summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/platform/ring0/pagetables/pagetables.go
blob: 87e88e97d379eaa3ffa7527f4be6d9c04f07911a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
// Copyright 2018 The gVisor Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Package pagetables provides a generic implementation of pagetables.
//
// The core functions must be safe to call from a nosplit context. Furthermore,
// this pagetables implementation goes to lengths to ensure that all functions
// are free from runtime allocation. Calls to NewPTEs/FreePTEs may be made
// during walks, but these can be cached elsewhere if required.
package pagetables

import (
	"gvisor.dev/gvisor/pkg/usermem"
)

// PageTables is a set of page tables.
type PageTables struct {
	// Allocator is used to allocate nodes.
	Allocator Allocator

	// root is the pagetable root.
	root *PTEs

	// rootPhysical is the cached physical address of the root.
	//
	// This is saved only to prevent constant translation.
	rootPhysical uintptr

	// archPageTables includes architecture-specific features.
	archPageTables
}

// New returns new PageTables.
func New(a Allocator) *PageTables {
	p := new(PageTables)
	p.Init(a)
	return p
}

// mapVisitor is used for map.
type mapVisitor struct {
	target   uintptr // Input.
	physical uintptr // Input.
	opts     MapOpts // Input.
	prev     bool    // Output.
}

// visit is used for map.
//
//go:nosplit
func (v *mapVisitor) visit(start uintptr, pte *PTE, align uintptr) {
	p := v.physical + (start - uintptr(v.target))
	if pte.Valid() && (pte.Address() != p || pte.Opts() != v.opts) {
		v.prev = true
	}
	if p&align != 0 {
		// We will install entries at a smaller granulaity if we don't
		// install a valid entry here, however we must zap any existing
		// entry to ensure this happens.
		pte.Clear()
		return
	}
	pte.Set(p, v.opts)
}

//go:nosplit
func (*mapVisitor) requiresAlloc() bool { return true }

//go:nosplit
func (*mapVisitor) requiresSplit() bool { return true }

// Map installs a mapping with the given physical address.
//
// True is returned iff there was a previous mapping in the range.
//
// Precondition: addr & length must be page-aligned, their sum must not overflow.
//
//go:nosplit
func (p *PageTables) Map(addr usermem.Addr, length uintptr, opts MapOpts, physical uintptr) bool {
	if !opts.AccessType.Any() {
		return p.Unmap(addr, length)
	}
	w := mapWalker{
		pageTables: p,
		visitor: mapVisitor{
			target:   uintptr(addr),
			physical: physical,
			opts:     opts,
		},
	}
	w.iterateRange(uintptr(addr), uintptr(addr)+length)
	return w.visitor.prev
}

// unmapVisitor is used for unmap.
type unmapVisitor struct {
	count int
}

//go:nosplit
func (*unmapVisitor) requiresAlloc() bool { return false }

//go:nosplit
func (*unmapVisitor) requiresSplit() bool { return true }

// visit unmaps the given entry.
//
//go:nosplit
func (v *unmapVisitor) visit(start uintptr, pte *PTE, align uintptr) {
	pte.Clear()
	v.count++
}

// Unmap unmaps the given range.
//
// True is returned iff there was a previous mapping in the range.
//
// Precondition: addr & length must be page-aligned.
//
//go:nosplit
func (p *PageTables) Unmap(addr usermem.Addr, length uintptr) bool {
	w := unmapWalker{
		pageTables: p,
		visitor: unmapVisitor{
			count: 0,
		},
	}
	w.iterateRange(uintptr(addr), uintptr(addr)+length)
	return w.visitor.count > 0
}

// emptyVisitor is used for emptiness checks.
type emptyVisitor struct {
	count int
}

//go:nosplit
func (*emptyVisitor) requiresAlloc() bool { return false }

//go:nosplit
func (*emptyVisitor) requiresSplit() bool { return false }

// visit unmaps the given entry.
//
//go:nosplit
func (v *emptyVisitor) visit(start uintptr, pte *PTE, align uintptr) {
	v.count++
}

// IsEmpty checks if the given range is empty.
//
// Precondition: addr & length must be page-aligned.
//
//go:nosplit
func (p *PageTables) IsEmpty(addr usermem.Addr, length uintptr) bool {
	w := emptyWalker{
		pageTables: p,
	}
	w.iterateRange(uintptr(addr), uintptr(addr)+length)
	return w.visitor.count == 0
}

// lookupVisitor is used for lookup.
type lookupVisitor struct {
	target   uintptr // Input.
	physical uintptr // Output.
	opts     MapOpts // Output.
}

// visit matches the given address.
//
//go:nosplit
func (v *lookupVisitor) visit(start uintptr, pte *PTE, align uintptr) {
	if !pte.Valid() {
		return
	}
	v.physical = pte.Address() + (start - uintptr(v.target))
	v.opts = pte.Opts()
}

//go:nosplit
func (*lookupVisitor) requiresAlloc() bool { return false }

//go:nosplit
func (*lookupVisitor) requiresSplit() bool { return false }

// Lookup returns the physical address for the given virtual address.
//
//go:nosplit
func (p *PageTables) Lookup(addr usermem.Addr) (physical uintptr, opts MapOpts) {
	mask := uintptr(usermem.PageSize - 1)
	offset := uintptr(addr) & mask
	w := lookupWalker{
		pageTables: p,
		visitor: lookupVisitor{
			target: uintptr(addr &^ usermem.Addr(mask)),
		},
	}
	w.iterateRange(uintptr(addr), uintptr(addr)+1)
	return w.visitor.physical + offset, w.visitor.opts
}