1 Kernel Memory Leak Detector 1 Kernel Memory Leak Detector
2 =========================== 2 ===========================
3 3
4 Kmemleak provides a way of detecting possible 4 Kmemleak provides a way of detecting possible kernel memory leaks in a
5 way similar to a `tracing garbage collector !! 5 way similar to a tracing garbage collector
6 <https://en.wikipedia.org/wiki/Tracing_garbage !! 6 (https://en.wikipedia.org/wiki/Garbage_collection_%28computer_science%29#Tracing_garbage_collectors),
7 with the difference that the orphan objects ar 7 with the difference that the orphan objects are not freed but only
8 reported via /sys/kernel/debug/kmemleak. A sim 8 reported via /sys/kernel/debug/kmemleak. A similar method is used by the
9 Valgrind tool (``memcheck --leak-check``) to d 9 Valgrind tool (``memcheck --leak-check``) to detect the memory leaks in
10 user-space applications. 10 user-space applications.
>> 11 Kmemleak is supported on x86, arm, powerpc, sparc, sh, microblaze, ppc, mips, s390 and tile.
11 12
12 Usage 13 Usage
13 ----- 14 -----
14 15
15 CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has 16 CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel
16 thread scans the memory every 10 minutes (by d 17 thread scans the memory every 10 minutes (by default) and prints the
17 number of new unreferenced objects found. If t !! 18 number of new unreferenced objects found. To display the details of all
18 mounted, mount with:: !! 19 the possible memory leaks::
19 20
20 # mount -t debugfs nodev /sys/kernel/debug/ 21 # mount -t debugfs nodev /sys/kernel/debug/
21 <<
22 To display the details of all the possible sca <<
23 <<
24 # cat /sys/kernel/debug/kmemleak 22 # cat /sys/kernel/debug/kmemleak
25 23
26 To trigger an intermediate memory scan:: 24 To trigger an intermediate memory scan::
27 25
28 # echo scan > /sys/kernel/debug/kmemleak 26 # echo scan > /sys/kernel/debug/kmemleak
29 27
30 To clear the list of all current possible memo 28 To clear the list of all current possible memory leaks::
31 29
32 # echo clear > /sys/kernel/debug/kmemleak 30 # echo clear > /sys/kernel/debug/kmemleak
33 31
34 New leaks will then come up upon reading ``/sy 32 New leaks will then come up upon reading ``/sys/kernel/debug/kmemleak``
35 again. 33 again.
36 34
37 Note that the orphan objects are listed in the 35 Note that the orphan objects are listed in the order they were allocated
38 and one object at the beginning of the list ma 36 and one object at the beginning of the list may cause other subsequent
39 objects to be reported as orphan. 37 objects to be reported as orphan.
40 38
41 Memory scanning parameters can be modified at 39 Memory scanning parameters can be modified at run-time by writing to the
42 ``/sys/kernel/debug/kmemleak`` file. The follo 40 ``/sys/kernel/debug/kmemleak`` file. The following parameters are supported:
43 41
44 - off 42 - off
45 disable kmemleak (irreversible) 43 disable kmemleak (irreversible)
46 - stack=on 44 - stack=on
47 enable the task stacks scanning (default) 45 enable the task stacks scanning (default)
48 - stack=off 46 - stack=off
49 disable the tasks stacks scanning 47 disable the tasks stacks scanning
50 - scan=on 48 - scan=on
51 start the automatic memory scanning thread 49 start the automatic memory scanning thread (default)
52 - scan=off 50 - scan=off
53 stop the automatic memory scanning thread 51 stop the automatic memory scanning thread
54 - scan=<secs> 52 - scan=<secs>
55 set the automatic memory scanning period i 53 set the automatic memory scanning period in seconds
56 (default 600, 0 to stop the automatic scan 54 (default 600, 0 to stop the automatic scanning)
57 - scan 55 - scan
58 trigger a memory scan 56 trigger a memory scan
59 - clear 57 - clear
60 clear list of current memory leak suspects 58 clear list of current memory leak suspects, done by
61 marking all current reported unreferenced 59 marking all current reported unreferenced objects grey,
62 or free all kmemleak objects if kmemleak h 60 or free all kmemleak objects if kmemleak has been disabled.
63 - dump=<addr> 61 - dump=<addr>
64 dump information about the object found at 62 dump information about the object found at <addr>
65 63
66 Kmemleak can also be disabled at boot-time by 64 Kmemleak can also be disabled at boot-time by passing ``kmemleak=off`` on
67 the kernel command line. 65 the kernel command line.
68 66
69 Memory may be allocated or freed before kmemle 67 Memory may be allocated or freed before kmemleak is initialised and
70 these actions are stored in an early log buffe 68 these actions are stored in an early log buffer. The size of this buffer
71 is configured via the CONFIG_DEBUG_KMEMLEAK_ME !! 69 is configured via the CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE option.
72 70
73 If CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF are enabl 71 If CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF are enabled, the kmemleak is
74 disabled by default. Passing ``kmemleak=on`` o 72 disabled by default. Passing ``kmemleak=on`` on the kernel command
75 line enables the function. 73 line enables the function.
76 74
77 If you are getting errors like "Error while wr <<
78 Invalid argument", make sure kmemleak is prope <<
79 <<
80 Basic Algorithm 75 Basic Algorithm
81 --------------- 76 ---------------
82 77
83 The memory allocations via :c:func:`kmalloc`, 78 The memory allocations via :c:func:`kmalloc`, :c:func:`vmalloc`,
84 :c:func:`kmem_cache_alloc` and 79 :c:func:`kmem_cache_alloc` and
85 friends are traced and the pointers, together 80 friends are traced and the pointers, together with additional
86 information like size and stack trace, are sto 81 information like size and stack trace, are stored in a rbtree.
87 The corresponding freeing function calls are t 82 The corresponding freeing function calls are tracked and the pointers
88 removed from the kmemleak data structures. 83 removed from the kmemleak data structures.
89 84
90 An allocated block of memory is considered orp 85 An allocated block of memory is considered orphan if no pointer to its
91 start address or to any location inside the bl 86 start address or to any location inside the block can be found by
92 scanning the memory (including saved registers 87 scanning the memory (including saved registers). This means that there
93 might be no way for the kernel to pass the add 88 might be no way for the kernel to pass the address of the allocated
94 block to a freeing function and therefore the 89 block to a freeing function and therefore the block is considered a
95 memory leak. 90 memory leak.
96 91
97 The scanning algorithm steps: 92 The scanning algorithm steps:
98 93
99 1. mark all objects as white (remaining whit 94 1. mark all objects as white (remaining white objects will later be
100 considered orphan) 95 considered orphan)
101 2. scan the memory starting with the data se 96 2. scan the memory starting with the data section and stacks, checking
102 the values against the addresses stored i 97 the values against the addresses stored in the rbtree. If
103 a pointer to a white object is found, the 98 a pointer to a white object is found, the object is added to the
104 gray list 99 gray list
105 3. scan the gray objects for matching addres 100 3. scan the gray objects for matching addresses (some white objects
106 can become gray and added at the end of t 101 can become gray and added at the end of the gray list) until the
107 gray set is finished 102 gray set is finished
108 4. the remaining white objects are considere 103 4. the remaining white objects are considered orphan and reported via
109 /sys/kernel/debug/kmemleak 104 /sys/kernel/debug/kmemleak
110 105
111 Some allocated memory blocks have pointers sto 106 Some allocated memory blocks have pointers stored in the kernel's
112 internal data structures and they cannot be de 107 internal data structures and they cannot be detected as orphans. To
113 avoid this, kmemleak can also store the number 108 avoid this, kmemleak can also store the number of values pointing to an
114 address inside the block address range that ne 109 address inside the block address range that need to be found so that the
115 block is not considered a leak. One example is 110 block is not considered a leak. One example is __vmalloc().
116 111
117 Testing specific sections with kmemleak 112 Testing specific sections with kmemleak
118 --------------------------------------- 113 ---------------------------------------
119 114
120 Upon initial bootup your /sys/kernel/debug/kme 115 Upon initial bootup your /sys/kernel/debug/kmemleak output page may be
121 quite extensive. This can also be the case if 116 quite extensive. This can also be the case if you have very buggy code
122 when doing development. To work around these s 117 when doing development. To work around these situations you can use the
123 'clear' command to clear all reported unrefere 118 'clear' command to clear all reported unreferenced objects from the
124 /sys/kernel/debug/kmemleak output. By issuing 119 /sys/kernel/debug/kmemleak output. By issuing a 'scan' after a 'clear'
125 you can find new unreferenced objects; this sh 120 you can find new unreferenced objects; this should help with testing
126 specific sections of code. 121 specific sections of code.
127 122
128 To test a critical section on demand with a cl 123 To test a critical section on demand with a clean kmemleak do::
129 124
130 # echo clear > /sys/kernel/debug/kmemleak 125 # echo clear > /sys/kernel/debug/kmemleak
131 ... test your kernel or modules ... 126 ... test your kernel or modules ...
132 # echo scan > /sys/kernel/debug/kmemleak 127 # echo scan > /sys/kernel/debug/kmemleak
133 128
134 Then as usual to get your report with:: 129 Then as usual to get your report with::
135 130
136 # cat /sys/kernel/debug/kmemleak 131 # cat /sys/kernel/debug/kmemleak
137 132
138 Freeing kmemleak internal objects 133 Freeing kmemleak internal objects
139 --------------------------------- 134 ---------------------------------
140 135
141 To allow access to previously found memory lea 136 To allow access to previously found memory leaks after kmemleak has been
142 disabled by the user or due to an fatal error, 137 disabled by the user or due to an fatal error, internal kmemleak objects
143 won't be freed when kmemleak is disabled, and 138 won't be freed when kmemleak is disabled, and those objects may occupy
144 a large part of physical memory. 139 a large part of physical memory.
145 140
146 In this situation, you may reclaim memory with 141 In this situation, you may reclaim memory with::
147 142
148 # echo clear > /sys/kernel/debug/kmemleak 143 # echo clear > /sys/kernel/debug/kmemleak
149 144
150 Kmemleak API 145 Kmemleak API
151 ------------ 146 ------------
152 147
153 See the include/linux/kmemleak.h header for th 148 See the include/linux/kmemleak.h header for the functions prototype.
154 149
155 - ``kmemleak_init`` - initialize 150 - ``kmemleak_init`` - initialize kmemleak
156 - ``kmemleak_alloc`` - notify of a 151 - ``kmemleak_alloc`` - notify of a memory block allocation
157 - ``kmemleak_alloc_percpu`` - notify of a 152 - ``kmemleak_alloc_percpu`` - notify of a percpu memory block allocation
158 - ``kmemleak_vmalloc`` - notify of a 153 - ``kmemleak_vmalloc`` - notify of a vmalloc() memory allocation
159 - ``kmemleak_free`` - notify of a 154 - ``kmemleak_free`` - notify of a memory block freeing
160 - ``kmemleak_free_part`` - notify of a 155 - ``kmemleak_free_part`` - notify of a partial memory block freeing
161 - ``kmemleak_free_percpu`` - notify of a 156 - ``kmemleak_free_percpu`` - notify of a percpu memory block freeing
162 - ``kmemleak_update_trace`` - update obje 157 - ``kmemleak_update_trace`` - update object allocation stack trace
163 - ``kmemleak_not_leak`` - mark an object as n 158 - ``kmemleak_not_leak`` - mark an object as not a leak
164 - ``kmemleak_ignore`` - do not scan 159 - ``kmemleak_ignore`` - do not scan or report an object as leak
165 - ``kmemleak_scan_area`` - add scan ar 160 - ``kmemleak_scan_area`` - add scan areas inside a memory block
166 - ``kmemleak_no_scan`` - do not scan a memor 161 - ``kmemleak_no_scan`` - do not scan a memory block
167 - ``kmemleak_erase`` - erase an ol 162 - ``kmemleak_erase`` - erase an old value in a pointer variable
168 - ``kmemleak_alloc_recursive`` - as kmemleak_a 163 - ``kmemleak_alloc_recursive`` - as kmemleak_alloc but checks the recursiveness
169 - ``kmemleak_free_recursive`` - as kmemleak 164 - ``kmemleak_free_recursive`` - as kmemleak_free but checks the recursiveness
170 165
171 The following functions take a physical addres 166 The following functions take a physical address as the object pointer
172 and only perform the corresponding action if t 167 and only perform the corresponding action if the address has a lowmem
173 mapping: 168 mapping:
174 169
175 - ``kmemleak_alloc_phys`` 170 - ``kmemleak_alloc_phys``
176 - ``kmemleak_free_part_phys`` 171 - ``kmemleak_free_part_phys``
>> 172 - ``kmemleak_not_leak_phys``
177 - ``kmemleak_ignore_phys`` 173 - ``kmemleak_ignore_phys``
178 174
179 Dealing with false positives/negatives 175 Dealing with false positives/negatives
180 -------------------------------------- 176 --------------------------------------
181 177
182 The false negatives are real memory leaks (orp 178 The false negatives are real memory leaks (orphan objects) but not
183 reported by kmemleak because values found duri 179 reported by kmemleak because values found during the memory scanning
184 point to such objects. To reduce the number of 180 point to such objects. To reduce the number of false negatives, kmemleak
185 provides the kmemleak_ignore, kmemleak_scan_ar 181 provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and
186 kmemleak_erase functions (see above). The task 182 kmemleak_erase functions (see above). The task stacks also increase the
187 amount of false negatives and their scanning i 183 amount of false negatives and their scanning is not enabled by default.
188 184
189 The false positives are objects wrongly report 185 The false positives are objects wrongly reported as being memory leaks
190 (orphan). For objects known not to be leaks, k 186 (orphan). For objects known not to be leaks, kmemleak provides the
191 kmemleak_not_leak function. The kmemleak_ignor 187 kmemleak_not_leak function. The kmemleak_ignore could also be used if
192 the memory block is known not to contain other 188 the memory block is known not to contain other pointers and it will no
193 longer be scanned. 189 longer be scanned.
194 190
195 Some of the reported leaks are only transient, 191 Some of the reported leaks are only transient, especially on SMP
196 systems, because of pointers temporarily store 192 systems, because of pointers temporarily stored in CPU registers or
197 stacks. Kmemleak defines MSECS_MIN_AGE (defaul 193 stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing
198 the minimum age of an object to be reported as 194 the minimum age of an object to be reported as a memory leak.
199 195
200 Limitations and Drawbacks 196 Limitations and Drawbacks
201 ------------------------- 197 -------------------------
202 198
203 The main drawback is the reduced performance o 199 The main drawback is the reduced performance of memory allocation and
204 freeing. To avoid other penalties, the memory 200 freeing. To avoid other penalties, the memory scanning is only performed
205 when the /sys/kernel/debug/kmemleak file is re 201 when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is
206 intended for debugging purposes where the perf 202 intended for debugging purposes where the performance might not be the
207 most important requirement. 203 most important requirement.
208 204
209 To keep the algorithm simple, kmemleak scans f 205 To keep the algorithm simple, kmemleak scans for values pointing to any
210 address inside a block's address range. This m 206 address inside a block's address range. This may lead to an increased
211 number of false negatives. However, it is like 207 number of false negatives. However, it is likely that a real memory leak
212 will eventually become visible. 208 will eventually become visible.
213 209
214 Another source of false negatives is the data 210 Another source of false negatives is the data stored in non-pointer
215 values. In a future version, kmemleak could on 211 values. In a future version, kmemleak could only scan the pointer
216 members in the allocated structures. This feat 212 members in the allocated structures. This feature would solve many of
217 the false negative cases described above. 213 the false negative cases described above.
218 214
219 The tool can report false positives. These are 215 The tool can report false positives. These are cases where an allocated
220 block doesn't need to be freed (some cases in 216 block doesn't need to be freed (some cases in the init_call functions),
221 the pointer is calculated by other methods tha 217 the pointer is calculated by other methods than the usual container_of
222 macro or the pointer is stored in a location n 218 macro or the pointer is stored in a location not scanned by kmemleak.
223 219
224 Page allocations and ioremap are not tracked. 220 Page allocations and ioremap are not tracked.
225 <<
226 Testing with kmemleak-test <<
227 -------------------------- <<
228 <<
229 To check if you have all set up to use kmemlea <<
230 module, a module that deliberately leaks memor <<
231 as module (it can't be used as built-in) and b <<
232 enabled. Load the module and perform a scan wi <<
233 <<
234 # modprobe kmemleak-test <<
235 # echo scan > /sys/kernel/debug/kmemle <<
236 <<
237 Note that the you may not get results instantl <<
238 kmemleak gets results, it'll log ``kmemleak: < <<
239 memory leaks``. Then read the file to see then <<
240 <<
241 # cat /sys/kernel/debug/kmemleak <<
242 unreferenced object 0xffff89862ca702e8 <<
243 comm "modprobe", pid 2088, jiffies 4 <<
244 hex dump (first 32 bytes): <<
245 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6 <<
246 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6 <<
247 backtrace: <<
248 [<00000000e0a73ec7>] 0xffffffffc01 <<
249 [<000000000c5d2a46>] do_one_initca <<
250 [<0000000046db7e0a>] do_init_modul <<
251 [<00000000542b9814>] load_module+0 <<
252 [<00000000c2850256>] __do_sys_fini <<
253 [<000000006564e7ef>] do_syscall_64 <<
254 [<000000007c873fa6>] entry_SYSCALL <<
255 ... <<
256 <<
257 Removing the module with ``rmmod kmemleak_test <<
258 kmemleak results. <<
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.