEST_RSI; error = vmx_getreg(vmx, vcpuid, reg, &val); KASSERT(error == 0, ("%s: vmx_getreg error %d", __func__, error)); return (val); } static uint64_t inout_str_count(struct vmx *vmx, int vcpuid, int rep) { uint64_t val; int error; if (rep) { error = vmx_getreg(vmx, vcpuid, VM_REG_GUEST_RCX, &val); KASSERT(!error, ("%s: vmx_getreg error %d", __func__, error)); } else { val = 1; } return (val); } static int inout_str_addrsize(uint32_t inst_info) { uint32_t size; size = (inst_info >> 7) & 0x7; switch (size) { case 0: return (2); /* 16 bit */ case 1: return (4); /* 32 bit */ case 2: return (8); /* 64 bit */ default: panic("%s: invalid size encoding %d", __func__, size); } } static void inout_str_seginfo(struct vmx *vmx, int vcpuid, uint32_t inst_info, int in, struct vm_inout_str *vis) { int error, s; if (in) { vis->seg_name = VM_REG_GUEST_ES; } else { s = (inst_info >> 15) & 0x7; vis->seg_name = vm_segment_name(s); } error = vmx_getdesc(vmx, vcpuid, vis->seg_name, &vis->seg_desc); KASSERT(error == 0, ("%s: vmx_getdesc error %d", __func__, error)); } static void vmx_paging_info(struct vm_guest_paging *paging) { paging->cr3 = vmcs_guest_cr3(); paging->cpl = vmx_cpl(); paging->cpu_mode = vmx_cpu_mode(); paging->paging_mode = vmx_paging_mode(); } static void vmexit_inst_emul(struct vm_exit *vmexit, uint64_t gpa, uint64_t gla) { struct vm_guest_paging *paging; uint32_t csar; paging = &vmexit->u.inst_emul.paging; vmexit->exitcode = VM_EXITCODE_INST_EMUL; vmexit->inst_length = 0; vmexit->u.inst_emul.gpa = gpa; vmexit->u.inst_emul.gla = gla; vmx_paging_info(paging); switch (paging->cpu_mode) { case CPU_MODE_REAL: vmexit->u.inst_emul.cs_base = vmcs_read(VMCS_GUEST_CS_BASE); vmexit->u.inst_emul.cs_d = 0; break; case CPU_MODE_PROTECTED: case CPU_MODE_COMPATIBILITY: vmexit->u.inst_emul.cs_base = vmcs_read(VMCS_GUEST_CS_BASE); csar = vmcs_read(VMCS_GUEST_CS_ACCESS_RIGHTS); vmexit->u.inst_emul.cs_d = SEG_DESC_DEF32(csar); break; default: vmexit->u.inst_emul.cs_base = 0; vmexit->u.inst_emul.cs_d = 0; break; } vie_init(&vmexit->u.inst_emul.vie, NULL, 0); } static int ept_fault_type(uint64_t ept_qual) { int fault_type; if (ept_qual & EPT_VIOLATION_DATA_WRITE) fault_type = VM_PROT_WRITE; else if (ept_qual & EPT_VIOLATION_INST_FETCH) fault_type = VM_PROT_EXECUTE; else fault_type= VM_PROT_READ; return (fault_type); } static boolean_t ept_emulation_fault(uint64_t ept_qual) { int read, write; /* EPT fault on an instruction fetch doesn't make sense here */ if (ept_qual & EPT_VIOLATION_INST_FETCH) return (FALSE); /* EPT fault must be a read fault or a write fault */ read = ept_qual & EPT_VIOLATION_DATA_READ ? 1 : 0; write = ept_qual & EPT_VIOLATION_DATA_WRITE ? 1 : 0; if ((read | write) == 0) return (FALSE); /* * The EPT violation must have been caused by accessing a * guest-physical address that is a translation of a guest-linear * address. */ if ((ept_qual & EPT_VIOLATION_GLA_VALID) == 0 || (ept_qual & EPT_VIOLATION_XLAT_VALID) == 0) { return (FALSE); } return (TRUE); } static __inline int apic_access_virtualization(struct vmx *vmx, int vcpuid) { uint32_t proc_ctls2; proc_ctls2 = vmx->cap[vcpuid].proc_ctls2; return ((proc_ctls2 & PROCBASED2_VIRTUALIZE_APIC_ACCESSES) ? 1 : 0); } static __inline int x2apic_virtualization(struct vmx *vmx, int vcpuid) { uint32_t proc_ctls2; proc_ctls2 = vmx->cap[vcpuid].proc_ctls2; return ((proc_ctls2 & PROCBASED2_VIRTUALIZE_X2APIC_MODE) ? 1 : 0); } static int vmx_handle_apic_write(struct vmx *vmx, int vcpuid, struct vlapic *vlapic, uint64_t qual) { int error, handled, offset; uint32_t *apic_regs, vector; bool retu; handled = HANDLED; offset = APIC_WRITE_OFFSET(qual); if (!apic_access_virtualization(vmx, vcpuid)) { /* * In general there should not be any APIC write VM-exits * unless APIC-access virtualization is enabled. * * However self-IPI virtualization can legitimately trigger * an APIC-write VM-exit so treat it specially. */ if (x2apic_virtualization(vmx, vcpuid) && offset == APIC_OFFSET_SELF_IPI) { apic_regs = (uint32_t *)(vlapic->apic_page); vector = apic_regs[APIC_OFFSET_SELF_IPI / 4]; vlapic_self_ipi_handler(vlapic, vector); return (HANDLED); } else return (UNHANDLED); } switch (offset) { case APIC_OFFSET_ID: vlapic_id_write_handler(vlapic); break; case APIC_OFFSET_LDR: vlapic_ldr_write_handler(vlapic); break; case APIC_OFFSET_DFR: vlapic_dfr_write_handler(vlapic); break; case APIC_OFFSET_SVR: vlapic_svr_write_handler(vlapic); break; case APIC_OFFSET_ESR: vlapic_esr_write_handler(vlapic); break; case APIC_OFFSET_ICR_LOW: retu = false; error = vlapic_icrlo_write_handler(vlapic, &retu); if (error != 0 || retu) handled = UNHANDLED; break; case APIC_OFFSET_CMCI_LVT: case APIC_OFFSET_TIMER_LVT ... APIC_OFFSET_ERROR_LVT: vlapic_lvt_write_handler(vlapic, offset); break; case APIC_OFFSET_TIMER_ICR: vlapic_icrtmr_write_handler(vlapic); break; case APIC_OFFSET_TIMER_DCR: vlapic_dcr_write_handler(vlapic); break; default: handled = UNHANDLED; break; } return (handled); } static bool apic_access_fault(struct vmx *vmx, int vcpuid, uint64_t gpa) { if (apic_access_virtualization(vmx, vcpuid) && (gpa >= DEFAULT_APIC_BASE && gpa < DEFAULT_APIC_BASE + PAGE_SIZE)) return (true); else return (false); } static int vmx_handle_apic_access(struct vmx *vmx, int vcpuid, struct vm_exit *vmexit) { uint64_t qual; int access_type, offset, allowed; if (!apic_access_virtualization(vmx, vcpuid)) return (UNHANDLED); qual = vmexit->u.vmx.exit_qualification; access_type = APIC_ACCESS_TYPE(qual); offset = APIC_ACCESS_OFFSET(qual); allowed = 0; if (access_type == 0) { /* * Read data access to the following registers is expected. */ switch (offset) { case APIC_OFFSET_APR: case APIC_OFFSET_PPR: case APIC_OFFSET_RRR: case APIC_OFFSET_CMCI_LVT: case APIC_OFFSET_TIMER_CCR: allowed = 1; break; default: break; } } else if (access_type == 1) { /* * Write data access to the following registers is expected. */ switch (offset) { case APIC_OFFSET_VER: case APIC_OFFSET_APR: case APIC_OFFSET_PPR: case APIC_OFFSET_RRR: case APIC_OFFSET_ISR0 ... APIC_OFFSET_ISR7: case APIC_OFFSET_TMR0 ... APIC_OFFSET_TMR7: case APIC_OFFSET_IRR0 ... APIC_OFFSET_IRR7: case APIC_OFFSET_CMCI_LVT: case APIC_OFFSET_TIMER_CCR: allowed = 1; break; default: break; } } if (allowed) { vmexit_inst_emul(vmexit, DEFAULT_APIC_BASE + offset, VIE_INVALID_GLA); } /* * Regardless of whether the APIC-access is allowed this handler * always returns UNHANDLED: * - if the access is allowed then it is handled by emulating the * instruction that caused the VM-exit (outside the critical section) * - if the access is not allowed then it will be converted to an * exitcode of VM_EXITCODE_VMX and will be dealt with in userland. */ return (UNHANDLED); } static enum task_switch_reason vmx_task_switch_reason(uint64_t qual) { int reason; reason = (qual >> 30) & 0x3; switch (reason) { case 0: return (TSR_CALL); case 1: return (TSR_IRET); case 2: return (TSR_JMP); case 3: return (TSR_IDT_GATE); default: panic("%s: invalid reason %d", __func__, reason); } } static int emulate_wrmsr(struct vmx *vmx, int vcpuid, u_int num, uint64_t val, bool *retu) { int error; if (lapic_msr(num)) error = lapic_wrmsr(vmx->vm, vcpuid, num, val, retu); else error = vmx_wrmsr(vmx, vcpuid, num, val, retu); return (error); } static int emulate_rdmsr(struct vmx *vmx, int vcpuid, u_int num, bool *retu) { struct vmxctx *vmxctx; uint64_t result; uint32_t eax, edx; int error; if (lapic_msr(num)) error = lapic_rdmsr(vmx->vm, vcpuid, num, &result, retu); else error = vmx_rdmsr(vmx, vcpuid, num, &result, retu); if (error == 0) { eax = result; vmxctx = &vmx->ctx[vcpuid]; error = vmxctx_setreg(vmxctx, VM_REG_GUEST_RAX, eax); KASSERT(error == 0, ("vmxctx_setreg(rax) error %d", error)); edx = result >> 32; error = vmxctx_setreg(vmxctx, VM_REG_GUEST_RDX, edx); KASSERT(error == 0, ("vmxctx_setreg(rdx) error %d", error)); } return (error); } static int vmx_exit_process(struct vmx *vmx, int vcpu, struct vm_exit *vmexit) { int error, errcode, errcode_valid, handled, in; struct vmxctx *vmxctx; struct vlapic *vlapic; struct vm_inout_str *vis; struct vm_task_switch *ts; uint32_t eax, ecx, edx, idtvec_info, idtvec_err, intr_info, inst_info; uint32_t intr_type, intr_vec, reason; uint64_t exitintinfo, qual, gpa; bool retu; CTASSERT((PINBASED_CTLS_ONE_SETTING & PINBASED_VIRTUAL_NMI) != 0); CTASSERT((PINBASED_CTLS_ONE_SETTING & PINBASED_NMI_EXITING) != 0); handled = UNHANDLED; vmxctx = &vmx->ctx[vcpu]; qual = vmexit->u.vmx.exit_qualification; reason = vmexit->u.vmx.exit_reason; vmexit->exitcode = VM_EXITCODE_BOGUS; vmm_stat_incr(vmx->vm, vcpu, VMEXIT_COUNT, 1); SDT_PROBE3(vmm, vmx, exit, entry, vmx, vcpu, vmexit); /* * VM-entry failures during or after loading guest state. * * These VM-exits are uncommon but must be handled specially * as most VM-exit fields are not populated as usual. */ if (__predict_false(reason == EXIT_REASON_MCE_DURING_ENTRY)) { VCPU_CTR0(vmx->vm, vcpu, "Handling MCE during VM-entry"); __asm __volatile("int $18"); return (1); } /* * VM exits that can be triggered during event delivery need to * be handled specially by re-injecting the event if the IDT * vectoring information field's valid bit is set. * * See "Information for VM Exits During Event Delivery" in Intel SDM * for details. */ idtvec_info = vmcs_idt_vectoring_info(); if (idtvec_info & VMCS_IDT_VEC_VALID) { idtvec_info &= ~(1 << 12); /* clear undefined bit */ exitintinfo = idtvec_info; if (idtvec_info & VMCS_IDT_VEC_ERRCODE_VALID) { idtvec_err = vmcs_idt_vectoring_err(); exitintinfo |= (uint64_t)idtvec_err << 32; } error = vm_exit_intinfo(vmx->vm, vcpu, exitintinfo); KASSERT(error == 0, ("%s: vm_set_intinfo error %d", __func__, error)); /* * If 'virtual NMIs' are being used and the VM-exit * happened while injecting an NMI during the previous * VM-entry, then clear "blocking by NMI" in the * Guest Interruptibility-State so the NMI can be * reinjected on the subsequent VM-entry. * * However, if the NMI was being delivered through a task * gate, then the new task must start execution with NMIs * blocked so don't clear NMI blocking in this case. */ intr_type = idtvec_info & VMCS_INTR_T_MASK; if (intr_type == VMCS_INTR_T_NMI) { if (reason != EXIT_REASON_TASK_SWITCH) vmx_clear_nmi_blocking(vmx, vcpu); else vmx_assert_nmi_blocking(vmx, vcpu); } /* * Update VM-entry instruction length if the event being * delivered was a software interrupt or software exception. */ if (intr_type == VMCS_INTR_T_SWINTR || intr_type == VMCS_INTR_T_PRIV_SWEXCEPTION || intr_type == VMCS_INTR_T_SWEXCEPTION) { vmcs_write(VMCS_ENTRY_INST_LENGTH, vmexit->inst_length); } } switch (reason) { case EXIT_REASON_TASK_SWITCH: ts = &vmexit->u.task_switch; ts->tsssel = qual & 0xffff; ts->reason = vmx_task_switch_reason(qual); ts->ext = 0; ts->errcode_valid = 0; vmx_paging_info(&ts->paging); /* * If the task switch was due to a CALL, JMP, IRET, software * interrupt (INT n) or software exception (INT3, INTO), * then the saved %rip references the instruction that caused * the task switch. The instruction length field in the VMCS * is valid in this case. * * In all other cases (e.g., NMI, hardware exception) the * saved %rip is one that would have been saved in the old TSS * had the task switch completed normally so the instruction * length field is not needed in this case and is explicitly * set to 0. */ if (ts->reason == TSR_IDT_GATE) { KASSERT(idtvec_info & VMCS_IDT_VEC_VALID, ("invalid idtvec_info %#x for IDT task switch", idtvec_info)); intr_type = idtvec_info & VMCS_INTR_T_MASK; if (intr_type != VMCS_INTR_T_SWINTR && intr_type != VMCS_INTR_T_SWEXCEPTION && intr_type != VMCS_INTR_T_PRIV_SWEXCEPTION) { /* Task switch triggered by external event */ ts->ext = 1; vmexit->inst_length = 0; if (idtvec_info & VMCS_IDT_VEC_ERRCODE_VALID) { ts->errcode_valid = 1; ts->errcode = vmcs_idt_vectoring_err(); } } } vmexit->exitcode = VM_EXITCODE_TASK_SWITCH; SDT_PROBE4(vmm, vmx, exit, taskswitch, vmx, vcpu, vmexit, ts); VCPU_CTR4(vmx->vm, vcpu, "task switch reason %d, tss 0x%04x, " "%s errcode 0x%016lx", ts->reason, ts->tsssel, ts->ext ? "external" : "internal", ((uint64_t)ts->errcode << 32) | ts->errcode_valid); break; case EXIT_REASON_CR_ACCESS: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_CR_ACCESS, 1); SDT_PROBE4(vmm, vmx, exit, craccess, vmx, vcpu, vmexit, qual); switch (qual & 0xf) { case 0: handled = vmx_emulate_cr0_access(vmx, vcpu, qual); break; case 4: handled = vmx_emulate_cr4_access(vmx, vcpu, qual); break; case 8: handled = vmx_emulate_cr8_access(vmx, vcpu, qual); break; } break; case EXIT_REASON_RDMSR: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_RDMSR, 1); retu = false; ecx = vmxctx->guest_rcx; VCPU_CTR1(vmx->vm, vcpu, "rdmsr 0x%08x", ecx); SDT_PROBE4(vmm, vmx, exit, rdmsr, vmx, vcpu, vmexit, ecx); error = emulate_rdmsr(vmx, vcpu, ecx, &retu); if (error) { vmexit->exitcode = VM_EXITCODE_RDMSR; vmexit->u.msr.code = ecx; } else if (!retu) { handled = HANDLED; } else { /* Return to userspace with a valid exitcode */ KASSERT(vmexit->exitcode != VM_EXITCODE_BOGUS, ("emulate_rdmsr retu with bogus exitcode")); } break; case EXIT_REASON_WRMSR: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_WRMSR, 1); retu = false; eax = vmxctx->guest_rax; ecx = vmxctx->guest_rcx; edx = vmxctx->guest_rdx; VCPU_CTR2(vmx->vm, vcpu, "wrmsr 0x%08x value 0x%016lx", ecx, (uint64_t)edx << 32 | eax); SDT_PROBE5(vmm, vmx, exit, wrmsr, vmx, vmexit, vcpu, ecx, (uint64_t)edx << 32 | eax); error = emulate_wrmsr(vmx, vcpu, ecx, (uint64_t)edx << 32 | eax, &retu); if (error) { vmexit->exitcode = VM_EXITCODE_WRMSR; vmexit->u.msr.code = ecx; vmexit->u.msr.wval = (uint64_t)edx << 32 | eax; } else if (!retu) { handled = HANDLED; } else { /* Return to userspace with a valid exitcode */ KASSERT(vmexit->exitcode != VM_EXITCODE_BOGUS, ("emulate_wrmsr retu with bogus exitcode")); } break; case EXIT_REASON_HLT: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_HLT, 1); SDT_PROBE3(vmm, vmx, exit, halt, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_HLT; vmexit->u.hlt.rflags = vmcs_read(VMCS_GUEST_RFLAGS); if (virtual_interrupt_delivery) vmexit->u.hlt.intr_status = vmcs_read(VMCS_GUEST_INTR_STATUS); else vmexit->u.hlt.intr_status = 0; break; case EXIT_REASON_MTF: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_MTRAP, 1); SDT_PROBE3(vmm, vmx, exit, mtrap, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_MTRAP; vmexit->inst_length = 0; break; case EXIT_REASON_PAUSE: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_PAUSE, 1); SDT_PROBE3(vmm, vmx, exit, pause, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_PAUSE; break; case EXIT_REASON_INTR_WINDOW: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_INTR_WINDOW, 1); SDT_PROBE3(vmm, vmx, exit, intrwindow, vmx, vcpu, vmexit); vmx_clear_int_window_exiting(vmx, vcpu); return (1); case EXIT_REASON_EXT_INTR: /* * External interrupts serve only to cause VM exits and allow * the host interrupt handler to run. * * If this external interrupt triggers a virtual interrupt * to a VM, then that state will be recorded by the * host interrupt handler in the VM's softc. We will inject * this virtual interrupt during the subsequent VM enter. */ intr_info = vmcs_read(VMCS_EXIT_INTR_INFO); SDT_PROBE4(vmm, vmx, exit, interrupt, vmx, vcpu, vmexit, intr_info); /* * XXX: Ignore this exit if VMCS_INTR_VALID is not set. * This appears to be a bug in VMware Fusion? */ if (!(intr_info & VMCS_INTR_VALID)) return (1); KASSERT((intr_info & VMCS_INTR_VALID) != 0 && (intr_info & VMCS_INTR_T_MASK) == VMCS_INTR_T_HWINTR, ("VM exit interruption info invalid: %#x", intr_info)); vmx_trigger_hostintr(intr_info & 0xff); /* * This is special. We want to treat this as an 'handled' * VM-exit but not increment the instruction pointer. */ vmm_stat_incr(vmx->vm, vcpu, VMEXIT_EXTINT, 1); return (1); case EXIT_REASON_NMI_WINDOW: SDT_PROBE3(vmm, vmx, exit, nmiwindow, vmx, vcpu, vmexit); /* Exit to allow the pending virtual NMI to be injected */ if (vm_nmi_pending(vmx->vm, vcpu)) vmx_inject_nmi(vmx, vcpu); vmx_clear_nmi_window_exiting(vmx, vcpu); vmm_stat_incr(vmx->vm, vcpu, VMEXIT_NMI_WINDOW, 1); return (1); case EXIT_REASON_INOUT: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_INOUT, 1); vmexit->exitcode = VM_EXITCODE_INOUT; vmexit->u.inout.bytes = (qual & 0x7) + 1; vmexit->u.inout.in = in = (qual & 0x8) ? 1 : 0; vmexit->u.inout.string = (qual & 0x10) ? 1 : 0; vmexit->u.inout.rep = (qual & 0x20) ? 1 : 0; vmexit->u.inout.port = (uint16_t)(qual >> 16); vmexit->u.inout.eax = (uint32_t)(vmxctx->guest_rax); if (vmexit->u.inout.string) { inst_info = vmcs_read(VMCS_EXIT_INSTRUCTION_INFO); vmexit->exitcode = VM_EXITCODE_INOUT_STR; vis = &vmexit->u.inout_str; vmx_paging_info(&vis->paging); vis->rflags = vmcs_read(VMCS_GUEST_RFLAGS); vis->cr0 = vmcs_read(VMCS_GUEST_CR0); vis->index = inout_str_index(vmx, vcpu, in); vis->count = inout_str_count(vmx, vcpu, vis->inout.rep); vis->addrsize = inout_str_addrsize(inst_info); inout_str_seginfo(vmx, vcpu, inst_info, in, vis); } SDT_PROBE3(vmm, vmx, exit, inout, vmx, vcpu, vmexit); break; case EXIT_REASON_CPUID: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_CPUID, 1); SDT_PROBE3(vmm, vmx, exit, cpuid, vmx, vcpu, vmexit); handled = vmx_handle_cpuid(vmx->vm, vcpu, vmxctx); break; case EXIT_REASON_EXCEPTION: vmm_stat_incr(vmx->vm, vcpu, VMEXIT_EXCEPTION, 1); intr_info = vmcs_read(VMCS_EXIT_INTR_INFO); KASSERT((intr_info & VMCS_INTR_VALID) != 0, ("VM exit interruption info invalid: %#x", intr_info)); intr_vec = intr_info & 0xff; intr_type = intr_info & VMCS_INTR_T_MASK; /* * If Virtual NMIs control is 1 and the VM-exit is due to a * fault encountered during the execution of IRET then we must * restore the state of "virtual-NMI blocking" before resuming * the guest. * * See "Resuming Guest Software after Handling an Exception". * See "Information for VM Exits Due to Vectored Events". */ if ((idtvec_info & VMCS_IDT_VEC_VALID) == 0 && (intr_vec != IDT_DF) && (intr_info & EXIT_QUAL_NMIUDTI) != 0) vmx_restore_nmi_blocking(vmx, vcpu); /* * The NMI has already been handled in vmx_exit_handle_nmi(). */ if (intr_type == VMCS_INTR_T_NMI) return (1); /* * Call the machine check handler by hand. Also don't reflect * the machine check back into the guest. */ if (intr_vec == IDT_MC) { VCPU_CTR0(vmx->vm, vcpu, "Vectoring to MCE handler"); __asm __volatile("int $18"); return (1); } if (intr_vec == IDT_PF) { error = vmxctx_setreg(vmxctx, VM_REG_GUEST_CR2, qual); KASSERT(error == 0, ("%s: vmxctx_setreg(cr2) error %d", __func__, error)); } /* * Software exceptions exhibit trap-like behavior. This in * turn requires populating the VM-entry instruction length * so that the %rip in the trap frame is past the INT3/INTO * instruction. */ if (intr_type == VMCS_INTR_T_SWEXCEPTION) vmcs_write(VMCS_ENTRY_INST_LENGTH, vmexit->inst_length); /* Reflect all other exceptions back into the guest */ errcode_valid = errcode = 0; if (intr_info & VMCS_INTR_DEL_ERRCODE) { errcode_valid = 1; errcode = vmcs_read(VMCS_EXIT_INTR_ERRCODE); } VCPU_CTR2(vmx->vm, vcpu, "Reflecting exception %d/%#x into " "the guest", intr_vec, errcode); SDT_PROBE5(vmm, vmx, exit, exception, vmx, vcpu, vmexit, intr_vec, errcode); error = vm_inject_exception(vmx->vm, vcpu, intr_vec, errcode_valid, errcode, 0); KASSERT(error == 0, ("%s: vm_inject_exception error %d", __func__, error)); return (1); case EXIT_REASON_EPT_FAULT: /* * If 'gpa' lies within the address space allocated to * memory then this must be a nested page fault otherwise * this must be an instruction that accesses MMIO space. */ gpa = vmcs_gpa(); if (vm_mem_allocated(vmx->vm, vcpu, gpa) || apic_access_fault(vmx, vcpu, gpa)) { vmexit->exitcode = VM_EXITCODE_PAGING; vmexit->inst_length = 0; vmexit->u.paging.gpa = gpa; vmexit->u.paging.fault_type = ept_fault_type(qual); vmm_stat_incr(vmx->vm, vcpu, VMEXIT_NESTED_FAULT, 1); SDT_PROBE5(vmm, vmx, exit, nestedfault, vmx, vcpu, vmexit, gpa, qual); } else if (ept_emulation_fault(qual)) { vmexit_inst_emul(vmexit, gpa, vmcs_gla()); vmm_stat_incr(vmx->vm, vcpu, VMEXIT_INST_EMUL, 1); SDT_PROBE4(vmm, vmx, exit, mmiofault, vmx, vcpu, vmexit, gpa); } /* * If Virtual NMIs control is 1 and the VM-exit is due to an * EPT fault during the execution of IRET then we must restore * the state of "virtual-NMI blocking" before resuming. * * See description of "NMI unblocking due to IRET" in * "Exit Qualification for EPT Violations". */ if ((idtvec_info & VMCS_IDT_VEC_VALID) == 0 && (qual & EXIT_QUAL_NMIUDTI) != 0) vmx_restore_nmi_blocking(vmx, vcpu); break; case EXIT_REASON_VIRTUALIZED_EOI: vmexit->exitcode = VM_EXITCODE_IOAPIC_EOI; vmexit->u.ioapic_eoi.vector = qual & 0xFF; SDT_PROBE3(vmm, vmx, exit, eoi, vmx, vcpu, vmexit); vmexit->inst_length = 0; /* trap-like */ break; case EXIT_REASON_APIC_ACCESS: SDT_PROBE3(vmm, vmx, exit, apicaccess, vmx, vcpu, vmexit); handled = vmx_handle_apic_access(vmx, vcpu, vmexit); break; case EXIT_REASON_APIC_WRITE: /* * APIC-write VM exit is trap-like so the %rip is already * pointing to the next instruction. */ vmexit->inst_length = 0; vlapic = vm_lapic(vmx->vm, vcpu); SDT_PROBE4(vmm, vmx, exit, apicwrite, vmx, vcpu, vmexit, vlapic); handled = vmx_handle_apic_write(vmx, vcpu, vlapic, qual); break; case EXIT_REASON_XSETBV: SDT_PROBE3(vmm, vmx, exit, xsetbv, vmx, vcpu, vmexit); handled = vmx_emulate_xsetbv(vmx, vcpu, vmexit); break; case EXIT_REASON_MONITOR: SDT_PROBE3(vmm, vmx, exit, monitor, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_MONITOR; break; case EXIT_REASON_MWAIT: SDT_PROBE3(vmm, vmx, exit, mwait, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_MWAIT; break; case EXIT_REASON_VMCALL: case EXIT_REASON_VMCLEAR: case EXIT_REASON_VMLAUNCH: case EXIT_REASON_VMPTRLD: case EXIT_REASON_VMPTRST: case EXIT_REASON_VMREAD: case EXIT_REASON_VMRESUME: case EXIT_REASON_VMWRITE: case EXIT_REASON_VMXOFF: case EXIT_REASON_VMXON: SDT_PROBE3(vmm, vmx, exit, vminsn, vmx, vcpu, vmexit); vmexit->exitcode = VM_EXITCODE_VMINSN; break; default: SDT_PROBE4(vmm, vmx, exit, unknown, vmx, vcpu, vmexit, reason); vmm_stat_incr(vmx->vm, vcpu, VMEXIT_UNKNOWN, 1); break; } if (handled) { /* * It is possible that control is returned to userland * even though we were able to handle the VM exit in the * kernel. * * In such a case we want to make sure that the userland * restarts guest execution at the instruction *after* * the one we just processed. Therefore we update the * guest rip in the VMCS and in 'vmexit'. */ vmexit->rip += vmexit->inst_length; vmexit->inst_length = 0; vmcs_write(VMCS_GUEST_RIP, vmexit->rip); } else { if (vmexit->exitcode == VM_EXITCODE_BOGUS) { /* * If this VM exit was not claimed by anybody then * treat it as a generic VMX exit. */ vmexit->exitcode = VM_EXITCODE_VMX; vmexit->u.vmx.status = VM_SUCCESS; vmexit->u.vmx.inst_type = 0; vmexit->u.vmx.inst_error = 0; } else { /* * The exitcode and collateral have been populated. * The VM exit will be processed further in userland. */ } } SDT_PROBE4(vmm, vmx, exit, return, vmx, vcpu, vmexit, handled); return (handled); } static __inline void vmx_exit_inst_error(struct vmxctx *vmxctx, int rc, struct vm_exit *vmexit) { KASSERT(vmxctx->inst_fail_status != VM_SUCCESS, ("vmx_exit_inst_error: invalid inst_fail_status %d", vmxctx->inst_fail_status)); vmexit->inst_length = 0; vmexit->exitcode = VM_EXITCODE_VMX; vmexit->u.vmx.status = vmxctx->inst_fail_status; vmexit->u.vmx.inst_error = vmcs_instruction_error(); vmexit->u.vmx.exit_reason = ~0; vmexit->u.vmx.exit_qualification = ~0; switch (rc) { case VMX_VMRESUME_ERROR: case VMX_VMLAUNCH_ERROR: case VMX_INVEPT_ERROR: vmexit->u.vmx.inst_type = rc; break; default: panic("vm_exit_inst_error: vmx_enter_guest returned %d", rc); } } /* * If the NMI-exiting VM execution control is set to '1' then an NMI in * non-root operation causes a VM-exit. NMI blocking is in effect so it is * sufficient to simply vector to the NMI handler via a software interrupt. * However, this must be done before maskable interrupts are enabled * otherwise the "iret" issued by an interrupt handler will incorrectly * clear NMI blocking. */ static __inline void vmx_exit_handle_nmi(struct vmx *vmx, int vcpuid, struct vm_exit *vmexit) { uint32_t intr_info; KASSERT((read_rflags() & PSL_I) == 0, ("interrupts enabled")); if (vmexit->u.vmx.exit_reason != EXIT_REASON_EXCEPTION) return; intr_info = vmcs_read(VMCS_EXIT_INTR_INFO); KASSERT((intr_info & VMCS_INTR_VALID) != 0, ("VM exit interruption info invalid: %#x", intr_info)); if ((intr_info & VMCS_INTR_T_MASK) == VMCS_INTR_T_NMI) { KASSERT((intr_info & 0xff) == IDT_NMI, ("VM exit due " "to NMI has invalid vector: %#x", intr_info)); VCPU_CTR0(vmx->vm, vcpuid, "Vectoring to NMI handler"); __asm __volatile("int $2"); } } static __inline void vmx_dr_enter_guest(struct vmxctx *vmxctx) { register_t rflags; /* Save host control debug registers. */ vmxctx->host_dr7 = rdr7(); vmxctx->host_debugctl = rdmsr(MSR_DEBUGCTLMSR); /* * Disable debugging in DR7 and DEBUGCTL to avoid triggering * exceptions in the host based on the guest DRx values. The * guest DR7 and DEBUGCTL are saved/restored in the VMCS. */ load_dr7(0); wrmsr(MSR_DEBUGCTLMSR, 0); /* * Disable single stepping the kernel to avoid corrupting the * guest DR6. A debugger might still be able to corrupt the * guest DR6 by setting a breakpoint after this point and then * single stepping. */ rflags = read_rflags(); vmxctx->host_tf = rflags & PSL_T; write_rflags(rflags & ~PSL_T); /* Save host debug registers. */ vmxctx->host_dr0 = rdr0(); vmxctx->host_dr1 = rdr1(); vmxctx->host_dr2 = rdr2(); vmxctx->host_dr3 = rdr3(); vmxctx->host_dr6 = rdr6(); /* Restore guest debug registers. */ load_dr0(vmxctx->guest_dr0); load_dr1(vmxctx->guest_dr1); load_dr2(vmxctx->guest_dr2); load_dr3(vmxctx->guest_dr3); load_dr6(vmxctx->guest_dr6); } static __inline void vmx_dr_leave_guest(struct vmxctx *vmxctx) { /* Save guest debug registers. */ vmxctx->guest_dr0 = rdr0(); vmxctx->guest_dr1 = rdr1(); vmxctx->guest_dr2 = rdr2(); vmxctx->guest_dr3 = rdr3(); vmxctx->guest_dr6 = rdr6(); /* * Restore host debug registers. Restore DR7, DEBUGCTL, and * PSL_T last. */ load_dr0(vmxctx->host_dr0); load_dr1(vmxctx->host_dr1); load_dr2(vmxctx->host_dr2); load_dr3(vmxctx->host_dr3); load_dr6(vmxctx->host_dr6); wrmsr(MSR_DEBUGCTLMSR, vmxctx->host_debugctl); load_dr7(vmxctx->host_dr7); write_rflags(read_rflags() | vmxctx->host_tf); } static int vmx_run(void *arg, int vcpu, register_t rip, pmap_t pmap, struct vm_eventinfo *evinfo) { int rc, handled, launched; struct vmx *vmx; struct vm *vm; struct vmxctx *vmxctx; struct vmcs *vmcs; struct vm_exit *vmexit; struct vlapic *vlapic; uint32_t exit_reason; struct region_descriptor gdtr, idtr; uint16_t ldt_sel; vmx = arg; vm = vmx->vm; vmcs = &vmx->vmcs[vcpu]; vmxctx = &vmx->ctx[vcpu]; vlapic = vm_lapic(vm, vcpu); vmexit = vm_exitinfo(vm, vcpu); launched = 0; KASSERT(vmxctx->pmap == pmap, ("pmap %p different than ctx pmap %p", pmap, vmxctx->pmap)); vmx_msr_guest_enter(vmx, vcpu); VMPTRLD(vmcs); /* * XXX * We do this every time because we may setup the virtual machine * from a different process than the one that actually runs it. * * If the life of a virtual machine was spent entirely in the context * of a single process we could do this once in vmx_vminit(). */ vmcs_write(VMCS_HOST_CR3, rcr3()); vmcs_write(VMCS_GUEST_RIP, rip); vmx_set_pcpu_defaults(vmx, vcpu, pmap); do { KASSERT(vmcs_guest_rip() == rip, ("%s: vmcs guest rip mismatch " "%#lx/%#lx", __func__, vmcs_guest_rip(), rip)); handled = UNHANDLED; /* * Interrupts are disabled from this point on until the * guest starts executing. This is done for the following * reasons: * * If an AST is asserted on this thread after the check below, * then the IPI_AST notification will not be lost, because it * will cause a VM exit due to external interrupt as soon as * the guest state is loaded. * * A posted interrupt after 'vmx_inject_interrupts()' will * not be "lost" because it will be held pending in the host * APIC because interrupts are disabled. The pending interrupt * will be recognized as soon as the guest state is loaded. * * The same reasoning applies to the IPI generated by * pmap_invalidate_ept(). */ disable_intr(); vmx_inject_interrupts(vmx, vcpu, vlapic, rip); /* * Check for vcpu suspension after injecting events because * vmx_inject_interrupts() can suspend the vcpu due to a * triple fault. */ if (vcpu_suspended(evinfo)) { enable_intr(); vm_exit_suspended(vmx->vm, vcpu, rip); break; } if (vcpu_rendezvous_pending(evinfo)) { enable_intr(); vm_exit_rendezvous(vmx->vm, vcpu, rip); break; } if (vcpu_reqidle(evinfo)) { enable_intr(); vm_exit_reqidle(vmx->vm, vcpu, rip); break; } if (vcpu_should_yield(vm, vcpu)) { enable_intr(); vm_exit_astpending(vmx->vm, vcpu, rip); vmx_astpending_trace(vmx, vcpu, rip); handled = HANDLED; break; } if (vcpu_debugged(vm, vcpu)) { enable_intr(); vm_exit_debug(vmx->vm, vcpu, rip); break; } /* * VM exits restore the base address but not the * limits of GDTR and IDTR. The VMCS only stores the * base address, so VM exits set the limits to 0xffff. * Save and restore the full GDTR and IDTR to restore * the limits. * * The VMCS does not save the LDTR at all, and VM * exits clear LDTR as if a NULL selector were loaded. * The userspace hypervisor probably doesn't use a * LDT, but save and restore it to be safe. */ sgdt(&gdtr); sidt(&idtr); ldt_sel = sldt(); vmx_run_trace(vmx, vcpu); vmx_dr_enter_guest(vmxctx); rc = vmx_enter_guest(vmxctx, vmx, launched); vmx_dr_leave_guest(vmxctx); bare_lgdt(&gdtr); lidt(&idtr); lldt(ldt_sel); /* Collect some information for VM exit processing */ vmexit->rip = rip = vmcs_guest_rip(); vmexit->inst_length = vmexit_instruction_length(); vmexit->u.vmx.exit_reason = exit_reason = vmcs_exit_reason(); vmexit->u.vmx.exit_qualification = vmcs_exit_qualification(); /* Update 'nextrip' */ vmx->state[vcpu].nextrip = rip; if (rc == VMX_GUEST_VMEXIT) { vmx_exit_handle_nmi(vmx, vcpu, vmexit); enable_intr(); handled = vmx_exit_process(vmx, vcpu, vmexit); } else { enable_intr(); vmx_exit_inst_error(vmxctx, rc, vmexit); } launched = 1; vmx_exit_trace(vmx, vcpu, rip, exit_reason, handled); rip = vmexit->rip; } while (handled); /* * If a VM exit has been handled then the exitcode must be BOGUS * If a VM exit is not handled then the exitcode must not be BOGUS */ if ((handled && vmexit->exitcode != VM_EXITCODE_BOGUS) || (!handled && vmexit->exitcode == VM_EXITCODE_BOGUS)) { panic("Mismatch between handled (%d) and exitcode (%d)", handled, vmexit->exitcode); } if (!handled) vmm_stat_incr(vm, vcpu, VMEXIT_USERSPACE, 1); VCPU_CTR1(vm, vcpu, "returning from vmx_run: exitcode %d", vmexit->exitcode); VMCLEAR(vmcs); vmx_msr_guest_exit(vmx, vcpu); return (0); } static void vmx_vmcleanup(void *arg) { int i; struct vmx *vmx = arg; if (apic_access_virtualization(vmx, 0)) vm_unmap_mmio(vmx->vm, DEFAULT_APIC_BASE, PAGE_SIZE); for (i = 0; i < VM_MAXCPU; i++) vpid_free(vmx->state[i].vpid); free(vmx, M_VMX); return; } static register_t * vmxctx_regptr(struct vmxctx *vmxctx, int reg) { switch (reg) { case VM_REG_GUEST_RAX: return (&vmxctx->guest_rax); case VM_REG_GUEST_RBX: return (&vmxctx->guest_rbx); case VM_REG_GUEST_RCX: return (&vmxctx->guest_rcx); case VM_REG_GUEST_RDX: return (&vmxctx->guest_rdx); case VM_REG_GUEST_RSI: return (&vmxctx->guest_rsi); case VM_REG_GUEST_RDI: return (&vmxctx->guest_rdi); case VM_REG_GUEST_RBP: return (&vmxctx->guest_rbp); case VM_REG_GUEST_R8: return (&vmxctx->guest_r8); case VM_REG_GUEST_R9: return (&vmxctx->guest_r9); case VM_REG_GUEST_R10: return (&vmxctx->guest_r10); case VM_REG_GUEST_R11: return (&vmxctx->guest_r11); case VM_REG_GUEST_R12: return (&vmxctx->guest_r12); case VM_REG_GUEST_R13: return (&vmxctx->guest_r13); case VM_REG_GUEST_R14: return (&vmxctx->guest_r14); case VM_REG_GUEST_R15: return (&vmxctx->guest_r15); case VM_REG_GUEST_CR2: return (&vmxctx->guest_cr2); case VM_REG_GUEST_DR0: return (&vmxctx->guest_dr0); case VM_REG_GUEST_DR1: return (&vmxctx->guest_dr1); case VM_REG_GUEST_DR2: return (&vmxctx->guest_dr2); case VM_REG_GUEST_DR3: return (&vmxctx->guest_dr3); case VM_REG_GUEST_DR6: return (&vmxctx->guest_dr6); default: break; } return (NULL); } static int vmxctx_getreg(struct vmxctx *vmxctx, int reg, uint64_t *retval) { register_t *regp; if ((regp = vmxctx_regptr(vmxctx, reg)) != NULL) { *retval = *regp; return (0); } else return (EINVAL); } static int vmxctx_setreg(struct vmxctx *vmxctx, int reg, uint64_t val) { register_t *regp; if ((regp = vmxctx_regptr(vmxctx, reg)) != NULL) { *regp = val; return (0); } else return (EINVAL); } static int vmx_get_intr_shadow(struct vmx *vmx, int vcpu, int running, uint64_t *retval) { uint64_t gi; int error; error = vmcs_getreg(&vmx->vmcs[vcpu], running, VMCS_IDENT(VMCS_GUEST_INTERRUPTIBILITY), &gi); *retval = (gi & HWINTR_BLOCKING) ? 1 : 0; return (error); } static int vmx_modify_intr_shadow(struct vmx *vmx, int vcpu, int running, uint64_t val) { struct vmcs *vmcs; uint64_t gi; int error, ident; /* * Forcing the vcpu into an interrupt shadow is not supported. */ if (val) { error = EINVAL; goto done; } vmcs = &vmx->vmcs[vcpu]; ident = VMCS_IDENT(VMCS_GUEST_INTERRUPTIBILITY); error = vmcs_getreg(vmcs, running, ident, &gi); if (error == 0) { gi &= ~HWINTR_BLOCKING; error = vmcs_setreg(vmcs, running, ident, gi); } done: VCPU_CTR2(vmx->vm, vcpu, "Setting intr_shadow to %#lx %s", val, error ? "failed" : "succeeded"); return (error); } static int vmx_shadow_reg(int reg) { int shreg; shreg = -1; switch (reg) { case VM_REG_GUEST_CR0: shreg = VMCS_CR0_SHADOW; break; case VM_REG_GUEST_CR4: shreg = VMCS_CR4_SHADOW; break; default: break; } return (shreg); } static int vmx_getreg(void *arg, int vcpu, int reg, uint64_t *retval) { int running, hostcpu; struct vmx *vmx = arg; running = vcpu_is_running(vmx->vm, vcpu, &hostcpu); if (running && hostcpu != curcpu) panic("vmx_getreg: %s%d is running", vm_name(vmx->vm), vcpu); if (reg == VM_REG_GUEST_INTR_SHADOW) return (vmx_get_intr_shadow(vmx, vcpu, running, retval)); if (vmxctx_getreg(&vmx->ctx[vcpu], reg, retval) == 0) return (0); return (vmcs_getreg(&vmx->vmcs[vcpu], running, reg, retval)); } static int vmx_setreg(void *arg, int vcpu, int reg, uint64_t val) { int error, hostcpu, running, shadow; uint64_t ctls; pmap_t pmap; struct vmx *vmx = arg; running = vcpu_is_running(vmx->vm, vcpu, &hostcpu); if (running && hostcpu != curcpu) panic("vmx_setreg: %s%d is running", vm_name(vmx->vm), vcpu); if (reg == VM_REG_GUEST_INTR_SHADOW) return (vmx_modify_intr_shadow(vmx, vcpu, running, val)); if (vmxctx_setreg(&vmx->ctx[vcpu], reg, val) == 0) return (0); error = vmcs_setreg(&vmx->vmcs[vcpu], running, reg, val); if (error == 0) { /* * If the "load EFER" VM-entry control is 1 then the * value of EFER.LMA must be identical to "IA-32e mode guest" * bit in the VM-entry control. */ if ((entry_ctls & VM_ENTRY_LOAD_EFER) != 0 && (reg == VM_REG_GUEST_EFER)) { vmcs_getreg(&vmx->vmcs[vcpu], running, VMCS_IDENT(VMCS_ENTRY_CTLS), &ctls); if (val & EFER_LMA) ctls |= VM_ENTRY_GUEST_LMA; else ctls &= ~VM_ENTRY_GUEST_LMA; vmcs_setreg(&vmx->vmcs[vcpu], running, VMCS_IDENT(VMCS_ENTRY_CTLS), ctls); } shadow = vmx_shadow_reg(reg); if (shadow > 0) { /* * Store the unmodified value in the shadow */ error = vmcs_setreg(&vmx->vmcs[vcpu], running, VMCS_IDENT(shadow), val); } if (reg == VM_REG_GUEST_CR3) { /* * Invalidate the guest vcpu's TLB mappings to emulate * the behavior of updating %cr3. * * XXX the processor retains global mappings when %cr3 * is updated but vmx_invvpid() does not. */ pmap = vmx->ctx[vcpu].pmap; vmx_invvpid(vmx, vcpu, pmap, running); } } return (error); } static int vmx_getdesc(void *arg, int vcpu, int reg, struct seg_desc *desc) { int hostcpu, running; struct vmx *vmx = arg; running = vcpu_is_running(vmx->vm, vcpu, &hostcpu); if (running && hostcpu != curcpu) panic("vmx_getdesc: %s%d is running", vm_name(vmx->vm), vcpu); return (vmcs_getdesc(&vmx->vmcs[vcpu], running, reg, desc)); } static int vmx_setdesc(void *arg, int vcpu, int reg, struct seg_desc *desc) { int hostcpu, running; struct vmx *vmx = arg; running = vcpu_is_running(vmx->vm, vcpu, &hostcpu); if (running && hostcpu != curcpu) panic("vmx_setdesc: %s%d is running", vm_name(vmx->vm), vcpu); return (vmcs_setdesc(&vmx->vmcs[vcpu], running, reg, desc)); } static int vmx_getcap(void *arg, int vcpu, int type, int *retval) { struct vmx *vmx = arg; int vcap; int ret; ret = ENOENT; vcap = vmx->cap[vcpu].set; switch (type) { case VM_CAP_HALT_EXIT: if (cap_halt_exit) ret = 0; break; case VM_CAP_PAUSE_EXIT: if (cap_pause_exit) ret = 0; break; case VM_CAP_MTRAP_EXIT: if (cap_monitor_trap) ret = 0; break; case VM_CAP_UNRESTRICTED_GUEST: if (cap_unrestricted_guest) ret = 0; break; case VM_CAP_ENABLE_INVPCID: if (cap_invpcid) ret = 0; break; default: break; } if (ret == 0) *retval = (vcap & (1 << type)) ? 1 : 0; return (ret); } static int vmx_setcap(void *arg, int vcpu, int type, int val) { struct vmx *vmx = arg; struct vmcs *vmcs = &vmx->vmcs[vcpu]; uint32_t baseval; uint32_t *pptr; int error; int flag; int reg; int retval; retval = ENOENT; pptr = NULL; switch (type) { case VM_CAP_HALT_EXIT: if (cap_halt_exit) { retval = 0; pptr = &vmx->cap[vcpu].proc_ctls; baseval = *pptr; flag = PROCBASED_HLT_EXITING; reg = VMCS_PRI_PROC_BASED_CTLS; } break; case VM_CAP_MTRAP_EXIT: if (cap_monitor_trap) { retval = 0; pptr = &vmx->cap[vcpu].proc_ctls; baseval = *pptr; flag = PROCBASED_MTF; reg = VMCS_PRI_PROC_BASED_CTLS; } break; case VM_CAP_PAUSE_EXIT: if (cap_pause_exit) { retval = 0; pptr = &vmx->cap[vcpu].proc_ctls; baseval = *pptr; flag = PROCBASED_PAUSE_EXITING; reg = VMCS_PRI_PROC_BASED_CTLS; } break; case VM_CAP_UNRESTRICTED_GUEST: if (cap_unrestricted_guest) { retval = 0; pptr = &vmx->cap[vcpu].proc_ctls2; baseval = *pptr; flag = PROCBASED2_UNRESTRICTED_GUEST; reg = VMCS_SEC_PROC_BASED_CTLS; } break; case VM_CAP_ENABLE_INVPCID: if (cap_invpcid) { retval = 0; pptr = &vmx->cap[vcpu].proc_ctls2; baseval = *pptr; flag = PROCBASED2_ENABLE_INVPCID; reg = VMCS_SEC_PROC_BASED_CTLS; } break; default: break; } if (retval == 0) { if (val) { baseval |= flag; } else { baseval &= ~flag; } VMPTRLD(vmcs); error = vmwrite(reg, baseval); VMCLEAR(vmcs); if (error) { retval = error; } else { /* * Update optional stored flags, and record * setting */ if (pptr != NULL) { *pptr = baseval; } if (val) { vmx->cap[vcpu].set |= (1 << type); } else { vmx->cap[vcpu].set &= ~(1 << type); } } } return (retval); } struct vlapic_vtx { struct vlapic vlapic; struct pir_desc *pir_desc; struct vmx *vmx; }; #define VMX_CTR_PIR(vm, vcpuid, pir_desc, notify, vector, level, msg) \ do { \ VCPU_CTR2(vm, vcpuid, msg " assert %s-triggered vector %d", \ level ? "level" : "edge", vector); \ VCPU_CTR1(vm, vcpuid, msg " pir0 0x%016lx", pir_desc->pir[0]); \ VCPU_CTR1(vm, vcpuid, msg " pir1 0x%016lx", pir_desc->pir[1]); \ VCPU_CTR1(vm, vcpuid, msg " pir2 0x%016lx", pir_desc->pir[2]); \ VCPU_CTR1(vm, vcpuid, msg " pir3 0x%016lx", pir_desc->pir[3]); \ VCPU_CTR1(vm, vcpuid, msg " notify: %s", notify ? "yes" : "no");\ } while (0) /* * vlapic->ops handlers that utilize the APICv hardware assist described in * Chapter 29 of the Intel SDM. */ static int vmx_set_intr_ready(struct vlapic *vlapic, int vector, bool level) { struct vlapic_vtx *vlapic_vtx; struct pir_desc *pir_desc; uint64_t mask; int idx, notify; vlapic_vtx = (struct vlapic_vtx *)vlapic; pir_desc = vlapic_vtx->pir_desc; /* * Keep track of interrupt requests in the PIR descriptor. This is * because the virtual APIC page pointed to by the VMCS cannot be * modified if the vcpu is running. */ idx = vector / 64; mask = 1UL << (vector % 64); atomic_set_long(&pir_desc->pir[idx], mask); notify = atomic_cmpset_long(&pir_desc->pending, 0, 1); VMX_CTR_PIR(vlapic->vm, vlapic->vcpuid, pir_desc, notify, vector, level, "vmx_set_intr_ready"); return (notify); } static int vmx_pending_intr(struct vlapic *vlapic, int *vecptr) { struct vlapic_vtx *vlapic_vtx; struct pir_desc *pir_desc; struct LAPIC *lapic; uint64_t pending, pirval; uint32_t ppr, vpr; int i; /* * This function is only expected to be called from the 'HLT' exit * handler which does not care about the vector that is pending. */ KASSERT(vecptr == NULL, ("vmx_pending_intr: vecptr must be NULL")); vlapic_vtx = (struct vlapic_vtx *)vlapic; pir_desc = vlapic_vtx->pir_desc; pending = atomic_load_acq_long(&pir_desc->pending); if (!pending) { /* * While a virtual interrupt may have already been * processed the actual delivery maybe pending the * interruptibility of the guest. Recognize a pending * interrupt by reevaluating virtual interrupts * following Section 29.2.1 in the Intel SDM Volume 3. */ struct vm_exit *vmexit; uint8_t rvi, ppr; vmexit = vm_exitinfo(vlapic->vm, vlapic->vcpuid); KASSERT(vmexit->exitcode == VM_EXITCODE_HLT, ("vmx_pending_intr: exitcode not 'HLT'")); rvi = vmexit->u.hlt.intr_status & APIC_TPR_INT; lapic = vlapic->apic_page; ppr = lapic->ppr & APIC_TPR_INT; if (rvi > ppr) { return (1); } return (0); } /* * If there is an interrupt pending then it will be recognized only * if its priority is greater than the processor priority. * * Special case: if the processor priority is zero then any pending * interrupt will be recognized. */ lapic = vlapic->apic_page; ppr = lapic->ppr & APIC_TPR_INT; if (ppr == 0) return (1); VCPU_CTR1(vlapic->vm, vlapic->vcpuid, "HLT with non-zero PPR %d", lapic->ppr); for (i = 3; i >= 0; i--) { pirval = pir_desc->pir[i]; if (pirval != 0) { vpr = (i * 64 + flsl(pirval) - 1) & APIC_TPR_INT; return (vpr > ppr); } } return (0); } static void vmx_intr_accepted(struct vlapic *vlapic, int vector) { panic("vmx_intr_accepted: not expected to be called"); } static void vmx_set_tmr(struct vlapic *vlapic, int vector, bool level) { struct vlapic_vtx *vlapic_vtx; struct vmx *vmx; struct vmcs *vmcs; uint64_t mask, val; KASSERT(vector >= 0 && vector <= 255, ("invalid vector %d", vector)); KASSERT(!vcpu_is_running(vlapic->vm, vlapic->vcpuid, NULL), ("vmx_set_tmr: vcpu cannot be running")); vlapic_vtx = (struct vlapic_vtx *)vlapic; vmx = vlapic_vtx->vmx; vmcs = &vmx->vmcs[vlapic->vcpuid]; mask = 1UL << (vector % 64); VMPTRLD(vmcs); val = vmcs_read(VMCS_EOI_EXIT(vector)); if (level) val |= mask; else val &= ~mask; vmcs_write(VMCS_EOI_EXIT(vector), val); VMCLEAR(vmcs); } static void vmx_enable_x2apic_mode(struct vlapic *vlapic) { struct vmx *vmx; struct vmcs *vmcs; uint32_t proc_ctls2; int vcpuid, error; vcpuid = vlapic->vcpuid; vmx = ((struct vlapic_vtx *)vlapic)->vmx; vmcs = &vmx->vmcs[vcpuid]; proc_ctls2 = vmx->cap[vcpuid].proc_ctls2; KASSERT((proc_ctls2 & PROCBASED2_VIRTUALIZE_APIC_ACCESSES) != 0, ("%s: invalid proc_ctls2 %#x", __func__, proc_ctls2)); proc_ctls2 &= ~PROCBASED2_VIRTUALIZE_APIC_ACCESSES; proc_ctls2 |= PROCBASED2_VIRTUALIZE_X2APIC_MODE; vmx->cap[vcpuid].proc_ctls2 = proc_ctls2; VMPTRLD(vmcs); vmcs_write(VMCS_SEC_PROC_BASED_CTLS, proc_ctls2); VMCLEAR(vmcs); if (vlapic->vcpuid == 0) { /* * The nested page table mappings are shared by all vcpus * so unmap the APIC access page just once. */ error = vm_unmap_mmio(vmx->vm, DEFAULT_APIC_BASE, PAGE_SIZE); KASSERT(error == 0, ("%s: vm_unmap_mmio error %d", __func__, error)); /* * The MSR bitmap is shared by all vcpus so modify it only * once in the context of vcpu 0. */ error = vmx_allow_x2apic_msrs(vmx); KASSERT(error == 0, ("%s: vmx_allow_x2apic_msrs error %d", __func__, error)); } } static void vmx_post_intr(struct vlapic *vlapic, int hostcpu) { ipi_cpu(hostcpu, pirvec); } /* * Transfer the pending interrupts in the PIR descriptor to the IRR * in the virtual APIC page. */ static void vmx_inject_pir(struct vlapic *vlapic) { struct vlapic_vtx *vlapic_vtx; struct pir_desc *pir_desc; struct LAPIC *lapic; uint64_t val, pirval; int rvi, pirbase = -1; uint16_t intr_status_old, intr_status_new; vlapic_vtx = (struct vlapic_vtx *)vlapic; pir_desc = vlapic_vtx->pir_desc; if (atomic_cmpset_long(&pir_desc->pending, 1, 0) == 0) { VCPU_CTR0(vlapic->vm, vlapic->vcpuid, "vmx_inject_pir: " "no posted interrupt pending"); return; } pirval = 0; pirbase = -1; lapic = vlapic->apic_page; val = atomic_readandclear_long(&pir_desc->pir[0]); if (val != 0) { lapic->irr0 |= val; lapic->irr1 |= val >> 32; pirbase = 0; pirval = val; } val = atomic_readandclear_long(&pir_desc->pir[1]); if (val != 0) { lapic->irr2 |= val; lapic->irr3 |= val >> 32; pirbase = 64; pirval = val; } val = atomic_readandclear_long(&pir_desc->pir[2]); if (val != 0) { lapic->irr4 |= val; lapic->irr5 |= val >> 32; pirbase = 128; pirval = val; } val = atomic_readandclear_long(&pir_desc->pir[3]); if (val != 0) { lapic->irr6 |= val; lapic->irr7 |= val >> 32; pirbase = 192; pirval = val; } VLAPIC_CTR_IRR(vlapic, "vmx_inject_pir"); /* * Update RVI so the processor can evaluate pending virtual * interrupts on VM-entry. * * It is possible for pirval to be 0 here, even though the * pending bit has been set. The scenario is: * CPU-Y is sending a posted interrupt to CPU-X, which * is running a guest and processing posted interrupts in h/w. * CPU-X will eventually exit and the state seen in s/w is * the pending bit set, but no PIR bits set. * * CPU-X CPU-Y * (vm running) (host running) * rx posted interrupt * CLEAR pending bit * SET PIR bit * READ/CLEAR PIR bits * SET pending bit * (vm exit) * pending bit set, PIR 0 */ if (pirval != 0) { rvi = pirbase + flsl(pirval) - 1; intr_status_old = vmcs_read(VMCS_GUEST_INTR_STATUS); intr_status_new = (intr_status_old & 0xFF00) | rvi; if (intr_status_new > intr_status_old) { vmcs_write(VMCS_GUEST_INTR_STATUS, intr_status_new); VCPU_CTR2(vlapic->vm, vlapic->vcpuid, "vmx_inject_pir: " "guest_intr_status changed from 0x%04x to 0x%04x", intr_status_old, intr_status_new); } } } static struct vlapic * vmx_vlapic_init(void *arg, int vcpuid) { struct vmx *vmx; struct vlapic *vlapic; struct vlapic_vtx *vlapic_vtx; vmx = arg; vlapic = malloc(sizeof(struct vlapic_vtx), M_VLAPIC, M_WAITOK | M_ZERO); vlapic->vm = vmx->vm; vlapic->vcpuid = vcpuid; vlapic->apic_page = (struct LAPIC *)&vmx->apic_page[vcpuid]; vlapic_vtx = (struct vlapic_vtx *)vlapic; vlapic_vtx->pir_desc = &vmx->pir_desc[vcpuid]; vlapic_vtx->vmx = vmx; if (virtual_interrupt_delivery) { vlapic->ops.set_intr_ready = vmx_set_intr_ready; vlapic->ops.pending_intr = vmx_pending_intr; vlapic->ops.intr_accepted = vmx_intr_accepted; vlapic->ops.set_tmr = vmx_set_tmr; vlapic->ops.enable_x2apic_mode = vmx_enable_x2apic_mode; } if (posted_interrupts) vlapic->ops.post_intr = vmx_post_intr; vlapic_init(vlapic); return (vlapic); } static void vmx_vlapic_cleanup(void *arg, struct vlapic *vlapic) { vlapic_cleanup(vlapic); free(vlapic, M_VLAPIC); } struct vmm_ops vmm_ops_intel = { vmx_init, vmx_cleanup, vmx_restore, vmx_vminit, vmx_run, vmx_vmcleanup, vmx_getreg, vmx_setreg, vmx_getdesc, vmx_setdesc, vmx_getcap, vmx_setcap, ept_vmspace_alloc, ept_vmspace_free, vmx_vlapic_init, vmx_vlapic_cleanup, }; Index: releng/12.0/sys/x86/include/specialreg.h =================================================================== --- releng/12.0/sys/x86/include/specialreg.h (revision 340626) +++ releng/12.0/sys/x86/include/specialreg.h (revision 340627) @@ -1,1059 +1,1061 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1991 The Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)specialreg.h 7.1 (Berkeley) 5/9/91 * $FreeBSD$ */ #ifndef _MACHINE_SPECIALREG_H_ #define _MACHINE_SPECIALREG_H_ /* * Bits in 386 special registers: */ #define CR0_PE 0x00000001 /* Protected mode Enable */ #define CR0_MP 0x00000002 /* "Math" (fpu) Present */ #define CR0_EM 0x00000004 /* EMulate FPU instructions. (trap ESC only) */ #define CR0_TS 0x00000008 /* Task Switched (if MP, trap ESC and WAIT) */ #define CR0_PG 0x80000000 /* PaGing enable */ /* * Bits in 486 special registers: */ #define CR0_NE 0x00000020 /* Numeric Error enable (EX16 vs IRQ13) */ #define CR0_WP 0x00010000 /* Write Protect (honor page protect in all modes) */ #define CR0_AM 0x00040000 /* Alignment Mask (set to enable AC flag) */ #define CR0_NW 0x20000000 /* Not Write-through */ #define CR0_CD 0x40000000 /* Cache Disable */ #define CR3_PCID_SAVE 0x8000000000000000 #define CR3_PCID_MASK 0xfff /* * Bits in PPro special registers */ #define CR4_VME 0x00000001 /* Virtual 8086 mode extensions */ #define CR4_PVI 0x00000002 /* Protected-mode virtual interrupts */ #define CR4_TSD 0x00000004 /* Time stamp disable */ #define CR4_DE 0x00000008 /* Debugging extensions */ #define CR4_PSE 0x00000010 /* Page size extensions */ #define CR4_PAE 0x00000020 /* Physical address extension */ #define CR4_MCE 0x00000040 /* Machine check enable */ #define CR4_PGE 0x00000080 /* Page global enable */ #define CR4_PCE 0x00000100 /* Performance monitoring counter enable */ #define CR4_FXSR 0x00000200 /* Fast FPU save/restore used by OS */ #define CR4_XMM 0x00000400 /* enable SIMD/MMX2 to use except 16 */ #define CR4_VMXE 0x00002000 /* enable VMX operation (Intel-specific) */ #define CR4_FSGSBASE 0x00010000 /* Enable FS/GS BASE accessing instructions */ #define CR4_PCIDE 0x00020000 /* Enable Context ID */ #define CR4_XSAVE 0x00040000 /* XSETBV/XGETBV */ #define CR4_SMEP 0x00100000 /* Supervisor-Mode Execution Prevention */ #define CR4_SMAP 0x00200000 /* Supervisor-Mode Access Prevention */ /* * Bits in AMD64 special registers. EFER is 64 bits wide. */ #define EFER_SCE 0x000000001 /* System Call Extensions (R/W) */ #define EFER_LME 0x000000100 /* Long mode enable (R/W) */ #define EFER_LMA 0x000000400 /* Long mode active (R) */ #define EFER_NXE 0x000000800 /* PTE No-Execute bit enable (R/W) */ #define EFER_SVM 0x000001000 /* SVM enable bit for AMD, reserved for Intel */ #define EFER_LMSLE 0x000002000 /* Long Mode Segment Limit Enable */ #define EFER_FFXSR 0x000004000 /* Fast FXSAVE/FSRSTOR */ #define EFER_TCE 0x000008000 /* Translation Cache Extension */ /* * Intel Extended Features registers */ #define XCR0 0 /* XFEATURE_ENABLED_MASK register */ #define XFEATURE_ENABLED_X87 0x00000001 #define XFEATURE_ENABLED_SSE 0x00000002 #define XFEATURE_ENABLED_YMM_HI128 0x00000004 #define XFEATURE_ENABLED_AVX XFEATURE_ENABLED_YMM_HI128 #define XFEATURE_ENABLED_BNDREGS 0x00000008 #define XFEATURE_ENABLED_BNDCSR 0x00000010 #define XFEATURE_ENABLED_OPMASK 0x00000020 #define XFEATURE_ENABLED_ZMM_HI256 0x00000040 #define XFEATURE_ENABLED_HI16_ZMM 0x00000080 #define XFEATURE_AVX \ (XFEATURE_ENABLED_X87 | XFEATURE_ENABLED_SSE | XFEATURE_ENABLED_AVX) #define XFEATURE_AVX512 \ (XFEATURE_ENABLED_OPMASK | XFEATURE_ENABLED_ZMM_HI256 | \ XFEATURE_ENABLED_HI16_ZMM) #define XFEATURE_MPX \ (XFEATURE_ENABLED_BNDREGS | XFEATURE_ENABLED_BNDCSR) /* * CPUID instruction features register */ #define CPUID_FPU 0x00000001 #define CPUID_VME 0x00000002 #define CPUID_DE 0x00000004 #define CPUID_PSE 0x00000008 #define CPUID_TSC 0x00000010 #define CPUID_MSR 0x00000020 #define CPUID_PAE 0x00000040 #define CPUID_MCE 0x00000080 #define CPUID_CX8 0x00000100 #define CPUID_APIC 0x00000200 #define CPUID_B10 0x00000400 #define CPUID_SEP 0x00000800 #define CPUID_MTRR 0x00001000 #define CPUID_PGE 0x00002000 #define CPUID_MCA 0x00004000 #define CPUID_CMOV 0x00008000 #define CPUID_PAT 0x00010000 #define CPUID_PSE36 0x00020000 #define CPUID_PSN 0x00040000 #define CPUID_CLFSH 0x00080000 #define CPUID_B20 0x00100000 #define CPUID_DS 0x00200000 #define CPUID_ACPI 0x00400000 #define CPUID_MMX 0x00800000 #define CPUID_FXSR 0x01000000 #define CPUID_SSE 0x02000000 #define CPUID_XMM 0x02000000 #define CPUID_SSE2 0x04000000 #define CPUID_SS 0x08000000 #define CPUID_HTT 0x10000000 #define CPUID_TM 0x20000000 #define CPUID_IA64 0x40000000 #define CPUID_PBE 0x80000000 #define CPUID2_SSE3 0x00000001 #define CPUID2_PCLMULQDQ 0x00000002 #define CPUID2_DTES64 0x00000004 #define CPUID2_MON 0x00000008 #define CPUID2_DS_CPL 0x00000010 #define CPUID2_VMX 0x00000020 #define CPUID2_SMX 0x00000040 #define CPUID2_EST 0x00000080 #define CPUID2_TM2 0x00000100 #define CPUID2_SSSE3 0x00000200 #define CPUID2_CNXTID 0x00000400 #define CPUID2_SDBG 0x00000800 #define CPUID2_FMA 0x00001000 #define CPUID2_CX16 0x00002000 #define CPUID2_XTPR 0x00004000 #define CPUID2_PDCM 0x00008000 #define CPUID2_PCID 0x00020000 #define CPUID2_DCA 0x00040000 #define CPUID2_SSE41 0x00080000 #define CPUID2_SSE42 0x00100000 #define CPUID2_X2APIC 0x00200000 #define CPUID2_MOVBE 0x00400000 #define CPUID2_POPCNT 0x00800000 #define CPUID2_TSCDLT 0x01000000 #define CPUID2_AESNI 0x02000000 #define CPUID2_XSAVE 0x04000000 #define CPUID2_OSXSAVE 0x08000000 #define CPUID2_AVX 0x10000000 #define CPUID2_F16C 0x20000000 #define CPUID2_RDRAND 0x40000000 #define CPUID2_HV 0x80000000 /* * Important bits in the Thermal and Power Management flags * CPUID.6 EAX and ECX. */ #define CPUTPM1_SENSOR 0x00000001 #define CPUTPM1_TURBO 0x00000002 #define CPUTPM1_ARAT 0x00000004 #define CPUTPM2_EFFREQ 0x00000001 /* Intel Processor Trace CPUID. */ /* Leaf 0 ebx. */ #define CPUPT_CR3 (1 << 0) /* CR3 Filtering Support */ #define CPUPT_PSB (1 << 1) /* Configurable PSB and Cycle-Accurate Mode Supported */ #define CPUPT_IPF (1 << 2) /* IP Filtering and TraceStop supported */ #define CPUPT_MTC (1 << 3) /* MTC Supported */ #define CPUPT_PRW (1 << 4) /* PTWRITE Supported */ #define CPUPT_PWR (1 << 5) /* Power Event Trace Supported */ /* Leaf 0 ecx. */ #define CPUPT_TOPA (1 << 0) /* ToPA Output Supported */ #define CPUPT_TOPA_MULTI (1 << 1) /* ToPA Tables Allow Multiple Output Entries */ #define CPUPT_SINGLE (1 << 2) /* Single-Range Output Supported */ #define CPUPT_TT_OUT (1 << 3) /* Output to Trace Transport Subsystem Supported */ #define CPUPT_LINEAR_IP (1 << 31) /* IP Payloads are Linear IP, otherwise IP is effective */ /* Leaf 1 eax. */ #define CPUPT_NADDR_S 0 /* Number of Address Ranges */ #define CPUPT_NADDR_M (0x7 << CPUPT_NADDR_S) #define CPUPT_MTC_BITMAP_S 16 /* Bitmap of supported MTC Period Encodings */ #define CPUPT_MTC_BITMAP_M (0xffff << CPUPT_MTC_BITMAP_S) /* Leaf 1 ebx. */ #define CPUPT_CT_BITMAP_S 0 /* Bitmap of supported Cycle Threshold values */ #define CPUPT_CT_BITMAP_M (0xffff << CPUPT_CT_BITMAP_S) #define CPUPT_PFE_BITMAP_S 16 /* Bitmap of supported Configurable PSB Frequency encoding */ #define CPUPT_PFE_BITMAP_M (0xffff << CPUPT_PFE_BITMAP_S) /* * Important bits in the AMD extended cpuid flags */ #define AMDID_SYSCALL 0x00000800 #define AMDID_MP 0x00080000 #define AMDID_NX 0x00100000 #define AMDID_EXT_MMX 0x00400000 #define AMDID_FFXSR 0x02000000 #define AMDID_PAGE1GB 0x04000000 #define AMDID_RDTSCP 0x08000000 #define AMDID_LM 0x20000000 #define AMDID_EXT_3DNOW 0x40000000 #define AMDID_3DNOW 0x80000000 #define AMDID2_LAHF 0x00000001 #define AMDID2_CMP 0x00000002 #define AMDID2_SVM 0x00000004 #define AMDID2_EXT_APIC 0x00000008 #define AMDID2_CR8 0x00000010 #define AMDID2_ABM 0x00000020 #define AMDID2_SSE4A 0x00000040 #define AMDID2_MAS 0x00000080 #define AMDID2_PREFETCH 0x00000100 #define AMDID2_OSVW 0x00000200 #define AMDID2_IBS 0x00000400 #define AMDID2_XOP 0x00000800 #define AMDID2_SKINIT 0x00001000 #define AMDID2_WDT 0x00002000 #define AMDID2_LWP 0x00008000 #define AMDID2_FMA4 0x00010000 #define AMDID2_TCE 0x00020000 #define AMDID2_NODE_ID 0x00080000 #define AMDID2_TBM 0x00200000 #define AMDID2_TOPOLOGY 0x00400000 #define AMDID2_PCXC 0x00800000 #define AMDID2_PNXC 0x01000000 #define AMDID2_DBE 0x04000000 #define AMDID2_PTSC 0x08000000 #define AMDID2_PTSCEL2I 0x10000000 #define AMDID2_MWAITX 0x20000000 /* * CPUID instruction 1 eax info */ #define CPUID_STEPPING 0x0000000f #define CPUID_MODEL 0x000000f0 #define CPUID_FAMILY 0x00000f00 #define CPUID_EXT_MODEL 0x000f0000 #define CPUID_EXT_FAMILY 0x0ff00000 #ifdef __i386__ #define CPUID_TO_MODEL(id) \ ((((id) & CPUID_MODEL) >> 4) | \ ((((id) & CPUID_FAMILY) >= 0x600) ? \ (((id) & CPUID_EXT_MODEL) >> 12) : 0)) #define CPUID_TO_FAMILY(id) \ ((((id) & CPUID_FAMILY) >> 8) + \ ((((id) & CPUID_FAMILY) == 0xf00) ? \ (((id) & CPUID_EXT_FAMILY) >> 20) : 0)) #else #define CPUID_TO_MODEL(id) \ ((((id) & CPUID_MODEL) >> 4) | \ (((id) & CPUID_EXT_MODEL) >> 12)) #define CPUID_TO_FAMILY(id) \ ((((id) & CPUID_FAMILY) >> 8) + \ (((id) & CPUID_EXT_FAMILY) >> 20)) #endif /* * CPUID instruction 1 ebx info */ #define CPUID_BRAND_INDEX 0x000000ff #define CPUID_CLFUSH_SIZE 0x0000ff00 #define CPUID_HTT_CORES 0x00ff0000 #define CPUID_LOCAL_APIC_ID 0xff000000 /* * CPUID instruction 5 info */ #define CPUID5_MON_MIN_SIZE 0x0000ffff /* eax */ #define CPUID5_MON_MAX_SIZE 0x0000ffff /* ebx */ #define CPUID5_MON_MWAIT_EXT 0x00000001 /* ecx */ #define CPUID5_MWAIT_INTRBREAK 0x00000002 /* ecx */ /* * MWAIT cpu power states. Lower 4 bits are sub-states. */ #define MWAIT_C0 0xf0 #define MWAIT_C1 0x00 #define MWAIT_C2 0x10 #define MWAIT_C3 0x20 #define MWAIT_C4 0x30 /* * MWAIT extensions. */ /* Interrupt breaks MWAIT even when masked. */ #define MWAIT_INTRBREAK 0x00000001 /* * CPUID instruction 6 ecx info */ #define CPUID_PERF_STAT 0x00000001 #define CPUID_PERF_BIAS 0x00000008 /* * CPUID instruction 0xb ebx info. */ #define CPUID_TYPE_INVAL 0 #define CPUID_TYPE_SMT 1 #define CPUID_TYPE_CORE 2 /* * CPUID instruction 0xd Processor Extended State Enumeration Sub-leaf 1 */ #define CPUID_EXTSTATE_XSAVEOPT 0x00000001 #define CPUID_EXTSTATE_XSAVEC 0x00000002 #define CPUID_EXTSTATE_XINUSE 0x00000004 #define CPUID_EXTSTATE_XSAVES 0x00000008 /* * AMD extended function 8000_0007h ebx info */ #define AMDRAS_MCA_OF_RECOV 0x00000001 #define AMDRAS_SUCCOR 0x00000002 #define AMDRAS_HW_ASSERT 0x00000004 #define AMDRAS_SCALABLE_MCA 0x00000008 #define AMDRAS_PFEH_SUPPORT 0x00000010 /* * AMD extended function 8000_0007h edx info */ #define AMDPM_TS 0x00000001 #define AMDPM_FID 0x00000002 #define AMDPM_VID 0x00000004 #define AMDPM_TTP 0x00000008 #define AMDPM_TM 0x00000010 #define AMDPM_STC 0x00000020 #define AMDPM_100MHZ_STEPS 0x00000040 #define AMDPM_HW_PSTATE 0x00000080 #define AMDPM_TSC_INVARIANT 0x00000100 #define AMDPM_CPB 0x00000200 /* * AMD extended function 8000_0008h ebx info (amd_extended_feature_extensions) */ #define AMDFEID_CLZERO 0x00000001 #define AMDFEID_IRPERF 0x00000002 #define AMDFEID_XSAVEERPTR 0x00000004 /* * AMD extended function 8000_0008h ecx info */ #define AMDID_CMP_CORES 0x000000ff #define AMDID_COREID_SIZE 0x0000f000 #define AMDID_COREID_SIZE_SHIFT 12 /* * CPUID instruction 7 Structured Extended Features, leaf 0 ebx info */ #define CPUID_STDEXT_FSGSBASE 0x00000001 #define CPUID_STDEXT_TSC_ADJUST 0x00000002 #define CPUID_STDEXT_SGX 0x00000004 #define CPUID_STDEXT_BMI1 0x00000008 #define CPUID_STDEXT_HLE 0x00000010 #define CPUID_STDEXT_AVX2 0x00000020 #define CPUID_STDEXT_FDP_EXC 0x00000040 #define CPUID_STDEXT_SMEP 0x00000080 #define CPUID_STDEXT_BMI2 0x00000100 #define CPUID_STDEXT_ERMS 0x00000200 #define CPUID_STDEXT_INVPCID 0x00000400 #define CPUID_STDEXT_RTM 0x00000800 #define CPUID_STDEXT_PQM 0x00001000 #define CPUID_STDEXT_NFPUSG 0x00002000 #define CPUID_STDEXT_MPX 0x00004000 #define CPUID_STDEXT_PQE 0x00008000 #define CPUID_STDEXT_AVX512F 0x00010000 #define CPUID_STDEXT_AVX512DQ 0x00020000 #define CPUID_STDEXT_RDSEED 0x00040000 #define CPUID_STDEXT_ADX 0x00080000 #define CPUID_STDEXT_SMAP 0x00100000 #define CPUID_STDEXT_AVX512IFMA 0x00200000 #define CPUID_STDEXT_PCOMMIT 0x00400000 #define CPUID_STDEXT_CLFLUSHOPT 0x00800000 #define CPUID_STDEXT_CLWB 0x01000000 #define CPUID_STDEXT_PROCTRACE 0x02000000 #define CPUID_STDEXT_AVX512PF 0x04000000 #define CPUID_STDEXT_AVX512ER 0x08000000 #define CPUID_STDEXT_AVX512CD 0x10000000 #define CPUID_STDEXT_SHA 0x20000000 #define CPUID_STDEXT_AVX512BW 0x40000000 #define CPUID_STDEXT_AVX512VL 0x80000000 /* * CPUID instruction 7 Structured Extended Features, leaf 0 ecx info */ #define CPUID_STDEXT2_PREFETCHWT1 0x00000001 #define CPUID_STDEXT2_UMIP 0x00000004 #define CPUID_STDEXT2_PKU 0x00000008 #define CPUID_STDEXT2_OSPKE 0x00000010 #define CPUID_STDEXT2_RDPID 0x00400000 #define CPUID_STDEXT2_SGXLC 0x40000000 /* * CPUID instruction 7 Structured Extended Features, leaf 0 edx info */ #define CPUID_STDEXT3_IBPB 0x04000000 #define CPUID_STDEXT3_STIBP 0x08000000 #define CPUID_STDEXT3_L1D_FLUSH 0x10000000 #define CPUID_STDEXT3_ARCH_CAP 0x20000000 #define CPUID_STDEXT3_SSBD 0x80000000 /* MSR IA32_ARCH_CAP(ABILITIES) bits */ #define IA32_ARCH_CAP_RDCL_NO 0x00000001 #define IA32_ARCH_CAP_IBRS_ALL 0x00000002 -#define IA32_ARCH_CAP_SSBD_NO 0x00000004 +#define IA32_ARCH_CAP_RSBA 0x00000004 +#define IA32_ARCH_CAP_SKIP_L1DFL_VMENTRY 0x00000008 +#define IA32_ARCH_CAP_SSB_NO 0x00000010 /* * CPUID manufacturers identifiers */ #define AMD_VENDOR_ID "AuthenticAMD" #define CENTAUR_VENDOR_ID "CentaurHauls" #define CYRIX_VENDOR_ID "CyrixInstead" #define INTEL_VENDOR_ID "GenuineIntel" #define NEXGEN_VENDOR_ID "NexGenDriven" #define NSC_VENDOR_ID "Geode by NSC" #define RISE_VENDOR_ID "RiseRiseRise" #define SIS_VENDOR_ID "SiS SiS SiS " #define TRANSMETA_VENDOR_ID "GenuineTMx86" #define UMC_VENDOR_ID "UMC UMC UMC " /* * Model-specific registers for the i386 family */ #define MSR_P5_MC_ADDR 0x000 #define MSR_P5_MC_TYPE 0x001 #define MSR_TSC 0x010 #define MSR_P5_CESR 0x011 #define MSR_P5_CTR0 0x012 #define MSR_P5_CTR1 0x013 #define MSR_IA32_PLATFORM_ID 0x017 #define MSR_APICBASE 0x01b #define MSR_EBL_CR_POWERON 0x02a #define MSR_TEST_CTL 0x033 #define MSR_IA32_FEATURE_CONTROL 0x03a #define MSR_IA32_SPEC_CTRL 0x048 #define MSR_IA32_PRED_CMD 0x049 #define MSR_BIOS_UPDT_TRIG 0x079 #define MSR_BBL_CR_D0 0x088 #define MSR_BBL_CR_D1 0x089 #define MSR_BBL_CR_D2 0x08a #define MSR_BIOS_SIGN 0x08b #define MSR_PERFCTR0 0x0c1 #define MSR_PERFCTR1 0x0c2 #define MSR_PLATFORM_INFO 0x0ce #define MSR_MPERF 0x0e7 #define MSR_APERF 0x0e8 #define MSR_IA32_EXT_CONFIG 0x0ee /* Undocumented. Core Solo/Duo only */ #define MSR_MTRRcap 0x0fe #define MSR_IA32_ARCH_CAP 0x10a #define MSR_IA32_FLUSH_CMD 0x10b #define MSR_BBL_CR_ADDR 0x116 #define MSR_BBL_CR_DECC 0x118 #define MSR_BBL_CR_CTL 0x119 #define MSR_BBL_CR_TRIG 0x11a #define MSR_BBL_CR_BUSY 0x11b #define MSR_BBL_CR_CTL3 0x11e #define MSR_SYSENTER_CS_MSR 0x174 #define MSR_SYSENTER_ESP_MSR 0x175 #define MSR_SYSENTER_EIP_MSR 0x176 #define MSR_MCG_CAP 0x179 #define MSR_MCG_STATUS 0x17a #define MSR_MCG_CTL 0x17b #define MSR_EVNTSEL0 0x186 #define MSR_EVNTSEL1 0x187 #define MSR_THERM_CONTROL 0x19a #define MSR_THERM_INTERRUPT 0x19b #define MSR_THERM_STATUS 0x19c #define MSR_IA32_MISC_ENABLE 0x1a0 #define MSR_IA32_TEMPERATURE_TARGET 0x1a2 #define MSR_TURBO_RATIO_LIMIT 0x1ad #define MSR_TURBO_RATIO_LIMIT1 0x1ae #define MSR_DEBUGCTLMSR 0x1d9 #define MSR_LASTBRANCHFROMIP 0x1db #define MSR_LASTBRANCHTOIP 0x1dc #define MSR_LASTINTFROMIP 0x1dd #define MSR_LASTINTTOIP 0x1de #define MSR_ROB_CR_BKUPTMPDR6 0x1e0 #define MSR_MTRRVarBase 0x200 #define MSR_MTRR64kBase 0x250 #define MSR_MTRR16kBase 0x258 #define MSR_MTRR4kBase 0x268 #define MSR_PAT 0x277 #define MSR_MC0_CTL2 0x280 #define MSR_MTRRdefType 0x2ff #define MSR_MC0_CTL 0x400 #define MSR_MC0_STATUS 0x401 #define MSR_MC0_ADDR 0x402 #define MSR_MC0_MISC 0x403 #define MSR_MC1_CTL 0x404 #define MSR_MC1_STATUS 0x405 #define MSR_MC1_ADDR 0x406 #define MSR_MC1_MISC 0x407 #define MSR_MC2_CTL 0x408 #define MSR_MC2_STATUS 0x409 #define MSR_MC2_ADDR 0x40a #define MSR_MC2_MISC 0x40b #define MSR_MC3_CTL 0x40c #define MSR_MC3_STATUS 0x40d #define MSR_MC3_ADDR 0x40e #define MSR_MC3_MISC 0x40f #define MSR_MC4_CTL 0x410 #define MSR_MC4_STATUS 0x411 #define MSR_MC4_ADDR 0x412 #define MSR_MC4_MISC 0x413 #define MSR_RAPL_POWER_UNIT 0x606 #define MSR_PKG_ENERGY_STATUS 0x611 #define MSR_DRAM_ENERGY_STATUS 0x619 #define MSR_PP0_ENERGY_STATUS 0x639 #define MSR_PP1_ENERGY_STATUS 0x641 #define MSR_TSC_DEADLINE 0x6e0 /* Writes are not serializing */ /* * VMX MSRs */ #define MSR_VMX_BASIC 0x480 #define MSR_VMX_PINBASED_CTLS 0x481 #define MSR_VMX_PROCBASED_CTLS 0x482 #define MSR_VMX_EXIT_CTLS 0x483 #define MSR_VMX_ENTRY_CTLS 0x484 #define MSR_VMX_CR0_FIXED0 0x486 #define MSR_VMX_CR0_FIXED1 0x487 #define MSR_VMX_CR4_FIXED0 0x488 #define MSR_VMX_CR4_FIXED1 0x489 #define MSR_VMX_PROCBASED_CTLS2 0x48b #define MSR_VMX_EPT_VPID_CAP 0x48c #define MSR_VMX_TRUE_PINBASED_CTLS 0x48d #define MSR_VMX_TRUE_PROCBASED_CTLS 0x48e #define MSR_VMX_TRUE_EXIT_CTLS 0x48f #define MSR_VMX_TRUE_ENTRY_CTLS 0x490 /* * X2APIC MSRs. * Writes are not serializing. */ #define MSR_APIC_000 0x800 #define MSR_APIC_ID 0x802 #define MSR_APIC_VERSION 0x803 #define MSR_APIC_TPR 0x808 #define MSR_APIC_EOI 0x80b #define MSR_APIC_LDR 0x80d #define MSR_APIC_SVR 0x80f #define MSR_APIC_ISR0 0x810 #define MSR_APIC_ISR1 0x811 #define MSR_APIC_ISR2 0x812 #define MSR_APIC_ISR3 0x813 #define MSR_APIC_ISR4 0x814 #define MSR_APIC_ISR5 0x815 #define MSR_APIC_ISR6 0x816 #define MSR_APIC_ISR7 0x817 #define MSR_APIC_TMR0 0x818 #define MSR_APIC_IRR0 0x820 #define MSR_APIC_ESR 0x828 #define MSR_APIC_LVT_CMCI 0x82F #define MSR_APIC_ICR 0x830 #define MSR_APIC_LVT_TIMER 0x832 #define MSR_APIC_LVT_THERMAL 0x833 #define MSR_APIC_LVT_PCINT 0x834 #define MSR_APIC_LVT_LINT0 0x835 #define MSR_APIC_LVT_LINT1 0x836 #define MSR_APIC_LVT_ERROR 0x837 #define MSR_APIC_ICR_TIMER 0x838 #define MSR_APIC_CCR_TIMER 0x839 #define MSR_APIC_DCR_TIMER 0x83e #define MSR_APIC_SELF_IPI 0x83f #define MSR_IA32_XSS 0xda0 /* * Intel Processor Trace (PT) MSRs. */ #define MSR_IA32_RTIT_OUTPUT_BASE 0x560 /* Trace Output Base Register (R/W) */ #define MSR_IA32_RTIT_OUTPUT_MASK_PTRS 0x561 /* Trace Output Mask Pointers Register (R/W) */ #define MSR_IA32_RTIT_CTL 0x570 /* Trace Control Register (R/W) */ #define RTIT_CTL_TRACEEN (1 << 0) #define RTIT_CTL_CYCEN (1 << 1) #define RTIT_CTL_OS (1 << 2) #define RTIT_CTL_USER (1 << 3) #define RTIT_CTL_PWREVTEN (1 << 4) #define RTIT_CTL_FUPONPTW (1 << 5) #define RTIT_CTL_FABRICEN (1 << 6) #define RTIT_CTL_CR3FILTER (1 << 7) #define RTIT_CTL_TOPA (1 << 8) #define RTIT_CTL_MTCEN (1 << 9) #define RTIT_CTL_TSCEN (1 << 10) #define RTIT_CTL_DISRETC (1 << 11) #define RTIT_CTL_PTWEN (1 << 12) #define RTIT_CTL_BRANCHEN (1 << 13) #define RTIT_CTL_MTC_FREQ_S 14 #define RTIT_CTL_MTC_FREQ(n) ((n) << RTIT_CTL_MTC_FREQ_S) #define RTIT_CTL_MTC_FREQ_M (0xf << RTIT_CTL_MTC_FREQ_S) #define RTIT_CTL_CYC_THRESH_S 19 #define RTIT_CTL_CYC_THRESH_M (0xf << RTIT_CTL_CYC_THRESH_S) #define RTIT_CTL_PSB_FREQ_S 24 #define RTIT_CTL_PSB_FREQ_M (0xf << RTIT_CTL_PSB_FREQ_S) #define RTIT_CTL_ADDR_CFG_S(n) (32 + (n) * 4) #define RTIT_CTL_ADDR0_CFG_S 32 #define RTIT_CTL_ADDR0_CFG_M (0xfULL << RTIT_CTL_ADDR0_CFG_S) #define RTIT_CTL_ADDR1_CFG_S 36 #define RTIT_CTL_ADDR1_CFG_M (0xfULL << RTIT_CTL_ADDR1_CFG_S) #define RTIT_CTL_ADDR2_CFG_S 40 #define RTIT_CTL_ADDR2_CFG_M (0xfULL << RTIT_CTL_ADDR2_CFG_S) #define RTIT_CTL_ADDR3_CFG_S 44 #define RTIT_CTL_ADDR3_CFG_M (0xfULL << RTIT_CTL_ADDR3_CFG_S) #define MSR_IA32_RTIT_STATUS 0x571 /* Tracing Status Register (R/W) */ #define RTIT_STATUS_FILTEREN (1 << 0) #define RTIT_STATUS_CONTEXTEN (1 << 1) #define RTIT_STATUS_TRIGGEREN (1 << 2) #define RTIT_STATUS_ERROR (1 << 4) #define RTIT_STATUS_STOPPED (1 << 5) #define RTIT_STATUS_PACKETBYTECNT_S 32 #define RTIT_STATUS_PACKETBYTECNT_M (0x1ffffULL << RTIT_STATUS_PACKETBYTECNT_S) #define MSR_IA32_RTIT_CR3_MATCH 0x572 /* Trace Filter CR3 Match Register (R/W) */ #define MSR_IA32_RTIT_ADDR_A(n) (0x580 + (n) * 2) #define MSR_IA32_RTIT_ADDR_B(n) (0x581 + (n) * 2) #define MSR_IA32_RTIT_ADDR0_A 0x580 /* Region 0 Start Address (R/W) */ #define MSR_IA32_RTIT_ADDR0_B 0x581 /* Region 0 End Address (R/W) */ #define MSR_IA32_RTIT_ADDR1_A 0x582 /* Region 1 Start Address (R/W) */ #define MSR_IA32_RTIT_ADDR1_B 0x583 /* Region 1 End Address (R/W) */ #define MSR_IA32_RTIT_ADDR2_A 0x584 /* Region 2 Start Address (R/W) */ #define MSR_IA32_RTIT_ADDR2_B 0x585 /* Region 2 End Address (R/W) */ #define MSR_IA32_RTIT_ADDR3_A 0x586 /* Region 3 Start Address (R/W) */ #define MSR_IA32_RTIT_ADDR3_B 0x587 /* Region 3 End Address (R/W) */ /* Intel Processor Trace Table of Physical Addresses (ToPA). */ #define TOPA_SIZE_S 6 #define TOPA_SIZE_M (0xf << TOPA_SIZE_S) #define TOPA_SIZE_4K (0 << TOPA_SIZE_S) #define TOPA_SIZE_8K (1 << TOPA_SIZE_S) #define TOPA_SIZE_16K (2 << TOPA_SIZE_S) #define TOPA_SIZE_32K (3 << TOPA_SIZE_S) #define TOPA_SIZE_64K (4 << TOPA_SIZE_S) #define TOPA_SIZE_128K (5 << TOPA_SIZE_S) #define TOPA_SIZE_256K (6 << TOPA_SIZE_S) #define TOPA_SIZE_512K (7 << TOPA_SIZE_S) #define TOPA_SIZE_1M (8 << TOPA_SIZE_S) #define TOPA_SIZE_2M (9 << TOPA_SIZE_S) #define TOPA_SIZE_4M (10 << TOPA_SIZE_S) #define TOPA_SIZE_8M (11 << TOPA_SIZE_S) #define TOPA_SIZE_16M (12 << TOPA_SIZE_S) #define TOPA_SIZE_32M (13 << TOPA_SIZE_S) #define TOPA_SIZE_64M (14 << TOPA_SIZE_S) #define TOPA_SIZE_128M (15 << TOPA_SIZE_S) #define TOPA_STOP (1 << 4) #define TOPA_INT (1 << 2) #define TOPA_END (1 << 0) /* * Constants related to MSR's. */ #define APICBASE_RESERVED 0x000002ff #define APICBASE_BSP 0x00000100 #define APICBASE_X2APIC 0x00000400 #define APICBASE_ENABLED 0x00000800 #define APICBASE_ADDRESS 0xfffff000 /* MSR_IA32_FEATURE_CONTROL related */ #define IA32_FEATURE_CONTROL_LOCK 0x01 /* lock bit */ #define IA32_FEATURE_CONTROL_SMX_EN 0x02 /* enable VMX inside SMX */ #define IA32_FEATURE_CONTROL_VMX_EN 0x04 /* enable VMX outside SMX */ /* MSR IA32_MISC_ENABLE */ #define IA32_MISC_EN_FASTSTR 0x0000000000000001ULL #define IA32_MISC_EN_ATCCE 0x0000000000000008ULL #define IA32_MISC_EN_PERFMON 0x0000000000000080ULL #define IA32_MISC_EN_PEBSU 0x0000000000001000ULL #define IA32_MISC_EN_ESSTE 0x0000000000010000ULL #define IA32_MISC_EN_MONE 0x0000000000040000ULL #define IA32_MISC_EN_LIMCPUID 0x0000000000400000ULL #define IA32_MISC_EN_xTPRD 0x0000000000800000ULL #define IA32_MISC_EN_XDD 0x0000000400000000ULL /* * IA32_SPEC_CTRL and IA32_PRED_CMD MSRs are described in the Intel' * document 336996-001 Speculative Execution Side Channel Mitigations. */ /* MSR IA32_SPEC_CTRL */ #define IA32_SPEC_CTRL_IBRS 0x00000001 #define IA32_SPEC_CTRL_STIBP 0x00000002 #define IA32_SPEC_CTRL_SSBD 0x00000004 /* MSR IA32_PRED_CMD */ #define IA32_PRED_CMD_IBPB_BARRIER 0x0000000000000001ULL /* MSR IA32_FLUSH_CMD */ #define IA32_FLUSH_CMD_L1D 0x00000001 /* * PAT modes. */ #define PAT_UNCACHEABLE 0x00 #define PAT_WRITE_COMBINING 0x01 #define PAT_WRITE_THROUGH 0x04 #define PAT_WRITE_PROTECTED 0x05 #define PAT_WRITE_BACK 0x06 #define PAT_UNCACHED 0x07 #define PAT_VALUE(i, m) ((long long)(m) << (8 * (i))) #define PAT_MASK(i) PAT_VALUE(i, 0xff) /* * Constants related to MTRRs */ #define MTRR_UNCACHEABLE 0x00 #define MTRR_WRITE_COMBINING 0x01 #define MTRR_WRITE_THROUGH 0x04 #define MTRR_WRITE_PROTECTED 0x05 #define MTRR_WRITE_BACK 0x06 #define MTRR_N64K 8 /* numbers of fixed-size entries */ #define MTRR_N16K 16 #define MTRR_N4K 64 #define MTRR_CAP_WC 0x0000000000000400 #define MTRR_CAP_FIXED 0x0000000000000100 #define MTRR_CAP_VCNT 0x00000000000000ff #define MTRR_DEF_ENABLE 0x0000000000000800 #define MTRR_DEF_FIXED_ENABLE 0x0000000000000400 #define MTRR_DEF_TYPE 0x00000000000000ff #define MTRR_PHYSBASE_PHYSBASE 0x000ffffffffff000 #define MTRR_PHYSBASE_TYPE 0x00000000000000ff #define MTRR_PHYSMASK_PHYSMASK 0x000ffffffffff000 #define MTRR_PHYSMASK_VALID 0x0000000000000800 /* * Cyrix configuration registers, accessible as IO ports. */ #define CCR0 0xc0 /* Configuration control register 0 */ #define CCR0_NC0 0x01 /* First 64K of each 1M memory region is non-cacheable */ #define CCR0_NC1 0x02 /* 640K-1M region is non-cacheable */ #define CCR0_A20M 0x04 /* Enables A20M# input pin */ #define CCR0_KEN 0x08 /* Enables KEN# input pin */ #define CCR0_FLUSH 0x10 /* Enables FLUSH# input pin */ #define CCR0_BARB 0x20 /* Flushes internal cache when entering hold state */ #define CCR0_CO 0x40 /* Cache org: 1=direct mapped, 0=2x set assoc */ #define CCR0_SUSPEND 0x80 /* Enables SUSP# and SUSPA# pins */ #define CCR1 0xc1 /* Configuration control register 1 */ #define CCR1_RPL 0x01 /* Enables RPLSET and RPLVAL# pins */ #define CCR1_SMI 0x02 /* Enables SMM pins */ #define CCR1_SMAC 0x04 /* System management memory access */ #define CCR1_MMAC 0x08 /* Main memory access */ #define CCR1_NO_LOCK 0x10 /* Negate LOCK# */ #define CCR1_SM3 0x80 /* SMM address space address region 3 */ #define CCR2 0xc2 #define CCR2_WB 0x02 /* Enables WB cache interface pins */ #define CCR2_SADS 0x02 /* Slow ADS */ #define CCR2_LOCK_NW 0x04 /* LOCK NW Bit */ #define CCR2_SUSP_HLT 0x08 /* Suspend on HALT */ #define CCR2_WT1 0x10 /* WT region 1 */ #define CCR2_WPR1 0x10 /* Write-protect region 1 */ #define CCR2_BARB 0x20 /* Flushes write-back cache when entering hold state. */ #define CCR2_BWRT 0x40 /* Enables burst write cycles */ #define CCR2_USE_SUSP 0x80 /* Enables suspend pins */ #define CCR3 0xc3 #define CCR3_SMILOCK 0x01 /* SMM register lock */ #define CCR3_NMI 0x02 /* Enables NMI during SMM */ #define CCR3_LINBRST 0x04 /* Linear address burst cycles */ #define CCR3_SMMMODE 0x08 /* SMM Mode */ #define CCR3_MAPEN0 0x10 /* Enables Map0 */ #define CCR3_MAPEN1 0x20 /* Enables Map1 */ #define CCR3_MAPEN2 0x40 /* Enables Map2 */ #define CCR3_MAPEN3 0x80 /* Enables Map3 */ #define CCR4 0xe8 #define CCR4_IOMASK 0x07 #define CCR4_MEM 0x08 /* Enables momory bypassing */ #define CCR4_DTE 0x10 /* Enables directory table entry cache */ #define CCR4_FASTFPE 0x20 /* Fast FPU exception */ #define CCR4_CPUID 0x80 /* Enables CPUID instruction */ #define CCR5 0xe9 #define CCR5_WT_ALLOC 0x01 /* Write-through allocate */ #define CCR5_SLOP 0x02 /* LOOP instruction slowed down */ #define CCR5_LBR1 0x10 /* Local bus region 1 */ #define CCR5_ARREN 0x20 /* Enables ARR region */ #define CCR6 0xea #define CCR7 0xeb /* Performance Control Register (5x86 only). */ #define PCR0 0x20 #define PCR0_RSTK 0x01 /* Enables return stack */ #define PCR0_BTB 0x02 /* Enables branch target buffer */ #define PCR0_LOOP 0x04 /* Enables loop */ #define PCR0_AIS 0x08 /* Enables all instrcutions stalled to serialize pipe. */ #define PCR0_MLR 0x10 /* Enables reordering of misaligned loads */ #define PCR0_BTBRT 0x40 /* Enables BTB test register. */ #define PCR0_LSSER 0x80 /* Disable reorder */ /* Device Identification Registers */ #define DIR0 0xfe #define DIR1 0xff /* * Machine Check register constants. */ #define MCG_CAP_COUNT 0x000000ff #define MCG_CAP_CTL_P 0x00000100 #define MCG_CAP_EXT_P 0x00000200 #define MCG_CAP_CMCI_P 0x00000400 #define MCG_CAP_TES_P 0x00000800 #define MCG_CAP_EXT_CNT 0x00ff0000 #define MCG_CAP_SER_P 0x01000000 #define MCG_STATUS_RIPV 0x00000001 #define MCG_STATUS_EIPV 0x00000002 #define MCG_STATUS_MCIP 0x00000004 #define MCG_CTL_ENABLE 0xffffffffffffffff #define MCG_CTL_DISABLE 0x0000000000000000 #define MSR_MC_CTL(x) (MSR_MC0_CTL + (x) * 4) #define MSR_MC_STATUS(x) (MSR_MC0_STATUS + (x) * 4) #define MSR_MC_ADDR(x) (MSR_MC0_ADDR + (x) * 4) #define MSR_MC_MISC(x) (MSR_MC0_MISC + (x) * 4) #define MSR_MC_CTL2(x) (MSR_MC0_CTL2 + (x)) /* If MCG_CAP_CMCI_P */ #define MC_STATUS_MCA_ERROR 0x000000000000ffff #define MC_STATUS_MODEL_ERROR 0x00000000ffff0000 #define MC_STATUS_OTHER_INFO 0x01ffffff00000000 #define MC_STATUS_COR_COUNT 0x001fffc000000000 /* If MCG_CAP_CMCI_P */ #define MC_STATUS_TES_STATUS 0x0060000000000000 /* If MCG_CAP_TES_P */ #define MC_STATUS_AR 0x0080000000000000 /* If MCG_CAP_TES_P */ #define MC_STATUS_S 0x0100000000000000 /* If MCG_CAP_TES_P */ #define MC_STATUS_PCC 0x0200000000000000 #define MC_STATUS_ADDRV 0x0400000000000000 #define MC_STATUS_MISCV 0x0800000000000000 #define MC_STATUS_EN 0x1000000000000000 #define MC_STATUS_UC 0x2000000000000000 #define MC_STATUS_OVER 0x4000000000000000 #define MC_STATUS_VAL 0x8000000000000000 #define MC_MISC_RA_LSB 0x000000000000003f /* If MCG_CAP_SER_P */ #define MC_MISC_ADDRESS_MODE 0x00000000000001c0 /* If MCG_CAP_SER_P */ #define MC_CTL2_THRESHOLD 0x0000000000007fff #define MC_CTL2_CMCI_EN 0x0000000040000000 #define MC_AMDNB_BANK 4 #define MC_MISC_AMD_VAL 0x8000000000000000 /* Counter presence valid */ #define MC_MISC_AMD_CNTP 0x4000000000000000 /* Counter present */ #define MC_MISC_AMD_LOCK 0x2000000000000000 /* Register locked */ #define MC_MISC_AMD_INTP 0x1000000000000000 /* Int. type can generate interrupts */ #define MC_MISC_AMD_LVT_MASK 0x00f0000000000000 /* Extended LVT offset */ #define MC_MISC_AMD_LVT_SHIFT 52 #define MC_MISC_AMD_CNTEN 0x0008000000000000 /* Counter enabled */ #define MC_MISC_AMD_INT_MASK 0x0006000000000000 /* Interrupt type */ #define MC_MISC_AMD_INT_LVT 0x0002000000000000 /* Interrupt via Extended LVT */ #define MC_MISC_AMD_INT_SMI 0x0004000000000000 /* SMI */ #define MC_MISC_AMD_OVERFLOW 0x0001000000000000 /* Counter overflow */ #define MC_MISC_AMD_CNT_MASK 0x00000fff00000000 /* Counter value */ #define MC_MISC_AMD_CNT_SHIFT 32 #define MC_MISC_AMD_CNT_MAX 0xfff #define MC_MISC_AMD_PTR_MASK 0x00000000ff000000 /* Pointer to additional registers */ #define MC_MISC_AMD_PTR_SHIFT 24 /* * The following four 3-byte registers control the non-cacheable regions. * These registers must be written as three separate bytes. * * NCRx+0: A31-A24 of starting address * NCRx+1: A23-A16 of starting address * NCRx+2: A15-A12 of starting address | NCR_SIZE_xx. * * The non-cacheable region's starting address must be aligned to the * size indicated by the NCR_SIZE_xx field. */ #define NCR1 0xc4 #define NCR2 0xc7 #define NCR3 0xca #define NCR4 0xcd #define NCR_SIZE_0K 0 #define NCR_SIZE_4K 1 #define NCR_SIZE_8K 2 #define NCR_SIZE_16K 3 #define NCR_SIZE_32K 4 #define NCR_SIZE_64K 5 #define NCR_SIZE_128K 6 #define NCR_SIZE_256K 7 #define NCR_SIZE_512K 8 #define NCR_SIZE_1M 9 #define NCR_SIZE_2M 10 #define NCR_SIZE_4M 11 #define NCR_SIZE_8M 12 #define NCR_SIZE_16M 13 #define NCR_SIZE_32M 14 #define NCR_SIZE_4G 15 /* * The address region registers are used to specify the location and * size for the eight address regions. * * ARRx + 0: A31-A24 of start address * ARRx + 1: A23-A16 of start address * ARRx + 2: A15-A12 of start address | ARR_SIZE_xx */ #define ARR0 0xc4 #define ARR1 0xc7 #define ARR2 0xca #define ARR3 0xcd #define ARR4 0xd0 #define ARR5 0xd3 #define ARR6 0xd6 #define ARR7 0xd9 #define ARR_SIZE_0K 0 #define ARR_SIZE_4K 1 #define ARR_SIZE_8K 2 #define ARR_SIZE_16K 3 #define ARR_SIZE_32K 4 #define ARR_SIZE_64K 5 #define ARR_SIZE_128K 6 #define ARR_SIZE_256K 7 #define ARR_SIZE_512K 8 #define ARR_SIZE_1M 9 #define ARR_SIZE_2M 10 #define ARR_SIZE_4M 11 #define ARR_SIZE_8M 12 #define ARR_SIZE_16M 13 #define ARR_SIZE_32M 14 #define ARR_SIZE_4G 15 /* * The region control registers specify the attributes associated with * the ARRx addres regions. */ #define RCR0 0xdc #define RCR1 0xdd #define RCR2 0xde #define RCR3 0xdf #define RCR4 0xe0 #define RCR5 0xe1 #define RCR6 0xe2 #define RCR7 0xe3 #define RCR_RCD 0x01 /* Disables caching for ARRx (x = 0-6). */ #define RCR_RCE 0x01 /* Enables caching for ARR7. */ #define RCR_WWO 0x02 /* Weak write ordering. */ #define RCR_WL 0x04 /* Weak locking. */ #define RCR_WG 0x08 /* Write gathering. */ #define RCR_WT 0x10 /* Write-through. */ #define RCR_NLB 0x20 /* LBA# pin is not asserted. */ /* AMD Write Allocate Top-Of-Memory and Control Register */ #define AMD_WT_ALLOC_TME 0x40000 /* top-of-memory enable */ #define AMD_WT_ALLOC_PRE 0x20000 /* programmable range enable */ #define AMD_WT_ALLOC_FRE 0x10000 /* fixed (A0000-FFFFF) range enable */ /* AMD64 MSR's */ #define MSR_EFER 0xc0000080 /* extended features */ #define MSR_STAR 0xc0000081 /* legacy mode SYSCALL target/cs/ss */ #define MSR_LSTAR 0xc0000082 /* long mode SYSCALL target rip */ #define MSR_CSTAR 0xc0000083 /* compat mode SYSCALL target rip */ #define MSR_SF_MASK 0xc0000084 /* syscall flags mask */ #define MSR_FSBASE 0xc0000100 /* base address of the %fs "segment" */ #define MSR_GSBASE 0xc0000101 /* base address of the %gs "segment" */ #define MSR_KGSBASE 0xc0000102 /* base address of the kernel %gs */ #define MSR_PERFEVSEL0 0xc0010000 #define MSR_PERFEVSEL1 0xc0010001 #define MSR_PERFEVSEL2 0xc0010002 #define MSR_PERFEVSEL3 0xc0010003 #define MSR_K7_PERFCTR0 0xc0010004 #define MSR_K7_PERFCTR1 0xc0010005 #define MSR_K7_PERFCTR2 0xc0010006 #define MSR_K7_PERFCTR3 0xc0010007 #define MSR_SYSCFG 0xc0010010 #define MSR_HWCR 0xc0010015 #define MSR_IORRBASE0 0xc0010016 #define MSR_IORRMASK0 0xc0010017 #define MSR_IORRBASE1 0xc0010018 #define MSR_IORRMASK1 0xc0010019 #define MSR_TOP_MEM 0xc001001a /* boundary for ram below 4G */ #define MSR_TOP_MEM2 0xc001001d /* boundary for ram above 4G */ #define MSR_NB_CFG1 0xc001001f /* NB configuration 1 */ #define MSR_K8_UCODE_UPDATE 0xc0010020 /* update microcode */ #define MSR_MC0_CTL_MASK 0xc0010044 #define MSR_P_STATE_LIMIT 0xc0010061 /* P-state Current Limit Register */ #define MSR_P_STATE_CONTROL 0xc0010062 /* P-state Control Register */ #define MSR_P_STATE_STATUS 0xc0010063 /* P-state Status Register */ #define MSR_P_STATE_CONFIG(n) (0xc0010064 + (n)) /* P-state Config */ #define MSR_SMM_ADDR 0xc0010112 /* SMM TSEG base address */ #define MSR_SMM_MASK 0xc0010113 /* SMM TSEG address mask */ #define MSR_VM_CR 0xc0010114 /* SVM: feature control */ #define MSR_VM_HSAVE_PA 0xc0010117 /* SVM: host save area address */ #define MSR_AMD_CPUID07 0xc0011002 /* CPUID 07 %ebx override */ #define MSR_EXTFEATURES 0xc0011005 /* Extended CPUID Features override */ #define MSR_IC_CFG 0xc0011021 /* Instruction Cache Configuration */ /* MSR_VM_CR related */ #define VM_CR_SVMDIS 0x10 /* SVM: disabled by BIOS */ /* VIA ACE crypto featureset: for via_feature_rng */ #define VIA_HAS_RNG 1 /* cpu has RNG */ /* VIA ACE crypto featureset: for via_feature_xcrypt */ #define VIA_HAS_AES 1 /* cpu has AES */ #define VIA_HAS_SHA 2 /* cpu has SHA1 & SHA256 */ #define VIA_HAS_MM 4 /* cpu has RSA instructions */ #define VIA_HAS_AESCTR 8 /* cpu has AES-CTR instructions */ /* Centaur Extended Feature flags */ #define VIA_CPUID_HAS_RNG 0x000004 #define VIA_CPUID_DO_RNG 0x000008 #define VIA_CPUID_HAS_ACE 0x000040 #define VIA_CPUID_DO_ACE 0x000080 #define VIA_CPUID_HAS_ACE2 0x000100 #define VIA_CPUID_DO_ACE2 0x000200 #define VIA_CPUID_HAS_PHE 0x000400 #define VIA_CPUID_DO_PHE 0x000800 #define VIA_CPUID_HAS_PMM 0x001000 #define VIA_CPUID_DO_PMM 0x002000 /* VIA ACE xcrypt-* instruction context control options */ #define VIA_CRYPT_CWLO_ROUND_M 0x0000000f #define VIA_CRYPT_CWLO_ALG_M 0x00000070 #define VIA_CRYPT_CWLO_ALG_AES 0x00000000 #define VIA_CRYPT_CWLO_KEYGEN_M 0x00000080 #define VIA_CRYPT_CWLO_KEYGEN_HW 0x00000000 #define VIA_CRYPT_CWLO_KEYGEN_SW 0x00000080 #define VIA_CRYPT_CWLO_NORMAL 0x00000000 #define VIA_CRYPT_CWLO_INTERMEDIATE 0x00000100 #define VIA_CRYPT_CWLO_ENCRYPT 0x00000000 #define VIA_CRYPT_CWLO_DECRYPT 0x00000200 #define VIA_CRYPT_CWLO_KEY128 0x0000000a /* 128bit, 10 rds */ #define VIA_CRYPT_CWLO_KEY192 0x0000040c /* 192bit, 12 rds */ #define VIA_CRYPT_CWLO_KEY256 0x0000080e /* 256bit, 15 rds */ #endif /* !_MACHINE_SPECIALREG_H_ */ Index: releng/12.0/sys/x86/x86/cpu_machdep.c =================================================================== --- releng/12.0/sys/x86/x86/cpu_machdep.c (revision 340626) +++ releng/12.0/sys/x86/x86/cpu_machdep.c (revision 340627) @@ -1,962 +1,962 @@ /*- * Copyright (c) 2003 Peter Wemm. * Copyright (c) 1992 Terrence R. Lambert. * Copyright (c) 1982, 1987, 1990 The Regents of the University of California. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)machdep.c 7.4 (Berkeley) 6/3/91 */ #include __FBSDID("$FreeBSD$"); #include "opt_acpi.h" #include "opt_atpic.h" #include "opt_cpu.h" #include "opt_ddb.h" #include "opt_inet.h" #include "opt_isa.h" #include "opt_kdb.h" #include "opt_kstack_pages.h" #include "opt_maxmem.h" #include "opt_mp_watchdog.h" #include "opt_platform.h" #ifdef __i386__ #include "opt_apic.h" #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SMP #include #endif #ifdef CPU_ELAN #include #endif #include #include #include #include #include #include #include #include #include #include #include #define STATE_RUNNING 0x0 #define STATE_MWAIT 0x1 #define STATE_SLEEPING 0x2 #ifdef SMP static u_int cpu_reset_proxyid; static volatile u_int cpu_reset_proxy_active; #endif /* * Machine dependent boot() routine * * I haven't seen anything to put here yet * Possibly some stuff might be grafted back here from boot() */ void cpu_boot(int howto) { } /* * Flush the D-cache for non-DMA I/O so that the I-cache can * be made coherent later. */ void cpu_flush_dcache(void *ptr, size_t len) { /* Not applicable */ } void acpi_cpu_c1(void) { __asm __volatile("sti; hlt"); } /* * Use mwait to pause execution while waiting for an interrupt or * another thread to signal that there is more work. * * NOTE: Interrupts will cause a wakeup; however, this function does * not enable interrupt handling. The caller is responsible to enable * interrupts. */ void acpi_cpu_idle_mwait(uint32_t mwait_hint) { int *state; uint64_t v; /* * A comment in Linux patch claims that 'CPUs run faster with * speculation protection disabled. All CPU threads in a core * must disable speculation protection for it to be * disabled. Disable it while we are idle so the other * hyperthread can run fast.' * * XXXKIB. Software coordination mode should be supported, * but all Intel CPUs provide hardware coordination. */ state = (int *)PCPU_PTR(monitorbuf); KASSERT(atomic_load_int(state) == STATE_SLEEPING, ("cpu_mwait_cx: wrong monitorbuf state")); atomic_store_int(state, STATE_MWAIT); if (PCPU_GET(ibpb_set) || hw_ssb_active) { v = rdmsr(MSR_IA32_SPEC_CTRL); wrmsr(MSR_IA32_SPEC_CTRL, v & ~(IA32_SPEC_CTRL_IBRS | IA32_SPEC_CTRL_STIBP | IA32_SPEC_CTRL_SSBD)); } else { v = 0; } cpu_monitor(state, 0, 0); if (atomic_load_int(state) == STATE_MWAIT) cpu_mwait(MWAIT_INTRBREAK, mwait_hint); /* * SSB cannot be disabled while we sleep, or rather, if it was * disabled, the sysctl thread will bind to our cpu to tweak * MSR. */ if (v != 0) wrmsr(MSR_IA32_SPEC_CTRL, v); /* * We should exit on any event that interrupts mwait, because * that event might be a wanted interrupt. */ atomic_store_int(state, STATE_RUNNING); } /* Get current clock frequency for the given cpu id. */ int cpu_est_clockrate(int cpu_id, uint64_t *rate) { uint64_t tsc1, tsc2; uint64_t acnt, mcnt, perf; register_t reg; if (pcpu_find(cpu_id) == NULL || rate == NULL) return (EINVAL); #ifdef __i386__ if ((cpu_feature & CPUID_TSC) == 0) return (EOPNOTSUPP); #endif /* * If TSC is P-state invariant and APERF/MPERF MSRs do not exist, * DELAY(9) based logic fails. */ if (tsc_is_invariant && !tsc_perf_stat) return (EOPNOTSUPP); #ifdef SMP if (smp_cpus > 1) { /* Schedule ourselves on the indicated cpu. */ thread_lock(curthread); sched_bind(curthread, cpu_id); thread_unlock(curthread); } #endif /* Calibrate by measuring a short delay. */ reg = intr_disable(); if (tsc_is_invariant) { wrmsr(MSR_MPERF, 0); wrmsr(MSR_APERF, 0); tsc1 = rdtsc(); DELAY(1000); mcnt = rdmsr(MSR_MPERF); acnt = rdmsr(MSR_APERF); tsc2 = rdtsc(); intr_restore(reg); perf = 1000 * acnt / mcnt; *rate = (tsc2 - tsc1) * perf; } else { tsc1 = rdtsc(); DELAY(1000); tsc2 = rdtsc(); intr_restore(reg); *rate = (tsc2 - tsc1) * 1000; } #ifdef SMP if (smp_cpus > 1) { thread_lock(curthread); sched_unbind(curthread); thread_unlock(curthread); } #endif return (0); } /* * Shutdown the CPU as much as possible */ void cpu_halt(void) { for (;;) halt(); } static void cpu_reset_real(void) { struct region_descriptor null_idt; int b; disable_intr(); #ifdef CPU_ELAN if (elan_mmcr != NULL) elan_mmcr->RESCFG = 1; #endif #ifdef __i386__ if (cpu == CPU_GEODE1100) { /* Attempt Geode's own reset */ outl(0xcf8, 0x80009044ul); outl(0xcfc, 0xf); } #endif #if !defined(BROKEN_KEYBOARD_RESET) /* * Attempt to do a CPU reset via the keyboard controller, * do not turn off GateA20, as any machine that fails * to do the reset here would then end up in no man's land. */ outb(IO_KBD + 4, 0xFE); DELAY(500000); /* wait 0.5 sec to see if that did it */ #endif /* * Attempt to force a reset via the Reset Control register at * I/O port 0xcf9. Bit 2 forces a system reset when it * transitions from 0 to 1. Bit 1 selects the type of reset * to attempt: 0 selects a "soft" reset, and 1 selects a * "hard" reset. We try a "hard" reset. The first write sets * bit 1 to select a "hard" reset and clears bit 2. The * second write forces a 0 -> 1 transition in bit 2 to trigger * a reset. */ outb(0xcf9, 0x2); outb(0xcf9, 0x6); DELAY(500000); /* wait 0.5 sec to see if that did it */ /* * Attempt to force a reset via the Fast A20 and Init register * at I/O port 0x92. Bit 1 serves as an alternate A20 gate. * Bit 0 asserts INIT# when set to 1. We are careful to only * preserve bit 1 while setting bit 0. We also must clear bit * 0 before setting it if it isn't already clear. */ b = inb(0x92); if (b != 0xff) { if ((b & 0x1) != 0) outb(0x92, b & 0xfe); outb(0x92, b | 0x1); DELAY(500000); /* wait 0.5 sec to see if that did it */ } printf("No known reset method worked, attempting CPU shutdown\n"); DELAY(1000000); /* wait 1 sec for printf to complete */ /* Wipe the IDT. */ null_idt.rd_limit = 0; null_idt.rd_base = 0; lidt(&null_idt); /* "good night, sweet prince .... " */ breakpoint(); /* NOTREACHED */ while(1); } #ifdef SMP static void cpu_reset_proxy(void) { cpu_reset_proxy_active = 1; while (cpu_reset_proxy_active == 1) ia32_pause(); /* Wait for other cpu to see that we've started */ printf("cpu_reset_proxy: Stopped CPU %d\n", cpu_reset_proxyid); DELAY(1000000); cpu_reset_real(); } #endif void cpu_reset(void) { #ifdef SMP cpuset_t map; u_int cnt; if (smp_started) { map = all_cpus; CPU_CLR(PCPU_GET(cpuid), &map); CPU_NAND(&map, &stopped_cpus); if (!CPU_EMPTY(&map)) { printf("cpu_reset: Stopping other CPUs\n"); stop_cpus(map); } if (PCPU_GET(cpuid) != 0) { cpu_reset_proxyid = PCPU_GET(cpuid); cpustop_restartfunc = cpu_reset_proxy; cpu_reset_proxy_active = 0; printf("cpu_reset: Restarting BSP\n"); /* Restart CPU #0. */ CPU_SETOF(0, &started_cpus); wmb(); cnt = 0; while (cpu_reset_proxy_active == 0 && cnt < 10000000) { ia32_pause(); cnt++; /* Wait for BSP to announce restart */ } if (cpu_reset_proxy_active == 0) { printf("cpu_reset: Failed to restart BSP\n"); } else { cpu_reset_proxy_active = 2; while (1) ia32_pause(); /* NOTREACHED */ } } DELAY(1000000); } #endif cpu_reset_real(); /* NOTREACHED */ } bool cpu_mwait_usable(void) { return ((cpu_feature2 & CPUID2_MON) != 0 && ((cpu_mon_mwait_flags & (CPUID5_MON_MWAIT_EXT | CPUID5_MWAIT_INTRBREAK)) == (CPUID5_MON_MWAIT_EXT | CPUID5_MWAIT_INTRBREAK))); } void (*cpu_idle_hook)(sbintime_t) = NULL; /* ACPI idle hook. */ static int cpu_ident_amdc1e = 0; /* AMD C1E supported. */ static int idle_mwait = 1; /* Use MONITOR/MWAIT for short idle. */ SYSCTL_INT(_machdep, OID_AUTO, idle_mwait, CTLFLAG_RWTUN, &idle_mwait, 0, "Use MONITOR/MWAIT for short idle"); static void cpu_idle_acpi(sbintime_t sbt) { int *state; state = (int *)PCPU_PTR(monitorbuf); atomic_store_int(state, STATE_SLEEPING); /* See comments in cpu_idle_hlt(). */ disable_intr(); if (sched_runnable()) enable_intr(); else if (cpu_idle_hook) cpu_idle_hook(sbt); else acpi_cpu_c1(); atomic_store_int(state, STATE_RUNNING); } static void cpu_idle_hlt(sbintime_t sbt) { int *state; state = (int *)PCPU_PTR(monitorbuf); atomic_store_int(state, STATE_SLEEPING); /* * Since we may be in a critical section from cpu_idle(), if * an interrupt fires during that critical section we may have * a pending preemption. If the CPU halts, then that thread * may not execute until a later interrupt awakens the CPU. * To handle this race, check for a runnable thread after * disabling interrupts and immediately return if one is * found. Also, we must absolutely guarentee that hlt is * the next instruction after sti. This ensures that any * interrupt that fires after the call to disable_intr() will * immediately awaken the CPU from hlt. Finally, please note * that on x86 this works fine because of interrupts enabled only * after the instruction following sti takes place, while IF is set * to 1 immediately, allowing hlt instruction to acknowledge the * interrupt. */ disable_intr(); if (sched_runnable()) enable_intr(); else acpi_cpu_c1(); atomic_store_int(state, STATE_RUNNING); } static void cpu_idle_mwait(sbintime_t sbt) { int *state; state = (int *)PCPU_PTR(monitorbuf); atomic_store_int(state, STATE_MWAIT); /* See comments in cpu_idle_hlt(). */ disable_intr(); if (sched_runnable()) { atomic_store_int(state, STATE_RUNNING); enable_intr(); return; } cpu_monitor(state, 0, 0); if (atomic_load_int(state) == STATE_MWAIT) __asm __volatile("sti; mwait" : : "a" (MWAIT_C1), "c" (0)); else enable_intr(); atomic_store_int(state, STATE_RUNNING); } static void cpu_idle_spin(sbintime_t sbt) { int *state; int i; state = (int *)PCPU_PTR(monitorbuf); atomic_store_int(state, STATE_RUNNING); /* * The sched_runnable() call is racy but as long as there is * a loop missing it one time will have just a little impact if any * (and it is much better than missing the check at all). */ for (i = 0; i < 1000; i++) { if (sched_runnable()) return; cpu_spinwait(); } } /* * C1E renders the local APIC timer dead, so we disable it by * reading the Interrupt Pending Message register and clearing * both C1eOnCmpHalt (bit 28) and SmiOnCmpHalt (bit 27). * * Reference: * "BIOS and Kernel Developer's Guide for AMD NPT Family 0Fh Processors" * #32559 revision 3.00+ */ #define MSR_AMDK8_IPM 0xc0010055 #define AMDK8_SMIONCMPHALT (1ULL << 27) #define AMDK8_C1EONCMPHALT (1ULL << 28) #define AMDK8_CMPHALT (AMDK8_SMIONCMPHALT | AMDK8_C1EONCMPHALT) void cpu_probe_amdc1e(void) { /* * Detect the presence of C1E capability mostly on latest * dual-cores (or future) k8 family. */ if (cpu_vendor_id == CPU_VENDOR_AMD && (cpu_id & 0x00000f00) == 0x00000f00 && (cpu_id & 0x0fff0000) >= 0x00040000) { cpu_ident_amdc1e = 1; } } void (*cpu_idle_fn)(sbintime_t) = cpu_idle_acpi; void cpu_idle(int busy) { uint64_t msr; sbintime_t sbt = -1; CTR2(KTR_SPARE2, "cpu_idle(%d) at %d", busy, curcpu); #ifdef MP_WATCHDOG ap_watchdog(PCPU_GET(cpuid)); #endif /* If we are busy - try to use fast methods. */ if (busy) { if ((cpu_feature2 & CPUID2_MON) && idle_mwait) { cpu_idle_mwait(busy); goto out; } } /* If we have time - switch timers into idle mode. */ if (!busy) { critical_enter(); sbt = cpu_idleclock(); } /* Apply AMD APIC timer C1E workaround. */ if (cpu_ident_amdc1e && cpu_disable_c3_sleep) { msr = rdmsr(MSR_AMDK8_IPM); if (msr & AMDK8_CMPHALT) wrmsr(MSR_AMDK8_IPM, msr & ~AMDK8_CMPHALT); } /* Call main idle method. */ cpu_idle_fn(sbt); /* Switch timers back into active mode. */ if (!busy) { cpu_activeclock(); critical_exit(); } out: CTR2(KTR_SPARE2, "cpu_idle(%d) at %d done", busy, curcpu); } static int cpu_idle_apl31_workaround; SYSCTL_INT(_machdep, OID_AUTO, idle_apl31, CTLFLAG_RW, &cpu_idle_apl31_workaround, 0, "Apollo Lake APL31 MWAIT bug workaround"); int cpu_idle_wakeup(int cpu) { int *state; state = (int *)pcpu_find(cpu)->pc_monitorbuf; switch (atomic_load_int(state)) { case STATE_SLEEPING: return (0); case STATE_MWAIT: atomic_store_int(state, STATE_RUNNING); return (cpu_idle_apl31_workaround ? 0 : 1); case STATE_RUNNING: return (1); default: panic("bad monitor state"); return (1); } } /* * Ordered by speed/power consumption. */ static struct { void *id_fn; char *id_name; int id_cpuid2_flag; } idle_tbl[] = { { .id_fn = cpu_idle_spin, .id_name = "spin" }, { .id_fn = cpu_idle_mwait, .id_name = "mwait", .id_cpuid2_flag = CPUID2_MON }, { .id_fn = cpu_idle_hlt, .id_name = "hlt" }, { .id_fn = cpu_idle_acpi, .id_name = "acpi" }, }; static int idle_sysctl_available(SYSCTL_HANDLER_ARGS) { char *avail, *p; int error; int i; avail = malloc(256, M_TEMP, M_WAITOK); p = avail; for (i = 0; i < nitems(idle_tbl); i++) { if (idle_tbl[i].id_cpuid2_flag != 0 && (cpu_feature2 & idle_tbl[i].id_cpuid2_flag) == 0) continue; if (strcmp(idle_tbl[i].id_name, "acpi") == 0 && cpu_idle_hook == NULL) continue; p += sprintf(p, "%s%s", p != avail ? ", " : "", idle_tbl[i].id_name); } error = sysctl_handle_string(oidp, avail, 0, req); free(avail, M_TEMP); return (error); } SYSCTL_PROC(_machdep, OID_AUTO, idle_available, CTLTYPE_STRING | CTLFLAG_RD, 0, 0, idle_sysctl_available, "A", "list of available idle functions"); static bool cpu_idle_selector(const char *new_idle_name) { int i; for (i = 0; i < nitems(idle_tbl); i++) { if (idle_tbl[i].id_cpuid2_flag != 0 && (cpu_feature2 & idle_tbl[i].id_cpuid2_flag) == 0) continue; if (strcmp(idle_tbl[i].id_name, "acpi") == 0 && cpu_idle_hook == NULL) continue; if (strcmp(idle_tbl[i].id_name, new_idle_name)) continue; cpu_idle_fn = idle_tbl[i].id_fn; if (bootverbose) printf("CPU idle set to %s\n", idle_tbl[i].id_name); return (true); } return (false); } static int cpu_idle_sysctl(SYSCTL_HANDLER_ARGS) { char buf[16], *p; int error, i; p = "unknown"; for (i = 0; i < nitems(idle_tbl); i++) { if (idle_tbl[i].id_fn == cpu_idle_fn) { p = idle_tbl[i].id_name; break; } } strncpy(buf, p, sizeof(buf)); error = sysctl_handle_string(oidp, buf, sizeof(buf), req); if (error != 0 || req->newptr == NULL) return (error); return (cpu_idle_selector(buf) ? 0 : EINVAL); } SYSCTL_PROC(_machdep, OID_AUTO, idle, CTLTYPE_STRING | CTLFLAG_RW, 0, 0, cpu_idle_sysctl, "A", "currently selected idle function"); static void cpu_idle_tun(void *unused __unused) { char tunvar[16]; if (TUNABLE_STR_FETCH("machdep.idle", tunvar, sizeof(tunvar))) cpu_idle_selector(tunvar); else if (cpu_vendor_id == CPU_VENDOR_AMD && CPUID_TO_FAMILY(cpu_id) == 0x17 && CPUID_TO_MODEL(cpu_id) == 0x1) { /* Ryzen erratas 1057, 1109. */ cpu_idle_selector("hlt"); idle_mwait = 0; } if (cpu_vendor_id == CPU_VENDOR_INTEL && cpu_id == 0x506c9) { /* * Apollo Lake errata APL31 (public errata APL30). * Stores to the armed address range may not trigger * MWAIT to resume execution. OS needs to use * interrupts to wake processors from MWAIT-induced * sleep states. */ cpu_idle_apl31_workaround = 1; } TUNABLE_INT_FETCH("machdep.idle_apl31", &cpu_idle_apl31_workaround); } SYSINIT(cpu_idle_tun, SI_SUB_CPU, SI_ORDER_MIDDLE, cpu_idle_tun, NULL); static int panic_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RWTUN, &panic_on_nmi, 0, "Panic on NMI raised by hardware failure"); int nmi_is_broadcast = 1; SYSCTL_INT(_machdep, OID_AUTO, nmi_is_broadcast, CTLFLAG_RWTUN, &nmi_is_broadcast, 0, "Chipset NMI is broadcast"); #ifdef KDB int kdb_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RWTUN, &kdb_on_nmi, 0, "Go to KDB on NMI with unknown source"); #endif void nmi_call_kdb(u_int cpu, u_int type, struct trapframe *frame) { bool claimed = false; #ifdef DEV_ISA /* machine/parity/power fail/"kitchen sink" faults */ if (isa_nmi(frame->tf_err)) { claimed = true; if (panic_on_nmi) panic("NMI indicates hardware failure"); } #endif /* DEV_ISA */ #ifdef KDB if (!claimed && kdb_on_nmi) { /* * NMI can be hooked up to a pushbutton for debugging. */ printf("NMI/cpu%d ... going to debugger\n", cpu); kdb_trap(type, 0, frame); } #endif /* KDB */ } void nmi_handle_intr(u_int type, struct trapframe *frame) { #ifdef SMP if (nmi_is_broadcast) { nmi_call_kdb_smp(type, frame); return; } #endif nmi_call_kdb(PCPU_GET(cpuid), type, frame); } int hw_ibrs_active; int hw_ibrs_disable = 1; SYSCTL_INT(_hw, OID_AUTO, ibrs_active, CTLFLAG_RD, &hw_ibrs_active, 0, "Indirect Branch Restricted Speculation active"); void hw_ibrs_recalculate(void) { uint64_t v; if ((cpu_ia32_arch_caps & IA32_ARCH_CAP_IBRS_ALL) != 0) { if (hw_ibrs_disable) { v = rdmsr(MSR_IA32_SPEC_CTRL); v &= ~(uint64_t)IA32_SPEC_CTRL_IBRS; wrmsr(MSR_IA32_SPEC_CTRL, v); } else { v = rdmsr(MSR_IA32_SPEC_CTRL); v |= IA32_SPEC_CTRL_IBRS; wrmsr(MSR_IA32_SPEC_CTRL, v); } return; } hw_ibrs_active = (cpu_stdext_feature3 & CPUID_STDEXT3_IBPB) != 0 && !hw_ibrs_disable; } static int hw_ibrs_disable_handler(SYSCTL_HANDLER_ARGS) { int error, val; val = hw_ibrs_disable; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); hw_ibrs_disable = val != 0; hw_ibrs_recalculate(); return (0); } SYSCTL_PROC(_hw, OID_AUTO, ibrs_disable, CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE, NULL, 0, hw_ibrs_disable_handler, "I", "Disable Indirect Branch Restricted Speculation"); int hw_ssb_active; int hw_ssb_disable; SYSCTL_INT(_hw, OID_AUTO, spec_store_bypass_disable_active, CTLFLAG_RD, &hw_ssb_active, 0, "Speculative Store Bypass Disable active"); static void hw_ssb_set_one(bool enable) { uint64_t v; v = rdmsr(MSR_IA32_SPEC_CTRL); if (enable) v |= (uint64_t)IA32_SPEC_CTRL_SSBD; else v &= ~(uint64_t)IA32_SPEC_CTRL_SSBD; wrmsr(MSR_IA32_SPEC_CTRL, v); } static void hw_ssb_set(bool enable, bool for_all_cpus) { struct thread *td; int bound_cpu, i, is_bound; if ((cpu_stdext_feature3 & CPUID_STDEXT3_SSBD) == 0) { hw_ssb_active = 0; return; } hw_ssb_active = enable; if (for_all_cpus) { td = curthread; thread_lock(td); is_bound = sched_is_bound(td); bound_cpu = td->td_oncpu; CPU_FOREACH(i) { sched_bind(td, i); hw_ssb_set_one(enable); } if (is_bound) sched_bind(td, bound_cpu); else sched_unbind(td); thread_unlock(td); } else { hw_ssb_set_one(enable); } } void hw_ssb_recalculate(bool all_cpus) { switch (hw_ssb_disable) { default: hw_ssb_disable = 0; /* FALLTHROUGH */ case 0: /* off */ hw_ssb_set(false, all_cpus); break; case 1: /* on */ hw_ssb_set(true, all_cpus); break; case 2: /* auto */ - hw_ssb_set((cpu_ia32_arch_caps & IA32_ARCH_CAP_SSBD_NO) != 0 ? + hw_ssb_set((cpu_ia32_arch_caps & IA32_ARCH_CAP_SSB_NO) != 0 ? false : true, all_cpus); break; } } static int hw_ssb_disable_handler(SYSCTL_HANDLER_ARGS) { int error, val; val = hw_ssb_disable; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); hw_ssb_disable = val; hw_ssb_recalculate(true); return (0); } SYSCTL_PROC(_hw, OID_AUTO, spec_store_bypass_disable, CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE, NULL, 0, hw_ssb_disable_handler, "I", "Speculative Store Bypass Disable (0 - off, 1 - on, 2 - auto"); /* * Enable and restore kernel text write permissions. * Callers must ensure that disable_wp()/restore_wp() are executed * without rescheduling on the same core. */ bool disable_wp(void) { u_int cr0; cr0 = rcr0(); if ((cr0 & CR0_WP) == 0) return (false); load_cr0(cr0 & ~CR0_WP); return (true); } void restore_wp(bool old_wp) { if (old_wp) load_cr0(rcr0() | CR0_WP); } bool acpi_get_fadt_bootflags(uint16_t *flagsp) { #ifdef DEV_ACPI ACPI_TABLE_FADT *fadt; vm_paddr_t physaddr; physaddr = acpi_find_table(ACPI_SIG_FADT); if (physaddr == 0) return (false); fadt = acpi_map_table(physaddr, ACPI_SIG_FADT); if (fadt == NULL) return (false); *flagsp = fadt->BootFlags; acpi_unmap_table(fadt); return (true); #else return (false); #endif } Index: releng/12.0/sys/x86/x86/identcpu.c =================================================================== --- releng/12.0/sys/x86/x86/identcpu.c (revision 340626) +++ releng/12.0/sys/x86/x86/identcpu.c (revision 340627) @@ -1,2503 +1,2506 @@ /*- * Copyright (c) 1992 Terrence R. Lambert. * Copyright (c) 1982, 1987, 1990 The Regents of the University of California. * Copyright (c) 1997 KATO Takenori. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: Id: machdep.c,v 1.193 1996/06/18 01:22:04 bde Exp */ #include __FBSDID("$FreeBSD$"); #include "opt_cpu.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __i386__ #define IDENTBLUE_CYRIX486 0 #define IDENTBLUE_IBMCPU 1 #define IDENTBLUE_CYRIXM2 2 static void identifycyrix(void); static void print_transmeta_info(void); #endif static u_int find_cpu_vendor_id(void); static void print_AMD_info(void); static void print_INTEL_info(void); static void print_INTEL_TLB(u_int data); static void print_hypervisor_info(void); static void print_svm_info(void); static void print_via_padlock_info(void); static void print_vmx_info(void); #ifdef __i386__ int cpu; /* Are we 386, 386sx, 486, etc? */ int cpu_class; #endif u_int cpu_feature; /* Feature flags */ u_int cpu_feature2; /* Feature flags */ u_int amd_feature; /* AMD feature flags */ u_int amd_feature2; /* AMD feature flags */ u_int amd_rascap; /* AMD RAS capabilities */ u_int amd_pminfo; /* AMD advanced power management info */ u_int amd_extended_feature_extensions; u_int via_feature_rng; /* VIA RNG features */ u_int via_feature_xcrypt; /* VIA ACE features */ u_int cpu_high; /* Highest arg to CPUID */ u_int cpu_exthigh; /* Highest arg to extended CPUID */ u_int cpu_id; /* Stepping ID */ u_int cpu_procinfo; /* HyperThreading Info / Brand Index / CLFUSH */ u_int cpu_procinfo2; /* Multicore info */ char cpu_vendor[20]; /* CPU Origin code */ u_int cpu_vendor_id; /* CPU vendor ID */ u_int cpu_fxsr; /* SSE enabled */ u_int cpu_mxcsr_mask; /* Valid bits in mxcsr */ u_int cpu_clflush_line_size = 32; u_int cpu_stdext_feature; /* %ebx */ u_int cpu_stdext_feature2; /* %ecx */ u_int cpu_stdext_feature3; /* %edx */ uint64_t cpu_ia32_arch_caps; u_int cpu_max_ext_state_size; u_int cpu_mon_mwait_flags; /* MONITOR/MWAIT flags (CPUID.05H.ECX) */ u_int cpu_mon_min_size; /* MONITOR minimum range size, bytes */ u_int cpu_mon_max_size; /* MONITOR minimum range size, bytes */ u_int cpu_maxphyaddr; /* Max phys addr width in bits */ char machine[] = MACHINE; SYSCTL_UINT(_hw, OID_AUTO, via_feature_rng, CTLFLAG_RD, &via_feature_rng, 0, "VIA RNG feature available in CPU"); SYSCTL_UINT(_hw, OID_AUTO, via_feature_xcrypt, CTLFLAG_RD, &via_feature_xcrypt, 0, "VIA xcrypt feature available in CPU"); #ifdef __amd64__ #ifdef SCTL_MASK32 extern int adaptive_machine_arch; #endif static int sysctl_hw_machine(SYSCTL_HANDLER_ARGS) { #ifdef SCTL_MASK32 static const char machine32[] = "i386"; #endif int error; #ifdef SCTL_MASK32 if ((req->flags & SCTL_MASK32) != 0 && adaptive_machine_arch) error = SYSCTL_OUT(req, machine32, sizeof(machine32)); else #endif error = SYSCTL_OUT(req, machine, sizeof(machine)); return (error); } SYSCTL_PROC(_hw, HW_MACHINE, machine, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, sysctl_hw_machine, "A", "Machine class"); #else SYSCTL_STRING(_hw, HW_MACHINE, machine, CTLFLAG_RD, machine, 0, "Machine class"); #endif static char cpu_model[128]; SYSCTL_STRING(_hw, HW_MODEL, model, CTLFLAG_RD | CTLFLAG_MPSAFE, cpu_model, 0, "Machine model"); static int hw_clockrate; SYSCTL_INT(_hw, OID_AUTO, clockrate, CTLFLAG_RD, &hw_clockrate, 0, "CPU instruction clock rate"); u_int hv_high; char hv_vendor[16]; SYSCTL_STRING(_hw, OID_AUTO, hv_vendor, CTLFLAG_RD | CTLFLAG_MPSAFE, hv_vendor, 0, "Hypervisor vendor"); static eventhandler_tag tsc_post_tag; static char cpu_brand[48]; #ifdef __i386__ #define MAX_BRAND_INDEX 8 static const char *cpu_brandtable[MAX_BRAND_INDEX + 1] = { NULL, /* No brand */ "Intel Celeron", "Intel Pentium III", "Intel Pentium III Xeon", NULL, NULL, NULL, NULL, "Intel Pentium 4" }; static struct { char *cpu_name; int cpu_class; } cpus[] = { { "Intel 80286", CPUCLASS_286 }, /* CPU_286 */ { "i386SX", CPUCLASS_386 }, /* CPU_386SX */ { "i386DX", CPUCLASS_386 }, /* CPU_386 */ { "i486SX", CPUCLASS_486 }, /* CPU_486SX */ { "i486DX", CPUCLASS_486 }, /* CPU_486 */ { "Pentium", CPUCLASS_586 }, /* CPU_586 */ { "Cyrix 486", CPUCLASS_486 }, /* CPU_486DLC */ { "Pentium Pro", CPUCLASS_686 }, /* CPU_686 */ { "Cyrix 5x86", CPUCLASS_486 }, /* CPU_M1SC */ { "Cyrix 6x86", CPUCLASS_486 }, /* CPU_M1 */ { "Blue Lightning", CPUCLASS_486 }, /* CPU_BLUE */ { "Cyrix 6x86MX", CPUCLASS_686 }, /* CPU_M2 */ { "NexGen 586", CPUCLASS_386 }, /* CPU_NX586 (XXX) */ { "Cyrix 486S/DX", CPUCLASS_486 }, /* CPU_CY486DX */ { "Pentium II", CPUCLASS_686 }, /* CPU_PII */ { "Pentium III", CPUCLASS_686 }, /* CPU_PIII */ { "Pentium 4", CPUCLASS_686 }, /* CPU_P4 */ }; #endif static struct { char *vendor; u_int vendor_id; } cpu_vendors[] = { { INTEL_VENDOR_ID, CPU_VENDOR_INTEL }, /* GenuineIntel */ { AMD_VENDOR_ID, CPU_VENDOR_AMD }, /* AuthenticAMD */ { CENTAUR_VENDOR_ID, CPU_VENDOR_CENTAUR }, /* CentaurHauls */ #ifdef __i386__ { NSC_VENDOR_ID, CPU_VENDOR_NSC }, /* Geode by NSC */ { CYRIX_VENDOR_ID, CPU_VENDOR_CYRIX }, /* CyrixInstead */ { TRANSMETA_VENDOR_ID, CPU_VENDOR_TRANSMETA }, /* GenuineTMx86 */ { SIS_VENDOR_ID, CPU_VENDOR_SIS }, /* SiS SiS SiS */ { UMC_VENDOR_ID, CPU_VENDOR_UMC }, /* UMC UMC UMC */ { NEXGEN_VENDOR_ID, CPU_VENDOR_NEXGEN }, /* NexGenDriven */ { RISE_VENDOR_ID, CPU_VENDOR_RISE }, /* RiseRiseRise */ #if 0 /* XXX CPUID 8000_0000h and 8086_0000h, not 0000_0000h */ { "TransmetaCPU", CPU_VENDOR_TRANSMETA }, #endif #endif }; void printcpuinfo(void) { u_int regs[4], i; char *brand; printf("CPU: "); #ifdef __i386__ cpu_class = cpus[cpu].cpu_class; strncpy(cpu_model, cpus[cpu].cpu_name, sizeof (cpu_model)); #else strncpy(cpu_model, "Hammer", sizeof (cpu_model)); #endif /* Check for extended CPUID information and a processor name. */ if (cpu_exthigh >= 0x80000004) { brand = cpu_brand; for (i = 0x80000002; i < 0x80000005; i++) { do_cpuid(i, regs); memcpy(brand, regs, sizeof(regs)); brand += sizeof(regs); } } switch (cpu_vendor_id) { case CPU_VENDOR_INTEL: #ifdef __i386__ if ((cpu_id & 0xf00) > 0x300) { u_int brand_index; cpu_model[0] = '\0'; switch (cpu_id & 0x3000) { case 0x1000: strcpy(cpu_model, "Overdrive "); break; case 0x2000: strcpy(cpu_model, "Dual "); break; } switch (cpu_id & 0xf00) { case 0x400: strcat(cpu_model, "i486 "); /* Check the particular flavor of 486 */ switch (cpu_id & 0xf0) { case 0x00: case 0x10: strcat(cpu_model, "DX"); break; case 0x20: strcat(cpu_model, "SX"); break; case 0x30: strcat(cpu_model, "DX2"); break; case 0x40: strcat(cpu_model, "SL"); break; case 0x50: strcat(cpu_model, "SX2"); break; case 0x70: strcat(cpu_model, "DX2 Write-Back Enhanced"); break; case 0x80: strcat(cpu_model, "DX4"); break; } break; case 0x500: /* Check the particular flavor of 586 */ strcat(cpu_model, "Pentium"); switch (cpu_id & 0xf0) { case 0x00: strcat(cpu_model, " A-step"); break; case 0x10: strcat(cpu_model, "/P5"); break; case 0x20: strcat(cpu_model, "/P54C"); break; case 0x30: strcat(cpu_model, "/P24T"); break; case 0x40: strcat(cpu_model, "/P55C"); break; case 0x70: strcat(cpu_model, "/P54C"); break; case 0x80: strcat(cpu_model, "/P55C (quarter-micron)"); break; default: /* nothing */ break; } #if defined(I586_CPU) && !defined(NO_F00F_HACK) /* * XXX - If/when Intel fixes the bug, this * should also check the version of the * CPU, not just that it's a Pentium. */ has_f00f_bug = 1; #endif break; case 0x600: /* Check the particular flavor of 686 */ switch (cpu_id & 0xf0) { case 0x00: strcat(cpu_model, "Pentium Pro A-step"); break; case 0x10: strcat(cpu_model, "Pentium Pro"); break; case 0x30: case 0x50: case 0x60: strcat(cpu_model, "Pentium II/Pentium II Xeon/Celeron"); cpu = CPU_PII; break; case 0x70: case 0x80: case 0xa0: case 0xb0: strcat(cpu_model, "Pentium III/Pentium III Xeon/Celeron"); cpu = CPU_PIII; break; default: strcat(cpu_model, "Unknown 80686"); break; } break; case 0xf00: strcat(cpu_model, "Pentium 4"); cpu = CPU_P4; break; default: strcat(cpu_model, "unknown"); break; } /* * If we didn't get a brand name from the extended * CPUID, try to look it up in the brand table. */ if (cpu_high > 0 && *cpu_brand == '\0') { brand_index = cpu_procinfo & CPUID_BRAND_INDEX; if (brand_index <= MAX_BRAND_INDEX && cpu_brandtable[brand_index] != NULL) strcpy(cpu_brand, cpu_brandtable[brand_index]); } } #else /* Please make up your mind folks! */ strcat(cpu_model, "EM64T"); #endif break; case CPU_VENDOR_AMD: /* * Values taken from AMD Processor Recognition * http://www.amd.com/K6/k6docs/pdf/20734g.pdf * (also describes ``Features'' encodings. */ strcpy(cpu_model, "AMD "); #ifdef __i386__ switch (cpu_id & 0xFF0) { case 0x410: strcat(cpu_model, "Standard Am486DX"); break; case 0x430: strcat(cpu_model, "Enhanced Am486DX2 Write-Through"); break; case 0x470: strcat(cpu_model, "Enhanced Am486DX2 Write-Back"); break; case 0x480: strcat(cpu_model, "Enhanced Am486DX4/Am5x86 Write-Through"); break; case 0x490: strcat(cpu_model, "Enhanced Am486DX4/Am5x86 Write-Back"); break; case 0x4E0: strcat(cpu_model, "Am5x86 Write-Through"); break; case 0x4F0: strcat(cpu_model, "Am5x86 Write-Back"); break; case 0x500: strcat(cpu_model, "K5 model 0"); break; case 0x510: strcat(cpu_model, "K5 model 1"); break; case 0x520: strcat(cpu_model, "K5 PR166 (model 2)"); break; case 0x530: strcat(cpu_model, "K5 PR200 (model 3)"); break; case 0x560: strcat(cpu_model, "K6"); break; case 0x570: strcat(cpu_model, "K6 266 (model 1)"); break; case 0x580: strcat(cpu_model, "K6-2"); break; case 0x590: strcat(cpu_model, "K6-III"); break; case 0x5a0: strcat(cpu_model, "Geode LX"); break; default: strcat(cpu_model, "Unknown"); break; } #else if ((cpu_id & 0xf00) == 0xf00) strcat(cpu_model, "AMD64 Processor"); else strcat(cpu_model, "Unknown"); #endif break; #ifdef __i386__ case CPU_VENDOR_CYRIX: strcpy(cpu_model, "Cyrix "); switch (cpu_id & 0xff0) { case 0x440: strcat(cpu_model, "MediaGX"); break; case 0x520: strcat(cpu_model, "6x86"); break; case 0x540: cpu_class = CPUCLASS_586; strcat(cpu_model, "GXm"); break; case 0x600: strcat(cpu_model, "6x86MX"); break; default: /* * Even though CPU supports the cpuid * instruction, it can be disabled. * Therefore, this routine supports all Cyrix * CPUs. */ switch (cyrix_did & 0xf0) { case 0x00: switch (cyrix_did & 0x0f) { case 0x00: strcat(cpu_model, "486SLC"); break; case 0x01: strcat(cpu_model, "486DLC"); break; case 0x02: strcat(cpu_model, "486SLC2"); break; case 0x03: strcat(cpu_model, "486DLC2"); break; case 0x04: strcat(cpu_model, "486SRx"); break; case 0x05: strcat(cpu_model, "486DRx"); break; case 0x06: strcat(cpu_model, "486SRx2"); break; case 0x07: strcat(cpu_model, "486DRx2"); break; case 0x08: strcat(cpu_model, "486SRu"); break; case 0x09: strcat(cpu_model, "486DRu"); break; case 0x0a: strcat(cpu_model, "486SRu2"); break; case 0x0b: strcat(cpu_model, "486DRu2"); break; default: strcat(cpu_model, "Unknown"); break; } break; case 0x10: switch (cyrix_did & 0x0f) { case 0x00: strcat(cpu_model, "486S"); break; case 0x01: strcat(cpu_model, "486S2"); break; case 0x02: strcat(cpu_model, "486Se"); break; case 0x03: strcat(cpu_model, "486S2e"); break; case 0x0a: strcat(cpu_model, "486DX"); break; case 0x0b: strcat(cpu_model, "486DX2"); break; case 0x0f: strcat(cpu_model, "486DX4"); break; default: strcat(cpu_model, "Unknown"); break; } break; case 0x20: if ((cyrix_did & 0x0f) < 8) strcat(cpu_model, "6x86"); /* Where did you get it? */ else strcat(cpu_model, "5x86"); break; case 0x30: strcat(cpu_model, "6x86"); break; case 0x40: if ((cyrix_did & 0xf000) == 0x3000) { cpu_class = CPUCLASS_586; strcat(cpu_model, "GXm"); } else strcat(cpu_model, "MediaGX"); break; case 0x50: strcat(cpu_model, "6x86MX"); break; case 0xf0: switch (cyrix_did & 0x0f) { case 0x0d: strcat(cpu_model, "Overdrive CPU"); break; case 0x0e: strcpy(cpu_model, "Texas Instruments 486SXL"); break; case 0x0f: strcat(cpu_model, "486SLC/DLC"); break; default: strcat(cpu_model, "Unknown"); break; } break; default: strcat(cpu_model, "Unknown"); break; } break; } break; case CPU_VENDOR_RISE: strcpy(cpu_model, "Rise "); switch (cpu_id & 0xff0) { case 0x500: /* 6401 and 6441 (Kirin) */ case 0x520: /* 6510 (Lynx) */ strcat(cpu_model, "mP6"); break; default: strcat(cpu_model, "Unknown"); } break; #endif case CPU_VENDOR_CENTAUR: #ifdef __i386__ switch (cpu_id & 0xff0) { case 0x540: strcpy(cpu_model, "IDT WinChip C6"); break; case 0x580: strcpy(cpu_model, "IDT WinChip 2"); break; case 0x590: strcpy(cpu_model, "IDT WinChip 3"); break; case 0x660: strcpy(cpu_model, "VIA C3 Samuel"); break; case 0x670: if (cpu_id & 0x8) strcpy(cpu_model, "VIA C3 Ezra"); else strcpy(cpu_model, "VIA C3 Samuel 2"); break; case 0x680: strcpy(cpu_model, "VIA C3 Ezra-T"); break; case 0x690: strcpy(cpu_model, "VIA C3 Nehemiah"); break; case 0x6a0: case 0x6d0: strcpy(cpu_model, "VIA C7 Esther"); break; case 0x6f0: strcpy(cpu_model, "VIA Nano"); break; default: strcpy(cpu_model, "VIA/IDT Unknown"); } #else strcpy(cpu_model, "VIA "); if ((cpu_id & 0xff0) == 0x6f0) strcat(cpu_model, "Nano Processor"); else strcat(cpu_model, "Unknown"); #endif break; #ifdef __i386__ case CPU_VENDOR_IBM: strcpy(cpu_model, "Blue Lightning CPU"); break; case CPU_VENDOR_NSC: switch (cpu_id & 0xff0) { case 0x540: strcpy(cpu_model, "Geode SC1100"); cpu = CPU_GEODE1100; break; default: strcpy(cpu_model, "Geode/NSC unknown"); break; } break; #endif default: strcat(cpu_model, "Unknown"); break; } /* * Replace cpu_model with cpu_brand minus leading spaces if * we have one. */ brand = cpu_brand; while (*brand == ' ') ++brand; if (*brand != '\0') strcpy(cpu_model, brand); printf("%s (", cpu_model); if (tsc_freq != 0) { hw_clockrate = (tsc_freq + 5000) / 1000000; printf("%jd.%02d-MHz ", (intmax_t)(tsc_freq + 4999) / 1000000, (u_int)((tsc_freq + 4999) / 10000) % 100); } #ifdef __i386__ switch(cpu_class) { case CPUCLASS_286: printf("286"); break; case CPUCLASS_386: printf("386"); break; #if defined(I486_CPU) case CPUCLASS_486: printf("486"); break; #endif #if defined(I586_CPU) case CPUCLASS_586: printf("586"); break; #endif #if defined(I686_CPU) case CPUCLASS_686: printf("686"); break; #endif default: printf("Unknown"); /* will panic below... */ } #else printf("K8"); #endif printf("-class CPU)\n"); if (*cpu_vendor) printf(" Origin=\"%s\"", cpu_vendor); if (cpu_id) printf(" Id=0x%x", cpu_id); if (cpu_vendor_id == CPU_VENDOR_INTEL || cpu_vendor_id == CPU_VENDOR_AMD || cpu_vendor_id == CPU_VENDOR_CENTAUR || #ifdef __i386__ cpu_vendor_id == CPU_VENDOR_TRANSMETA || cpu_vendor_id == CPU_VENDOR_RISE || cpu_vendor_id == CPU_VENDOR_NSC || (cpu_vendor_id == CPU_VENDOR_CYRIX && ((cpu_id & 0xf00) > 0x500)) || #endif 0) { printf(" Family=0x%x", CPUID_TO_FAMILY(cpu_id)); printf(" Model=0x%x", CPUID_TO_MODEL(cpu_id)); printf(" Stepping=%u", cpu_id & CPUID_STEPPING); #ifdef __i386__ if (cpu_vendor_id == CPU_VENDOR_CYRIX) printf("\n DIR=0x%04x", cyrix_did); #endif /* * AMD CPUID Specification * http://support.amd.com/us/Embedded_TechDocs/25481.pdf * * Intel Processor Identification and CPUID Instruction * http://www.intel.com/assets/pdf/appnote/241618.pdf */ if (cpu_high > 0) { /* * Here we should probably set up flags indicating * whether or not various features are available. * The interesting ones are probably VME, PSE, PAE, * and PGE. The code already assumes without bothering * to check that all CPUs >= Pentium have a TSC and * MSRs. */ printf("\n Features=0x%b", cpu_feature, "\020" "\001FPU" /* Integral FPU */ "\002VME" /* Extended VM86 mode support */ "\003DE" /* Debugging Extensions (CR4.DE) */ "\004PSE" /* 4MByte page tables */ "\005TSC" /* Timestamp counter */ "\006MSR" /* Machine specific registers */ "\007PAE" /* Physical address extension */ "\010MCE" /* Machine Check support */ "\011CX8" /* CMPEXCH8 instruction */ "\012APIC" /* SMP local APIC */ "\013oldMTRR" /* Previous implementation of MTRR */ "\014SEP" /* Fast System Call */ "\015MTRR" /* Memory Type Range Registers */ "\016PGE" /* PG_G (global bit) support */ "\017MCA" /* Machine Check Architecture */ "\020CMOV" /* CMOV instruction */ "\021PAT" /* Page attributes table */ "\022PSE36" /* 36 bit address space support */ "\023PN" /* Processor Serial number */ "\024CLFLUSH" /* Has the CLFLUSH instruction */ "\025" "\026DTS" /* Debug Trace Store */ "\027ACPI" /* ACPI support */ "\030MMX" /* MMX instructions */ "\031FXSR" /* FXSAVE/FXRSTOR */ "\032SSE" /* Streaming SIMD Extensions */ "\033SSE2" /* Streaming SIMD Extensions #2 */ "\034SS" /* Self snoop */ "\035HTT" /* Hyperthreading (see EBX bit 16-23) */ "\036TM" /* Thermal Monitor clock slowdown */ "\037IA64" /* CPU can execute IA64 instructions */ "\040PBE" /* Pending Break Enable */ ); if (cpu_feature2 != 0) { printf("\n Features2=0x%b", cpu_feature2, "\020" "\001SSE3" /* SSE3 */ "\002PCLMULQDQ" /* Carry-Less Mul Quadword */ "\003DTES64" /* 64-bit Debug Trace */ "\004MON" /* MONITOR/MWAIT Instructions */ "\005DS_CPL" /* CPL Qualified Debug Store */ "\006VMX" /* Virtual Machine Extensions */ "\007SMX" /* Safer Mode Extensions */ "\010EST" /* Enhanced SpeedStep */ "\011TM2" /* Thermal Monitor 2 */ "\012SSSE3" /* SSSE3 */ "\013CNXT-ID" /* L1 context ID available */ "\014SDBG" /* IA32 silicon debug */ "\015FMA" /* Fused Multiply Add */ "\016CX16" /* CMPXCHG16B Instruction */ "\017xTPR" /* Send Task Priority Messages*/ "\020PDCM" /* Perf/Debug Capability MSR */ "\021" "\022PCID" /* Process-context Identifiers*/ "\023DCA" /* Direct Cache Access */ "\024SSE4.1" /* SSE 4.1 */ "\025SSE4.2" /* SSE 4.2 */ "\026x2APIC" /* xAPIC Extensions */ "\027MOVBE" /* MOVBE Instruction */ "\030POPCNT" /* POPCNT Instruction */ "\031TSCDLT" /* TSC-Deadline Timer */ "\032AESNI" /* AES Crypto */ "\033XSAVE" /* XSAVE/XRSTOR States */ "\034OSXSAVE" /* OS-Enabled State Management*/ "\035AVX" /* Advanced Vector Extensions */ "\036F16C" /* Half-precision conversions */ "\037RDRAND" /* RDRAND Instruction */ "\040HV" /* Hypervisor */ ); } if (amd_feature != 0) { printf("\n AMD Features=0x%b", amd_feature, "\020" /* in hex */ "\001" /* Same */ "\002" /* Same */ "\003" /* Same */ "\004" /* Same */ "\005" /* Same */ "\006" /* Same */ "\007" /* Same */ "\010" /* Same */ "\011" /* Same */ "\012" /* Same */ "\013" /* Undefined */ "\014SYSCALL" /* Have SYSCALL/SYSRET */ "\015" /* Same */ "\016" /* Same */ "\017" /* Same */ "\020" /* Same */ "\021" /* Same */ "\022" /* Same */ "\023" /* Reserved, unknown */ "\024MP" /* Multiprocessor Capable */ "\025NX" /* Has EFER.NXE, NX */ "\026" /* Undefined */ "\027MMX+" /* AMD MMX Extensions */ "\030" /* Same */ "\031" /* Same */ "\032FFXSR" /* Fast FXSAVE/FXRSTOR */ "\033Page1GB" /* 1-GB large page support */ "\034RDTSCP" /* RDTSCP */ "\035" /* Undefined */ "\036LM" /* 64 bit long mode */ "\0373DNow!+" /* AMD 3DNow! Extensions */ "\0403DNow!" /* AMD 3DNow! */ ); } if (amd_feature2 != 0) { printf("\n AMD Features2=0x%b", amd_feature2, "\020" "\001LAHF" /* LAHF/SAHF in long mode */ "\002CMP" /* CMP legacy */ "\003SVM" /* Secure Virtual Mode */ "\004ExtAPIC" /* Extended APIC register */ "\005CR8" /* CR8 in legacy mode */ "\006ABM" /* LZCNT instruction */ "\007SSE4A" /* SSE4A */ "\010MAS" /* Misaligned SSE mode */ "\011Prefetch" /* 3DNow! Prefetch/PrefetchW */ "\012OSVW" /* OS visible workaround */ "\013IBS" /* Instruction based sampling */ "\014XOP" /* XOP extended instructions */ "\015SKINIT" /* SKINIT/STGI */ "\016WDT" /* Watchdog timer */ "\017" "\020LWP" /* Lightweight Profiling */ "\021FMA4" /* 4-operand FMA instructions */ "\022TCE" /* Translation Cache Extension */ "\023" "\024NodeId" /* NodeId MSR support */ "\025" "\026TBM" /* Trailing Bit Manipulation */ "\027Topology" /* Topology Extensions */ "\030PCXC" /* Core perf count */ "\031PNXC" /* NB perf count */ "\032" "\033DBE" /* Data Breakpoint extension */ "\034PTSC" /* Performance TSC */ "\035PL2I" /* L2I perf count */ "\036MWAITX" /* MONITORX/MWAITX instructions */ "\037" "\040" ); } if (cpu_stdext_feature != 0) { printf("\n Structured Extended Features=0x%b", cpu_stdext_feature, "\020" /* RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE */ "\001FSGSBASE" "\002TSCADJ" "\003SGX" /* Bit Manipulation Instructions */ "\004BMI1" /* Hardware Lock Elision */ "\005HLE" /* Advanced Vector Instructions 2 */ "\006AVX2" /* FDP_EXCPTN_ONLY */ "\007FDPEXC" /* Supervisor Mode Execution Prot. */ "\010SMEP" /* Bit Manipulation Instructions */ "\011BMI2" "\012ERMS" /* Invalidate Processor Context ID */ "\013INVPCID" /* Restricted Transactional Memory */ "\014RTM" "\015PQM" "\016NFPUSG" /* Intel Memory Protection Extensions */ "\017MPX" "\020PQE" /* AVX512 Foundation */ "\021AVX512F" "\022AVX512DQ" /* Enhanced NRBG */ "\023RDSEED" /* ADCX + ADOX */ "\024ADX" /* Supervisor Mode Access Prevention */ "\025SMAP" "\026AVX512IFMA" "\027PCOMMIT" "\030CLFLUSHOPT" "\031CLWB" "\032PROCTRACE" "\033AVX512PF" "\034AVX512ER" "\035AVX512CD" "\036SHA" "\037AVX512BW" "\040AVX512VL" ); } if (cpu_stdext_feature2 != 0) { printf("\n Structured Extended Features2=0x%b", cpu_stdext_feature2, "\020" "\001PREFETCHWT1" "\002AVX512VBMI" "\003UMIP" "\004PKU" "\005OSPKE" "\027RDPID" "\037SGXLC" ); } if (cpu_stdext_feature3 != 0) { printf("\n Structured Extended Features3=0x%b", cpu_stdext_feature3, "\020" "\033IBPB" "\034STIBP" "\035L1DFL" "\036ARCH_CAP" "\040SSBD" ); } if ((cpu_feature2 & CPUID2_XSAVE) != 0) { cpuid_count(0xd, 0x1, regs); if (regs[0] != 0) { printf("\n XSAVE Features=0x%b", regs[0], "\020" "\001XSAVEOPT" "\002XSAVEC" "\003XINUSE" "\004XSAVES"); } } if (cpu_ia32_arch_caps != 0) { printf("\n IA32_ARCH_CAPS=0x%b", (u_int)cpu_ia32_arch_caps, "\020" "\001RDCL_NO" "\002IBRS_ALL" + "\003RSBA" + "\004SKIP_L1DFL_VME" + "\005SSB_NO" ); } if (amd_extended_feature_extensions != 0) { printf("\n " "AMD Extended Feature Extensions ID EBX=" "0x%b", amd_extended_feature_extensions, "\020" "\001CLZERO" "\002IRPerf" "\003XSaveErPtr"); } if (via_feature_rng != 0 || via_feature_xcrypt != 0) print_via_padlock_info(); if (cpu_feature2 & CPUID2_VMX) print_vmx_info(); if (amd_feature2 & AMDID2_SVM) print_svm_info(); if ((cpu_feature & CPUID_HTT) && cpu_vendor_id == CPU_VENDOR_AMD) cpu_feature &= ~CPUID_HTT; /* * If this CPU supports P-state invariant TSC then * mention the capability. */ if (tsc_is_invariant) { printf("\n TSC: P-state invariant"); if (tsc_perf_stat) printf(", performance statistics"); } } #ifdef __i386__ } else if (cpu_vendor_id == CPU_VENDOR_CYRIX) { printf(" DIR=0x%04x", cyrix_did); printf(" Stepping=%u", (cyrix_did & 0xf000) >> 12); printf(" Revision=%u", (cyrix_did & 0x0f00) >> 8); #ifndef CYRIX_CACHE_REALLY_WORKS if (cpu == CPU_M1 && (cyrix_did & 0xff00) < 0x1700) printf("\n CPU cache: write-through mode"); #endif #endif } /* Avoid ugly blank lines: only print newline when we have to. */ if (*cpu_vendor || cpu_id) printf("\n"); if (bootverbose) { if (cpu_vendor_id == CPU_VENDOR_AMD) print_AMD_info(); else if (cpu_vendor_id == CPU_VENDOR_INTEL) print_INTEL_info(); #ifdef __i386__ else if (cpu_vendor_id == CPU_VENDOR_TRANSMETA) print_transmeta_info(); #endif } print_hypervisor_info(); } #ifdef __i386__ void panicifcpuunsupported(void) { #if !defined(lint) #if !defined(I486_CPU) && !defined(I586_CPU) && !defined(I686_CPU) #error This kernel is not configured for one of the supported CPUs #endif #else /* lint */ #endif /* lint */ /* * Now that we have told the user what they have, * let them know if that machine type isn't configured. */ switch (cpu_class) { case CPUCLASS_286: /* a 286 should not make it this far, anyway */ case CPUCLASS_386: #if !defined(I486_CPU) case CPUCLASS_486: #endif #if !defined(I586_CPU) case CPUCLASS_586: #endif #if !defined(I686_CPU) case CPUCLASS_686: #endif panic("CPU class not configured"); default: break; } } static volatile u_int trap_by_rdmsr; /* * Special exception 6 handler. * The rdmsr instruction generates invalid opcodes fault on 486-class * Cyrix CPU. Stacked eip register points the rdmsr instruction in the * function identblue() when this handler is called. Stacked eip should * be advanced. */ inthand_t bluetrap6; #ifdef __GNUCLIKE_ASM __asm (" \n\ .text \n\ .p2align 2,0x90 \n\ .type " __XSTRING(CNAME(bluetrap6)) ",@function \n\ " __XSTRING(CNAME(bluetrap6)) ": \n\ ss \n\ movl $0xa8c1d," __XSTRING(CNAME(trap_by_rdmsr)) " \n\ addl $2, (%esp) /* rdmsr is a 2-byte instruction */ \n\ iret \n\ "); #endif /* * Special exception 13 handler. * Accessing non-existent MSR generates general protection fault. */ inthand_t bluetrap13; #ifdef __GNUCLIKE_ASM __asm (" \n\ .text \n\ .p2align 2,0x90 \n\ .type " __XSTRING(CNAME(bluetrap13)) ",@function \n\ " __XSTRING(CNAME(bluetrap13)) ": \n\ ss \n\ movl $0xa89c4," __XSTRING(CNAME(trap_by_rdmsr)) " \n\ popl %eax /* discard error code */ \n\ addl $2, (%esp) /* rdmsr is a 2-byte instruction */ \n\ iret \n\ "); #endif /* * Distinguish IBM Blue Lightning CPU from Cyrix CPUs that does not * support cpuid instruction. This function should be called after * loading interrupt descriptor table register. * * I don't like this method that handles fault, but I couldn't get * information for any other methods. Does blue giant know? */ static int identblue(void) { trap_by_rdmsr = 0; /* * Cyrix 486-class CPU does not support rdmsr instruction. * The rdmsr instruction generates invalid opcode fault, and exception * will be trapped by bluetrap6() on Cyrix 486-class CPU. The * bluetrap6() set the magic number to trap_by_rdmsr. */ setidt(IDT_UD, bluetrap6, SDT_SYS386TGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); /* * Certain BIOS disables cpuid instruction of Cyrix 6x86MX CPU. * In this case, rdmsr generates general protection fault, and * exception will be trapped by bluetrap13(). */ setidt(IDT_GP, bluetrap13, SDT_SYS386TGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL)); rdmsr(0x1002); /* Cyrix CPU generates fault. */ if (trap_by_rdmsr == 0xa8c1d) return IDENTBLUE_CYRIX486; else if (trap_by_rdmsr == 0xa89c4) return IDENTBLUE_CYRIXM2; return IDENTBLUE_IBMCPU; } /* * identifycyrix() set lower 16 bits of cyrix_did as follows: * * F E D C B A 9 8 7 6 5 4 3 2 1 0 * +-------+-------+---------------+ * | SID | RID | Device ID | * | (DIR 1) | (DIR 0) | * +-------+-------+---------------+ */ static void identifycyrix(void) { register_t saveintr; int ccr2_test = 0, dir_test = 0; u_char ccr2, ccr3; saveintr = intr_disable(); ccr2 = read_cyrix_reg(CCR2); write_cyrix_reg(CCR2, ccr2 ^ CCR2_LOCK_NW); read_cyrix_reg(CCR2); if (read_cyrix_reg(CCR2) != ccr2) ccr2_test = 1; write_cyrix_reg(CCR2, ccr2); ccr3 = read_cyrix_reg(CCR3); write_cyrix_reg(CCR3, ccr3 ^ CCR3_MAPEN3); read_cyrix_reg(CCR3); if (read_cyrix_reg(CCR3) != ccr3) dir_test = 1; /* CPU supports DIRs. */ write_cyrix_reg(CCR3, ccr3); if (dir_test) { /* Device ID registers are available. */ cyrix_did = read_cyrix_reg(DIR1) << 8; cyrix_did += read_cyrix_reg(DIR0); } else if (ccr2_test) cyrix_did = 0x0010; /* 486S A-step */ else cyrix_did = 0x00ff; /* Old 486SLC/DLC and TI486SXLC/SXL */ intr_restore(saveintr); } #endif /* Update TSC freq with the value indicated by the caller. */ static void tsc_freq_changed(void *arg __unused, const struct cf_level *level, int status) { /* If there was an error during the transition, don't do anything. */ if (status != 0) return; /* Total setting for this level gives the new frequency in MHz. */ hw_clockrate = level->total_set.freq; } static void hook_tsc_freq(void *arg __unused) { if (tsc_is_invariant) return; tsc_post_tag = EVENTHANDLER_REGISTER(cpufreq_post_change, tsc_freq_changed, NULL, EVENTHANDLER_PRI_ANY); } SYSINIT(hook_tsc_freq, SI_SUB_CONFIGURE, SI_ORDER_ANY, hook_tsc_freq, NULL); static const char *const vm_bnames[] = { "QEMU", /* QEMU */ "Plex86", /* Plex86 */ "Bochs", /* Bochs */ "Xen", /* Xen */ "BHYVE", /* bhyve */ "Seabios", /* KVM */ NULL }; static const char *const vm_pnames[] = { "VMware Virtual Platform", /* VMWare VM */ "Virtual Machine", /* Microsoft VirtualPC */ "VirtualBox", /* Sun xVM VirtualBox */ "Parallels Virtual Platform", /* Parallels VM */ "KVM", /* KVM */ NULL }; void identify_hypervisor(void) { u_int regs[4]; char *p; int i; /* * [RFC] CPUID usage for interaction between Hypervisors and Linux. * http://lkml.org/lkml/2008/10/1/246 * * KB1009458: Mechanisms to determine if software is running in * a VMware virtual machine * http://kb.vmware.com/kb/1009458 */ if (cpu_feature2 & CPUID2_HV) { vm_guest = VM_GUEST_VM; do_cpuid(0x40000000, regs); /* * KVM from Linux kernels prior to commit * 57c22e5f35aa4b9b2fe11f73f3e62bbf9ef36190 set %eax * to 0 rather than a valid hv_high value. Check for * the KVM signature bytes and fixup %eax to the * highest supported leaf in that case. */ if (regs[0] == 0 && regs[1] == 0x4b4d564b && regs[2] == 0x564b4d56 && regs[3] == 0x0000004d) regs[0] = 0x40000001; if (regs[0] >= 0x40000000) { hv_high = regs[0]; ((u_int *)&hv_vendor)[0] = regs[1]; ((u_int *)&hv_vendor)[1] = regs[2]; ((u_int *)&hv_vendor)[2] = regs[3]; hv_vendor[12] = '\0'; if (strcmp(hv_vendor, "VMwareVMware") == 0) vm_guest = VM_GUEST_VMWARE; else if (strcmp(hv_vendor, "Microsoft Hv") == 0) vm_guest = VM_GUEST_HV; else if (strcmp(hv_vendor, "KVMKVMKVM") == 0) vm_guest = VM_GUEST_KVM; else if (strcmp(hv_vendor, "bhyve bhyve") == 0) vm_guest = VM_GUEST_BHYVE; } return; } /* * Examine SMBIOS strings for older hypervisors. */ p = kern_getenv("smbios.system.serial"); if (p != NULL) { if (strncmp(p, "VMware-", 7) == 0 || strncmp(p, "VMW", 3) == 0) { vmware_hvcall(VMW_HVCMD_GETVERSION, regs); if (regs[1] == VMW_HVMAGIC) { vm_guest = VM_GUEST_VMWARE; freeenv(p); return; } } freeenv(p); } /* * XXX: Some of these entries may not be needed since they were * added to FreeBSD before the checks above. */ p = kern_getenv("smbios.bios.vendor"); if (p != NULL) { for (i = 0; vm_bnames[i] != NULL; i++) if (strcmp(p, vm_bnames[i]) == 0) { vm_guest = VM_GUEST_VM; freeenv(p); return; } freeenv(p); } p = kern_getenv("smbios.system.product"); if (p != NULL) { for (i = 0; vm_pnames[i] != NULL; i++) if (strcmp(p, vm_pnames[i]) == 0) { vm_guest = VM_GUEST_VM; freeenv(p); return; } freeenv(p); } } bool fix_cpuid(void) { uint64_t msr; /* * Clear "Limit CPUID Maxval" bit and return true if the caller should * get the largest standard CPUID function number again if it is set * from BIOS. It is necessary for probing correct CPU topology later * and for the correct operation of the AVX-aware userspace. */ if (cpu_vendor_id == CPU_VENDOR_INTEL && ((CPUID_TO_FAMILY(cpu_id) == 0xf && CPUID_TO_MODEL(cpu_id) >= 0x3) || (CPUID_TO_FAMILY(cpu_id) == 0x6 && CPUID_TO_MODEL(cpu_id) >= 0xe))) { msr = rdmsr(MSR_IA32_MISC_ENABLE); if ((msr & IA32_MISC_EN_LIMCPUID) != 0) { msr &= ~IA32_MISC_EN_LIMCPUID; wrmsr(MSR_IA32_MISC_ENABLE, msr); return (true); } } /* * Re-enable AMD Topology Extension that could be disabled by BIOS * on some notebook processors. Without the extension it's really * hard to determine the correct CPU cache topology. * See BIOS and Kernel Developer’s Guide (BKDG) for AMD Family 15h * Models 60h-6Fh Processors, Publication # 50742. */ if (vm_guest == VM_GUEST_NO && cpu_vendor_id == CPU_VENDOR_AMD && CPUID_TO_FAMILY(cpu_id) == 0x15) { msr = rdmsr(MSR_EXTFEATURES); if ((msr & ((uint64_t)1 << 54)) == 0) { msr |= (uint64_t)1 << 54; wrmsr(MSR_EXTFEATURES, msr); return (true); } } return (false); } void identify_cpu1(void) { u_int regs[4]; do_cpuid(0, regs); cpu_high = regs[0]; ((u_int *)&cpu_vendor)[0] = regs[1]; ((u_int *)&cpu_vendor)[1] = regs[3]; ((u_int *)&cpu_vendor)[2] = regs[2]; cpu_vendor[12] = '\0'; do_cpuid(1, regs); cpu_id = regs[0]; cpu_procinfo = regs[1]; cpu_feature = regs[3]; cpu_feature2 = regs[2]; } void identify_cpu2(void) { u_int regs[4], cpu_stdext_disable; if (cpu_high >= 7) { cpuid_count(7, 0, regs); cpu_stdext_feature = regs[1]; /* * Some hypervisors failed to filter out unsupported * extended features. Allow to disable the * extensions, activation of which requires setting a * bit in CR4, and which VM monitors do not support. */ cpu_stdext_disable = 0; TUNABLE_INT_FETCH("hw.cpu_stdext_disable", &cpu_stdext_disable); cpu_stdext_feature &= ~cpu_stdext_disable; cpu_stdext_feature2 = regs[2]; cpu_stdext_feature3 = regs[3]; if ((cpu_stdext_feature3 & CPUID_STDEXT3_ARCH_CAP) != 0) cpu_ia32_arch_caps = rdmsr(MSR_IA32_ARCH_CAP); } } /* * Final stage of CPU identification. */ void finishidentcpu(void) { u_int regs[4]; #ifdef __i386__ u_char ccr3; #endif cpu_vendor_id = find_cpu_vendor_id(); if (fix_cpuid()) { do_cpuid(0, regs); cpu_high = regs[0]; } if (cpu_high >= 5 && (cpu_feature2 & CPUID2_MON) != 0) { do_cpuid(5, regs); cpu_mon_mwait_flags = regs[2]; cpu_mon_min_size = regs[0] & CPUID5_MON_MIN_SIZE; cpu_mon_max_size = regs[1] & CPUID5_MON_MAX_SIZE; } identify_cpu2(); #ifdef __i386__ if (cpu_high > 0 && (cpu_vendor_id == CPU_VENDOR_INTEL || cpu_vendor_id == CPU_VENDOR_AMD || cpu_vendor_id == CPU_VENDOR_TRANSMETA || cpu_vendor_id == CPU_VENDOR_CENTAUR || cpu_vendor_id == CPU_VENDOR_NSC)) { do_cpuid(0x80000000, regs); if (regs[0] >= 0x80000000) cpu_exthigh = regs[0]; } #else if (cpu_vendor_id == CPU_VENDOR_INTEL || cpu_vendor_id == CPU_VENDOR_AMD || cpu_vendor_id == CPU_VENDOR_CENTAUR) { do_cpuid(0x80000000, regs); cpu_exthigh = regs[0]; } #endif if (cpu_exthigh >= 0x80000001) { do_cpuid(0x80000001, regs); amd_feature = regs[3] & ~(cpu_feature & 0x0183f3ff); amd_feature2 = regs[2]; } if (cpu_exthigh >= 0x80000007) { do_cpuid(0x80000007, regs); amd_rascap = regs[1]; amd_pminfo = regs[3]; } if (cpu_exthigh >= 0x80000008) { do_cpuid(0x80000008, regs); cpu_maxphyaddr = regs[0] & 0xff; amd_extended_feature_extensions = regs[1]; cpu_procinfo2 = regs[2]; } else { cpu_maxphyaddr = (cpu_feature & CPUID_PAE) != 0 ? 36 : 32; } #ifdef __i386__ if (cpu_vendor_id == CPU_VENDOR_CYRIX) { if (cpu == CPU_486) { /* * These conditions are equivalent to: * - CPU does not support cpuid instruction. * - Cyrix/IBM CPU is detected. */ if (identblue() == IDENTBLUE_IBMCPU) { strcpy(cpu_vendor, "IBM"); cpu_vendor_id = CPU_VENDOR_IBM; cpu = CPU_BLUE; return; } } switch (cpu_id & 0xf00) { case 0x600: /* * Cyrix's datasheet does not describe DIRs. * Therefor, I assume it does not have them * and use the result of the cpuid instruction. * XXX they seem to have it for now at least. -Peter */ identifycyrix(); cpu = CPU_M2; break; default: identifycyrix(); /* * This routine contains a trick. * Don't check (cpu_id & 0x00f0) == 0x50 to detect M2, now. */ switch (cyrix_did & 0x00f0) { case 0x00: case 0xf0: cpu = CPU_486DLC; break; case 0x10: cpu = CPU_CY486DX; break; case 0x20: if ((cyrix_did & 0x000f) < 8) cpu = CPU_M1; else cpu = CPU_M1SC; break; case 0x30: cpu = CPU_M1; break; case 0x40: /* MediaGX CPU */ cpu = CPU_M1SC; break; default: /* M2 and later CPUs are treated as M2. */ cpu = CPU_M2; /* * enable cpuid instruction. */ ccr3 = read_cyrix_reg(CCR3); write_cyrix_reg(CCR3, CCR3_MAPEN0); write_cyrix_reg(CCR4, read_cyrix_reg(CCR4) | CCR4_CPUID); write_cyrix_reg(CCR3, ccr3); do_cpuid(0, regs); cpu_high = regs[0]; /* eax */ do_cpuid(1, regs); cpu_id = regs[0]; /* eax */ cpu_feature = regs[3]; /* edx */ break; } } } else if (cpu == CPU_486 && *cpu_vendor == '\0') { /* * There are BlueLightning CPUs that do not change * undefined flags by dividing 5 by 2. In this case, * the CPU identification routine in locore.s leaves * cpu_vendor null string and puts CPU_486 into the * cpu. */ if (identblue() == IDENTBLUE_IBMCPU) { strcpy(cpu_vendor, "IBM"); cpu_vendor_id = CPU_VENDOR_IBM; cpu = CPU_BLUE; return; } } #endif } int pti_get_default(void) { if (strcmp(cpu_vendor, AMD_VENDOR_ID) == 0) return (0); if ((cpu_ia32_arch_caps & IA32_ARCH_CAP_RDCL_NO) != 0) return (0); return (1); } static u_int find_cpu_vendor_id(void) { int i; for (i = 0; i < nitems(cpu_vendors); i++) if (strcmp(cpu_vendor, cpu_vendors[i].vendor) == 0) return (cpu_vendors[i].vendor_id); return (0); } static void print_AMD_assoc(int i) { if (i == 255) printf(", fully associative\n"); else printf(", %d-way associative\n", i); } static void print_AMD_l2_assoc(int i) { switch (i & 0x0f) { case 0: printf(", disabled/not present\n"); break; case 1: printf(", direct mapped\n"); break; case 2: printf(", 2-way associative\n"); break; case 4: printf(", 4-way associative\n"); break; case 6: printf(", 8-way associative\n"); break; case 8: printf(", 16-way associative\n"); break; case 15: printf(", fully associative\n"); break; default: printf(", reserved configuration\n"); break; } } static void print_AMD_info(void) { #ifdef __i386__ uint64_t amd_whcr; #endif u_int regs[4]; if (cpu_exthigh >= 0x80000005) { do_cpuid(0x80000005, regs); printf("L1 2MB data TLB: %d entries", (regs[0] >> 16) & 0xff); print_AMD_assoc(regs[0] >> 24); printf("L1 2MB instruction TLB: %d entries", regs[0] & 0xff); print_AMD_assoc((regs[0] >> 8) & 0xff); printf("L1 4KB data TLB: %d entries", (regs[1] >> 16) & 0xff); print_AMD_assoc(regs[1] >> 24); printf("L1 4KB instruction TLB: %d entries", regs[1] & 0xff); print_AMD_assoc((regs[1] >> 8) & 0xff); printf("L1 data cache: %d kbytes", regs[2] >> 24); printf(", %d bytes/line", regs[2] & 0xff); printf(", %d lines/tag", (regs[2] >> 8) & 0xff); print_AMD_assoc((regs[2] >> 16) & 0xff); printf("L1 instruction cache: %d kbytes", regs[3] >> 24); printf(", %d bytes/line", regs[3] & 0xff); printf(", %d lines/tag", (regs[3] >> 8) & 0xff); print_AMD_assoc((regs[3] >> 16) & 0xff); } if (cpu_exthigh >= 0x80000006) { do_cpuid(0x80000006, regs); if ((regs[0] >> 16) != 0) { printf("L2 2MB data TLB: %d entries", (regs[0] >> 16) & 0xfff); print_AMD_l2_assoc(regs[0] >> 28); printf("L2 2MB instruction TLB: %d entries", regs[0] & 0xfff); print_AMD_l2_assoc((regs[0] >> 28) & 0xf); } else { printf("L2 2MB unified TLB: %d entries", regs[0] & 0xfff); print_AMD_l2_assoc((regs[0] >> 28) & 0xf); } if ((regs[1] >> 16) != 0) { printf("L2 4KB data TLB: %d entries", (regs[1] >> 16) & 0xfff); print_AMD_l2_assoc(regs[1] >> 28); printf("L2 4KB instruction TLB: %d entries", (regs[1] >> 16) & 0xfff); print_AMD_l2_assoc((regs[1] >> 28) & 0xf); } else { printf("L2 4KB unified TLB: %d entries", (regs[1] >> 16) & 0xfff); print_AMD_l2_assoc((regs[1] >> 28) & 0xf); } printf("L2 unified cache: %d kbytes", regs[2] >> 16); printf(", %d bytes/line", regs[2] & 0xff); printf(", %d lines/tag", (regs[2] >> 8) & 0x0f); print_AMD_l2_assoc((regs[2] >> 12) & 0x0f); } #ifdef __i386__ if (((cpu_id & 0xf00) == 0x500) && (((cpu_id & 0x0f0) > 0x80) || (((cpu_id & 0x0f0) == 0x80) && (cpu_id & 0x00f) > 0x07))) { /* K6-2(new core [Stepping 8-F]), K6-III or later */ amd_whcr = rdmsr(0xc0000082); if (!(amd_whcr & (0x3ff << 22))) { printf("Write Allocate Disable\n"); } else { printf("Write Allocate Enable Limit: %dM bytes\n", (u_int32_t)((amd_whcr & (0x3ff << 22)) >> 22) * 4); printf("Write Allocate 15-16M bytes: %s\n", (amd_whcr & (1 << 16)) ? "Enable" : "Disable"); } } else if (((cpu_id & 0xf00) == 0x500) && ((cpu_id & 0x0f0) > 0x50)) { /* K6, K6-2(old core) */ amd_whcr = rdmsr(0xc0000082); if (!(amd_whcr & (0x7f << 1))) { printf("Write Allocate Disable\n"); } else { printf("Write Allocate Enable Limit: %dM bytes\n", (u_int32_t)((amd_whcr & (0x7f << 1)) >> 1) * 4); printf("Write Allocate 15-16M bytes: %s\n", (amd_whcr & 0x0001) ? "Enable" : "Disable"); printf("Hardware Write Allocate Control: %s\n", (amd_whcr & 0x0100) ? "Enable" : "Disable"); } } #endif /* * Opteron Rev E shows a bug as in very rare occasions a read memory * barrier is not performed as expected if it is followed by a * non-atomic read-modify-write instruction. * As long as that bug pops up very rarely (intensive machine usage * on other operating systems generally generates one unexplainable * crash any 2 months) and as long as a model specific fix would be * impractical at this stage, print out a warning string if the broken * model and family are identified. */ if (CPUID_TO_FAMILY(cpu_id) == 0xf && CPUID_TO_MODEL(cpu_id) >= 0x20 && CPUID_TO_MODEL(cpu_id) <= 0x3f) printf("WARNING: This architecture revision has known SMP " "hardware bugs which may cause random instability\n"); } static void print_INTEL_info(void) { u_int regs[4]; u_int rounds, regnum; u_int nwaycode, nway; if (cpu_high >= 2) { rounds = 0; do { do_cpuid(0x2, regs); if (rounds == 0 && (rounds = (regs[0] & 0xff)) == 0) break; /* we have a buggy CPU */ for (regnum = 0; regnum <= 3; ++regnum) { if (regs[regnum] & (1<<31)) continue; if (regnum != 0) print_INTEL_TLB(regs[regnum] & 0xff); print_INTEL_TLB((regs[regnum] >> 8) & 0xff); print_INTEL_TLB((regs[regnum] >> 16) & 0xff); print_INTEL_TLB((regs[regnum] >> 24) & 0xff); } } while (--rounds > 0); } if (cpu_exthigh >= 0x80000006) { do_cpuid(0x80000006, regs); nwaycode = (regs[2] >> 12) & 0x0f; if (nwaycode >= 0x02 && nwaycode <= 0x08) nway = 1 << (nwaycode / 2); else nway = 0; printf("L2 cache: %u kbytes, %u-way associative, %u bytes/line\n", (regs[2] >> 16) & 0xffff, nway, regs[2] & 0xff); } } static void print_INTEL_TLB(u_int data) { switch (data) { case 0x0: case 0x40: default: break; case 0x1: printf("Instruction TLB: 4 KB pages, 4-way set associative, 32 entries\n"); break; case 0x2: printf("Instruction TLB: 4 MB pages, fully associative, 2 entries\n"); break; case 0x3: printf("Data TLB: 4 KB pages, 4-way set associative, 64 entries\n"); break; case 0x4: printf("Data TLB: 4 MB Pages, 4-way set associative, 8 entries\n"); break; case 0x6: printf("1st-level instruction cache: 8 KB, 4-way set associative, 32 byte line size\n"); break; case 0x8: printf("1st-level instruction cache: 16 KB, 4-way set associative, 32 byte line size\n"); break; case 0x9: printf("1st-level instruction cache: 32 KB, 4-way set associative, 64 byte line size\n"); break; case 0xa: printf("1st-level data cache: 8 KB, 2-way set associative, 32 byte line size\n"); break; case 0xb: printf("Instruction TLB: 4 MByte pages, 4-way set associative, 4 entries\n"); break; case 0xc: printf("1st-level data cache: 16 KB, 4-way set associative, 32 byte line size\n"); break; case 0xd: printf("1st-level data cache: 16 KBytes, 4-way set associative, 64 byte line size"); break; case 0xe: printf("1st-level data cache: 24 KBytes, 6-way set associative, 64 byte line size\n"); break; case 0x1d: printf("2nd-level cache: 128 KBytes, 2-way set associative, 64 byte line size\n"); break; case 0x21: printf("2nd-level cache: 256 KBytes, 8-way set associative, 64 byte line size\n"); break; case 0x22: printf("3rd-level cache: 512 KB, 4-way set associative, sectored cache, 64 byte line size\n"); break; case 0x23: printf("3rd-level cache: 1 MB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x24: printf("2nd-level cache: 1 MBytes, 16-way set associative, 64 byte line size\n"); break; case 0x25: printf("3rd-level cache: 2 MB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x29: printf("3rd-level cache: 4 MB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x2c: printf("1st-level data cache: 32 KB, 8-way set associative, 64 byte line size\n"); break; case 0x30: printf("1st-level instruction cache: 32 KB, 8-way set associative, 64 byte line size\n"); break; case 0x39: /* De-listed in SDM rev. 54 */ printf("2nd-level cache: 128 KB, 4-way set associative, sectored cache, 64 byte line size\n"); break; case 0x3b: /* De-listed in SDM rev. 54 */ printf("2nd-level cache: 128 KB, 2-way set associative, sectored cache, 64 byte line size\n"); break; case 0x3c: /* De-listed in SDM rev. 54 */ printf("2nd-level cache: 256 KB, 4-way set associative, sectored cache, 64 byte line size\n"); break; case 0x41: printf("2nd-level cache: 128 KB, 4-way set associative, 32 byte line size\n"); break; case 0x42: printf("2nd-level cache: 256 KB, 4-way set associative, 32 byte line size\n"); break; case 0x43: printf("2nd-level cache: 512 KB, 4-way set associative, 32 byte line size\n"); break; case 0x44: printf("2nd-level cache: 1 MB, 4-way set associative, 32 byte line size\n"); break; case 0x45: printf("2nd-level cache: 2 MB, 4-way set associative, 32 byte line size\n"); break; case 0x46: printf("3rd-level cache: 4 MB, 4-way set associative, 64 byte line size\n"); break; case 0x47: printf("3rd-level cache: 8 MB, 8-way set associative, 64 byte line size\n"); break; case 0x48: printf("2nd-level cache: 3MByte, 12-way set associative, 64 byte line size\n"); break; case 0x49: if (CPUID_TO_FAMILY(cpu_id) == 0xf && CPUID_TO_MODEL(cpu_id) == 0x6) printf("3rd-level cache: 4MB, 16-way set associative, 64-byte line size\n"); else printf("2nd-level cache: 4 MByte, 16-way set associative, 64 byte line size"); break; case 0x4a: printf("3rd-level cache: 6MByte, 12-way set associative, 64 byte line size\n"); break; case 0x4b: printf("3rd-level cache: 8MByte, 16-way set associative, 64 byte line size\n"); break; case 0x4c: printf("3rd-level cache: 12MByte, 12-way set associative, 64 byte line size\n"); break; case 0x4d: printf("3rd-level cache: 16MByte, 16-way set associative, 64 byte line size\n"); break; case 0x4e: printf("2nd-level cache: 6MByte, 24-way set associative, 64 byte line size\n"); break; case 0x4f: printf("Instruction TLB: 4 KByte pages, 32 entries\n"); break; case 0x50: printf("Instruction TLB: 4 KB, 2 MB or 4 MB pages, fully associative, 64 entries\n"); break; case 0x51: printf("Instruction TLB: 4 KB, 2 MB or 4 MB pages, fully associative, 128 entries\n"); break; case 0x52: printf("Instruction TLB: 4 KB, 2 MB or 4 MB pages, fully associative, 256 entries\n"); break; case 0x55: printf("Instruction TLB: 2-MByte or 4-MByte pages, fully associative, 7 entries\n"); break; case 0x56: printf("Data TLB0: 4 MByte pages, 4-way set associative, 16 entries\n"); break; case 0x57: printf("Data TLB0: 4 KByte pages, 4-way associative, 16 entries\n"); break; case 0x59: printf("Data TLB0: 4 KByte pages, fully associative, 16 entries\n"); break; case 0x5a: printf("Data TLB0: 2-MByte or 4 MByte pages, 4-way set associative, 32 entries\n"); break; case 0x5b: printf("Data TLB: 4 KB or 4 MB pages, fully associative, 64 entries\n"); break; case 0x5c: printf("Data TLB: 4 KB or 4 MB pages, fully associative, 128 entries\n"); break; case 0x5d: printf("Data TLB: 4 KB or 4 MB pages, fully associative, 256 entries\n"); break; case 0x60: printf("1st-level data cache: 16 KB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x61: printf("Instruction TLB: 4 KByte pages, fully associative, 48 entries\n"); break; case 0x63: printf("Data TLB: 2 MByte or 4 MByte pages, 4-way set associative, 32 entries and a separate array with 1 GByte pages, 4-way set associative, 4 entries\n"); break; case 0x64: printf("Data TLB: 4 KBytes pages, 4-way set associative, 512 entries\n"); break; case 0x66: printf("1st-level data cache: 8 KB, 4-way set associative, sectored cache, 64 byte line size\n"); break; case 0x67: printf("1st-level data cache: 16 KB, 4-way set associative, sectored cache, 64 byte line size\n"); break; case 0x68: printf("1st-level data cache: 32 KB, 4 way set associative, sectored cache, 64 byte line size\n"); break; case 0x6a: printf("uTLB: 4KByte pages, 8-way set associative, 64 entries\n"); break; case 0x6b: printf("DTLB: 4KByte pages, 8-way set associative, 256 entries\n"); break; case 0x6c: printf("DTLB: 2M/4M pages, 8-way set associative, 128 entries\n"); break; case 0x6d: printf("DTLB: 1 GByte pages, fully associative, 16 entries\n"); break; case 0x70: printf("Trace cache: 12K-uops, 8-way set associative\n"); break; case 0x71: printf("Trace cache: 16K-uops, 8-way set associative\n"); break; case 0x72: printf("Trace cache: 32K-uops, 8-way set associative\n"); break; case 0x76: printf("Instruction TLB: 2M/4M pages, fully associative, 8 entries\n"); break; case 0x78: printf("2nd-level cache: 1 MB, 4-way set associative, 64-byte line size\n"); break; case 0x79: printf("2nd-level cache: 128 KB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x7a: printf("2nd-level cache: 256 KB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x7b: printf("2nd-level cache: 512 KB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x7c: printf("2nd-level cache: 1 MB, 8-way set associative, sectored cache, 64 byte line size\n"); break; case 0x7d: printf("2nd-level cache: 2-MB, 8-way set associative, 64-byte line size\n"); break; case 0x7f: printf("2nd-level cache: 512-KB, 2-way set associative, 64-byte line size\n"); break; case 0x80: printf("2nd-level cache: 512 KByte, 8-way set associative, 64-byte line size\n"); break; case 0x82: printf("2nd-level cache: 256 KB, 8-way set associative, 32 byte line size\n"); break; case 0x83: printf("2nd-level cache: 512 KB, 8-way set associative, 32 byte line size\n"); break; case 0x84: printf("2nd-level cache: 1 MB, 8-way set associative, 32 byte line size\n"); break; case 0x85: printf("2nd-level cache: 2 MB, 8-way set associative, 32 byte line size\n"); break; case 0x86: printf("2nd-level cache: 512 KB, 4-way set associative, 64 byte line size\n"); break; case 0x87: printf("2nd-level cache: 1 MB, 8-way set associative, 64 byte line size\n"); break; case 0xa0: printf("DTLB: 4k pages, fully associative, 32 entries\n"); break; case 0xb0: printf("Instruction TLB: 4 KB Pages, 4-way set associative, 128 entries\n"); break; case 0xb1: printf("Instruction TLB: 2M pages, 4-way, 8 entries or 4M pages, 4-way, 4 entries\n"); break; case 0xb2: printf("Instruction TLB: 4KByte pages, 4-way set associative, 64 entries\n"); break; case 0xb3: printf("Data TLB: 4 KB Pages, 4-way set associative, 128 entries\n"); break; case 0xb4: printf("Data TLB1: 4 KByte pages, 4-way associative, 256 entries\n"); break; case 0xb5: printf("Instruction TLB: 4KByte pages, 8-way set associative, 64 entries\n"); break; case 0xb6: printf("Instruction TLB: 4KByte pages, 8-way set associative, 128 entries\n"); break; case 0xba: printf("Data TLB1: 4 KByte pages, 4-way associative, 64 entries\n"); break; case 0xc0: printf("Data TLB: 4 KByte and 4 MByte pages, 4-way associative, 8 entries\n"); break; case 0xc1: printf("Shared 2nd-Level TLB: 4 KByte/2MByte pages, 8-way associative, 1024 entries\n"); break; case 0xc2: printf("DTLB: 4 KByte/2 MByte pages, 4-way associative, 16 entries\n"); break; case 0xc3: printf("Shared 2nd-Level TLB: 4 KByte /2 MByte pages, 6-way associative, 1536 entries. Also 1GBbyte pages, 4-way, 16 entries\n"); break; case 0xc4: printf("DTLB: 2M/4M Byte pages, 4-way associative, 32 entries\n"); break; case 0xca: printf("Shared 2nd-Level TLB: 4 KByte pages, 4-way associative, 512 entries\n"); break; case 0xd0: printf("3rd-level cache: 512 KByte, 4-way set associative, 64 byte line size\n"); break; case 0xd1: printf("3rd-level cache: 1 MByte, 4-way set associative, 64 byte line size\n"); break; case 0xd2: printf("3rd-level cache: 2 MByte, 4-way set associative, 64 byte line size\n"); break; case 0xd6: printf("3rd-level cache: 1 MByte, 8-way set associative, 64 byte line size\n"); break; case 0xd7: printf("3rd-level cache: 2 MByte, 8-way set associative, 64 byte line size\n"); break; case 0xd8: printf("3rd-level cache: 4 MByte, 8-way set associative, 64 byte line size\n"); break; case 0xdc: printf("3rd-level cache: 1.5 MByte, 12-way set associative, 64 byte line size\n"); break; case 0xdd: printf("3rd-level cache: 3 MByte, 12-way set associative, 64 byte line size\n"); break; case 0xde: printf("3rd-level cache: 6 MByte, 12-way set associative, 64 byte line size\n"); break; case 0xe2: printf("3rd-level cache: 2 MByte, 16-way set associative, 64 byte line size\n"); break; case 0xe3: printf("3rd-level cache: 4 MByte, 16-way set associative, 64 byte line size\n"); break; case 0xe4: printf("3rd-level cache: 8 MByte, 16-way set associative, 64 byte line size\n"); break; case 0xea: printf("3rd-level cache: 12MByte, 24-way set associative, 64 byte line size\n"); break; case 0xeb: printf("3rd-level cache: 18MByte, 24-way set associative, 64 byte line size\n"); break; case 0xec: printf("3rd-level cache: 24MByte, 24-way set associative, 64 byte line size\n"); break; case 0xf0: printf("64-Byte prefetching\n"); break; case 0xf1: printf("128-Byte prefetching\n"); break; } } static void print_svm_info(void) { u_int features, regs[4]; uint64_t msr; int comma; printf("\n SVM: "); do_cpuid(0x8000000A, regs); features = regs[3]; msr = rdmsr(MSR_VM_CR); if ((msr & VM_CR_SVMDIS) == VM_CR_SVMDIS) printf("(disabled in BIOS) "); if (!bootverbose) { comma = 0; if (features & (1 << 0)) { printf("%sNP", comma ? "," : ""); comma = 1; } if (features & (1 << 3)) { printf("%sNRIP", comma ? "," : ""); comma = 1; } if (features & (1 << 5)) { printf("%sVClean", comma ? "," : ""); comma = 1; } if (features & (1 << 6)) { printf("%sAFlush", comma ? "," : ""); comma = 1; } if (features & (1 << 7)) { printf("%sDAssist", comma ? "," : ""); comma = 1; } printf("%sNAsids=%d", comma ? "," : "", regs[1]); return; } printf("Features=0x%b", features, "\020" "\001NP" /* Nested paging */ "\002LbrVirt" /* LBR virtualization */ "\003SVML" /* SVM lock */ "\004NRIPS" /* NRIP save */ "\005TscRateMsr" /* MSR based TSC rate control */ "\006VmcbClean" /* VMCB clean bits */ "\007FlushByAsid" /* Flush by ASID */ "\010DecodeAssist" /* Decode assist */ "\011" "\012" "\013PauseFilter" /* PAUSE intercept filter */ "\014EncryptedMcodePatch" "\015PauseFilterThreshold" /* PAUSE filter threshold */ "\016AVIC" /* virtual interrupt controller */ "\017" "\020V_VMSAVE_VMLOAD" "\021vGIF" "\022" "\023" "\024" "\025" "\026" "\027" "\030" "\031" "\032" "\033" "\034" "\035" "\036" "\037" "\040" ); printf("\nRevision=%d, ASIDs=%d", regs[0] & 0xff, regs[1]); } #ifdef __i386__ static void print_transmeta_info(void) { u_int regs[4], nreg = 0; do_cpuid(0x80860000, regs); nreg = regs[0]; if (nreg >= 0x80860001) { do_cpuid(0x80860001, regs); printf(" Processor revision %u.%u.%u.%u\n", (regs[1] >> 24) & 0xff, (regs[1] >> 16) & 0xff, (regs[1] >> 8) & 0xff, regs[1] & 0xff); } if (nreg >= 0x80860002) { do_cpuid(0x80860002, regs); printf(" Code Morphing Software revision %u.%u.%u-%u-%u\n", (regs[1] >> 24) & 0xff, (regs[1] >> 16) & 0xff, (regs[1] >> 8) & 0xff, regs[1] & 0xff, regs[2]); } if (nreg >= 0x80860006) { char info[65]; do_cpuid(0x80860003, (u_int*) &info[0]); do_cpuid(0x80860004, (u_int*) &info[16]); do_cpuid(0x80860005, (u_int*) &info[32]); do_cpuid(0x80860006, (u_int*) &info[48]); info[64] = 0; printf(" %s\n", info); } } #endif static void print_via_padlock_info(void) { u_int regs[4]; do_cpuid(0xc0000001, regs); printf("\n VIA Padlock Features=0x%b", regs[3], "\020" "\003RNG" /* RNG */ "\007AES" /* ACE */ "\011AES-CTR" /* ACE2 */ "\013SHA1,SHA256" /* PHE */ "\015RSA" /* PMM */ ); } static uint32_t vmx_settable(uint64_t basic, int msr, int true_msr) { uint64_t val; if (basic & (1ULL << 55)) val = rdmsr(true_msr); else val = rdmsr(msr); /* Just report the controls that can be set to 1. */ return (val >> 32); } static void print_vmx_info(void) { uint64_t basic, msr; uint32_t entry, exit, mask, pin, proc, proc2; int comma; printf("\n VT-x: "); msr = rdmsr(MSR_IA32_FEATURE_CONTROL); if (!(msr & IA32_FEATURE_CONTROL_VMX_EN)) printf("(disabled in BIOS) "); basic = rdmsr(MSR_VMX_BASIC); pin = vmx_settable(basic, MSR_VMX_PINBASED_CTLS, MSR_VMX_TRUE_PINBASED_CTLS); proc = vmx_settable(basic, MSR_VMX_PROCBASED_CTLS, MSR_VMX_TRUE_PROCBASED_CTLS); if (proc & PROCBASED_SECONDARY_CONTROLS) proc2 = vmx_settable(basic, MSR_VMX_PROCBASED_CTLS2, MSR_VMX_PROCBASED_CTLS2); else proc2 = 0; exit = vmx_settable(basic, MSR_VMX_EXIT_CTLS, MSR_VMX_TRUE_EXIT_CTLS); entry = vmx_settable(basic, MSR_VMX_ENTRY_CTLS, MSR_VMX_TRUE_ENTRY_CTLS); if (!bootverbose) { comma = 0; if (exit & VM_EXIT_SAVE_PAT && exit & VM_EXIT_LOAD_PAT && entry & VM_ENTRY_LOAD_PAT) { printf("%sPAT", comma ? "," : ""); comma = 1; } if (proc & PROCBASED_HLT_EXITING) { printf("%sHLT", comma ? "," : ""); comma = 1; } if (proc & PROCBASED_MTF) { printf("%sMTF", comma ? "," : ""); comma = 1; } if (proc & PROCBASED_PAUSE_EXITING) { printf("%sPAUSE", comma ? "," : ""); comma = 1; } if (proc2 & PROCBASED2_ENABLE_EPT) { printf("%sEPT", comma ? "," : ""); comma = 1; } if (proc2 & PROCBASED2_UNRESTRICTED_GUEST) { printf("%sUG", comma ? "," : ""); comma = 1; } if (proc2 & PROCBASED2_ENABLE_VPID) { printf("%sVPID", comma ? "," : ""); comma = 1; } if (proc & PROCBASED_USE_TPR_SHADOW && proc2 & PROCBASED2_VIRTUALIZE_APIC_ACCESSES && proc2 & PROCBASED2_VIRTUALIZE_X2APIC_MODE && proc2 & PROCBASED2_APIC_REGISTER_VIRTUALIZATION && proc2 & PROCBASED2_VIRTUAL_INTERRUPT_DELIVERY) { printf("%sVID", comma ? "," : ""); comma = 1; if (pin & PINBASED_POSTED_INTERRUPT) printf(",PostIntr"); } return; } mask = basic >> 32; printf("Basic Features=0x%b", mask, "\020" "\02132PA" /* 32-bit physical addresses */ "\022SMM" /* SMM dual-monitor */ "\027INS/OUTS" /* VM-exit info for INS and OUTS */ "\030TRUE" /* TRUE_CTLS MSRs */ ); printf("\n Pin-Based Controls=0x%b", pin, "\020" "\001ExtINT" /* External-interrupt exiting */ "\004NMI" /* NMI exiting */ "\006VNMI" /* Virtual NMIs */ "\007PreTmr" /* Activate VMX-preemption timer */ "\010PostIntr" /* Process posted interrupts */ ); printf("\n Primary Processor Controls=0x%b", proc, "\020" "\003INTWIN" /* Interrupt-window exiting */ "\004TSCOff" /* Use TSC offsetting */ "\010HLT" /* HLT exiting */ "\012INVLPG" /* INVLPG exiting */ "\013MWAIT" /* MWAIT exiting */ "\014RDPMC" /* RDPMC exiting */ "\015RDTSC" /* RDTSC exiting */ "\020CR3-LD" /* CR3-load exiting */ "\021CR3-ST" /* CR3-store exiting */ "\024CR8-LD" /* CR8-load exiting */ "\025CR8-ST" /* CR8-store exiting */ "\026TPR" /* Use TPR shadow */ "\027NMIWIN" /* NMI-window exiting */ "\030MOV-DR" /* MOV-DR exiting */ "\031IO" /* Unconditional I/O exiting */ "\032IOmap" /* Use I/O bitmaps */ "\034MTF" /* Monitor trap flag */ "\035MSRmap" /* Use MSR bitmaps */ "\036MONITOR" /* MONITOR exiting */ "\037PAUSE" /* PAUSE exiting */ ); if (proc & PROCBASED_SECONDARY_CONTROLS) printf("\n Secondary Processor Controls=0x%b", proc2, "\020" "\001APIC" /* Virtualize APIC accesses */ "\002EPT" /* Enable EPT */ "\003DT" /* Descriptor-table exiting */ "\004RDTSCP" /* Enable RDTSCP */ "\005x2APIC" /* Virtualize x2APIC mode */ "\006VPID" /* Enable VPID */ "\007WBINVD" /* WBINVD exiting */ "\010UG" /* Unrestricted guest */ "\011APIC-reg" /* APIC-register virtualization */ "\012VID" /* Virtual-interrupt delivery */ "\013PAUSE-loop" /* PAUSE-loop exiting */ "\014RDRAND" /* RDRAND exiting */ "\015INVPCID" /* Enable INVPCID */ "\016VMFUNC" /* Enable VM functions */ "\017VMCS" /* VMCS shadowing */ "\020EPT#VE" /* EPT-violation #VE */ "\021XSAVES" /* Enable XSAVES/XRSTORS */ ); printf("\n Exit Controls=0x%b", mask, "\020" "\003DR" /* Save debug controls */ /* Ignore Host address-space size */ "\015PERF" /* Load MSR_PERF_GLOBAL_CTRL */ "\020AckInt" /* Acknowledge interrupt on exit */ "\023PAT-SV" /* Save MSR_PAT */ "\024PAT-LD" /* Load MSR_PAT */ "\025EFER-SV" /* Save MSR_EFER */ "\026EFER-LD" /* Load MSR_EFER */ "\027PTMR-SV" /* Save VMX-preemption timer value */ ); printf("\n Entry Controls=0x%b", mask, "\020" "\003DR" /* Save debug controls */ /* Ignore IA-32e mode guest */ /* Ignore Entry to SMM */ /* Ignore Deactivate dual-monitor treatment */ "\016PERF" /* Load MSR_PERF_GLOBAL_CTRL */ "\017PAT" /* Load MSR_PAT */ "\020EFER" /* Load MSR_EFER */ ); if (proc & PROCBASED_SECONDARY_CONTROLS && (proc2 & (PROCBASED2_ENABLE_EPT | PROCBASED2_ENABLE_VPID)) != 0) { msr = rdmsr(MSR_VMX_EPT_VPID_CAP); mask = msr; printf("\n EPT Features=0x%b", mask, "\020" "\001XO" /* Execute-only translations */ "\007PW4" /* Page-walk length of 4 */ "\011UC" /* EPT paging-structure mem can be UC */ "\017WB" /* EPT paging-structure mem can be WB */ "\0212M" /* EPT PDE can map a 2-Mbyte page */ "\0221G" /* EPT PDPTE can map a 1-Gbyte page */ "\025INVEPT" /* INVEPT is supported */ "\026AD" /* Accessed and dirty flags for EPT */ "\032single" /* INVEPT single-context type */ "\033all" /* INVEPT all-context type */ ); mask = msr >> 32; printf("\n VPID Features=0x%b", mask, "\020" "\001INVVPID" /* INVVPID is supported */ "\011individual" /* INVVPID individual-address type */ "\012single" /* INVVPID single-context type */ "\013all" /* INVVPID all-context type */ /* INVVPID single-context-retaining-globals type */ "\014single-globals" ); } } static void print_hypervisor_info(void) { if (*hv_vendor) printf("Hypervisor: Origin = \"%s\"\n", hv_vendor); } Index: releng/12.0 =================================================================== --- releng/12.0 (revision 340626) +++ releng/12.0 (revision 340627) Property changes on: releng/12.0 ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,2 ## Merged /stable/12:r340605 Merged /head:r340487