2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/types.h>
35 #include <sys/systm.h>
37 #include <sys/sysctl.h>
41 #include <vm/vm_extern.h>
43 #include <machine/vmm.h>
45 #include "vmx_cpufunc.h"
48 #define EPT_SUPPORTS_EXEC_ONLY(cap) ((cap) & (1UL << 0))
49 #define EPT_PWL4(cap) ((cap) & (1UL << 6))
50 #define EPT_MEMORY_TYPE_WB(cap) ((cap) & (1UL << 14))
51 #define EPT_PDE_SUPERPAGE(cap) ((cap) & (1UL << 16)) /* 2MB pages */
52 #define EPT_PDPTE_SUPERPAGE(cap) ((cap) & (1UL << 17)) /* 1GB pages */
53 #define INVEPT_SUPPORTED(cap) ((cap) & (1UL << 20))
54 #define AD_BITS_SUPPORTED(cap) ((cap) & (1UL << 21))
55 #define INVVPID_SUPPORTED(cap) ((cap) & (1UL << 32))
57 #define INVVPID_ALL_TYPES_MASK 0xF0000000000UL
58 #define INVVPID_ALL_TYPES_SUPPORTED(cap) \
59 (((cap) & INVVPID_ALL_TYPES_MASK) == INVVPID_ALL_TYPES_MASK)
61 #define INVEPT_ALL_TYPES_MASK 0x6000000UL
62 #define INVEPT_ALL_TYPES_SUPPORTED(cap) \
63 (((cap) & INVEPT_ALL_TYPES_MASK) == INVEPT_ALL_TYPES_MASK)
65 #define EPT_PWLEVELS 4 /* page walk levels */
66 #define EPT_ENABLE_AD_BITS (1 << 6)
69 SYSCTL_NODE(_hw_vmm, OID_AUTO, ept, CTLFLAG_RW, NULL, NULL);
71 static int ept_enable_ad_bits;
73 static int ept_pmap_flags;
74 SYSCTL_INT(_hw_vmm_ept, OID_AUTO, pmap_flags, CTLFLAG_RD,
75 &ept_pmap_flags, 0, NULL);
80 int use_hw_ad_bits, use_superpages, use_exec_only;
83 cap = rdmsr(MSR_VMX_EPT_VPID_CAP);
87 * - page walk length is 4 steps
88 * - extended page tables can be laid out in write-back memory
89 * - invvpid instruction with all possible types is supported
90 * - invept instruction with all possible types is supported
93 !EPT_MEMORY_TYPE_WB(cap) ||
94 !INVVPID_SUPPORTED(cap) ||
95 !INVVPID_ALL_TYPES_SUPPORTED(cap) ||
96 !INVEPT_SUPPORTED(cap) ||
97 !INVEPT_ALL_TYPES_SUPPORTED(cap))
100 ept_pmap_flags = ipinum & PMAP_NESTED_IPIMASK;
103 TUNABLE_INT_FETCH("hw.vmm.ept.use_superpages", &use_superpages);
104 if (use_superpages && EPT_PDE_SUPERPAGE(cap))
105 ept_pmap_flags |= PMAP_PDE_SUPERPAGE; /* 2MB superpage */
108 TUNABLE_INT_FETCH("hw.vmm.ept.use_hw_ad_bits", &use_hw_ad_bits);
109 if (use_hw_ad_bits && AD_BITS_SUPPORTED(cap))
110 ept_enable_ad_bits = 1;
112 ept_pmap_flags |= PMAP_EMULATE_AD_BITS;
115 TUNABLE_INT_FETCH("hw.vmm.ept.use_exec_only", &use_exec_only);
116 if (use_exec_only && EPT_SUPPORTS_EXEC_ONLY(cap))
117 ept_pmap_flags |= PMAP_SUPPORTS_EXEC_ONLY;
124 ept_dump(uint64_t *ptp, int nlevels)
127 uint64_t *ptpnext, ptpval;
133 for (t = 0; t < tabs; t++)
135 printf("PTP = %p\n", ptp);
137 for (i = 0; i < 512; i++) {
143 for (t = 0; t < tabs; t++)
145 printf("%3d 0x%016lx\n", i, ptpval);
147 if (nlevels != 0 && (ptpval & EPT_PG_SUPERPAGE) == 0) {
148 ptpnext = (uint64_t *)
149 PHYS_TO_DMAP(ptpval & EPT_ADDR_MASK);
150 ept_dump(ptpnext, nlevels);
157 invept_single_context(void *arg)
159 struct invept_desc desc = *(struct invept_desc *)arg;
161 invept(INVEPT_TYPE_SINGLE_CONTEXT, desc);
165 ept_invalidate_mappings(u_long eptp)
167 struct invept_desc invept_desc = { 0 };
169 invept_desc.eptp = eptp;
171 smp_rendezvous(NULL, invept_single_context, NULL, &invept_desc);
175 ept_pinit(pmap_t pmap)
178 return (pmap_pinit_type(pmap, PT_EPT, ept_pmap_flags));
182 ept_vmspace_alloc(vm_offset_t min, vm_offset_t max)
185 return (vmspace_alloc(min, max, ept_pinit));
189 ept_vmspace_free(struct vmspace *vmspace)
192 vmspace_free(vmspace);
200 eptp_val = pml4 | (EPT_PWLEVELS - 1) << 3 | PAT_WRITE_BACK;
201 if (ept_enable_ad_bits)
202 eptp_val |= EPT_ENABLE_AD_BITS;