ept.c revision 276349 
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: stable/10/sys/amd64/vmm/intel/ept.c 276349 2014-12-28 21:27:13Z neel $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/sys/amd64/vmm/intel/ept.c 276349 2014-12-28 21:27:13Z neel $");
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 "ept.h"
48
49#define	EPT_SUPPORTS_EXEC_ONLY(cap)	((cap) & (1UL << 0))
50#define	EPT_PWL4(cap)			((cap) & (1UL << 6))
51#define	EPT_MEMORY_TYPE_WB(cap)		((cap) & (1UL << 14))
52#define	EPT_PDE_SUPERPAGE(cap)		((cap) & (1UL << 16))	/* 2MB pages */
53#define	EPT_PDPTE_SUPERPAGE(cap)	((cap) & (1UL << 17))	/* 1GB pages */
54#define	INVEPT_SUPPORTED(cap)		((cap) & (1UL << 20))
55#define	AD_BITS_SUPPORTED(cap)		((cap) & (1UL << 21))
56#define	INVVPID_SUPPORTED(cap)		((cap) & (1UL << 32))
57
58#define	INVVPID_ALL_TYPES_MASK		0xF0000000000UL
59#define	INVVPID_ALL_TYPES_SUPPORTED(cap)	\
60	(((cap) & INVVPID_ALL_TYPES_MASK) == INVVPID_ALL_TYPES_MASK)
61
62#define	INVEPT_ALL_TYPES_MASK		0x6000000UL
63#define	INVEPT_ALL_TYPES_SUPPORTED(cap)		\
64	(((cap) & INVEPT_ALL_TYPES_MASK) == INVEPT_ALL_TYPES_MASK)
65
66#define	EPT_PWLEVELS		4		/* page walk levels */
67#define	EPT_ENABLE_AD_BITS	(1 << 6)
68
69SYSCTL_DECL(_hw_vmm);
70SYSCTL_NODE(_hw_vmm, OID_AUTO, ept, CTLFLAG_RW, NULL, NULL);
71
72static int ept_enable_ad_bits;
73
74static int ept_pmap_flags;
75SYSCTL_INT(_hw_vmm_ept, OID_AUTO, pmap_flags, CTLFLAG_RD,
76    &ept_pmap_flags, 0, NULL);
77
78int
79ept_init(int ipinum)
80{
81	int use_hw_ad_bits, use_superpages, use_exec_only;
82	uint64_t cap;
83
84	cap = rdmsr(MSR_VMX_EPT_VPID_CAP);
85
86	/*
87	 * Verify that:
88	 * - page walk length is 4 steps
89	 * - extended page tables can be laid out in write-back memory
90	 * - invvpid instruction with all possible types is supported
91	 * - invept instruction with all possible types is supported
92	 */
93	if (!EPT_PWL4(cap) ||
94	    !EPT_MEMORY_TYPE_WB(cap) ||
95	    !INVVPID_SUPPORTED(cap) ||
96	    !INVVPID_ALL_TYPES_SUPPORTED(cap) ||
97	    !INVEPT_SUPPORTED(cap) ||
98	    !INVEPT_ALL_TYPES_SUPPORTED(cap))
99		return (EINVAL);
100
101	ept_pmap_flags = ipinum & PMAP_NESTED_IPIMASK;
102
103	use_superpages = 1;
104	TUNABLE_INT_FETCH("hw.vmm.ept.use_superpages", &use_superpages);
105	if (use_superpages && EPT_PDE_SUPERPAGE(cap))
106		ept_pmap_flags |= PMAP_PDE_SUPERPAGE;	/* 2MB superpage */
107
108	use_hw_ad_bits = 1;
109	TUNABLE_INT_FETCH("hw.vmm.ept.use_hw_ad_bits", &use_hw_ad_bits);
110	if (use_hw_ad_bits && AD_BITS_SUPPORTED(cap))
111		ept_enable_ad_bits = 1;
112	else
113		ept_pmap_flags |= PMAP_EMULATE_AD_BITS;
114
115	use_exec_only = 1;
116	TUNABLE_INT_FETCH("hw.vmm.ept.use_exec_only", &use_exec_only);
117	if (use_exec_only && EPT_SUPPORTS_EXEC_ONLY(cap))
118		ept_pmap_flags |= PMAP_SUPPORTS_EXEC_ONLY;
119
120	return (0);
121}
122
123#if 0
124static void
125ept_dump(uint64_t *ptp, int nlevels)
126{
127	int i, t, tabs;
128	uint64_t *ptpnext, ptpval;
129
130	if (--nlevels < 0)
131		return;
132
133	tabs = 3 - nlevels;
134	for (t = 0; t < tabs; t++)
135		printf("\t");
136	printf("PTP = %p\n", ptp);
137
138	for (i = 0; i < 512; i++) {
139		ptpval = ptp[i];
140
141		if (ptpval == 0)
142			continue;
143
144		for (t = 0; t < tabs; t++)
145			printf("\t");
146		printf("%3d 0x%016lx\n", i, ptpval);
147
148		if (nlevels != 0 && (ptpval & EPT_PG_SUPERPAGE) == 0) {
149			ptpnext = (uint64_t *)
150				  PHYS_TO_DMAP(ptpval & EPT_ADDR_MASK);
151			ept_dump(ptpnext, nlevels);
152		}
153	}
154}
155#endif
156
157static void
158invept_single_context(void *arg)
159{
160	struct invept_desc desc = *(struct invept_desc *)arg;
161
162	invept(INVEPT_TYPE_SINGLE_CONTEXT, desc);
163}
164
165void
166ept_invalidate_mappings(u_long eptp)
167{
168	struct invept_desc invept_desc = { 0 };
169
170	invept_desc.eptp = eptp;
171
172	smp_rendezvous(NULL, invept_single_context, NULL, &invept_desc);
173}
174
175static int
176ept_pinit(pmap_t pmap)
177{
178
179	return (pmap_pinit_type(pmap, PT_EPT, ept_pmap_flags));
180}
181
182struct vmspace *
183ept_vmspace_alloc(vm_offset_t min, vm_offset_t max)
184{
185
186	return (vmspace_alloc(min, max, ept_pinit));
187}
188
189void
190ept_vmspace_free(struct vmspace *vmspace)
191{
192
193	vmspace_free(vmspace);
194}
195
196uint64_t
197eptp(uint64_t pml4)
198{
199	uint64_t eptp_val;
200
201	eptp_val = pml4 | (EPT_PWLEVELS - 1) << 3 | PAT_WRITE_BACK;
202	if (ept_enable_ad_bits)
203		eptp_val |= EPT_ENABLE_AD_BITS;
204
205	return (eptp_val);
206}
207