Commit | Line | Data |
---|---|---|
04f70336 CM |
1 | Kernel Memory Leak Detector |
2 | =========================== | |
3 | ||
4 | Introduction | |
5 | ------------ | |
6 | ||
7 | Kmemleak provides a way of detecting possible kernel memory leaks in a | |
8 | way similar to a tracing garbage collector | |
9 | (http://en.wikipedia.org/wiki/Garbage_collection_%28computer_science%29#Tracing_garbage_collectors), | |
10 | with the difference that the orphan objects are not freed but only | |
11 | reported via /sys/kernel/debug/kmemleak. A similar method is used by the | |
12 | Valgrind tool (memcheck --leak-check) to detect the memory leaks in | |
13 | user-space applications. | |
14 | ||
15 | Usage | |
16 | ----- | |
17 | ||
18 | CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel | |
bab4a34a CM |
19 | thread scans the memory every 10 minutes (by default) and prints the |
20 | number of new unreferenced objects found. To trigger an intermediate | |
21 | scan and display the details of all the possible memory leaks: | |
04f70336 CM |
22 | |
23 | # mount -t debugfs nodev /sys/kernel/debug/ | |
24 | # cat /sys/kernel/debug/kmemleak | |
25 | ||
26 | Note that the orphan objects are listed in the order they were allocated | |
27 | and one object at the beginning of the list may cause other subsequent | |
28 | objects to be reported as orphan. | |
29 | ||
30 | Memory scanning parameters can be modified at run-time by writing to the | |
31 | /sys/kernel/debug/kmemleak file. The following parameters are supported: | |
32 | ||
33 | off - disable kmemleak (irreversible) | |
e0a2a160 | 34 | stack=on - enable the task stacks scanning (default) |
04f70336 | 35 | stack=off - disable the tasks stacks scanning |
e0a2a160 | 36 | scan=on - start the automatic memory scanning thread (default) |
04f70336 | 37 | scan=off - stop the automatic memory scanning thread |
e0a2a160 CM |
38 | scan=<secs> - set the automatic memory scanning period in seconds |
39 | (default 600, 0 to stop the automatic scanning) | |
04f70336 CM |
40 | |
41 | Kmemleak can also be disabled at boot-time by passing "kmemleak=off" on | |
42 | the kernel command line. | |
43 | ||
a9d9058a CM |
44 | Memory may be allocated or freed before kmemleak is initialised and |
45 | these actions are stored in an early log buffer. The size of this buffer | |
46 | is configured via the CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE option. | |
47 | ||
04f70336 CM |
48 | Basic Algorithm |
49 | --------------- | |
50 | ||
51 | The memory allocations via kmalloc, vmalloc, kmem_cache_alloc and | |
52 | friends are traced and the pointers, together with additional | |
53 | information like size and stack trace, are stored in a prio search tree. | |
54 | The corresponding freeing function calls are tracked and the pointers | |
55 | removed from the kmemleak data structures. | |
56 | ||
57 | An allocated block of memory is considered orphan if no pointer to its | |
58 | start address or to any location inside the block can be found by | |
59 | scanning the memory (including saved registers). This means that there | |
60 | might be no way for the kernel to pass the address of the allocated | |
61 | block to a freeing function and therefore the block is considered a | |
62 | memory leak. | |
63 | ||
64 | The scanning algorithm steps: | |
65 | ||
66 | 1. mark all objects as white (remaining white objects will later be | |
67 | considered orphan) | |
68 | 2. scan the memory starting with the data section and stacks, checking | |
69 | the values against the addresses stored in the prio search tree. If | |
70 | a pointer to a white object is found, the object is added to the | |
71 | gray list | |
72 | 3. scan the gray objects for matching addresses (some white objects | |
73 | can become gray and added at the end of the gray list) until the | |
74 | gray set is finished | |
75 | 4. the remaining white objects are considered orphan and reported via | |
76 | /sys/kernel/debug/kmemleak | |
77 | ||
78 | Some allocated memory blocks have pointers stored in the kernel's | |
79 | internal data structures and they cannot be detected as orphans. To | |
80 | avoid this, kmemleak can also store the number of values pointing to an | |
81 | address inside the block address range that need to be found so that the | |
82 | block is not considered a leak. One example is __vmalloc(). | |
83 | ||
84 | Kmemleak API | |
85 | ------------ | |
86 | ||
87 | See the include/linux/kmemleak.h header for the functions prototype. | |
88 | ||
89 | kmemleak_init - initialize kmemleak | |
90 | kmemleak_alloc - notify of a memory block allocation | |
91 | kmemleak_free - notify of a memory block freeing | |
92 | kmemleak_not_leak - mark an object as not a leak | |
93 | kmemleak_ignore - do not scan or report an object as leak | |
94 | kmemleak_scan_area - add scan areas inside a memory block | |
95 | kmemleak_no_scan - do not scan a memory block | |
96 | kmemleak_erase - erase an old value in a pointer variable | |
97 | kmemleak_alloc_recursive - as kmemleak_alloc but checks the recursiveness | |
98 | kmemleak_free_recursive - as kmemleak_free but checks the recursiveness | |
99 | ||
100 | Dealing with false positives/negatives | |
101 | -------------------------------------- | |
102 | ||
103 | The false negatives are real memory leaks (orphan objects) but not | |
104 | reported by kmemleak because values found during the memory scanning | |
105 | point to such objects. To reduce the number of false negatives, kmemleak | |
106 | provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and | |
107 | kmemleak_erase functions (see above). The task stacks also increase the | |
108 | amount of false negatives and their scanning is not enabled by default. | |
109 | ||
110 | The false positives are objects wrongly reported as being memory leaks | |
111 | (orphan). For objects known not to be leaks, kmemleak provides the | |
112 | kmemleak_not_leak function. The kmemleak_ignore could also be used if | |
113 | the memory block is known not to contain other pointers and it will no | |
114 | longer be scanned. | |
115 | ||
116 | Some of the reported leaks are only transient, especially on SMP | |
117 | systems, because of pointers temporarily stored in CPU registers or | |
118 | stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing | |
119 | the minimum age of an object to be reported as a memory leak. | |
120 | ||
121 | Limitations and Drawbacks | |
122 | ------------------------- | |
123 | ||
124 | The main drawback is the reduced performance of memory allocation and | |
125 | freeing. To avoid other penalties, the memory scanning is only performed | |
126 | when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is | |
127 | intended for debugging purposes where the performance might not be the | |
128 | most important requirement. | |
129 | ||
130 | To keep the algorithm simple, kmemleak scans for values pointing to any | |
131 | address inside a block's address range. This may lead to an increased | |
132 | number of false negatives. However, it is likely that a real memory leak | |
133 | will eventually become visible. | |
134 | ||
135 | Another source of false negatives is the data stored in non-pointer | |
136 | values. In a future version, kmemleak could only scan the pointer | |
137 | members in the allocated structures. This feature would solve many of | |
138 | the false negative cases described above. | |
139 | ||
140 | The tool can report false positives. These are cases where an allocated | |
141 | block doesn't need to be freed (some cases in the init_call functions), | |
142 | the pointer is calculated by other methods than the usual container_of | |
143 | macro or the pointer is stored in a location not scanned by kmemleak. | |
144 | ||
145 | Page allocations and ioremap are not tracked. Only the ARM and x86 | |
146 | architectures are currently supported. |