From owner-svn-src-head@FreeBSD.ORG Tue Oct 21 07:10:45 2014 Return-Path: Delivered-To: svn-src-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id C176EBB4; Tue, 21 Oct 2014 07:10:45 +0000 (UTC) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id A9F7FCD3; Tue, 21 Oct 2014 07:10:45 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.9/8.14.9) with ESMTP id s9L7AjL9086943; Tue, 21 Oct 2014 07:10:45 GMT (envelope-from neel@FreeBSD.org) Received: (from neel@localhost) by svn.freebsd.org (8.14.9/8.14.9/Submit) id s9L7Ai8S086934; Tue, 21 Oct 2014 07:10:44 GMT (envelope-from neel@FreeBSD.org) Message-Id: <201410210710.s9L7Ai8S086934@svn.freebsd.org> X-Authentication-Warning: svn.freebsd.org: neel set sender to neel@FreeBSD.org using -f From: Neel Natu Date: Tue, 21 Oct 2014 07:10:44 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r273375 - in head: sys/amd64/include sys/amd64/vmm sys/amd64/vmm/amd sys/amd64/vmm/intel sys/amd64/vmm/io sys/modules/vmm usr.sbin/bhyve usr.sbin/bhyvectl X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 21 Oct 2014 07:10:46 -0000 Author: neel Date: Tue Oct 21 07:10:43 2014 New Revision: 273375 URL: https://svnweb.freebsd.org/changeset/base/273375 Log: Merge projects/bhyve_svm into HEAD. After this change bhyve supports AMD processors with the SVM/AMD-V hardware extensions. More details available here: https://lists.freebsd.org/pipermail/freebsd-virtualization/2014-October/002905.html Submitted by: Anish Gupta (akgupt3@gmail.com) Tested by: Benjamin Perrault (ben.perrault@gmail.com) Tested by: Willem Jan Withagen (wjw@digiware.nl) Added: - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.c - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.h - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm.c - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm.h - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm_genassym.c - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm_msr.c - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm_msr.h - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm_softc.h - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm_support.S - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/vmcb.c - copied unchanged from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/vmcb.h Directory Properties: head/sys/amd64/vmm/amd/npt.c (props changed) head/sys/amd64/vmm/amd/npt.h (props changed) head/sys/amd64/vmm/amd/svm.c (props changed) head/sys/amd64/vmm/amd/svm.h (props changed) head/sys/amd64/vmm/amd/svm_genassym.c (props changed) head/sys/amd64/vmm/amd/svm_msr.c (props changed) head/sys/amd64/vmm/amd/svm_msr.h (props changed) head/sys/amd64/vmm/amd/svm_softc.h (props changed) head/sys/amd64/vmm/amd/svm_support.S (props changed) head/sys/amd64/vmm/amd/vmcb.c (props changed) head/sys/amd64/vmm/amd/vmcb.h (props changed) Modified: head/sys/amd64/include/vmm.h head/sys/amd64/include/vmm_instruction_emul.h head/sys/amd64/vmm/amd/amdv.c head/sys/amd64/vmm/intel/vmx.c head/sys/amd64/vmm/io/vlapic.c head/sys/amd64/vmm/vmm.c head/sys/amd64/vmm/vmm_instruction_emul.c head/sys/amd64/vmm/x86.c head/sys/modules/vmm/Makefile head/usr.sbin/bhyve/bhyverun.c head/usr.sbin/bhyve/xmsr.c head/usr.sbin/bhyvectl/bhyvectl.c Modified: head/sys/amd64/include/vmm.h ============================================================================== --- head/sys/amd64/include/vmm.h Tue Oct 21 06:25:41 2014 (r273374) +++ head/sys/amd64/include/vmm.h Tue Oct 21 07:10:43 2014 (r273375) @@ -487,6 +487,7 @@ enum vm_exitcode { VM_EXITCODE_TASK_SWITCH, VM_EXITCODE_MONITOR, VM_EXITCODE_MWAIT, + VM_EXITCODE_SVM, VM_EXITCODE_MAX }; @@ -564,6 +565,14 @@ struct vm_exit { int inst_type; int inst_error; } vmx; + /* + * SVM specific payload. + */ + struct { + uint64_t exitcode; + uint64_t exitinfo1; + uint64_t exitinfo2; + } svm; struct { uint32_t code; /* ecx value */ uint64_t wval; Modified: head/sys/amd64/include/vmm_instruction_emul.h ============================================================================== --- head/sys/amd64/include/vmm_instruction_emul.h Tue Oct 21 06:25:41 2014 (r273374) +++ head/sys/amd64/include/vmm_instruction_emul.h Tue Oct 21 07:10:43 2014 (r273375) @@ -93,7 +93,7 @@ int vmm_fetch_instruction(struct vm *vm, int vmm_gla2gpa(struct vm *vm, int vcpuid, struct vm_guest_paging *paging, uint64_t gla, int prot, uint64_t *gpa); -void vie_init(struct vie *vie); +void vie_init(struct vie *vie, const char *inst_bytes, int inst_length); /* * Decode the instruction fetched into 'vie' so it can be emulated. Modified: head/sys/amd64/vmm/amd/amdv.c ============================================================================== --- head/sys/amd64/vmm/amd/amdv.c Tue Oct 21 06:25:41 2014 (r273374) +++ head/sys/amd64/vmm/amd/amdv.c Tue Oct 21 07:10:43 2014 (r273375) @@ -38,149 +38,6 @@ __FBSDID("$FreeBSD$"); #include "io/iommu.h" static int -amdv_init(int ipinum) -{ - - printf("amdv_init: not implemented\n"); - return (ENXIO); -} - -static int -amdv_cleanup(void) -{ - - printf("amdv_cleanup: not implemented\n"); - return (ENXIO); -} - -static void -amdv_resume(void) -{ -} - -static void * -amdv_vminit(struct vm *vm, struct pmap *pmap) -{ - - printf("amdv_vminit: not implemented\n"); - return (NULL); -} - -static int -amdv_vmrun(void *arg, int vcpu, register_t rip, struct pmap *pmap, - void *rptr, void *sptr) -{ - - printf("amdv_vmrun: not implemented\n"); - return (ENXIO); -} - -static void -amdv_vmcleanup(void *arg) -{ - - printf("amdv_vmcleanup: not implemented\n"); - return; -} - -static int -amdv_getreg(void *arg, int vcpu, int regnum, uint64_t *retval) -{ - - printf("amdv_getreg: not implemented\n"); - return (EINVAL); -} - -static int -amdv_setreg(void *arg, int vcpu, int regnum, uint64_t val) -{ - - printf("amdv_setreg: not implemented\n"); - return (EINVAL); -} - -static int -amdv_getdesc(void *vmi, int vcpu, int num, struct seg_desc *desc) -{ - - printf("amdv_get_desc: not implemented\n"); - return (EINVAL); -} - -static int -amdv_setdesc(void *vmi, int vcpu, int num, struct seg_desc *desc) -{ - - printf("amdv_get_desc: not implemented\n"); - return (EINVAL); -} - -static int -amdv_getcap(void *arg, int vcpu, int type, int *retval) -{ - - printf("amdv_getcap: not implemented\n"); - return (EINVAL); -} - -static int -amdv_setcap(void *arg, int vcpu, int type, int val) -{ - - printf("amdv_setcap: not implemented\n"); - return (EINVAL); -} - -static struct vmspace * -amdv_vmspace_alloc(vm_offset_t min, vm_offset_t max) -{ - - printf("amdv_vmspace_alloc: not implemented\n"); - return (NULL); -} - -static void -amdv_vmspace_free(struct vmspace *vmspace) -{ - - printf("amdv_vmspace_free: not implemented\n"); - return; -} - -static struct vlapic * -amdv_vlapic_init(void *arg, int vcpuid) -{ - - panic("amdv_vlapic_init: not implmented"); -} - -static void -amdv_vlapic_cleanup(void *arg, struct vlapic *vlapic) -{ - - panic("amdv_vlapic_cleanup: not implemented"); -} - -struct vmm_ops vmm_ops_amd = { - amdv_init, - amdv_cleanup, - amdv_resume, - amdv_vminit, - amdv_vmrun, - amdv_vmcleanup, - amdv_getreg, - amdv_setreg, - amdv_getdesc, - amdv_setdesc, - amdv_getcap, - amdv_setcap, - amdv_vmspace_alloc, - amdv_vmspace_free, - amdv_vlapic_init, - amdv_vlapic_cleanup, -}; - -static int amd_iommu_init(void) { Copied: head/sys/amd64/vmm/amd/npt.c (from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.c) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/sys/amd64/vmm/amd/npt.c Tue Oct 21 07:10:43 2014 (r273375, copy of r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.c) @@ -0,0 +1,87 @@ +/*- + * Copyright (c) 2013 Anish Gupta (akgupt3@gmail.com) + * 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 unmodified, 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 THE AUTHOR ``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 AUTHOR 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. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include +#include +#include +#include + +#include +#include +#include + +#include + +#include "npt.h" + +SYSCTL_DECL(_hw_vmm); +SYSCTL_NODE(_hw_vmm, OID_AUTO, npt, CTLFLAG_RW, NULL, NULL); + +static int npt_flags; +SYSCTL_INT(_hw_vmm_npt, OID_AUTO, pmap_flags, CTLFLAG_RD, + &npt_flags, 0, NULL); + +#define NPT_IPIMASK 0xFF + +/* + * AMD nested page table init. + */ +int +svm_npt_init(int ipinum) +{ + int enable_superpage = 1; + + npt_flags = ipinum & NPT_IPIMASK; + TUNABLE_INT_FETCH("hw.vmm.npt.enable_superpage", &enable_superpage); + if (enable_superpage) + npt_flags |= PMAP_PDE_SUPERPAGE; + + return (0); +} + +static int +npt_pinit(pmap_t pmap) +{ + + return (pmap_pinit_type(pmap, PT_RVI, npt_flags)); +} + +struct vmspace * +svm_npt_alloc(vm_offset_t min, vm_offset_t max) +{ + + return (vmspace_alloc(min, max, npt_pinit)); +} + +void +svm_npt_free(struct vmspace *vmspace) +{ + + vmspace_free(vmspace); +} Copied: head/sys/amd64/vmm/amd/npt.h (from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.h) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/sys/amd64/vmm/amd/npt.h Tue Oct 21 07:10:43 2014 (r273375, copy of r273369, projects/bhyve_svm/sys/amd64/vmm/amd/npt.h) @@ -0,0 +1,36 @@ +/*- + * Copyright (c) 2013 Anish Gupta (akgupt3@gmail.com) + * 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 unmodified, 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 THE AUTHOR ``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 AUTHOR 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. + * + * $FreeBSD$ + */ + +#ifndef _SVM_NPT_H_ +#define _SVM_NPT_H_ + +int svm_npt_init(int ipinum); +struct vmspace *svm_npt_alloc(vm_offset_t min, vm_offset_t max); +void svm_npt_free(struct vmspace *vmspace); + +#endif /* _SVM_NPT_H_ */ Copied: head/sys/amd64/vmm/amd/svm.c (from r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm.c) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/sys/amd64/vmm/amd/svm.c Tue Oct 21 07:10:43 2014 (r273375, copy of r273369, projects/bhyve_svm/sys/amd64/vmm/amd/svm.c) @@ -0,0 +1,2092 @@ +/*- + * Copyright (c) 2013, Anish Gupta (akgupt3@gmail.com) + * 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 unmodified, 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 THE AUTHOR ``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 AUTHOR 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. + */ + +#include +__FBSDID("$FreeBSD$"); + +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include + +#include "vmm_lapic.h" +#include "vmm_stat.h" +#include "vmm_ktr.h" +#include "vmm_ioport.h" +#include "vatpic.h" +#include "vlapic.h" +#include "vlapic_priv.h" + +#include "x86.h" +#include "vmcb.h" +#include "svm.h" +#include "svm_softc.h" +#include "svm_msr.h" +#include "npt.h" + +SYSCTL_DECL(_hw_vmm); +SYSCTL_NODE(_hw_vmm, OID_AUTO, svm, CTLFLAG_RW, NULL, NULL); + +/* + * SVM CPUID function 0x8000_000A, edx bit decoding. + */ +#define AMD_CPUID_SVM_NP BIT(0) /* Nested paging or RVI */ +#define AMD_CPUID_SVM_LBR BIT(1) /* Last branch virtualization */ +#define AMD_CPUID_SVM_SVML BIT(2) /* SVM lock */ +#define AMD_CPUID_SVM_NRIP_SAVE BIT(3) /* Next RIP is saved */ +#define AMD_CPUID_SVM_TSC_RATE BIT(4) /* TSC rate control. */ +#define AMD_CPUID_SVM_VMCB_CLEAN BIT(5) /* VMCB state caching */ +#define AMD_CPUID_SVM_FLUSH_BY_ASID BIT(6) /* Flush by ASID */ +#define AMD_CPUID_SVM_DECODE_ASSIST BIT(7) /* Decode assist */ +#define AMD_CPUID_SVM_PAUSE_INC BIT(10) /* Pause intercept filter. */ +#define AMD_CPUID_SVM_PAUSE_FTH BIT(12) /* Pause filter threshold */ + +#define VMCB_CACHE_DEFAULT (VMCB_CACHE_ASID | \ + VMCB_CACHE_IOPM | \ + VMCB_CACHE_I | \ + VMCB_CACHE_TPR | \ + VMCB_CACHE_CR2 | \ + VMCB_CACHE_CR | \ + VMCB_CACHE_DT | \ + VMCB_CACHE_SEG | \ + VMCB_CACHE_NP) + +static uint32_t vmcb_clean = VMCB_CACHE_DEFAULT; +SYSCTL_INT(_hw_vmm_svm, OID_AUTO, vmcb_clean, CTLFLAG_RDTUN, &vmcb_clean, + 0, NULL); + +static MALLOC_DEFINE(M_SVM, "svm", "svm"); +static MALLOC_DEFINE(M_SVM_VLAPIC, "svm-vlapic", "svm-vlapic"); + +/* Per-CPU context area. */ +extern struct pcpu __pcpu[]; + +static uint32_t svm_feature; /* AMD SVM features. */ +SYSCTL_UINT(_hw_vmm_svm, OID_AUTO, features, CTLFLAG_RD, &svm_feature, 0, + "SVM features advertised by CPUID.8000000AH:EDX"); + +static int disable_npf_assist; +SYSCTL_INT(_hw_vmm_svm, OID_AUTO, disable_npf_assist, CTLFLAG_RWTUN, + &disable_npf_assist, 0, NULL); + +/* Maximum ASIDs supported by the processor */ +static uint32_t nasid; +SYSCTL_UINT(_hw_vmm_svm, OID_AUTO, num_asids, CTLFLAG_RD, &nasid, 0, + "Number of ASIDs supported by this processor"); + +/* Current ASID generation for each host cpu */ +static struct asid asid[MAXCPU]; + +/* + * SVM host state saved area of size 4KB for each core. + */ +static uint8_t hsave[MAXCPU][PAGE_SIZE] __aligned(PAGE_SIZE); + +static VMM_STAT_AMD(VCPU_EXITINTINFO, "VM exits during event delivery"); +static VMM_STAT_AMD(VCPU_INTINFO_INJECTED, "Events pending at VM entry"); +static VMM_STAT_AMD(VMEXIT_VINTR, "VM exits due to interrupt window"); + +static int svm_setreg(void *arg, int vcpu, int ident, uint64_t val); + +static __inline int +flush_by_asid(void) +{ + + return (svm_feature & AMD_CPUID_SVM_FLUSH_BY_ASID); +} + +static __inline int +decode_assist(void) +{ + + return (svm_feature & AMD_CPUID_SVM_DECODE_ASSIST); +} + +static void +svm_disable(void *arg __unused) +{ + uint64_t efer; + + efer = rdmsr(MSR_EFER); + efer &= ~EFER_SVM; + wrmsr(MSR_EFER, efer); +} + +/* + * Disable SVM on all CPUs. + */ +static int +svm_cleanup(void) +{ + + smp_rendezvous(NULL, svm_disable, NULL, NULL); + return (0); +} + +/* + * Verify that all the features required by bhyve are available. + */ +static int +check_svm_features(void) +{ + u_int regs[4]; + + /* CPUID Fn8000_000A is for SVM */ + do_cpuid(0x8000000A, regs); + svm_feature = regs[3]; + + printf("SVM: Revision %d\n", regs[0] & 0xFF); + printf("SVM: NumASID %u\n", regs[1]); + + nasid = regs[1]; + KASSERT(nasid > 1, ("Insufficient ASIDs for guests: %#x", nasid)); + + printf("SVM: Features 0x%b\n", svm_feature, + "\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" + "\014" + "\015PauseFilterThreshold" + "\016AVIC" + ); + + /* bhyve requires the Nested Paging feature */ + if (!(svm_feature & AMD_CPUID_SVM_NP)) { + printf("SVM: Nested Paging feature not available.\n"); + return (ENXIO); + } + + /* bhyve requires the NRIP Save feature */ + if (!(svm_feature & AMD_CPUID_SVM_NRIP_SAVE)) { + printf("SVM: NRIP Save feature not available.\n"); + return (ENXIO); + } + + return (0); +} + +static void +svm_enable(void *arg __unused) +{ + uint64_t efer; + + efer = rdmsr(MSR_EFER); + efer |= EFER_SVM; + wrmsr(MSR_EFER, efer); + + wrmsr(MSR_VM_HSAVE_PA, vtophys(hsave[curcpu])); +} + +/* + * Return 1 if SVM is enabled on this processor and 0 otherwise. + */ +static int +svm_available(void) +{ + uint64_t msr; + + /* Section 15.4 Enabling SVM from APM2. */ + if ((amd_feature2 & AMDID2_SVM) == 0) { + printf("SVM: not available.\n"); + return (0); + } + + msr = rdmsr(MSR_VM_CR); + if ((msr & VM_CR_SVMDIS) != 0) { + printf("SVM: disabled by BIOS.\n"); + return (0); + } + + return (1); +} + +static int +svm_init(int ipinum) +{ + int error, cpu; + + if (!svm_available()) + return (ENXIO); + + error = check_svm_features(); + if (error) + return (error); + + vmcb_clean &= VMCB_CACHE_DEFAULT; + + for (cpu = 0; cpu < MAXCPU; cpu++) { + /* + * Initialize the host ASIDs to their "highest" valid values. + * + * The next ASID allocation will rollover both 'gen' and 'num' + * and start off the sequence at {1,1}. + */ + asid[cpu].gen = ~0UL; + asid[cpu].num = nasid - 1; + } + + svm_msr_init(); + svm_npt_init(ipinum); + + /* Enable SVM on all CPUs */ + smp_rendezvous(NULL, svm_enable, NULL, NULL); + + return (0); +} + +static void +svm_restore(void) +{ + + svm_enable(NULL); +} + +/* Pentium compatible MSRs */ +#define MSR_PENTIUM_START 0 +#define MSR_PENTIUM_END 0x1FFF +/* AMD 6th generation and Intel compatible MSRs */ +#define MSR_AMD6TH_START 0xC0000000UL +#define MSR_AMD6TH_END 0xC0001FFFUL +/* AMD 7th and 8th generation compatible MSRs */ +#define MSR_AMD7TH_START 0xC0010000UL +#define MSR_AMD7TH_END 0xC0011FFFUL + +/* + * Get the index and bit position for a MSR in permission bitmap. + * Two bits are used for each MSR: lower bit for read and higher bit for write. + */ +static int +svm_msr_index(uint64_t msr, int *index, int *bit) +{ + uint32_t base, off; + + *index = -1; + *bit = (msr % 4) * 2; + base = 0; + + if (msr >= MSR_PENTIUM_START && msr <= MSR_PENTIUM_END) { + *index = msr / 4; + return (0); + } + + base += (MSR_PENTIUM_END - MSR_PENTIUM_START + 1); + if (msr >= MSR_AMD6TH_START && msr <= MSR_AMD6TH_END) { + off = (msr - MSR_AMD6TH_START); + *index = (off + base) / 4; + return (0); + } + + base += (MSR_AMD6TH_END - MSR_AMD6TH_START + 1); + if (msr >= MSR_AMD7TH_START && msr <= MSR_AMD7TH_END) { + off = (msr - MSR_AMD7TH_START); + *index = (off + base) / 4; + return (0); + } + + return (EINVAL); +} + +/* + * Allow vcpu to read or write the 'msr' without trapping into the hypervisor. + */ +static void +svm_msr_perm(uint8_t *perm_bitmap, uint64_t msr, bool read, bool write) +{ + int index, bit, error; + + error = svm_msr_index(msr, &index, &bit); + KASSERT(error == 0, ("%s: invalid msr %#lx", __func__, msr)); + KASSERT(index >= 0 && index < SVM_MSR_BITMAP_SIZE, + ("%s: invalid index %d for msr %#lx", __func__, index, msr)); + KASSERT(bit >= 0 && bit <= 6, ("%s: invalid bit position %d " + "msr %#lx", __func__, bit, msr)); + + if (read) + perm_bitmap[index] &= ~(1UL << bit); + + if (write) + perm_bitmap[index] &= ~(2UL << bit); +} + +static void +svm_msr_rw_ok(uint8_t *perm_bitmap, uint64_t msr) +{ + + svm_msr_perm(perm_bitmap, msr, true, true); +} + +static void +svm_msr_rd_ok(uint8_t *perm_bitmap, uint64_t msr) +{ + + svm_msr_perm(perm_bitmap, msr, true, false); +} + +static __inline int +svm_get_intercept(struct svm_softc *sc, int vcpu, int idx, uint32_t bitmask) +{ + struct vmcb_ctrl *ctrl; + + KASSERT(idx >=0 && idx < 5, ("invalid intercept index %d", idx)); + + ctrl = svm_get_vmcb_ctrl(sc, vcpu); + return (ctrl->intercept[idx] & bitmask ? 1 : 0); +} + +static __inline void +svm_set_intercept(struct svm_softc *sc, int vcpu, int idx, uint32_t bitmask, + int enabled) +{ + struct vmcb_ctrl *ctrl; + uint32_t oldval; + + KASSERT(idx >=0 && idx < 5, ("invalid intercept index %d", idx)); + + ctrl = svm_get_vmcb_ctrl(sc, vcpu); + oldval = ctrl->intercept[idx]; + + if (enabled) + ctrl->intercept[idx] |= bitmask; + else + ctrl->intercept[idx] &= ~bitmask; + + if (ctrl->intercept[idx] != oldval) { + svm_set_dirty(sc, vcpu, VMCB_CACHE_I); + VCPU_CTR3(sc->vm, vcpu, "intercept[%d] modified " + "from %#x to %#x", idx, oldval, ctrl->intercept[idx]); + } +} + +static __inline void +svm_disable_intercept(struct svm_softc *sc, int vcpu, int off, uint32_t bitmask) +{ + + svm_set_intercept(sc, vcpu, off, bitmask, 0); +} + +static __inline void +svm_enable_intercept(struct svm_softc *sc, int vcpu, int off, uint32_t bitmask) +{ + + svm_set_intercept(sc, vcpu, off, bitmask, 1); +} + +static void +vmcb_init(struct svm_softc *sc, int vcpu, uint64_t iopm_base_pa, + uint64_t msrpm_base_pa, uint64_t np_pml4) +{ + struct vmcb_ctrl *ctrl; + struct vmcb_state *state; + uint32_t mask; + int n; + + ctrl = svm_get_vmcb_ctrl(sc, vcpu); + state = svm_get_vmcb_state(sc, vcpu); + + ctrl->iopm_base_pa = iopm_base_pa; + ctrl->msrpm_base_pa = msrpm_base_pa; + + /* Enable nested paging */ + ctrl->np_enable = 1; + ctrl->n_cr3 = np_pml4; + + /* + * Intercept accesses to the control registers that are not shadowed + * in the VMCB - i.e. all except cr0, cr2, cr3, cr4 and cr8. + */ + for (n = 0; n < 16; n++) { + mask = (BIT(n) << 16) | BIT(n); + if (n == 0 || n == 2 || n == 3 || n == 4 || n == 8) + svm_disable_intercept(sc, vcpu, VMCB_CR_INTCPT, mask); + else + svm_enable_intercept(sc, vcpu, VMCB_CR_INTCPT, mask); + } + + /* Intercept Machine Check exceptions. */ + svm_enable_intercept(sc, vcpu, VMCB_EXC_INTCPT, BIT(IDT_MC)); + + /* Intercept various events (for e.g. I/O, MSR and CPUID accesses) */ + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_IO); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_MSR); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_CPUID); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_INTR); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_INIT); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_NMI); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_SMI); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, VMCB_INTCPT_SHUTDOWN); + svm_enable_intercept(sc, vcpu, VMCB_CTRL1_INTCPT, + VMCB_INTCPT_FERR_FREEZE); + + /* + * From section "Canonicalization and Consistency Checks" in APMv2 + * the VMRUN intercept bit must be set to pass the consistency check. + */ + svm_enable_intercept(sc, vcpu, VMCB_CTRL2_INTCPT, VMCB_INTCPT_VMRUN); + + /* + * The ASID will be set to a non-zero value just before VMRUN. + */ + ctrl->asid = 0; + + /* + * Section 15.21.1, Interrupt Masking in EFLAGS + * Section 15.21.2, Virtualizing APIC.TPR + * + * This must be set for %rflag and %cr8 isolation of guest and host. + */ + ctrl->v_intr_masking = 1; + + /* Enable Last Branch Record aka LBR for debugging */ + ctrl->lbr_virt_en = 1; + state->dbgctl = BIT(0); + + /* EFER_SVM must always be set when the guest is executing */ + state->efer = EFER_SVM; + + /* Set up the PAT to power-on state */ + state->g_pat = PAT_VALUE(0, PAT_WRITE_BACK) | + PAT_VALUE(1, PAT_WRITE_THROUGH) | + PAT_VALUE(2, PAT_UNCACHED) | + PAT_VALUE(3, PAT_UNCACHEABLE) | + PAT_VALUE(4, PAT_WRITE_BACK) | + PAT_VALUE(5, PAT_WRITE_THROUGH) | + PAT_VALUE(6, PAT_UNCACHED) | + PAT_VALUE(7, PAT_UNCACHEABLE); +} + +/* + * Initialize a virtual machine. + */ +static void * +svm_vminit(struct vm *vm, pmap_t pmap) +{ + struct svm_softc *svm_sc; + struct svm_vcpu *vcpu; + vm_paddr_t msrpm_pa, iopm_pa, pml4_pa; + int i; + + svm_sc = malloc(sizeof (struct svm_softc), M_SVM, M_WAITOK | M_ZERO); + svm_sc->vm = vm; + svm_sc->nptp = (vm_offset_t)vtophys(pmap->pm_pml4); + + /* + * Intercept read and write accesses to all MSRs. + */ + memset(svm_sc->msr_bitmap, 0xFF, sizeof(svm_sc->msr_bitmap)); + + /* + * Access to the following MSRs is redirected to the VMCB when the + * guest is executing. Therefore it is safe to allow the guest to + * read/write these MSRs directly without hypervisor involvement. + */ + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_GSBASE); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_FSBASE); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_KGSBASE); + + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_STAR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_LSTAR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_CSTAR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SF_MASK); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_CS_MSR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_ESP_MSR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_SYSENTER_EIP_MSR); + svm_msr_rw_ok(svm_sc->msr_bitmap, MSR_PAT); + + svm_msr_rd_ok(svm_sc->msr_bitmap, MSR_TSC); + + /* + * Intercept writes to make sure that the EFER_SVM bit is not cleared. + */ + svm_msr_rd_ok(svm_sc->msr_bitmap, MSR_EFER); + + /* Intercept access to all I/O ports. */ + memset(svm_sc->iopm_bitmap, 0xFF, sizeof(svm_sc->iopm_bitmap)); + + iopm_pa = vtophys(svm_sc->iopm_bitmap); + msrpm_pa = vtophys(svm_sc->msr_bitmap); + pml4_pa = svm_sc->nptp; + for (i = 0; i < VM_MAXCPU; i++) { + vcpu = svm_get_vcpu(svm_sc, i); + vcpu->lastcpu = NOCPU; + vcpu->vmcb_pa = vtophys(&vcpu->vmcb); + vmcb_init(svm_sc, i, iopm_pa, msrpm_pa, pml4_pa); + svm_msr_guest_init(svm_sc, i); + } + return (svm_sc); +} + +static int +svm_cpl(struct vmcb_state *state) +{ + + /* + * From APMv2: + * "Retrieve the CPL from the CPL field in the VMCB, not + * from any segment DPL" + */ + return (state->cpl); +} + +static enum vm_cpu_mode +svm_vcpu_mode(struct vmcb *vmcb) +{ + struct vmcb_segment seg; + struct vmcb_state *state; + int error; + + state = &vmcb->state; + + if (state->efer & EFER_LMA) { + error = vmcb_seg(vmcb, VM_REG_GUEST_CS, &seg); + KASSERT(error == 0, ("%s: vmcb_seg(cs) error %d", __func__, + error)); + + /* + * Section 4.8.1 for APM2, check if Code Segment has + * Long attribute set in descriptor. + */ + if (seg.attrib & VMCB_CS_ATTRIB_L) + return (CPU_MODE_64BIT); + else + return (CPU_MODE_COMPATIBILITY); + } else if (state->cr0 & CR0_PE) { + return (CPU_MODE_PROTECTED); + } else { + return (CPU_MODE_REAL); + } +} + +static enum vm_paging_mode +svm_paging_mode(uint64_t cr0, uint64_t cr4, uint64_t efer) +{ + + if ((cr0 & CR0_PG) == 0) + return (PAGING_MODE_FLAT); + if ((cr4 & CR4_PAE) == 0) + return (PAGING_MODE_32); + if (efer & EFER_LME) + return (PAGING_MODE_64); + else + return (PAGING_MODE_PAE); +} + +/* + * ins/outs utility routines + */ +static uint64_t +svm_inout_str_index(struct svm_regctx *regs, int in) +{ + uint64_t val; + + val = in ? regs->sctx_rdi : regs->sctx_rsi; + + return (val); +} + +static uint64_t +svm_inout_str_count(struct svm_regctx *regs, int rep) +{ + uint64_t val; + + val = rep ? regs->sctx_rcx : 1; + + return (val); +} + +static void +svm_inout_str_seginfo(struct svm_softc *svm_sc, int vcpu, int64_t info1, + int in, struct vm_inout_str *vis) +{ + int error, s; + + if (in) { + vis->seg_name = VM_REG_GUEST_ES; + } else { + /* The segment field has standard encoding */ + s = (info1 >> 10) & 0x7; + vis->seg_name = vm_segment_name(s); + } + + error = vmcb_getdesc(svm_sc, vcpu, vis->seg_name, &vis->seg_desc); + KASSERT(error == 0, ("%s: svm_getdesc error %d", __func__, error)); +} + +static int +svm_inout_str_addrsize(uint64_t info1) +{ + uint32_t size; + + size = (info1 >> 7) & 0x7; + switch (size) { + case 1: + return (2); /* 16 bit */ + case 2: + return (4); /* 32 bit */ + case 4: + return (8); /* 64 bit */ + default: + panic("%s: invalid size encoding %d", __func__, size); + } +} + +static void *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***