1 ============================================== 1 ==================================================
2 page owner: Tracking about who allocated each 2 page owner: Tracking about who allocated each page
3 ============================================== 3 ==================================================
4 4
5 Introduction 5 Introduction
6 ============ 6 ============
7 7
8 page owner is for the tracking about who alloc 8 page owner is for the tracking about who allocated each page.
9 It can be used to debug memory leak or to find 9 It can be used to debug memory leak or to find a memory hogger.
10 When allocation happens, information about all 10 When allocation happens, information about allocation such as call stack
11 and order of pages is stored into certain stor 11 and order of pages is stored into certain storage for each page.
12 When we need to know about status of all pages 12 When we need to know about status of all pages, we can get and analyze
13 this information. 13 this information.
14 14
15 Although we already have tracepoint for tracin 15 Although we already have tracepoint for tracing page allocation/free,
16 using it for analyzing who allocate each page 16 using it for analyzing who allocate each page is rather complex. We need
17 to enlarge the trace buffer for preventing ove 17 to enlarge the trace buffer for preventing overlapping until userspace
18 program launched. And, launched program contin 18 program launched. And, launched program continually dump out the trace
19 buffer for later analysis and it would change 19 buffer for later analysis and it would change system behaviour with more
20 possibility rather than just keeping it in mem 20 possibility rather than just keeping it in memory, so bad for debugging.
21 21
22 page owner can also be used for various purpos 22 page owner can also be used for various purposes. For example, accurate
23 fragmentation statistics can be obtained throu 23 fragmentation statistics can be obtained through gfp flag information of
24 each page. It is already implemented and activ 24 each page. It is already implemented and activated if page owner is
25 enabled. Other usages are more than welcome. 25 enabled. Other usages are more than welcome.
26 26
27 It can also be used to show all the stacks and <<
28 allocated base pages, which gives us a quick o <<
29 is going without the need to screen through al <<
30 allocation and free operation. <<
31 <<
32 page owner is disabled by default. So, if you' 27 page owner is disabled by default. So, if you'd like to use it, you need
33 to add "page_owner=on" to your boot cmdline. I 28 to add "page_owner=on" to your boot cmdline. If the kernel is built
34 with page owner and page owner is disabled in 29 with page owner and page owner is disabled in runtime due to not enabling
35 boot option, runtime overhead is marginal. If 30 boot option, runtime overhead is marginal. If disabled in runtime, it
36 doesn't require memory to store owner informat 31 doesn't require memory to store owner information, so there is no runtime
37 memory overhead. And, page owner inserts just 32 memory overhead. And, page owner inserts just two unlikely branches into
38 the page allocator hotpath and if not enabled, 33 the page allocator hotpath and if not enabled, then allocation is done
39 like as the kernel without page owner. These t 34 like as the kernel without page owner. These two unlikely branches should
40 not affect to allocation performance, especial 35 not affect to allocation performance, especially if the static keys jump
41 label patching functionality is available. Fol 36 label patching functionality is available. Following is the kernel's code
42 size change due to this facility. 37 size change due to this facility.
43 38
44 Although enabling page owner increases kernel 39 Although enabling page owner increases kernel size by several kilobytes,
45 most of this code is outside page allocator an 40 most of this code is outside page allocator and its hot path. Building
46 the kernel with page owner and turning it on i 41 the kernel with page owner and turning it on if needed would be great
47 option to debug kernel memory problem. 42 option to debug kernel memory problem.
48 43
49 There is one notice that is caused by implemen 44 There is one notice that is caused by implementation detail. page owner
50 stores information into the memory from struct 45 stores information into the memory from struct page extension. This memory
51 is initialized some time later than that page 46 is initialized some time later than that page allocator starts in sparse
52 memory system, so, until initialization, many 47 memory system, so, until initialization, many pages can be allocated and
53 they would have no owner information. To fix i 48 they would have no owner information. To fix it up, these early allocated
54 pages are investigated and marked as allocated 49 pages are investigated and marked as allocated in initialization phase.
55 Although it doesn't mean that they have the ri 50 Although it doesn't mean that they have the right owner information,
56 at least, we can tell whether the page is allo 51 at least, we can tell whether the page is allocated or not,
57 more accurately. On 2GB memory x86-64 VM box, 52 more accurately. On 2GB memory x86-64 VM box, 13343 early allocated pages
58 are caught and marked, although they are mostl 53 are caught and marked, although they are mostly allocated from struct
59 page extension feature. Anyway, after that, no 54 page extension feature. Anyway, after that, no page is left in
60 un-tracking state. 55 un-tracking state.
61 56
62 Usage 57 Usage
63 ===== 58 =====
64 59
65 1) Build user-space helper:: 60 1) Build user-space helper::
66 61
67 cd tools/mm 62 cd tools/mm
68 make page_owner_sort 63 make page_owner_sort
69 64
70 2) Enable page owner: add "page_owner=on" to b 65 2) Enable page owner: add "page_owner=on" to boot cmdline.
71 66
72 3) Do the job that you want to debug. 67 3) Do the job that you want to debug.
73 68
74 4) Analyze information from page owner:: 69 4) Analyze information from page owner::
75 <<
76 cat /sys/kernel/debug/page_owner_stack <<
77 cat stacks.txt <<
78 post_alloc_hook+0x177/0x1a0 <<
79 get_page_from_freelist+0xd01/0xd80 <<
80 __alloc_pages+0x39e/0x7e0 <<
81 allocate_slab+0xbc/0x3f0 <<
82 ___slab_alloc+0x528/0x8a0 <<
83 kmem_cache_alloc+0x224/0x3b0 <<
84 sk_prot_alloc+0x58/0x1a0 <<
85 sk_alloc+0x32/0x4f0 <<
86 inet_create+0x427/0xb50 <<
87 __sock_create+0x2e4/0x650 <<
88 inet_ctl_sock_create+0x30/0x180 <<
89 igmp_net_init+0xc1/0x130 <<
90 ops_init+0x167/0x410 <<
91 setup_net+0x304/0xa60 <<
92 copy_net_ns+0x29b/0x4a0 <<
93 create_new_namespaces+0x4a1/0x820 <<
94 nr_base_pages: 16 <<
95 ... <<
96 ... <<
97 echo 7000 > /sys/kernel/debug/page_own <<
98 cat /sys/kernel/debug/page_owner_stack <<
99 cat stacks_7000.txt <<
100 post_alloc_hook+0x177/0x1a0 <<
101 get_page_from_freelist+0xd01/0xd80 <<
102 __alloc_pages+0x39e/0x7e0 <<
103 alloc_pages_mpol+0x22e/0x490 <<
104 folio_alloc+0xd5/0x110 <<
105 filemap_alloc_folio+0x78/0x230 <<
106 page_cache_ra_order+0x287/0x6f0 <<
107 filemap_get_pages+0x517/0x1160 <<
108 filemap_read+0x304/0x9f0 <<
109 xfs_file_buffered_read+0xe6/0x1d0 [xf <<
110 xfs_file_read_iter+0x1f0/0x380 [xfs] <<
111 __kernel_read+0x3b9/0x730 <<
112 kernel_read_file+0x309/0x4d0 <<
113 __do_sys_finit_module+0x381/0x730 <<
114 do_syscall_64+0x8d/0x150 <<
115 entry_SYSCALL_64_after_hwframe+0x62/0 <<
116 nr_base_pages: 20824 <<
117 ... <<
118 70
119 cat /sys/kernel/debug/page_owner > pag 71 cat /sys/kernel/debug/page_owner > page_owner_full.txt
120 ./page_owner_sort page_owner_full.txt 72 ./page_owner_sort page_owner_full.txt sorted_page_owner.txt
121 73
122 The general output of ``page_owner_full.txt 74 The general output of ``page_owner_full.txt`` is as follows::
123 75
124 Page allocated via order XXX, ... 76 Page allocated via order XXX, ...
125 PFN XXX ... 77 PFN XXX ...
126 // Detailed stack 78 // Detailed stack
127 79
128 Page allocated via order XXX, ... 80 Page allocated via order XXX, ...
129 PFN XXX ... 81 PFN XXX ...
130 // Detailed stack 82 // Detailed stack
131 By default, it will do full pfn dump, to s 83 By default, it will do full pfn dump, to start with a given pfn,
132 page_owner supports fseek. 84 page_owner supports fseek.
133 85
134 FILE *fp = fopen("/sys/kernel/debug/page_o 86 FILE *fp = fopen("/sys/kernel/debug/page_owner", "r");
135 fseek(fp, pfn_start, SEEK_SET); 87 fseek(fp, pfn_start, SEEK_SET);
136 88
137 The ``page_owner_sort`` tool ignores ``PFN` 89 The ``page_owner_sort`` tool ignores ``PFN`` rows, puts the remaining rows
138 in buf, uses regexp to extract the page ord 90 in buf, uses regexp to extract the page order value, counts the times
139 and pages of buf, and finally sorts them ac 91 and pages of buf, and finally sorts them according to the parameter(s).
140 92
141 See the result about who allocated each pag 93 See the result about who allocated each page
142 in the ``sorted_page_owner.txt``. General o 94 in the ``sorted_page_owner.txt``. General output::
143 95
144 XXX times, XXX pages: 96 XXX times, XXX pages:
145 Page allocated via order XXX, ... 97 Page allocated via order XXX, ...
146 // Detailed stack 98 // Detailed stack
147 99
148 By default, ``page_owner_sort`` is sorted a 100 By default, ``page_owner_sort`` is sorted according to the times of buf.
149 If you want to sort by the page nums of buf 101 If you want to sort by the page nums of buf, use the ``-m`` parameter.
150 The detailed parameters are: 102 The detailed parameters are:
151 103
152 fundamental function:: 104 fundamental function::
153 105
154 Sort: 106 Sort:
155 -a Sort by memory 107 -a Sort by memory allocation time.
156 -m Sort by total 108 -m Sort by total memory.
157 -p Sort by pid. 109 -p Sort by pid.
158 -P Sort by tgid. 110 -P Sort by tgid.
159 -n Sort by task c 111 -n Sort by task command name.
160 -r Sort by memory 112 -r Sort by memory release time.
161 -s Sort by stack 113 -s Sort by stack trace.
162 -t Sort by times 114 -t Sort by times (default).
163 --sort <order> Specify sortin 115 --sort <order> Specify sorting order. Sorting syntax is [+|-]key[,[+|-]key[,...]].
164 Choose a key f 116 Choose a key from the **STANDARD FORMAT SPECIFIERS** section. The "+" is
165 optional since 117 optional since default direction is increasing numerical or lexicographic
166 order. Mixed u 118 order. Mixed use of abbreviated and complete-form of keys is allowed.
167 119
168 Examples: 120 Examples:
169 ./page_owner_s 121 ./page_owner_sort <input> <output> --sort=n,+pid,-tgid
170 ./page_owner_s 122 ./page_owner_sort <input> <output> --sort=at
171 123
172 additional function:: 124 additional function::
173 125
174 Cull: 126 Cull:
175 --cull <rules> 127 --cull <rules>
176 Specify cullin 128 Specify culling rules.Culling syntax is key[,key[,...]].Choose a
177 multi-letter k 129 multi-letter key from the **STANDARD FORMAT SPECIFIERS** section.
178 130
179 <rules> is a single argument i 131 <rules> is a single argument in the form of a comma-separated list,
180 which offers a way to specify 132 which offers a way to specify individual culling rules. The recognized
181 keywords are described in the 133 keywords are described in the **STANDARD FORMAT SPECIFIERS** section below.
182 <rules> can be specified by th 134 <rules> can be specified by the sequence of keys k1,k2, ..., as described in
183 the STANDARD SORT KEYS section 135 the STANDARD SORT KEYS section below. Mixed use of abbreviated and
184 complete-form of keys is allow 136 complete-form of keys is allowed.
185 137
186 Examples: 138 Examples:
187 ./page_owner_s 139 ./page_owner_sort <input> <output> --cull=stacktrace
188 ./page_owner_s 140 ./page_owner_sort <input> <output> --cull=st,pid,name
189 ./page_owner_s 141 ./page_owner_sort <input> <output> --cull=n,f
190 142
191 Filter: 143 Filter:
192 -f Filter out the 144 -f Filter out the information of blocks whose memory has been released.
193 145
194 Select: 146 Select:
195 --pid <pidlist> Select 147 --pid <pidlist> Select by pid. This selects the blocks whose process ID
196 number 148 numbers appear in <pidlist>.
197 --tgid <tgidlist> Select 149 --tgid <tgidlist> Select by tgid. This selects the blocks whose thread
198 group 150 group ID numbers appear in <tgidlist>.
199 --name <cmdlist> Select 151 --name <cmdlist> Select by task command name. This selects the blocks whose
200 task c 152 task command name appear in <cmdlist>.
201 153
202 <pidlist>, <tgidlist>, <cmdlis 154 <pidlist>, <tgidlist>, <cmdlist> are single arguments in the form of a comma-separated list,
203 which offers a way to specify 155 which offers a way to specify individual selecting rules.
204 156
205 157
206 Examples: 158 Examples:
207 ./page_owner_s 159 ./page_owner_sort <input> <output> --pid=1
208 ./page_owner_s 160 ./page_owner_sort <input> <output> --tgid=1,2,3
209 ./page_owner_s 161 ./page_owner_sort <input> <output> --name name1,name2
210 162
211 STANDARD FORMAT SPECIFIERS 163 STANDARD FORMAT SPECIFIERS
212 ========================== 164 ==========================
213 :: 165 ::
214 166
215 For --sort option: 167 For --sort option:
216 168
217 KEY LONG DESCRI 169 KEY LONG DESCRIPTION
218 p pid proces 170 p pid process ID
219 tg tgid thread 171 tg tgid thread group ID
220 n name task c 172 n name task command name
221 st stacktrace stack 173 st stacktrace stack trace of the page allocation
222 T txt full t 174 T txt full text of block
223 ft free_ts timest 175 ft free_ts timestamp of the page when it was released
224 at alloc_ts timest 176 at alloc_ts timestamp of the page when it was allocated
225 ator allocator memory 177 ator allocator memory allocator for pages
226 178
227 For --cull option: 179 For --cull option:
228 180
229 KEY LONG DESCRI 181 KEY LONG DESCRIPTION
230 p pid proces 182 p pid process ID
231 tg tgid thread 183 tg tgid thread group ID
232 n name task c 184 n name task command name
233 f free whethe 185 f free whether the page has been released or not
234 st stacktrace stack 186 st stacktrace stack trace of the page allocation
235 ator allocator memory 187 ator allocator memory allocator for pages
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.