[linux-2.6-block.git] / Documentation / vm / page_owner.txt

page owner: Tracking about who allocated each page
-----------------------------------------------------------

* Introduction

page owner is for the tracking about who allocated each page.
It can be used to debug memory leak or to find a memory hogger.
When allocation happens, information about allocation such as call stack
and order of pages is stored into certain storage for each page.
When we need to know about status of all pages, we can get and analyze
this information.

Although we already have tracepoint for tracing page allocation/free,
using it for analyzing who allocate each page is rather complex. We need
to enlarge the trace buffer for preventing overlapping until userspace
program launched. And, launched program continually dump out the trace
buffer for later analysis and it would change system behviour with more
possibility rather than just keeping it in memory, so bad for debugging.

page owner can also be used for various purposes. For example, accurate
fragmentation statistics can be obtained through gfp flag information of
each page. It is already implemented and activated if page owner is
enabled. Other usages are more than welcome.

page owner is disabled in default. So, if you'd like to use it, you need
to add "page_owner=on" into your boot cmdline. If the kernel is built
with page owner and page owner is disabled in runtime due to no enabling
boot option, runtime overhead is marginal. If disabled in runtime, it
doesn't require memory to store owner information, so there is no runtime
memory overhead. And, page owner inserts just two unlikely branches into
the page allocator hotpath and if it returns false then allocation is
done like as the kernel without page owner. These two unlikely branches
would not affect to allocation performance. Following is the kernel's
code size change due to this facility.

- Without page owner
   text    data     bss     dec     hex filename
  40662    1493     644   42799    a72f mm/page_alloc.o

- With page owner
   text    data     bss     dec     hex filename
  40892    1493     644   43029    a815 mm/page_alloc.o
   1427      24       8    1459     5b3 mm/page_ext.o
   2722      50       0    2772     ad4 mm/page_owner.o

Although, roughly, 4 KB code is added in total, page_alloc.o increase by
230 bytes and only half of it is in hotpath. Building the kernel with
page owner and turning it on if needed would be great option to debug
kernel memory problem.

There is one notice that is caused by implementation detail. page owner
stores information into the memory from struct page extension. This memory
is initialized some time later than that page allocator starts in sparse
memory system, so, until initialization, many pages can be allocated and
they would have no owner information. To fix it up, these early allocated
pages are investigated and marked as allocated in initialization phase.
Although it doesn't mean that they have the right owner information,
at least, we can tell whether the page is allocated or not,
more accurately. On 2GB memory x86-64 VM box, 13343 early allocated pages
are catched and marked, although they are mostly allocated from struct
page extension feature. Anyway, after that, no page is left in
un-tracking state.

* Usage

1) Build user-space helper
	cd tools/vm
	make page_owner_sort

2) Enable page owner
	Add "page_owner=on" to boot cmdline.

3) Do the job what you want to debug

4) Analyze information from page owner
	cat /sys/kernel/debug/page_owner > page_owner_full.txt
	grep -v ^PFN page_owner_full.txt > page_owner.txt
	./page_owner_sort page_owner.txt sorted_page_owner.txt

	See the result about who allocated each page
	in the sorted_page_owner.txt.
Commit	Line	Data
16a7ade8 JK	1	page owner: Tracking about who allocated each page
	2	-----------------------------------------------------------
	3
	4	* Introduction
	5
	6	page owner is for the tracking about who allocated each page.
	7	It can be used to debug memory leak or to find a memory hogger.
	8	When allocation happens, information about allocation such as call stack
	9	and order of pages is stored into certain storage for each page.
	10	When we need to know about status of all pages, we can get and analyze
	11	this information.
	12
	13	Although we already have tracepoint for tracing page allocation/free,
	14	using it for analyzing who allocate each page is rather complex. We need
	15	to enlarge the trace buffer for preventing overlapping until userspace
	16	program launched. And, launched program continually dump out the trace
	17	buffer for later analysis and it would change system behviour with more
	18	possibility rather than just keeping it in memory, so bad for debugging.
	19
	20	page owner can also be used for various purposes. For example, accurate
	21	fragmentation statistics can be obtained through gfp flag information of
	22	each page. It is already implemented and activated if page owner is
	23	enabled. Other usages are more than welcome.
	24
	25	page owner is disabled in default. So, if you'd like to use it, you need
	26	to add "page_owner=on" into your boot cmdline. If the kernel is built
	27	with page owner and page owner is disabled in runtime due to no enabling
	28	boot option, runtime overhead is marginal. If disabled in runtime, it
	29	doesn't require memory to store owner information, so there is no runtime
	30	memory overhead. And, page owner inserts just two unlikely branches into
	31	the page allocator hotpath and if it returns false then allocation is
	32	done like as the kernel without page owner. These two unlikely branches
	33	would not affect to allocation performance. Following is the kernel's
	34	code size change due to this facility.
	35
	36	- Without page owner
	37	text data bss dec hex filename
	38	40662 1493 644 42799 a72f mm/page_alloc.o
	39
	40	- With page owner
	41	text data bss dec hex filename
	42	40892 1493 644 43029 a815 mm/page_alloc.o
	43	1427 24 8 1459 5b3 mm/page_ext.o
	44	2722 50 0 2772 ad4 mm/page_owner.o
	45
	46	Although, roughly, 4 KB code is added in total, page_alloc.o increase by
	47	230 bytes and only half of it is in hotpath. Building the kernel with
	48	page owner and turning it on if needed would be great option to debug
	49	kernel memory problem.
	50
	51	There is one notice that is caused by implementation detail. page owner
	52	stores information into the memory from struct page extension. This memory
	53	is initialized some time later than that page allocator starts in sparse
	54	memory system, so, until initialization, many pages can be allocated and
	55	they would have no owner information. To fix it up, these early allocated
	56	pages are investigated and marked as allocated in initialization phase.
	57	Although it doesn't mean that they have the right owner information,
	58	at least, we can tell whether the page is allocated or not,
	59	more accurately. On 2GB memory x86-64 VM box, 13343 early allocated pages
	60	are catched and marked, although they are mostly allocated from struct
	61	page extension feature. Anyway, after that, no page is left in
	62	un-tracking state.
	63
	64	* Usage
65
66	1) Build user-space helper
67	cd tools/vm
68	make page_owner_sort
69
70	2) Enable page owner
71	Add "page_owner=on" to boot cmdline.
72
73	3) Do the job what you want to debug
74
75	4) Analyze information from page owner
76	cat /sys/kernel/debug/page_owner > page_owner_full.txt
77	grep -v ^PFN page_owner_full.txt > page_owner.txt
78	./page_owner_sort page_owner.txt sorted_page_owner.txt
79
80	See the result about who allocated each page
81	in the sorted_page_owner.txt.