blob: 34f6638aa5aceec30d290812fdc7fcebf3b86621 [file] [log] [blame]
Catalin Marinas04f70332009-06-11 13:22:39 +01001Kernel Memory Leak Detector
2===========================
3
4Introduction
5------------
6
7Kmemleak provides a way of detecting possible kernel memory leaks in a
8way similar to a tracing garbage collector
9(http://en.wikipedia.org/wiki/Garbage_collection_%28computer_science%29#Tracing_garbage_collectors),
10with the difference that the orphan objects are not freed but only
11reported via /sys/kernel/debug/kmemleak. A similar method is used by the
12Valgrind tool (memcheck --leak-check) to detect the memory leaks in
13user-space applications.
14
15Usage
16-----
17
18CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel
Catalin Marinasbab4a342009-06-26 17:38:26 +010019thread scans the memory every 10 minutes (by default) and prints the
Catalin Marinas4698c1f2009-06-26 17:38:27 +010020number of new unreferenced objects found. To display the details of all
21the possible memory leaks:
Catalin Marinas04f70332009-06-11 13:22:39 +010022
23 # mount -t debugfs nodev /sys/kernel/debug/
24 # cat /sys/kernel/debug/kmemleak
25
Catalin Marinas4698c1f2009-06-26 17:38:27 +010026To trigger an intermediate memory scan:
27
28 # echo scan > /sys/kernel/debug/kmemleak
29
Luis R. Rodriguez30b37102009-09-04 17:44:51 -070030To clear the list of all current possible memory leaks:
31
32 # echo clear > /sys/kernel/debug/kmemleak
33
34New leaks will then come up upon reading /sys/kernel/debug/kmemleak
35again.
36
Catalin Marinas04f70332009-06-11 13:22:39 +010037Note that the orphan objects are listed in the order they were allocated
38and one object at the beginning of the list may cause other subsequent
39objects to be reported as orphan.
40
41Memory scanning parameters can be modified at run-time by writing to the
42/sys/kernel/debug/kmemleak file. The following parameters are supported:
43
44 off - disable kmemleak (irreversible)
Catalin Marinase0a2a162009-06-26 17:38:25 +010045 stack=on - enable the task stacks scanning (default)
Catalin Marinas04f70332009-06-11 13:22:39 +010046 stack=off - disable the tasks stacks scanning
Catalin Marinase0a2a162009-06-26 17:38:25 +010047 scan=on - start the automatic memory scanning thread (default)
Catalin Marinas04f70332009-06-11 13:22:39 +010048 scan=off - stop the automatic memory scanning thread
Catalin Marinase0a2a162009-06-26 17:38:25 +010049 scan=<secs> - set the automatic memory scanning period in seconds
50 (default 600, 0 to stop the automatic scanning)
Catalin Marinas4698c1f2009-06-26 17:38:27 +010051 scan - trigger a memory scan
Luis R. Rodriguez30b37102009-09-04 17:44:51 -070052 clear - clear list of current memory leak suspects, done by
53 marking all current reported unreferenced objects grey
Catalin Marinas189d84e2009-08-27 14:29:15 +010054 dump=<addr> - dump information about the object found at <addr>
Catalin Marinas04f70332009-06-11 13:22:39 +010055
56Kmemleak can also be disabled at boot-time by passing "kmemleak=off" on
57the kernel command line.
58
Catalin Marinasa9d90582009-06-25 10:16:11 +010059Memory may be allocated or freed before kmemleak is initialised and
60these actions are stored in an early log buffer. The size of this buffer
61is configured via the CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE option.
62
Catalin Marinas04f70332009-06-11 13:22:39 +010063Basic Algorithm
64---------------
65
66The memory allocations via kmalloc, vmalloc, kmem_cache_alloc and
67friends are traced and the pointers, together with additional
68information like size and stack trace, are stored in a prio search tree.
69The corresponding freeing function calls are tracked and the pointers
70removed from the kmemleak data structures.
71
72An allocated block of memory is considered orphan if no pointer to its
73start address or to any location inside the block can be found by
74scanning the memory (including saved registers). This means that there
75might be no way for the kernel to pass the address of the allocated
76block to a freeing function and therefore the block is considered a
77memory leak.
78
79The scanning algorithm steps:
80
81 1. mark all objects as white (remaining white objects will later be
82 considered orphan)
83 2. scan the memory starting with the data section and stacks, checking
84 the values against the addresses stored in the prio search tree. If
85 a pointer to a white object is found, the object is added to the
86 gray list
87 3. scan the gray objects for matching addresses (some white objects
88 can become gray and added at the end of the gray list) until the
89 gray set is finished
90 4. the remaining white objects are considered orphan and reported via
91 /sys/kernel/debug/kmemleak
92
93Some allocated memory blocks have pointers stored in the kernel's
94internal data structures and they cannot be detected as orphans. To
95avoid this, kmemleak can also store the number of values pointing to an
96address inside the block address range that need to be found so that the
97block is not considered a leak. One example is __vmalloc().
98
Luis R. Rodriguez30b37102009-09-04 17:44:51 -070099Testing specific sections with kmemleak
100---------------------------------------
101
102Upon initial bootup your /sys/kernel/debug/kmemleak output page may be
103quite extensive. This can also be the case if you have very buggy code
104when doing development. To work around these situations you can use the
105'clear' command to clear all reported unreferenced objects from the
106/sys/kernel/debug/kmemleak output. By issuing a 'scan' after a 'clear'
107you can find new unreferenced objects; this should help with testing
108specific sections of code.
109
110To test a critical section on demand with a clean kmemleak do:
111
112 # echo clear > /sys/kernel/debug/kmemleak
113 ... test your kernel or modules ...
114 # echo scan > /sys/kernel/debug/kmemleak
115
116Then as usual to get your report with:
117
118 # cat /sys/kernel/debug/kmemleak
119
Catalin Marinas04f70332009-06-11 13:22:39 +0100120Kmemleak API
121------------
122
123See the include/linux/kmemleak.h header for the functions prototype.
124
125kmemleak_init - initialize kmemleak
126kmemleak_alloc - notify of a memory block allocation
127kmemleak_free - notify of a memory block freeing
128kmemleak_not_leak - mark an object as not a leak
129kmemleak_ignore - do not scan or report an object as leak
130kmemleak_scan_area - add scan areas inside a memory block
131kmemleak_no_scan - do not scan a memory block
132kmemleak_erase - erase an old value in a pointer variable
133kmemleak_alloc_recursive - as kmemleak_alloc but checks the recursiveness
134kmemleak_free_recursive - as kmemleak_free but checks the recursiveness
135
136Dealing with false positives/negatives
137--------------------------------------
138
139The false negatives are real memory leaks (orphan objects) but not
140reported by kmemleak because values found during the memory scanning
141point to such objects. To reduce the number of false negatives, kmemleak
142provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and
143kmemleak_erase functions (see above). The task stacks also increase the
144amount of false negatives and their scanning is not enabled by default.
145
146The false positives are objects wrongly reported as being memory leaks
147(orphan). For objects known not to be leaks, kmemleak provides the
148kmemleak_not_leak function. The kmemleak_ignore could also be used if
149the memory block is known not to contain other pointers and it will no
150longer be scanned.
151
152Some of the reported leaks are only transient, especially on SMP
153systems, because of pointers temporarily stored in CPU registers or
154stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing
155the minimum age of an object to be reported as a memory leak.
156
157Limitations and Drawbacks
158-------------------------
159
160The main drawback is the reduced performance of memory allocation and
161freeing. To avoid other penalties, the memory scanning is only performed
162when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is
163intended for debugging purposes where the performance might not be the
164most important requirement.
165
166To keep the algorithm simple, kmemleak scans for values pointing to any
167address inside a block's address range. This may lead to an increased
168number of false negatives. However, it is likely that a real memory leak
169will eventually become visible.
170
171Another source of false negatives is the data stored in non-pointer
172values. In a future version, kmemleak could only scan the pointer
173members in the allocated structures. This feature would solve many of
174the false negative cases described above.
175
176The tool can report false positives. These are cases where an allocated
177block doesn't need to be freed (some cases in the init_call functions),
178the pointer is calculated by other methods than the usual container_of
179macro or the pointer is stored in a location not scanned by kmemleak.
180
181Page allocations and ioremap are not tracked. Only the ARM and x86
182architectures are currently supported.