Rev 17 | Rev 26 | Go to most recent revision | Only display areas with differences | Ignore whitespace | Details | Blame | Last modification | View Log | RSS feed
Rev 17 | Rev 24 | ||
---|---|---|---|
1 | <?xml version="1.0" encoding="UTF-8"?> |
1 | <?xml version="1.0" encoding="UTF-8"?> |
2 | <chapter id="mm"> |
2 | <chapter id="mm"> |
3 | <?dbhtml filename="mm.html"?> |
3 | <?dbhtml filename="mm.html"?> |
4 | 4 | ||
5 | <title>Memory management</title> |
5 | <title>Memory management</title> |
6 | 6 | ||
- | 7 | ||
- | 8 | ||
7 | <section> |
9 | <section><!-- VM --> |
8 | <title>Virtual memory management</title> |
10 | <title>Virtual memory management</title> |
9 | 11 | ||
10 | <section> |
12 | <section> |
11 | <title>Address spaces</title> |
13 | <title>Address spaces</title> |
12 | 14 | ||
13 | <para></para> |
15 | <para></para> |
14 | </section> |
16 | </section> |
15 | 17 | ||
16 | <section> |
18 | <section> |
17 | <title>Virtual address translation</title> |
19 | <title>Virtual address translation</title> |
18 | 20 | ||
19 | <para></para> |
21 | <para></para> |
20 | </section> |
22 | </section> |
21 | </section> |
23 | </section><!-- End of VM --> |
- | 24 | ||
22 | 25 | ||
23 | <section> |
26 | <section><!-- Phys mem --> |
24 | <title>Physical memory management</title> |
27 | <title>Physical memory management</title> |
25 | 28 | ||
- | 29 | ||
- | 30 | <section id="zones_and_frames"> |
|
- | 31 | <title>Zones and frames</title> |
|
- | 32 | <para> <graphic fileref="images/mm2.png" /> </para> |
|
- | 33 | ||
- | 34 | ||
- | 35 | <para>On some architectures not whole physical memory is available for conventional usage. This limitations |
|
- | 36 | require from kernel to maintain a table of available and unavailable ranges of physical memory addresses. |
|
- | 37 | Main idea of zones is in creating memory zone entity, that is a continuous chunk of memory available for allocation. |
|
- | 38 | If some chunk is not available, we simply do not put it in any zone. |
|
- | 39 | </para> |
|
- | 40 | ||
- | 41 | <para> |
|
- | 42 | Zone is also serves for informational purposes, containing information about number of free and busy frames. Physical memory |
|
- | 43 | allocation is also done inside the certain zone. Allocation of zone frame must be organized by the |
|
- | 44 | <link linkend="frame_allocator">frame allocator</link> associated with the zone. |
|
- | 45 | </para> |
|
- | 46 | ||
- | 47 | <para>Some of the architectures (mips32, ppc32) have only one zone, that covers whole |
|
- | 48 | physical memory, and the others (like ia32) may have multiple zones. Information about zones on current machine is stored |
|
- | 49 | in BIOS hardware tables or can be hardcoded into kernel during compile time.</para> |
|
- | 50 | ||
- | 51 | </section> |
|
- | 52 | ||
- | 53 | <section id="frame_allocator"> |
|
- | 54 | <title>Frame allocator</title> |
|
- | 55 | ||
- | 56 | <formalpara> |
|
- | 57 | <title>Overview</title> |
|
- | 58 | <para>Frame allocator provides physical memory allocation for the kernel. Because of zonal organization of physical memory, |
|
- | 59 | frame allocator is always working in context of some zone, thus making impossible to allocate a piece of memory, which lays in different zone, which |
|
- | 60 | cannot happen, because two adjacent zones can be merged into one. Frame allocator is also being responsible to update information on |
|
- | 61 | the number of free/busy frames in zone. |
|
- | 62 | Physical memory allocation inside one <link |
|
- | 63 | linkend="zones_and_frames">memory zone</link> is being handled by an |
|
- | 64 | instance of <link linkend="buddy_allocator">buddy allocator</link> |
|
- | 65 | tailored to allocate blocks of physical memory frames. |
|
- | 66 | </para> |
|
- | 67 | </formalpara> |
|
- | 68 | ||
- | 69 | ||
- | 70 | ||
- | 71 | ||
- | 72 | <formalpara> |
|
- | 73 | <title>Allocation / deallocation</title> |
|
- | 74 | <para> |
|
- | 75 | Upon allocation request, frame allocator tries to find first zone, that can satisfy the incoming request (has required amount of free frames to allocate). |
|
- | 76 | During deallocation, frame allocator needs to find zone, that contain deallocated frame. |
|
- | 77 | ||
- | 78 | This approach could bring up two potential problems: |
|
- | 79 | <itemizedlist> |
|
- | 80 | <listitem> |
|
- | 81 | Linear search of zones does not any good to performance, but number of zones is not expected to be high. And if yes, list of zones can be replaced with more time-efficient B-tree. |
|
- | 82 | </listitem> |
|
- | 83 | <listitem> |
|
- | 84 | Quickly find out if zone contains required number of frames to allocate and if this chunk of memory is properly aligned. This issue is perfectly solved bu the buddy allocator. |
|
- | 85 | </listitem> |
|
- | 86 | </itemizedlist> |
|
- | 87 | ||
- | 88 | ||
- | 89 | </para> |
|
- | 90 | </formalpara> |
|
- | 91 | ||
- | 92 | </section> |
|
- | 93 | ||
- | 94 | </section> |
|
- | 95 | ||
- | 96 | ||
- | 97 | ||
26 | <section id="buddy_allocator"> |
98 | <section id="buddy_allocator"> |
27 | <title>Buddy allocator</title> |
99 | <title>Buddy allocator</title> |
28 | 100 | ||
29 | <section> |
101 | <section> |
30 | <title>Overview</title> |
102 | <title>Overview</title> |
31 | 103 | ||
32 | <para>In buddy allocator, memory is broken down into power-of-two |
104 | <para>In buddy allocator, memory is broken down into power-of-two |
33 | sized naturally aligned blocks. These blocks are organized in an array |
105 | sized naturally aligned blocks. These blocks are organized in an array |
34 | of lists in which list with index i contains all unallocated blocks of |
106 | of lists in which list with index i contains all unallocated blocks of |
35 | the size <mathphrase>2<superscript>i</superscript></mathphrase>. The |
107 | the size <mathphrase>2<superscript>i</superscript></mathphrase>. The |
36 | index i is called the order of block. Should there be two adjacent |
108 | index i is called the order of block. Should there be two adjacent |
37 | equally sized blocks in list <mathphrase>i</mathphrase> (i.e. |
109 | equally sized blocks in list <mathphrase>i</mathphrase> (i.e. |
38 | buddies), the buddy allocator would coalesce them and put the |
110 | buddies), the buddy allocator would coalesce them and put the |
39 | resulting block in list <mathphrase>i + 1</mathphrase>, provided that |
111 | resulting block in list <mathphrase>i + 1</mathphrase>, provided that |
40 | the resulting block would be naturally aligned. Similarily, when the |
112 | the resulting block would be naturally aligned. Similarily, when the |
41 | allocator is asked to allocate a block of size |
113 | allocator is asked to allocate a block of size |
42 | <mathphrase>2<superscript>i</superscript></mathphrase>, it first tries |
114 | <mathphrase>2<superscript>i</superscript></mathphrase>, it first tries |
43 | to satisfy the request from list with index i. If the request cannot |
115 | to satisfy the request from list with index i. If the request cannot |
44 | be satisfied (i.e. the list i is empty), the buddy allocator will try |
116 | be satisfied (i.e. the list i is empty), the buddy allocator will try |
45 | to allocate and split larger block from list with index i + 1. Both of |
117 | to allocate and split larger block from list with index i + 1. Both of |
46 | these algorithms are recursive. The recursion ends either when there |
118 | these algorithms are recursive. The recursion ends either when there |
47 | are no blocks to coalesce in the former case or when there are no |
119 | are no blocks to coalesce in the former case or when there are no |
48 | blocks that can be split in the latter case.</para> |
120 | blocks that can be split in the latter case.</para> |
49 | 121 | ||
50 | <graphic fileref="images/buddy_alloc.eps" format="EPS" /> |
122 | <graphic fileref="images/mm1.png" format="EPS" /> |
51 | 123 | ||
52 | <para>This approach greatly reduces external fragmentation of memory |
124 | <para>This approach greatly reduces external fragmentation of memory |
53 | and helps in allocating bigger continuous blocks of memory aligned to |
125 | and helps in allocating bigger continuous blocks of memory aligned to |
54 | their size. On the other hand, the buddy allocator suffers increased |
126 | their size. On the other hand, the buddy allocator suffers increased |
55 | internal fragmentation of memory and is not suitable for general |
127 | internal fragmentation of memory and is not suitable for general |
56 | kernel allocations. This purpose is better addressed by the <link |
128 | kernel allocations. This purpose is better addressed by the <link |
57 | linkend="slab">slab allocator</link>.</para> |
129 | linkend="slab">slab allocator</link>.</para> |
58 | </section> |
130 | </section> |
59 | 131 | ||
60 | <section> |
132 | <section> |
61 | <title>Implementation</title> |
133 | <title>Implementation</title> |
62 | 134 | ||
63 | <para>The buddy allocator is, in fact, an abstract framework wich can |
135 | <para>The buddy allocator is, in fact, an abstract framework wich can |
64 | be easily specialized to serve one particular task. It knows nothing |
136 | be easily specialized to serve one particular task. It knows nothing |
65 | about the nature of memory it helps to allocate. In order to beat the |
137 | about the nature of memory it helps to allocate. In order to beat the |
66 | lack of this knowledge, the buddy allocator exports an interface that |
138 | lack of this knowledge, the buddy allocator exports an interface that |
67 | each of its clients is required to implement. When supplied an |
139 | each of its clients is required to implement. When supplied an |
68 | implementation of this interface, the buddy allocator can use |
140 | implementation of this interface, the buddy allocator can use |
69 | specialized external functions to find buddy for a block, split and |
141 | specialized external functions to find buddy for a block, split and |
70 | coalesce blocks, manipulate block order and mark blocks busy or |
142 | coalesce blocks, manipulate block order and mark blocks busy or |
71 | available. For precize documentation of this interface, refer to <link |
143 | available. For precize documentation of this interface, refer to <link |
72 | linkend="???">HelenOS Generic Kernel Reference Manual</link>.</para> |
144 | linkend="???">HelenOS Generic Kernel Reference Manual</link>.</para> |
73 | 145 | ||
74 | <formalpara> |
146 | <formalpara> |
75 | <title>Data organization</title> |
147 | <title>Data organization</title> |
76 | 148 | ||
77 | <para>Each entity allocable by the buddy allocator is required to |
149 | <para>Each entity allocable by the buddy allocator is required to |
78 | contain space for storing block order number and a link variable |
150 | contain space for storing block order number and a link variable |
79 | used to interconnect blocks within the same order.</para> |
151 | used to interconnect blocks within the same order.</para> |
80 | 152 | ||
81 | <para>Whatever entities are allocated by the buddy allocator, the |
153 | <para>Whatever entities are allocated by the buddy allocator, the |
82 | first entity within a block is used to represent the entire block. |
154 | first entity within a block is used to represent the entire block. |
83 | The first entity keeps the order of the whole block. Other entities |
155 | The first entity keeps the order of the whole block. Other entities |
84 | within the block are assigned the magic value |
156 | within the block are assigned the magic value |
85 | <constant>BUDDY_INNER_BLOCK</constant>. This is especially important |
157 | <constant>BUDDY_INNER_BLOCK</constant>. This is especially important |
86 | for effective identification of buddies in one-dimensional array |
158 | for effective identification of buddies in one-dimensional array |
87 | because the entity that represents a potential buddy cannot be associated |
159 | because the entity that represents a potential buddy cannot be |
88 | with <constant>BUDDY_INNER_BLOCK</constant> (i.e. if it is associated |
160 | associated with <constant>BUDDY_INNER_BLOCK</constant> (i.e. if it |
89 | with <constant>BUDDY_INNER_BLOCK</constant> then it is not a |
161 | is associated with <constant>BUDDY_INNER_BLOCK</constant> then it is |
90 | buddy).</para> |
162 | not a buddy).</para> |
91 | </formalpara> |
163 | </formalpara> |
92 | </section> |
- | |
93 | </section> |
- | |
94 | 164 | ||
95 | <section id="zones_and_frames"> |
- | |
96 | <title>Zones and frames</title> |
- | |
97 | - | ||
98 | <para>Physical memory is divided into zones. Each zone represents |
- | |
99 | continuous area of physical memory frames. Allocation of frames is |
- | |
100 | handled by the <link linkend="frame_allocator">frame allocator</link> |
- | |
101 | associated with the zone. Zone also contains information about free and |
- | |
102 | occupied frames and its base addresss in the memory. Some of the |
- | |
103 | architectures (mips32, ppc32) have only one zone, that covers whole |
- | |
104 | physical memory. Other architectures (ia32) have multiple zones.</para> |
- | |
105 | </section> |
- | |
106 | - | ||
107 | <section id="frame_allocator"> |
- | |
108 | <title>Frame allocator</title> |
- | |
109 | - | ||
110 | <section> |
- | |
111 | <title>Overview</title> |
- | |
112 | - | ||
113 | <para>Physical memory allocation inside one <link |
- | |
114 | linkend="zones_and_frames">memory zone</link> is being handled by an |
- | |
115 | instance of <link linkend="buddy_allocator">buddy allocator</link> |
- | |
116 | tailored to allocate blocks of physical memory frames.</para> |
- | |
117 | - | ||
118 | <graphic fileref="images/mm1.png" /> |
- | |
119 | </section> |
- | |
120 | - | ||
121 | <section> |
- | |
122 | <title>Implementation</title> |
- | |
123 | - | ||
124 | <formalpara> |
165 | <formalpara> |
125 | <title>Data organization</title> |
166 | <title>Data organization</title> |
126 | 167 | ||
127 | <para>Buddy allocator always uses first frame to represent frame |
168 | <para>Buddy allocator always uses first frame to represent frame |
128 | block. This frame contains <varname>buddy_order</varname> variable |
169 | block. This frame contains <varname>buddy_order</varname> variable |
129 | to provide information about the block size it actually represents ( |
170 | to provide information about the block size it actually represents ( |
130 | <mathphrase>2<superscript>buddy_order</superscript></mathphrase> |
171 | <mathphrase>2<superscript>buddy_order</superscript></mathphrase> |
131 | frames block). Other frames in block have this value set to magic |
172 | frames block). Other frames in block have this value set to magic |
132 | <constant>BUDDY_INNER_BLOCK</constant> that is much greater than |
173 | <constant>BUDDY_INNER_BLOCK</constant> that is much greater than |
133 | buddy <varname>max_order</varname> value.</para> |
174 | buddy <varname>max_order</varname> value.</para> |
134 | 175 | ||
135 | <para>Each <varname>frame_t</varname> also contains pointer member |
176 | <para>Each <varname>frame_t</varname> also contains pointer member |
136 | to hold frame structure in the linked list inside one order.</para> |
177 | to hold frame structure in the linked list inside one order.</para> |
137 | </formalpara> |
178 | </formalpara> |
138 | 179 | ||
139 | <formalpara> |
180 | <formalpara> |
140 | <title>Allocation algorithm</title> |
181 | <title>Allocation algorithm</title> |
141 | 182 | ||
142 | <para>Upon <mathphrase>2<superscript>i</superscript></mathphrase> |
183 | <para>Upon <mathphrase>2<superscript>i</superscript></mathphrase> |
143 | frames block allocation request, allocator checks if there are any |
184 | frames block allocation request, allocator checks if there are any |
144 | blocks available at the order list <varname>i</varname>. If yes, |
185 | blocks available at the order list <varname>i</varname>. If yes, |
145 | removes block from order list and returns its address. If no, |
186 | removes block from order list and returns its address. If no, |
146 | recursively allocates |
187 | recursively allocates |
147 | <mathphrase>2<superscript>i+1</superscript></mathphrase> frame |
188 | <mathphrase>2<superscript>i+1</superscript></mathphrase> frame |
148 | block, splits it into two |
189 | block, splits it into two |
149 | <mathphrase>2<superscript>i</superscript></mathphrase> frame blocks. |
190 | <mathphrase>2<superscript>i</superscript></mathphrase> frame blocks. |
150 | Then adds one of the blocks to the <varname>i</varname> order list |
191 | Then adds one of the blocks to the <varname>i</varname> order list |
151 | and returns address of another.</para> |
192 | and returns address of another.</para> |
152 | </formalpara> |
193 | </formalpara> |
153 | 194 | ||
154 | <formalpara> |
195 | <formalpara> |
155 | <title>Deallocation algorithm</title> |
196 | <title>Deallocation algorithm</title> |
156 | 197 | ||
157 | <para>Check if block has so called buddy (another free |
198 | <para>Check if block has so called buddy (another free |
158 | <mathphrase>2<superscript>i</superscript></mathphrase> frame block |
199 | <mathphrase>2<superscript>i</superscript></mathphrase> frame block |
159 | that can be linked with freed block into the |
200 | that can be linked with freed block into the |
160 | <mathphrase>2<superscript>i+1</superscript></mathphrase> block). |
201 | <mathphrase>2<superscript>i+1</superscript></mathphrase> block). |
161 | Technically, buddy is a odd/even block for even/odd block |
202 | Technically, buddy is a odd/even block for even/odd block |
162 | respectively. Plus we can put an extra requirement, that resulting |
203 | respectively. Plus we can put an extra requirement, that resulting |
163 | block must be aligned to its size. This requirement guarantees |
204 | block must be aligned to its size. This requirement guarantees |
164 | natural block alignment for the blocks coming out the allocation |
205 | natural block alignment for the blocks coming out the allocation |
165 | system.</para> |
206 | system.</para> |
166 | 207 | ||
167 | <para>Using direct pointer arithmetics, |
208 | <para>Using direct pointer arithmetics, |
168 | <varname>frame_t::ref_count</varname> and |
209 | <varname>frame_t::ref_count</varname> and |
169 | <varname>frame_t::buddy_order</varname> variables, finding buddy is |
210 | <varname>frame_t::buddy_order</varname> variables, finding buddy is |
170 | done at constant time.</para> |
211 | done at constant time.</para> |
171 | </formalpara> |
212 | </formalpara> |
- | 213 | ||
172 | </section> |
214 | </section> |
173 | </section> |
215 | |
174 | 216 | ||
175 | <section id="slab"> |
217 | <section id="slab"> |
176 | <title>Slab allocator</title> |
218 | <title>Slab allocator</title> |
177 | 219 | ||
178 | <para>Kernel memory allocation is handled by slab.</para> |
220 | <para>Kernel memory allocation is handled by slab.</para> |
- | 221 | </section><!-- End of Physmem --> |
|
- | 222 | ||
179 | </section> |
223 | </section> |
- | 224 | ||
180 | 225 | ||
181 | <section> |
226 | <section> |
182 | <title>Memory sharing</title> |
227 | <title>Memory sharing</title> |
183 | 228 | ||
184 | <para>Not implemented yet(?)</para> |
229 | <para>Not implemented yet(?)</para> |
185 | </section> |
230 | </section> |
186 | </section> |
- | |
187 | </chapter> |
231 | </chapter> |
188 | 232 |