blob: 090e6ee04536285398a81257e1f9a1f9d34e398d [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.
Daniel Baluta21b86bd2011-04-04 14:58:03 -070014Kmemleak is supported on x86, arm, powerpc, sparc, sh, microblaze and tile.
Catalin Marinas04f70332009-06-11 13:22:39 +010015
16Usage
17-----
18
19CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel
Catalin Marinasbab4a342009-06-26 17:38:26 +010020thread scans the memory every 10 minutes (by default) and prints the
Catalin Marinas4698c1f2009-06-26 17:38:27 +010021number of new unreferenced objects found. To display the details of all
22the possible memory leaks:
Catalin Marinas04f70332009-06-11 13:22:39 +010023
24 # mount -t debugfs nodev /sys/kernel/debug/
25 # cat /sys/kernel/debug/kmemleak
26
Catalin Marinas4698c1f2009-06-26 17:38:27 +010027To trigger an intermediate memory scan:
28
29 # echo scan > /sys/kernel/debug/kmemleak
30
Luis R. Rodriguez30b37102009-09-04 17:44:51 -070031To clear the list of all current possible memory leaks:
32
33 # echo clear > /sys/kernel/debug/kmemleak
34
35New leaks will then come up upon reading /sys/kernel/debug/kmemleak
36again.
37
Catalin Marinas04f70332009-06-11 13:22:39 +010038Note that the orphan objects are listed in the order they were allocated
39and one object at the beginning of the list may cause other subsequent
40objects to be reported as orphan.
41
42Memory scanning parameters can be modified at run-time by writing to the
43/sys/kernel/debug/kmemleak file. The following parameters are supported:
44
45 off - disable kmemleak (irreversible)
Catalin Marinase0a2a162009-06-26 17:38:25 +010046 stack=on - enable the task stacks scanning (default)
Catalin Marinas04f70332009-06-11 13:22:39 +010047 stack=off - disable the tasks stacks scanning
Catalin Marinase0a2a162009-06-26 17:38:25 +010048 scan=on - start the automatic memory scanning thread (default)
Catalin Marinas04f70332009-06-11 13:22:39 +010049 scan=off - stop the automatic memory scanning thread
Catalin Marinase0a2a162009-06-26 17:38:25 +010050 scan=<secs> - set the automatic memory scanning period in seconds
51 (default 600, 0 to stop the automatic scanning)
Catalin Marinas4698c1f2009-06-26 17:38:27 +010052 scan - trigger a memory scan
Luis R. Rodriguez30b37102009-09-04 17:44:51 -070053 clear - clear list of current memory leak suspects, done by
54 marking all current reported unreferenced objects grey
Catalin Marinas189d84e2009-08-27 14:29:15 +010055 dump=<addr> - dump information about the object found at <addr>
Catalin Marinas04f70332009-06-11 13:22:39 +010056
57Kmemleak can also be disabled at boot-time by passing "kmemleak=off" on
58the kernel command line.
59
Catalin Marinasa9d90582009-06-25 10:16:11 +010060Memory may be allocated or freed before kmemleak is initialised and
61these actions are stored in an early log buffer. The size of this buffer
62is configured via the CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE option.
63
Catalin Marinas04f70332009-06-11 13:22:39 +010064Basic Algorithm
65---------------
66
67The memory allocations via kmalloc, vmalloc, kmem_cache_alloc and
68friends are traced and the pointers, together with additional
69information like size and stack trace, are stored in a prio search tree.
70The corresponding freeing function calls are tracked and the pointers
71removed from the kmemleak data structures.
72
73An allocated block of memory is considered orphan if no pointer to its
74start address or to any location inside the block can be found by
75scanning the memory (including saved registers). This means that there
76might be no way for the kernel to pass the address of the allocated
77block to a freeing function and therefore the block is considered a
78memory leak.
79
80The scanning algorithm steps:
81
82 1. mark all objects as white (remaining white objects will later be
83 considered orphan)
84 2. scan the memory starting with the data section and stacks, checking
85 the values against the addresses stored in the prio search tree. If
86 a pointer to a white object is found, the object is added to the
87 gray list
88 3. scan the gray objects for matching addresses (some white objects
89 can become gray and added at the end of the gray list) until the
90 gray set is finished
91 4. the remaining white objects are considered orphan and reported via
92 /sys/kernel/debug/kmemleak
93
94Some allocated memory blocks have pointers stored in the kernel's
95internal data structures and they cannot be detected as orphans. To
96avoid this, kmemleak can also store the number of values pointing to an
97address inside the block address range that need to be found so that the
98block is not considered a leak. One example is __vmalloc().
99
Luis R. Rodriguez30b37102009-09-04 17:44:51 -0700100Testing specific sections with kmemleak
101---------------------------------------
102
103Upon initial bootup your /sys/kernel/debug/kmemleak output page may be
104quite extensive. This can also be the case if you have very buggy code
105when doing development. To work around these situations you can use the
106'clear' command to clear all reported unreferenced objects from the
107/sys/kernel/debug/kmemleak output. By issuing a 'scan' after a 'clear'
108you can find new unreferenced objects; this should help with testing
109specific sections of code.
110
111To test a critical section on demand with a clean kmemleak do:
112
113 # echo clear > /sys/kernel/debug/kmemleak
114 ... test your kernel or modules ...
115 # echo scan > /sys/kernel/debug/kmemleak
116
117Then as usual to get your report with:
118
119 # cat /sys/kernel/debug/kmemleak
120
Catalin Marinas04f70332009-06-11 13:22:39 +0100121Kmemleak API
122------------
123
124See the include/linux/kmemleak.h header for the functions prototype.
125
126kmemleak_init - initialize kmemleak
127kmemleak_alloc - notify of a memory block allocation
128kmemleak_free - notify of a memory block freeing
129kmemleak_not_leak - mark an object as not a leak
130kmemleak_ignore - do not scan or report an object as leak
131kmemleak_scan_area - add scan areas inside a memory block
132kmemleak_no_scan - do not scan a memory block
133kmemleak_erase - erase an old value in a pointer variable
134kmemleak_alloc_recursive - as kmemleak_alloc but checks the recursiveness
135kmemleak_free_recursive - as kmemleak_free but checks the recursiveness
136
137Dealing with false positives/negatives
138--------------------------------------
139
140The false negatives are real memory leaks (orphan objects) but not
141reported by kmemleak because values found during the memory scanning
142point to such objects. To reduce the number of false negatives, kmemleak
143provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and
144kmemleak_erase functions (see above). The task stacks also increase the
145amount of false negatives and their scanning is not enabled by default.
146
147The false positives are objects wrongly reported as being memory leaks
148(orphan). For objects known not to be leaks, kmemleak provides the
149kmemleak_not_leak function. The kmemleak_ignore could also be used if
150the memory block is known not to contain other pointers and it will no
151longer be scanned.
152
153Some of the reported leaks are only transient, especially on SMP
154systems, because of pointers temporarily stored in CPU registers or
155stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing
156the minimum age of an object to be reported as a memory leak.
157
158Limitations and Drawbacks
159-------------------------
160
161The main drawback is the reduced performance of memory allocation and
162freeing. To avoid other penalties, the memory scanning is only performed
163when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is
164intended for debugging purposes where the performance might not be the
165most important requirement.
166
167To keep the algorithm simple, kmemleak scans for values pointing to any
168address inside a block's address range. This may lead to an increased
169number of false negatives. However, it is likely that a real memory leak
170will eventually become visible.
171
172Another source of false negatives is the data stored in non-pointer
173values. In a future version, kmemleak could only scan the pointer
174members in the allocated structures. This feature would solve many of
175the false negative cases described above.
176
177The tool can report false positives. These are cases where an allocated
178block doesn't need to be freed (some cases in the init_call functions),
179the pointer is calculated by other methods than the usual container_of
180macro or the pointer is stored in a location not scanned by kmemleak.
181
Daniel Baluta21b86bd2011-04-04 14:58:03 -0700182Page allocations and ioremap are not tracked.