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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
|
package pagetables
// Walker walks page tables.
type emptyWalker struct {
// pageTables are the tables to walk.
pageTables *PageTables
// Visitor is the set of arguments.
visitor emptyVisitor
}
// iterateRange iterates over all appropriate levels of page tables for the given range.
//
// If requiresAlloc is true, then Set _must_ be called on all given PTEs. The
// exception is super pages. If a valid super page (huge or jumbo) cannot be
// installed, then the walk will continue to individual entries.
//
// This algorithm will attempt to maximize the use of super pages whenever
// possible. Whether a super page is provided will be clear through the range
// provided in the callback.
//
// Note that if requiresAlloc is true, then no gaps will be present. However,
// if alloc is not set, then the iteration will likely be full of gaps.
//
// Note that this function should generally be avoided in favor of Map, Unmap,
// etc. when not necessary.
//
// Precondition: start must be page-aligned.
//
// Precondition: start must be less than end.
//
// Precondition: If requiresAlloc is true, then start and end should not span
// non-canonical ranges. If they do, a panic will result.
//
//go:nosplit
func (w *emptyWalker) iterateRange(start, end uintptr) {
if start%pteSize != 0 {
panic("unaligned start")
}
if end < start {
panic("start > end")
}
if start < lowerTop {
if end <= lowerTop {
w.iterateRangeCanonical(start, end)
} else if end > lowerTop && end <= upperBottom {
if w.visitor.requiresAlloc() {
panic("alloc spans non-canonical range")
}
w.iterateRangeCanonical(start, lowerTop)
} else {
if w.visitor.requiresAlloc() {
panic("alloc spans non-canonical range")
}
w.iterateRangeCanonical(start, lowerTop)
w.iterateRangeCanonical(upperBottom, end)
}
} else if start < upperBottom {
if end <= upperBottom {
if w.visitor.requiresAlloc() {
panic("alloc spans non-canonical range")
}
} else {
if w.visitor.requiresAlloc() {
panic("alloc spans non-canonical range")
}
w.iterateRangeCanonical(upperBottom, end)
}
} else {
w.iterateRangeCanonical(start, end)
}
}
// next returns the next address quantized by the given size.
//
//go:nosplit
func emptynext(start uintptr, size uintptr) uintptr {
start &= ^(size - 1)
start += size
return start
}
// iterateRangeCanonical walks a canonical range.
//
//go:nosplit
func (w *emptyWalker) iterateRangeCanonical(start, end uintptr) {
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() {
start = emptynext(start, pgdSize)
continue
}
pudEntries = w.pageTables.Allocator.NewPTEs()
pgdEntry.setPageTable(w.pageTables, pudEntries)
} else {
pudEntries = w.pageTables.Allocator.LookupPTEs(pgdEntry.Address())
}
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() {
clearPUDEntries++
start = emptynext(start, pudSize)
continue
}
if start&(pudSize-1) == 0 && end-start >= pudSize {
pudEntry.SetSuper()
w.visitor.visit(uintptr(start), pudEntry, pudSize-1)
if pudEntry.Valid() {
start = emptynext(start, pudSize)
continue
}
}
pmdEntries = w.pageTables.Allocator.NewPTEs()
pudEntry.setPageTable(w.pageTables, pmdEntries)
} else if pudEntry.IsSuper() {
if w.visitor.requiresSplit() && (start&(pudSize-1) != 0 || end < emptynext(start, pudSize)) {
pmdEntries = w.pageTables.Allocator.NewPTEs()
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 {
w.visitor.visit(uintptr(start), pudEntry, pudSize-1)
if !pudEntry.Valid() {
clearPUDEntries++
}
start = emptynext(start, pudSize)
continue
}
} else {
pmdEntries = w.pageTables.Allocator.LookupPTEs(pudEntry.Address())
}
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() {
clearPMDEntries++
start = emptynext(start, pmdSize)
continue
}
if start&(pmdSize-1) == 0 && end-start >= pmdSize {
pmdEntry.SetSuper()
w.visitor.visit(uintptr(start), pmdEntry, pmdSize-1)
if pmdEntry.Valid() {
start = emptynext(start, pmdSize)
continue
}
}
pteEntries = w.pageTables.Allocator.NewPTEs()
pmdEntry.setPageTable(w.pageTables, pteEntries)
} else if pmdEntry.IsSuper() {
if w.visitor.requiresSplit() && (start&(pmdSize-1) != 0 || end < emptynext(start, pmdSize)) {
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 {
w.visitor.visit(uintptr(start), pmdEntry, pmdSize-1)
if !pmdEntry.Valid() {
clearPMDEntries++
}
start = emptynext(start, pmdSize)
continue
}
} else {
pteEntries = w.pageTables.Allocator.LookupPTEs(pmdEntry.Address())
}
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
}
w.visitor.visit(uintptr(start), pteEntry, pteSize-1)
if !pteEntry.Valid() {
if w.visitor.requiresAlloc() {
panic("PTE not set after iteration with requiresAlloc!")
}
clearPTEEntries++
}
start += pteSize
continue
}
if clearPTEEntries == entriesPerPage {
pmdEntry.Clear()
w.pageTables.Allocator.FreePTEs(pteEntries)
clearPMDEntries++
}
}
if clearPMDEntries == entriesPerPage {
pudEntry.Clear()
w.pageTables.Allocator.FreePTEs(pmdEntries)
clearPUDEntries++
}
}
if clearPUDEntries == entriesPerPage {
pgdEntry.Clear()
w.pageTables.Allocator.FreePTEs(pudEntries)
}
}
}
|
