1 /*- 2 * Copyright (c) 2011 NetApp, Inc. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD$ 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 #include <sys/kernel.h> 34 #include <sys/types.h> 35 #include <sys/systm.h> 36 #include <sys/smp.h> 37 #include <sys/sysctl.h> 38 39 #include <vm/vm.h> 40 #include <vm/pmap.h> 41 #include <vm/vm_extern.h> 42 43 #include <machine/vmm.h> 44 45 #include "vmx_cpufunc.h" 46 #include "vmm_ipi.h" 47 #include "vmx_msr.h" 48 #include "ept.h" 49 50 #define EPT_SUPPORTS_EXEC_ONLY(cap) ((cap) & (1UL << 0)) 51 #define EPT_PWL4(cap) ((cap) & (1UL << 6)) 52 #define EPT_MEMORY_TYPE_WB(cap) ((cap) & (1UL << 14)) 53 #define EPT_PDE_SUPERPAGE(cap) ((cap) & (1UL << 16)) /* 2MB pages */ 54 #define EPT_PDPTE_SUPERPAGE(cap) ((cap) & (1UL << 17)) /* 1GB pages */ 55 #define INVEPT_SUPPORTED(cap) ((cap) & (1UL << 20)) 56 #define AD_BITS_SUPPORTED(cap) ((cap) & (1UL << 21)) 57 #define INVVPID_SUPPORTED(cap) ((cap) & (1UL << 32)) 58 59 #define INVVPID_ALL_TYPES_MASK 0xF0000000000UL 60 #define INVVPID_ALL_TYPES_SUPPORTED(cap) \ 61 (((cap) & INVVPID_ALL_TYPES_MASK) == INVVPID_ALL_TYPES_MASK) 62 63 #define INVEPT_ALL_TYPES_MASK 0x6000000UL 64 #define INVEPT_ALL_TYPES_SUPPORTED(cap) \ 65 (((cap) & INVEPT_ALL_TYPES_MASK) == INVEPT_ALL_TYPES_MASK) 66 67 #define EPT_PWLEVELS 4 /* page walk levels */ 68 #define EPT_ENABLE_AD_BITS (1 << 6) 69 70 SYSCTL_DECL(_hw_vmm); 71 SYSCTL_NODE(_hw_vmm, OID_AUTO, ept, CTLFLAG_RW, NULL, NULL); 72 73 static int ept_enable_ad_bits; 74 75 static int ept_pmap_flags; 76 SYSCTL_INT(_hw_vmm_ept, OID_AUTO, pmap_flags, CTLFLAG_RD, 77 &ept_pmap_flags, 0, NULL); 78 79 int 80 ept_init(int ipinum) 81 { 82 int use_hw_ad_bits, use_superpages, use_exec_only; 83 uint64_t cap; 84 85 cap = rdmsr(MSR_VMX_EPT_VPID_CAP); 86 87 /* 88 * Verify that: 89 * - page walk length is 4 steps 90 * - extended page tables can be laid out in write-back memory 91 * - invvpid instruction with all possible types is supported 92 * - invept instruction with all possible types is supported 93 */ 94 if (!EPT_PWL4(cap) || 95 !EPT_MEMORY_TYPE_WB(cap) || 96 !INVVPID_SUPPORTED(cap) || 97 !INVVPID_ALL_TYPES_SUPPORTED(cap) || 98 !INVEPT_SUPPORTED(cap) || 99 !INVEPT_ALL_TYPES_SUPPORTED(cap)) 100 return (EINVAL); 101 102 ept_pmap_flags = ipinum & PMAP_NESTED_IPIMASK; 103 104 use_superpages = 1; 105 TUNABLE_INT_FETCH("hw.vmm.ept.use_superpages", &use_superpages); 106 if (use_superpages && EPT_PDE_SUPERPAGE(cap)) 107 ept_pmap_flags |= PMAP_PDE_SUPERPAGE; /* 2MB superpage */ 108 109 use_hw_ad_bits = 1; 110 TUNABLE_INT_FETCH("hw.vmm.ept.use_hw_ad_bits", &use_hw_ad_bits); 111 if (use_hw_ad_bits && AD_BITS_SUPPORTED(cap)) 112 ept_enable_ad_bits = 1; 113 else 114 ept_pmap_flags |= PMAP_EMULATE_AD_BITS; 115 116 use_exec_only = 1; 117 TUNABLE_INT_FETCH("hw.vmm.ept.use_exec_only", &use_exec_only); 118 if (use_exec_only && EPT_SUPPORTS_EXEC_ONLY(cap)) 119 ept_pmap_flags |= PMAP_SUPPORTS_EXEC_ONLY; 120 121 return (0); 122 } 123 124 #if 0 125 static void 126 ept_dump(uint64_t *ptp, int nlevels) 127 { 128 int i, t, tabs; 129 uint64_t *ptpnext, ptpval; 130 131 if (--nlevels < 0) 132 return; 133 134 tabs = 3 - nlevels; 135 for (t = 0; t < tabs; t++) 136 printf("\t"); 137 printf("PTP = %p\n", ptp); 138 139 for (i = 0; i < 512; i++) { 140 ptpval = ptp[i]; 141 142 if (ptpval == 0) 143 continue; 144 145 for (t = 0; t < tabs; t++) 146 printf("\t"); 147 printf("%3d 0x%016lx\n", i, ptpval); 148 149 if (nlevels != 0 && (ptpval & EPT_PG_SUPERPAGE) == 0) { 150 ptpnext = (uint64_t *) 151 PHYS_TO_DMAP(ptpval & EPT_ADDR_MASK); 152 ept_dump(ptpnext, nlevels); 153 } 154 } 155 } 156 #endif 157 158 static void 159 invept_single_context(void *arg) 160 { 161 struct invept_desc desc = *(struct invept_desc *)arg; 162 163 invept(INVEPT_TYPE_SINGLE_CONTEXT, desc); 164 } 165 166 void 167 ept_invalidate_mappings(u_long eptp) 168 { 169 struct invept_desc invept_desc = { 0 }; 170 171 invept_desc.eptp = eptp; 172 173 smp_rendezvous(NULL, invept_single_context, NULL, &invept_desc); 174 } 175 176 static int 177 ept_pinit(pmap_t pmap) 178 { 179 180 return (pmap_pinit_type(pmap, PT_EPT, ept_pmap_flags)); 181 } 182 183 struct vmspace * 184 ept_vmspace_alloc(vm_offset_t min, vm_offset_t max) 185 { 186 187 return (vmspace_alloc(min, max, ept_pinit)); 188 } 189 190 void 191 ept_vmspace_free(struct vmspace *vmspace) 192 { 193 194 vmspace_free(vmspace); 195 } 196 197 uint64_t 198 eptp(uint64_t pml4) 199 { 200 uint64_t eptp_val; 201 202 eptp_val = pml4 | (EPT_PWLEVELS - 1) << 3 | PAT_WRITE_BACK; 203 if (ept_enable_ad_bits) 204 eptp_val |= EPT_ENABLE_AD_BITS; 205 206 return (eptp_val); 207 } 208