summaryrefslogtreecommitdiffhomepage
path: root/pkg/ring0/pagetables/walker_amd64.go
blob: eb4fbcc31813206556ce309b22090a400cc8d76d (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
213
214
215
216
217
218
219
220
221
// 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.

// +build amd64

package pagetables

// iterateRangeCanonical walks a canonical range.
//
//go:nosplit
func (w *Walker) iterateRangeCanonical(start, end uintptr) bool {
	for pgdIndex := uint16((start & pgdMask) >> pgdShift); start < end && pgdIndex < entriesPerPage; pgdIndex++ {
		var (
			pgdEntry   = &w.pageTables.root[pgdIndex]
			pudEntries *PTEs
		)
		if !pgdEntry.Valid() {
			if !w.visitor.requiresAlloc() {
				// Skip over this entry.
				start = next(start, pgdSize)
				continue
			}

			// Allocate a new pgd.
			pudEntries = w.pageTables.Allocator.NewPTEs() // escapes: depends on allocator.
			pgdEntry.setPageTable(w.pageTables, pudEntries)
		} else {
			pudEntries = w.pageTables.Allocator.LookupPTEs(pgdEntry.Address()) // escapes: see above.
		}

		// Map the next level.
		clearPUDEntries := uint16(0)

		for pudIndex := uint16((start & pudMask) >> pudShift); start < end && pudIndex < entriesPerPage; pudIndex++ {
			var (
				pudEntry   = &pudEntries[pudIndex]
				pmdEntries *PTEs
			)
			if !pudEntry.Valid() {
				if !w.visitor.requiresAlloc() {
					// Skip over this entry.
					clearPUDEntries++
					start = next(start, pudSize)
					continue
				}

				// This level has 1-GB super pages. Is this
				// entire region at least as large as a single
				// PUD entry?  If so, we can skip allocating a
				// new page for the pmd.
				if start&(pudSize-1) == 0 && end-start >= pudSize {
					pudEntry.SetSuper()
					if !w.visitor.visit(uintptr(start&^(pudSize-1)), pudEntry, pudSize-1) {
						return false
					}
					if pudEntry.Valid() {
						start = next(start, pudSize)
						continue
					}
				}

				// Allocate a new pud.
				pmdEntries = w.pageTables.Allocator.NewPTEs() // escapes: see above.
				pudEntry.setPageTable(w.pageTables, pmdEntries)

			} else if pudEntry.IsSuper() {
				// Does this page need to be split?
				if w.visitor.requiresSplit() && (start&(pudSize-1) != 0 || end < next(start, pudSize)) {
					// Install the relevant entries.
					pmdEntries = w.pageTables.Allocator.NewPTEs() // escapes: see above.
					for index := uint16(0); index < entriesPerPage; index++ {
						pmdEntries[index].SetSuper()
						pmdEntries[index].Set(
							pudEntry.Address()+(pmdSize*uintptr(index)),
							pudEntry.Opts())
					}
					pudEntry.setPageTable(w.pageTables, pmdEntries)
				} else {
					// A super page to be checked directly.
					if !w.visitor.visit(uintptr(start&^(pudSize-1)), pudEntry, pudSize-1) {
						return false
					}

					// Might have been cleared.
					if !pudEntry.Valid() {
						clearPUDEntries++
					}

					// Note that the super page was changed.
					start = next(start, pudSize)
					continue
				}
			} else {
				pmdEntries = w.pageTables.Allocator.LookupPTEs(pudEntry.Address()) // escapes: see above.
			}

			// Map the next level, since this is valid.
			clearPMDEntries := uint16(0)

			for pmdIndex := uint16((start & pmdMask) >> pmdShift); start < end && pmdIndex < entriesPerPage; pmdIndex++ {
				var (
					pmdEntry   = &pmdEntries[pmdIndex]
					pteEntries *PTEs
				)
				if !pmdEntry.Valid() {
					if !w.visitor.requiresAlloc() {
						// Skip over this entry.
						clearPMDEntries++
						start = next(start, pmdSize)
						continue
					}

					// This level has 2-MB huge pages. If this
					// region is contined in a single PMD entry?
					// As above, we can skip allocating a new page.
					if start&(pmdSize-1) == 0 && end-start >= pmdSize {
						pmdEntry.SetSuper()
						if !w.visitor.visit(uintptr(start&^(pmdSize-1)), pmdEntry, pmdSize-1) {
							return false
						}
						if pmdEntry.Valid() {
							start = next(start, pmdSize)
							continue
						}
					}

					// Allocate a new pmd.
					pteEntries = w.pageTables.Allocator.NewPTEs() // escapes: see above.
					pmdEntry.setPageTable(w.pageTables, pteEntries)

				} else if pmdEntry.IsSuper() {
					// Does this page need to be split?
					if w.visitor.requiresSplit() && (start&(pmdSize-1) != 0 || end < next(start, pmdSize)) {
						// Install the relevant entries.
						pteEntries = w.pageTables.Allocator.NewPTEs()
						for index := uint16(0); index < entriesPerPage; index++ {
							pteEntries[index].Set(
								pmdEntry.Address()+(pteSize*uintptr(index)),
								pmdEntry.Opts())
						}
						pmdEntry.setPageTable(w.pageTables, pteEntries)
					} else {
						// A huge page to be checked directly.
						if !w.visitor.visit(uintptr(start&^(pmdSize-1)), pmdEntry, pmdSize-1) {
							return false
						}

						// Might have been cleared.
						if !pmdEntry.Valid() {
							clearPMDEntries++
						}

						// Note that the huge page was changed.
						start = next(start, pmdSize)
						continue
					}
				} else {
					pteEntries = w.pageTables.Allocator.LookupPTEs(pmdEntry.Address()) // escapes: see above.
				}

				// Map the next level, since this is valid.
				clearPTEEntries := uint16(0)

				for pteIndex := uint16((start & pteMask) >> pteShift); start < end && pteIndex < entriesPerPage; pteIndex++ {
					var (
						pteEntry = &pteEntries[pteIndex]
					)
					if !pteEntry.Valid() && !w.visitor.requiresAlloc() {
						clearPTEEntries++
						start += pteSize
						continue
					}

					// At this point, we are guaranteed that start%pteSize == 0.
					if !w.visitor.visit(uintptr(start&^(pteSize-1)), pteEntry, pteSize-1) {
						return false
					}
					if !pteEntry.Valid() && !w.visitor.requiresAlloc() {
						clearPTEEntries++
					}

					// Note that the pte was changed.
					start += pteSize
					continue
				}

				// Check if we no longer need this page.
				if clearPTEEntries == entriesPerPage {
					pmdEntry.Clear()
					w.pageTables.Allocator.FreePTEs(pteEntries) // escapes: see above.
					clearPMDEntries++
				}
			}

			// Check if we no longer need this page.
			if clearPMDEntries == entriesPerPage {
				pudEntry.Clear()
				w.pageTables.Allocator.FreePTEs(pmdEntries) // escapes: see above.
				clearPUDEntries++
			}
		}

		// Check if we no longer need this page.
		if clearPUDEntries == entriesPerPage {
			pgdEntry.Clear()
			w.pageTables.Allocator.FreePTEs(pudEntries) // escapes: see above.
		}
	}
	return true
}