xref: /linux/Documentation/mm/page_owner.rst (revision 7203ca412fc8e8a0588e9adc0f777d3163f8dff3)
1==================================================
2page owner: Tracking about who allocated each page
3==================================================
4
5Introduction
6============
7
8page owner is for the tracking about who allocated each page.
9It can be used to debug memory leak or to find a memory hogger.
10When allocation happens, information about allocation such as call stack
11and order of pages is stored into certain storage for each page.
12When we need to know about status of all pages, we can get and analyze
13this information.
14
15Although we already have tracepoint for tracing page allocation/free,
16using it for analyzing who allocate each page is rather complex. We need
17to enlarge the trace buffer for preventing overlapping until userspace
18program launched. And, launched program continually dump out the trace
19buffer for later analysis and it would change system behaviour with more
20possibility rather than just keeping it in memory, so bad for debugging.
21
22page owner can also be used for various purposes. For example, accurate
23fragmentation statistics can be obtained through gfp flag information of
24each page. It is already implemented and activated if page owner is
25enabled. Other usages are more than welcome.
26
27It can also be used to show all the stacks and their current number of
28allocated base pages, which gives us a quick overview of where the memory
29is going without the need to screen through all the pages and match the
30allocation and free operation. It's also possible to show only a numeric
31identifier of all the stacks (without stack traces) and their number of
32allocated base pages (faster to read and parse, eg, for monitoring) that
33can be matched with stacks later (show_handles and show_stacks_handles).
34
35page owner is disabled by default. So, if you'd like to use it, you need
36to add "page_owner=on" to your boot cmdline. If the kernel is built
37with page owner and page owner is disabled in runtime due to not enabling
38boot option, runtime overhead is marginal. If disabled in runtime, it
39doesn't require memory to store owner information, so there is no runtime
40memory overhead. And, page owner inserts just two unlikely branches into
41the page allocator hotpath and if not enabled, then allocation is done
42like as the kernel without page owner. These two unlikely branches should
43not affect to allocation performance, especially if the static keys jump
44label patching functionality is available. Following is the kernel's code
45size change due to this facility.
46
47Although enabling page owner increases kernel size by several kilobytes,
48most of this code is outside page allocator and its hot path. Building
49the kernel with page owner and turning it on if needed would be great
50option to debug kernel memory problem.
51
52There is one notice that is caused by implementation detail. page owner
53stores information into the memory from struct page extension. This memory
54is initialized some time later than that page allocator starts in sparse
55memory system, so, until initialization, many pages can be allocated and
56they would have no owner information. To fix it up, these early allocated
57pages are investigated and marked as allocated in initialization phase.
58Although it doesn't mean that they have the right owner information,
59at least, we can tell whether the page is allocated or not,
60more accurately. On 2GB memory x86-64 VM box, 13343 early allocated pages
61are caught and marked, although they are mostly allocated from struct
62page extension feature. Anyway, after that, no page is left in
63un-tracking state.
64
65Usage
66=====
67
681) Build user-space helper::
69
70	cd tools/mm
71	make page_owner_sort
72
732) Enable page owner: add "page_owner=on" to boot cmdline.
74
753) Do the job that you want to debug.
76
774) Analyze information from page owner::
78
79	cat /sys/kernel/debug/page_owner_stacks/show_stacks > stacks.txt
80	cat stacks.txt
81	 post_alloc_hook+0x177/0x1a0
82	 get_page_from_freelist+0xd01/0xd80
83	 __alloc_pages+0x39e/0x7e0
84	 allocate_slab+0xbc/0x3f0
85	 ___slab_alloc+0x528/0x8a0
86	 kmem_cache_alloc+0x224/0x3b0
87	 sk_prot_alloc+0x58/0x1a0
88	 sk_alloc+0x32/0x4f0
89	 inet_create+0x427/0xb50
90	 __sock_create+0x2e4/0x650
91	 inet_ctl_sock_create+0x30/0x180
92	 igmp_net_init+0xc1/0x130
93	 ops_init+0x167/0x410
94	 setup_net+0x304/0xa60
95	 copy_net_ns+0x29b/0x4a0
96	 create_new_namespaces+0x4a1/0x820
97	nr_base_pages: 16
98	...
99	...
100	echo 7000 > /sys/kernel/debug/page_owner_stacks/count_threshold
101	cat /sys/kernel/debug/page_owner_stacks/show_stacks> stacks_7000.txt
102	cat stacks_7000.txt
103	 post_alloc_hook+0x177/0x1a0
104	 get_page_from_freelist+0xd01/0xd80
105	 __alloc_pages+0x39e/0x7e0
106	 alloc_pages_mpol+0x22e/0x490
107	 folio_alloc+0xd5/0x110
108	 filemap_alloc_folio+0x78/0x230
109	 page_cache_ra_order+0x287/0x6f0
110	 filemap_get_pages+0x517/0x1160
111	 filemap_read+0x304/0x9f0
112	 xfs_file_buffered_read+0xe6/0x1d0 [xfs]
113	 xfs_file_read_iter+0x1f0/0x380 [xfs]
114	 __kernel_read+0x3b9/0x730
115	 kernel_read_file+0x309/0x4d0
116	 __do_sys_finit_module+0x381/0x730
117	 do_syscall_64+0x8d/0x150
118	 entry_SYSCALL_64_after_hwframe+0x62/0x6a
119	nr_base_pages: 20824
120	...
121
122	cat /sys/kernel/debug/page_owner_stacks/show_handles > handles_7000.txt
123	cat handles_7000.txt
124	handle: 42
125	nr_base_pages: 20824
126	...
127
128	cat /sys/kernel/debug/page_owner_stacks/show_stacks_handles > stacks_handles.txt
129	cat stacks_handles.txt
130	 post_alloc_hook+0x177/0x1a0
131	 get_page_from_freelist+0xd01/0xd80
132	 __alloc_pages+0x39e/0x7e0
133	 alloc_pages_mpol+0x22e/0x490
134	 folio_alloc+0xd5/0x110
135	 filemap_alloc_folio+0x78/0x230
136	 page_cache_ra_order+0x287/0x6f0
137	 filemap_get_pages+0x517/0x1160
138	 filemap_read+0x304/0x9f0
139	 xfs_file_buffered_read+0xe6/0x1d0 [xfs]
140	 xfs_file_read_iter+0x1f0/0x380 [xfs]
141	 __kernel_read+0x3b9/0x730
142	 kernel_read_file+0x309/0x4d0
143	 __do_sys_finit_module+0x381/0x730
144	 do_syscall_64+0x8d/0x150
145	 entry_SYSCALL_64_after_hwframe+0x62/0x6a
146	handle: 42
147	...
148
149	cat /sys/kernel/debug/page_owner > page_owner_full.txt
150	./page_owner_sort page_owner_full.txt sorted_page_owner.txt
151
152   The general output of ``page_owner_full.txt`` is as follows::
153
154	Page allocated via order XXX, ...
155	PFN XXX ...
156	// Detailed stack
157
158	Page allocated via order XXX, ...
159	PFN XXX ...
160	// Detailed stack
161    By default, it will do full pfn dump, to start with a given pfn,
162    page_owner supports fseek.
163
164    FILE *fp = fopen("/sys/kernel/debug/page_owner", "r");
165    fseek(fp, pfn_start, SEEK_SET);
166
167   The ``page_owner_sort`` tool ignores ``PFN`` rows, puts the remaining rows
168   in buf, uses regexp to extract the page order value, counts the times
169   and pages of buf, and finally sorts them according to the parameter(s).
170
171   See the result about who allocated each page
172   in the ``sorted_page_owner.txt``. General output::
173
174	XXX times, XXX pages:
175	Page allocated via order XXX, ...
176	// Detailed stack
177
178   By default, ``page_owner_sort`` is sorted according to the times of buf.
179   If you want to sort by the page nums of buf, use the ``-m`` parameter.
180   The detailed parameters are:
181
182   fundamental function::
183
184	Sort:
185		-a		Sort by memory allocation time.
186		-m		Sort by total memory.
187		-p		Sort by pid.
188		-P		Sort by tgid.
189		-n		Sort by task command name.
190		-r		Sort by memory release time.
191		-s		Sort by stack trace.
192		-t		Sort by times (default).
193		--sort <order>	Specify sorting order.  Sorting syntax is [+|-]key[,[+|-]key[,...]].
194				Choose a key from the **STANDARD FORMAT SPECIFIERS** section. The "+" is
195				optional since default direction is increasing numerical or lexicographic
196				order. Mixed use of abbreviated and complete-form of keys is allowed.
197
198		Examples:
199				./page_owner_sort <input> <output> --sort=n,+pid,-tgid
200				./page_owner_sort <input> <output> --sort=at
201
202   additional function::
203
204	Cull:
205		--cull <rules>
206				Specify culling rules.Culling syntax is key[,key[,...]].Choose a
207				multi-letter key from the **STANDARD FORMAT SPECIFIERS** section.
208
209		<rules> is a single argument in the form of a comma-separated list,
210		which offers a way to specify individual culling rules.  The recognized
211		keywords are described in the **STANDARD FORMAT SPECIFIERS** section below.
212		<rules> can be specified by the sequence of keys k1,k2, ..., as described in
213		the STANDARD SORT KEYS section below. Mixed use of abbreviated and
214		complete-form of keys is allowed.
215
216		Examples:
217				./page_owner_sort <input> <output> --cull=stacktrace
218				./page_owner_sort <input> <output> --cull=st,pid,name
219				./page_owner_sort <input> <output> --cull=n,f
220
221	Filter:
222		-f		Filter out the information of blocks whose memory has been released.
223
224	Select:
225		--pid <pidlist>		Select by pid. This selects the blocks whose process ID
226					numbers appear in <pidlist>.
227		--tgid <tgidlist>	Select by tgid. This selects the blocks whose thread
228					group ID numbers appear in <tgidlist>.
229		--name <cmdlist>	Select by task command name. This selects the blocks whose
230					task command name appear in <cmdlist>.
231
232		<pidlist>, <tgidlist>, <cmdlist> are single arguments in the form of a comma-separated list,
233		which offers a way to specify individual selecting rules.
234
235
236		Examples:
237				./page_owner_sort <input> <output> --pid=1
238				./page_owner_sort <input> <output> --tgid=1,2,3
239				./page_owner_sort <input> <output> --name name1,name2
240
241STANDARD FORMAT SPECIFIERS
242==========================
243::
244
245  For --sort option:
246
247	KEY		LONG		DESCRIPTION
248	p		pid		process ID
249	tg		tgid		thread group ID
250	n		name		task command name
251	st		stacktrace	stack trace of the page allocation
252	T		txt		full text of block
253	ft		free_ts		timestamp of the page when it was released
254	at		alloc_ts	timestamp of the page when it was allocated
255	ator		allocator	memory allocator for pages
256
257  For --cull option:
258
259	KEY		LONG		DESCRIPTION
260	p		pid		process ID
261	tg		tgid		thread group ID
262	n		name		task command name
263	f		free		whether the page has been released or not
264	st		stacktrace	stack trace of the page allocation
265	ator		allocator	memory allocator for pages
266