1 <<
2 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
3 2
4 ========================================= 3 =========================================
5 A vmemmap diet for HugeTLB and Device DAX 4 A vmemmap diet for HugeTLB and Device DAX
6 ========================================= 5 =========================================
7 6
8 HugeTLB 7 HugeTLB
9 ======= 8 =======
10 9
11 This section is to explain how HugeTLB Vmemmap 10 This section is to explain how HugeTLB Vmemmap Optimization (HVO) works.
12 11
13 The ``struct page`` structures are used to des 12 The ``struct page`` structures are used to describe a physical page frame. By
14 default, there is a one-to-one mapping from a !! 13 default, there is a one-to-one mapping from a page frame to it's corresponding
15 ``struct page``. 14 ``struct page``.
16 15
17 HugeTLB pages consist of multiple base page si 16 HugeTLB pages consist of multiple base page size pages and is supported by many
18 architectures. See Documentation/admin-guide/m 17 architectures. See Documentation/admin-guide/mm/hugetlbpage.rst for more
19 details. On the x86-64 architecture, HugeTLB p 18 details. On the x86-64 architecture, HugeTLB pages of size 2MB and 1GB are
20 currently supported. Since the base page size 19 currently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB page
21 consists of 512 base pages and a 1GB HugeTLB p !! 20 consists of 512 base pages and a 1GB HugeTLB page consists of 4096 base pages.
22 For each base page, there is a corresponding ` 21 For each base page, there is a corresponding ``struct page``.
23 22
24 Within the HugeTLB subsystem, only the first 4 23 Within the HugeTLB subsystem, only the first 4 ``struct page`` are used to
25 contain unique information about a HugeTLB pag 24 contain unique information about a HugeTLB page. ``__NR_USED_SUBPAGE`` provides
26 this upper limit. The only 'useful' informatio 25 this upper limit. The only 'useful' information in the remaining ``struct page``
27 is the compound_head field, and this field is 26 is the compound_head field, and this field is the same for all tail pages.
28 27
29 By removing redundant ``struct page`` for Huge 28 By removing redundant ``struct page`` for HugeTLB pages, memory can be returned
30 to the buddy allocator for other uses. 29 to the buddy allocator for other uses.
31 30
32 Different architectures support different Huge 31 Different architectures support different HugeTLB pages. For example, the
33 following table is the HugeTLB page size suppo 32 following table is the HugeTLB page size supported by x86 and arm64
34 architectures. Because arm64 supports 4k, 16k, 33 architectures. Because arm64 supports 4k, 16k, and 64k base pages and
35 supports contiguous entries, so it supports ma 34 supports contiguous entries, so it supports many kinds of sizes of HugeTLB
36 page. 35 page.
37 36
38 +--------------+-----------+------------------ 37 +--------------+-----------+-----------------------------------------------+
39 | Architecture | Page Size | Hu 38 | Architecture | Page Size | HugeTLB Page Size |
40 +--------------+-----------+-----------+------ 39 +--------------+-----------+-----------+-----------+-----------+-----------+
41 | x86-64 | 4KB | 2MB | 1G 40 | x86-64 | 4KB | 2MB | 1GB | | |
42 +--------------+-----------+-----------+------ 41 +--------------+-----------+-----------+-----------+-----------+-----------+
43 | | 4KB | 64KB | 2M 42 | | 4KB | 64KB | 2MB | 32MB | 1GB |
44 | +-----------+-----------+------ 43 | +-----------+-----------+-----------+-----------+-----------+
45 | arm64 | 16KB | 2MB | 32M 44 | arm64 | 16KB | 2MB | 32MB | 1GB | |
46 | +-----------+-----------+------ 45 | +-----------+-----------+-----------+-----------+-----------+
47 | | 64KB | 2MB | 512M 46 | | 64KB | 2MB | 512MB | 16GB | |
48 +--------------+-----------+-----------+------ 47 +--------------+-----------+-----------+-----------+-----------+-----------+
49 48
50 When the system boot up, every HugeTLB page ha 49 When the system boot up, every HugeTLB page has more than one ``struct page``
51 structs which size is (unit: pages):: 50 structs which size is (unit: pages)::
52 51
53 struct_size = HugeTLB_Size / PAGE_SIZE * si 52 struct_size = HugeTLB_Size / PAGE_SIZE * sizeof(struct page) / PAGE_SIZE
54 53
55 Where HugeTLB_Size is the size of the HugeTLB 54 Where HugeTLB_Size is the size of the HugeTLB page. We know that the size
56 of the HugeTLB page is always n times PAGE_SIZ 55 of the HugeTLB page is always n times PAGE_SIZE. So we can get the following
57 relationship:: 56 relationship::
58 57
59 HugeTLB_Size = n * PAGE_SIZE 58 HugeTLB_Size = n * PAGE_SIZE
60 59
61 Then:: 60 Then::
62 61
63 struct_size = n * PAGE_SIZE / PAGE_SIZE * s 62 struct_size = n * PAGE_SIZE / PAGE_SIZE * sizeof(struct page) / PAGE_SIZE
64 = n * sizeof(struct page) / PAG 63 = n * sizeof(struct page) / PAGE_SIZE
65 64
66 We can use huge mapping at the pud/pmd level f 65 We can use huge mapping at the pud/pmd level for the HugeTLB page.
67 66
68 For the HugeTLB page of the pmd level mapping, 67 For the HugeTLB page of the pmd level mapping, then::
69 68
70 struct_size = n * sizeof(struct page) / PAG 69 struct_size = n * sizeof(struct page) / PAGE_SIZE
71 = PAGE_SIZE / sizeof(pte_t) * s 70 = PAGE_SIZE / sizeof(pte_t) * sizeof(struct page) / PAGE_SIZE
72 = sizeof(struct page) / sizeof( 71 = sizeof(struct page) / sizeof(pte_t)
73 = 64 / 8 72 = 64 / 8
74 = 8 (pages) 73 = 8 (pages)
75 74
76 Where n is how many pte entries which one page 75 Where n is how many pte entries which one page can contains. So the value of
77 n is (PAGE_SIZE / sizeof(pte_t)). 76 n is (PAGE_SIZE / sizeof(pte_t)).
78 77
79 This optimization only supports 64-bit system, 78 This optimization only supports 64-bit system, so the value of sizeof(pte_t)
80 is 8. And this optimization also applicable on 79 is 8. And this optimization also applicable only when the size of ``struct page``
81 is a power of two. In most cases, the size of 80 is a power of two. In most cases, the size of ``struct page`` is 64 bytes (e.g.
82 x86-64 and arm64). So if we use pmd level mapp 81 x86-64 and arm64). So if we use pmd level mapping for a HugeTLB page, the
83 size of ``struct page`` structs of it is 8 pag 82 size of ``struct page`` structs of it is 8 page frames which size depends on the
84 size of the base page. 83 size of the base page.
85 84
86 For the HugeTLB page of the pud level mapping, 85 For the HugeTLB page of the pud level mapping, then::
87 86
88 struct_size = PAGE_SIZE / sizeof(pmd_t) * s 87 struct_size = PAGE_SIZE / sizeof(pmd_t) * struct_size(pmd)
89 = PAGE_SIZE / 8 * 8 (pages) 88 = PAGE_SIZE / 8 * 8 (pages)
90 = PAGE_SIZE (pages) 89 = PAGE_SIZE (pages)
91 90
92 Where the struct_size(pmd) is the size of the 91 Where the struct_size(pmd) is the size of the ``struct page`` structs of a
93 HugeTLB page of the pmd level mapping. 92 HugeTLB page of the pmd level mapping.
94 93
95 E.g.: A 2MB HugeTLB page on x86_64 consists in 94 E.g.: A 2MB HugeTLB page on x86_64 consists in 8 page frames while 1GB
96 HugeTLB page consists in 4096. 95 HugeTLB page consists in 4096.
97 96
98 Next, we take the pmd level mapping of the Hug 97 Next, we take the pmd level mapping of the HugeTLB page as an example to
99 show the internal implementation of this optim 98 show the internal implementation of this optimization. There are 8 pages
100 ``struct page`` structs associated with a Huge 99 ``struct page`` structs associated with a HugeTLB page which is pmd mapped.
101 100
102 Here is how things look before optimization:: 101 Here is how things look before optimization::
103 102
104 HugeTLB struct pages(8 pa 103 HugeTLB struct pages(8 pages) page frame(8 pages)
105 +-----------+ ---virt_to_page---> +---------- 104 +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+
106 | | | 0 105 | | | 0 | -------------> | 0 |
107 | | +---------- 106 | | +-----------+ +-----------+
108 | | | 1 107 | | | 1 | -------------> | 1 |
109 | | +---------- 108 | | +-----------+ +-----------+
110 | | | 2 109 | | | 2 | -------------> | 2 |
111 | | +---------- 110 | | +-----------+ +-----------+
112 | | | 3 111 | | | 3 | -------------> | 3 |
113 | | +---------- 112 | | +-----------+ +-----------+
114 | | | 4 113 | | | 4 | -------------> | 4 |
115 | PMD | +---------- 114 | PMD | +-----------+ +-----------+
116 | level | | 5 115 | level | | 5 | -------------> | 5 |
117 | mapping | +---------- 116 | mapping | +-----------+ +-----------+
118 | | | 6 117 | | | 6 | -------------> | 6 |
119 | | +---------- 118 | | +-----------+ +-----------+
120 | | | 7 119 | | | 7 | -------------> | 7 |
121 | | +---------- 120 | | +-----------+ +-----------+
122 | | 121 | |
123 | | 122 | |
124 | | 123 | |
125 +-----------+ 124 +-----------+
126 125
127 The value of page->compound_head is the same f 126 The value of page->compound_head is the same for all tail pages. The first
128 page of ``struct page`` (page 0) associated wi 127 page of ``struct page`` (page 0) associated with the HugeTLB page contains the 4
129 ``struct page`` necessary to describe the Huge 128 ``struct page`` necessary to describe the HugeTLB. The only use of the remaining
130 pages of ``struct page`` (page 1 to page 7) is 129 pages of ``struct page`` (page 1 to page 7) is to point to page->compound_head.
131 Therefore, we can remap pages 1 to 7 to page 0 130 Therefore, we can remap pages 1 to 7 to page 0. Only 1 page of ``struct page``
132 will be used for each HugeTLB page. This will 131 will be used for each HugeTLB page. This will allow us to free the remaining
133 7 pages to the buddy allocator. 132 7 pages to the buddy allocator.
134 133
135 Here is how things look after remapping:: 134 Here is how things look after remapping::
136 135
137 HugeTLB struct pages(8 pa 136 HugeTLB struct pages(8 pages) page frame(8 pages)
138 +-----------+ ---virt_to_page---> +---------- 137 +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+
139 | | | 0 138 | | | 0 | -------------> | 0 |
140 | | +---------- 139 | | +-----------+ +-----------+
141 | | | 1 140 | | | 1 | ---------------^ ^ ^ ^ ^ ^ ^
142 | | +---------- 141 | | +-----------+ | | | | | |
143 | | | 2 142 | | | 2 | -----------------+ | | | | |
144 | | +---------- 143 | | +-----------+ | | | | |
145 | | | 3 144 | | | 3 | -------------------+ | | | |
146 | | +---------- 145 | | +-----------+ | | | |
147 | | | 4 146 | | | 4 | ---------------------+ | | |
148 | PMD | +---------- 147 | PMD | +-----------+ | | |
149 | level | | 5 148 | level | | 5 | -----------------------+ | |
150 | mapping | +---------- 149 | mapping | +-----------+ | |
151 | | | 6 150 | | | 6 | -------------------------+ |
152 | | +---------- 151 | | +-----------+ |
153 | | | 7 152 | | | 7 | ---------------------------+
154 | | +---------- 153 | | +-----------+
155 | | 154 | |
156 | | 155 | |
157 | | 156 | |
158 +-----------+ 157 +-----------+
159 158
160 When a HugeTLB is freed to the buddy system, w 159 When a HugeTLB is freed to the buddy system, we should allocate 7 pages for
161 vmemmap pages and restore the previous mapping 160 vmemmap pages and restore the previous mapping relationship.
162 161
163 For the HugeTLB page of the pud level mapping. 162 For the HugeTLB page of the pud level mapping. It is similar to the former.
164 We also can use this approach to free (PAGE_SI 163 We also can use this approach to free (PAGE_SIZE - 1) vmemmap pages.
165 164
166 Apart from the HugeTLB page of the pmd/pud lev 165 Apart from the HugeTLB page of the pmd/pud level mapping, some architectures
167 (e.g. aarch64) provides a contiguous bit in th 166 (e.g. aarch64) provides a contiguous bit in the translation table entries
168 that hints to the MMU to indicate that it is o 167 that hints to the MMU to indicate that it is one of a contiguous set of
169 entries that can be cached in a single TLB ent 168 entries that can be cached in a single TLB entry.
170 169
171 The contiguous bit is used to increase the map 170 The contiguous bit is used to increase the mapping size at the pmd and pte
172 (last) level. So this type of HugeTLB page can 171 (last) level. So this type of HugeTLB page can be optimized only when its
173 size of the ``struct page`` structs is greater 172 size of the ``struct page`` structs is greater than **1** page.
174 173
175 Notice: The head vmemmap page is not freed to 174 Notice: The head vmemmap page is not freed to the buddy allocator and all
176 tail vmemmap pages are mapped to the head vmem 175 tail vmemmap pages are mapped to the head vmemmap page frame. So we can see
177 more than one ``struct page`` struct with ``PG 176 more than one ``struct page`` struct with ``PG_head`` (e.g. 8 per 2 MB HugeTLB
178 page) associated with each HugeTLB page. The ` 177 page) associated with each HugeTLB page. The ``compound_head()`` can handle
179 this correctly. There is only **one** head ``s 178 this correctly. There is only **one** head ``struct page``, the tail
180 ``struct page`` with ``PG_head`` are fake head 179 ``struct page`` with ``PG_head`` are fake head ``struct page``. We need an
181 approach to distinguish between those two diff 180 approach to distinguish between those two different types of ``struct page`` so
182 that ``compound_head()`` can return the real h 181 that ``compound_head()`` can return the real head ``struct page`` when the
183 parameter is the tail ``struct page`` but with !! 182 parameter is the tail ``struct page`` but with ``PG_head``. The following code
>> 183 snippet describes how to distinguish between real and fake head ``struct page``.
>> 184
>> 185 .. code-block:: c
>> 186
>> 187 if (test_bit(PG_head, &page->flags)) {
>> 188 unsigned long head = READ_ONCE(page[1].compound_head);
>> 189
>> 190 if (head & 1) {
>> 191 if (head == (unsigned long)page + 1)
>> 192 /* head struct page */
>> 193 else
>> 194 /* tail struct page */
>> 195 } else {
>> 196 /* head struct page */
>> 197 }
>> 198 }
>> 199
>> 200 We can safely access the field of the **page[1]** with ``PG_head`` because the
>> 201 page is a compound page composed with at least two contiguous pages.
>> 202 The implementation refers to ``page_fixed_fake_head()``.
184 203
185 Device DAX 204 Device DAX
186 ========== 205 ==========
187 206
188 The device-dax interface uses the same tail de 207 The device-dax interface uses the same tail deduplication technique explained
189 in the previous chapter, except when used with 208 in the previous chapter, except when used with the vmemmap in
190 the device (altmap). 209 the device (altmap).
191 210
192 The following page sizes are supported in DAX: 211 The following page sizes are supported in DAX: PAGE_SIZE (4K on x86_64),
193 PMD_SIZE (2M on x86_64) and PUD_SIZE (1G on x8 212 PMD_SIZE (2M on x86_64) and PUD_SIZE (1G on x86_64).
194 For powerpc equivalent details see Documentati <<
195 213
196 The differences with HugeTLB are relatively mi 214 The differences with HugeTLB are relatively minor.
197 215
198 It only use 3 ``struct page`` for storing all 216 It only use 3 ``struct page`` for storing all information as opposed
199 to 4 on HugeTLB pages. 217 to 4 on HugeTLB pages.
200 218
201 There's no remapping of vmemmap given that dev 219 There's no remapping of vmemmap given that device-dax memory is not part of
202 System RAM ranges initialized at boot. Thus th 220 System RAM ranges initialized at boot. Thus the tail page deduplication
203 happens at a later stage when we populate the 221 happens at a later stage when we populate the sections. HugeTLB reuses the
204 the head vmemmap page representing, whereas de 222 the head vmemmap page representing, whereas device-dax reuses the tail
205 vmemmap page. This results in only half of the 223 vmemmap page. This results in only half of the savings compared to HugeTLB.
206 224
207 Deduplicated tail pages are not mapped read-on 225 Deduplicated tail pages are not mapped read-only.
208 226
209 Here's how things look like on device-dax afte 227 Here's how things look like on device-dax after the sections are populated::
210 228
211 +-----------+ ---virt_to_page---> +---------- 229 +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+
212 | | | 0 230 | | | 0 | -------------> | 0 |
213 | | +---------- 231 | | +-----------+ +-----------+
214 | | | 1 232 | | | 1 | -------------> | 1 |
215 | | +---------- 233 | | +-----------+ +-----------+
216 | | | 2 234 | | | 2 | ----------------^ ^ ^ ^ ^ ^
217 | | +---------- 235 | | +-----------+ | | | | |
218 | | | 3 236 | | | 3 | ------------------+ | | | |
219 | | +---------- 237 | | +-----------+ | | | |
220 | | | 4 238 | | | 4 | --------------------+ | | |
221 | PMD | +---------- 239 | PMD | +-----------+ | | |
222 | level | | 5 240 | level | | 5 | ----------------------+ | |
223 | mapping | +---------- 241 | mapping | +-----------+ | |
224 | | | 6 242 | | | 6 | ------------------------+ |
225 | | +---------- 243 | | +-----------+ |
226 | | | 7 244 | | | 7 | --------------------------+
227 | | +---------- 245 | | +-----------+
228 | | 246 | |
229 | | 247 | |
230 | | 248 | |
231 +-----------+ 249 +-----------+
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.