Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 1 | Kernel Memory Leak Detector |
| 2 | =========================== |
| 3 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 4 | Kmemleak provides a way of detecting possible kernel memory leaks in a |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 5 | way similar to a `tracing garbage collector |
| 6 | <https://en.wikipedia.org/wiki/Tracing_garbage_collection>`_, |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 7 | with the difference that the orphan objects are not freed but only |
| 8 | reported via /sys/kernel/debug/kmemleak. A similar method is used by the |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 9 | Valgrind tool (``memcheck --leak-check``) to detect the memory leaks in |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 10 | user-space applications. |
| 11 | |
| 12 | Usage |
| 13 | ----- |
| 14 | |
| 15 | CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel |
Catalin Marinas | bab4a34 | 2009-06-26 17:38:26 +0100 | [diff] [blame] | 16 | thread scans the memory every 10 minutes (by default) and prints the |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 17 | number of new unreferenced objects found. If the ``debugfs`` isn't already |
| 18 | mounted, mount with:: |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 19 | |
| 20 | # mount -t debugfs nodev /sys/kernel/debug/ |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 21 | |
| 22 | To display the details of all the possible scanned memory leaks:: |
| 23 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 24 | # cat /sys/kernel/debug/kmemleak |
| 25 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 26 | To trigger an intermediate memory scan:: |
Catalin Marinas | 4698c1f | 2009-06-26 17:38:27 +0100 | [diff] [blame] | 27 | |
| 28 | # echo scan > /sys/kernel/debug/kmemleak |
| 29 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 30 | To clear the list of all current possible memory leaks:: |
Luis R. Rodriguez | 30b3710 | 2009-09-04 17:44:51 -0700 | [diff] [blame] | 31 | |
| 32 | # echo clear > /sys/kernel/debug/kmemleak |
| 33 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 34 | New leaks will then come up upon reading ``/sys/kernel/debug/kmemleak`` |
Luis R. Rodriguez | 30b3710 | 2009-09-04 17:44:51 -0700 | [diff] [blame] | 35 | again. |
| 36 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 37 | Note that the orphan objects are listed in the order they were allocated |
| 38 | and one object at the beginning of the list may cause other subsequent |
| 39 | objects to be reported as orphan. |
| 40 | |
| 41 | Memory scanning parameters can be modified at run-time by writing to the |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 42 | ``/sys/kernel/debug/kmemleak`` file. The following parameters are supported: |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 43 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 44 | - off |
| 45 | disable kmemleak (irreversible) |
| 46 | - stack=on |
| 47 | enable the task stacks scanning (default) |
| 48 | - stack=off |
| 49 | disable the tasks stacks scanning |
| 50 | - scan=on |
| 51 | start the automatic memory scanning thread (default) |
| 52 | - scan=off |
| 53 | stop the automatic memory scanning thread |
| 54 | - scan=<secs> |
| 55 | set the automatic memory scanning period in seconds |
| 56 | (default 600, 0 to stop the automatic scanning) |
| 57 | - scan |
| 58 | trigger a memory scan |
| 59 | - clear |
| 60 | clear list of current memory leak suspects, done by |
| 61 | marking all current reported unreferenced objects grey, |
| 62 | or free all kmemleak objects if kmemleak has been disabled. |
| 63 | - dump=<addr> |
| 64 | dump information about the object found at <addr> |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 65 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 66 | Kmemleak can also be disabled at boot-time by passing ``kmemleak=off`` on |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 67 | the kernel command line. |
| 68 | |
Catalin Marinas | a9d9058 | 2009-06-25 10:16:11 +0100 | [diff] [blame] | 69 | Memory may be allocated or freed before kmemleak is initialised and |
| 70 | these actions are stored in an early log buffer. The size of this buffer |
Jeremy Cline | 2c861bf | 2019-09-25 14:31:14 +0000 | [diff] [blame] | 71 | is configured via the CONFIG_DEBUG_KMEMLEAK_MEM_POOL_SIZE option. |
Catalin Marinas | a9d9058 | 2009-06-25 10:16:11 +0100 | [diff] [blame] | 72 | |
Masanari Iida | 6808a40 | 2014-10-24 21:25:00 +0900 | [diff] [blame] | 73 | If CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF are enabled, the kmemleak is |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 74 | disabled by default. Passing ``kmemleak=on`` on the kernel command |
Masanari Iida | 6808a40 | 2014-10-24 21:25:00 +0900 | [diff] [blame] | 75 | line enables the function. |
| 76 | |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 77 | If you are getting errors like "Error while writing to stdout" or "write_loop: |
| 78 | Invalid argument", make sure kmemleak is properly enabled. |
| 79 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 80 | Basic Algorithm |
| 81 | --------------- |
| 82 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 83 | The memory allocations via :c:func:`kmalloc`, :c:func:`vmalloc`, |
| 84 | :c:func:`kmem_cache_alloc` and |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 85 | friends are traced and the pointers, together with additional |
Wang YanQing | 4762c98 | 2014-04-03 14:50:38 -0700 | [diff] [blame] | 86 | information like size and stack trace, are stored in a rbtree. |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 87 | The corresponding freeing function calls are tracked and the pointers |
| 88 | removed from the kmemleak data structures. |
| 89 | |
| 90 | An allocated block of memory is considered orphan if no pointer to its |
| 91 | start address or to any location inside the block can be found by |
| 92 | scanning the memory (including saved registers). This means that there |
| 93 | might be no way for the kernel to pass the address of the allocated |
| 94 | block to a freeing function and therefore the block is considered a |
| 95 | memory leak. |
| 96 | |
| 97 | The scanning algorithm steps: |
| 98 | |
| 99 | 1. mark all objects as white (remaining white objects will later be |
| 100 | considered orphan) |
| 101 | 2. scan the memory starting with the data section and stacks, checking |
Wang YanQing | 4762c98 | 2014-04-03 14:50:38 -0700 | [diff] [blame] | 102 | the values against the addresses stored in the rbtree. If |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 103 | a pointer to a white object is found, the object is added to the |
| 104 | gray list |
| 105 | 3. scan the gray objects for matching addresses (some white objects |
| 106 | can become gray and added at the end of the gray list) until the |
| 107 | gray set is finished |
| 108 | 4. the remaining white objects are considered orphan and reported via |
| 109 | /sys/kernel/debug/kmemleak |
| 110 | |
| 111 | Some allocated memory blocks have pointers stored in the kernel's |
| 112 | internal data structures and they cannot be detected as orphans. To |
| 113 | avoid this, kmemleak can also store the number of values pointing to an |
| 114 | address inside the block address range that need to be found so that the |
| 115 | block is not considered a leak. One example is __vmalloc(). |
| 116 | |
Luis R. Rodriguez | 30b3710 | 2009-09-04 17:44:51 -0700 | [diff] [blame] | 117 | Testing specific sections with kmemleak |
| 118 | --------------------------------------- |
| 119 | |
| 120 | Upon initial bootup your /sys/kernel/debug/kmemleak output page may be |
| 121 | quite extensive. This can also be the case if you have very buggy code |
| 122 | when doing development. To work around these situations you can use the |
| 123 | 'clear' command to clear all reported unreferenced objects from the |
| 124 | /sys/kernel/debug/kmemleak output. By issuing a 'scan' after a 'clear' |
| 125 | you can find new unreferenced objects; this should help with testing |
| 126 | specific sections of code. |
| 127 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 128 | To test a critical section on demand with a clean kmemleak do:: |
Luis R. Rodriguez | 30b3710 | 2009-09-04 17:44:51 -0700 | [diff] [blame] | 129 | |
| 130 | # echo clear > /sys/kernel/debug/kmemleak |
| 131 | ... test your kernel or modules ... |
| 132 | # echo scan > /sys/kernel/debug/kmemleak |
| 133 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 134 | Then as usual to get your report with:: |
Luis R. Rodriguez | 30b3710 | 2009-09-04 17:44:51 -0700 | [diff] [blame] | 135 | |
| 136 | # cat /sys/kernel/debug/kmemleak |
| 137 | |
Li Zefan | c89da70 | 2014-04-03 14:46:27 -0700 | [diff] [blame] | 138 | Freeing kmemleak internal objects |
| 139 | --------------------------------- |
| 140 | |
Rahul Bedarkar | abb3b1f | 2014-07-31 23:50:19 +0530 | [diff] [blame] | 141 | To allow access to previously found memory leaks after kmemleak has been |
Li Zefan | c89da70 | 2014-04-03 14:46:27 -0700 | [diff] [blame] | 142 | disabled by the user or due to an fatal error, internal kmemleak objects |
| 143 | won't be freed when kmemleak is disabled, and those objects may occupy |
| 144 | a large part of physical memory. |
| 145 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 146 | In this situation, you may reclaim memory with:: |
Li Zefan | c89da70 | 2014-04-03 14:46:27 -0700 | [diff] [blame] | 147 | |
| 148 | # echo clear > /sys/kernel/debug/kmemleak |
| 149 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 150 | Kmemleak API |
| 151 | ------------ |
| 152 | |
| 153 | See the include/linux/kmemleak.h header for the functions prototype. |
| 154 | |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 155 | - ``kmemleak_init`` - initialize kmemleak |
| 156 | - ``kmemleak_alloc`` - notify of a memory block allocation |
| 157 | - ``kmemleak_alloc_percpu`` - notify of a percpu memory block allocation |
Catalin Marinas | 94f4a16 | 2017-07-06 15:40:22 -0700 | [diff] [blame] | 158 | - ``kmemleak_vmalloc`` - notify of a vmalloc() memory allocation |
Jonathan Corbet | ca90a7a3 | 2016-08-07 15:46:10 -0600 | [diff] [blame] | 159 | - ``kmemleak_free`` - notify of a memory block freeing |
| 160 | - ``kmemleak_free_part`` - notify of a partial memory block freeing |
| 161 | - ``kmemleak_free_percpu`` - notify of a percpu memory block freeing |
| 162 | - ``kmemleak_update_trace`` - update object allocation stack trace |
| 163 | - ``kmemleak_not_leak`` - mark an object as not a leak |
| 164 | - ``kmemleak_ignore`` - do not scan or report an object as leak |
| 165 | - ``kmemleak_scan_area`` - add scan areas inside a memory block |
| 166 | - ``kmemleak_no_scan`` - do not scan a memory block |
| 167 | - ``kmemleak_erase`` - erase an old value in a pointer variable |
| 168 | - ``kmemleak_alloc_recursive`` - as kmemleak_alloc but checks the recursiveness |
| 169 | - ``kmemleak_free_recursive`` - as kmemleak_free but checks the recursiveness |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 170 | |
Catalin Marinas | 9099dae | 2016-10-11 13:55:11 -0700 | [diff] [blame] | 171 | The following functions take a physical address as the object pointer |
| 172 | and only perform the corresponding action if the address has a lowmem |
| 173 | mapping: |
| 174 | |
| 175 | - ``kmemleak_alloc_phys`` |
| 176 | - ``kmemleak_free_part_phys`` |
| 177 | - ``kmemleak_not_leak_phys`` |
| 178 | - ``kmemleak_ignore_phys`` |
| 179 | |
Catalin Marinas | 04f7033 | 2009-06-11 13:22:39 +0100 | [diff] [blame] | 180 | Dealing with false positives/negatives |
| 181 | -------------------------------------- |
| 182 | |
| 183 | The false negatives are real memory leaks (orphan objects) but not |
| 184 | reported by kmemleak because values found during the memory scanning |
| 185 | point to such objects. To reduce the number of false negatives, kmemleak |
| 186 | provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and |
| 187 | kmemleak_erase functions (see above). The task stacks also increase the |
| 188 | amount of false negatives and their scanning is not enabled by default. |
| 189 | |
| 190 | The false positives are objects wrongly reported as being memory leaks |
| 191 | (orphan). For objects known not to be leaks, kmemleak provides the |
| 192 | kmemleak_not_leak function. The kmemleak_ignore could also be used if |
| 193 | the memory block is known not to contain other pointers and it will no |
| 194 | longer be scanned. |
| 195 | |
| 196 | Some of the reported leaks are only transient, especially on SMP |
| 197 | systems, because of pointers temporarily stored in CPU registers or |
| 198 | stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing |
| 199 | the minimum age of an object to be reported as a memory leak. |
| 200 | |
| 201 | Limitations and Drawbacks |
| 202 | ------------------------- |
| 203 | |
| 204 | The main drawback is the reduced performance of memory allocation and |
| 205 | freeing. To avoid other penalties, the memory scanning is only performed |
| 206 | when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is |
| 207 | intended for debugging purposes where the performance might not be the |
| 208 | most important requirement. |
| 209 | |
| 210 | To keep the algorithm simple, kmemleak scans for values pointing to any |
| 211 | address inside a block's address range. This may lead to an increased |
| 212 | number of false negatives. However, it is likely that a real memory leak |
| 213 | will eventually become visible. |
| 214 | |
| 215 | Another source of false negatives is the data stored in non-pointer |
| 216 | values. In a future version, kmemleak could only scan the pointer |
| 217 | members in the allocated structures. This feature would solve many of |
| 218 | the false negative cases described above. |
| 219 | |
| 220 | The tool can report false positives. These are cases where an allocated |
| 221 | block doesn't need to be freed (some cases in the init_call functions), |
| 222 | the pointer is calculated by other methods than the usual container_of |
| 223 | macro or the pointer is stored in a location not scanned by kmemleak. |
| 224 | |
Daniel Baluta | 21b86bd | 2011-04-04 14:58:03 -0700 | [diff] [blame] | 225 | Page allocations and ioremap are not tracked. |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 226 | |
| 227 | Testing with kmemleak-test |
| 228 | -------------------------- |
| 229 | |
| 230 | To check if you have all set up to use kmemleak, you can use the kmemleak-test |
| 231 | module, a module that deliberately leaks memory. Set CONFIG_DEBUG_KMEMLEAK_TEST |
Hui Su | 1abbef4 | 2020-10-13 16:48:53 -0700 | [diff] [blame] | 232 | as module (it can't be used as built-in) and boot the kernel with kmemleak |
André Almeida | b7c3613 | 2019-07-11 20:53:46 -0700 | [diff] [blame] | 233 | enabled. Load the module and perform a scan with:: |
| 234 | |
| 235 | # modprobe kmemleak-test |
| 236 | # echo scan > /sys/kernel/debug/kmemleak |
| 237 | |
| 238 | Note that the you may not get results instantly or on the first scanning. When |
| 239 | kmemleak gets results, it'll log ``kmemleak: <count of leaks> new suspected |
| 240 | memory leaks``. Then read the file to see then:: |
| 241 | |
| 242 | # cat /sys/kernel/debug/kmemleak |
| 243 | unreferenced object 0xffff89862ca702e8 (size 32): |
| 244 | comm "modprobe", pid 2088, jiffies 4294680594 (age 375.486s) |
| 245 | hex dump (first 32 bytes): |
| 246 | 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b kkkkkkkkkkkkkkkk |
| 247 | 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b a5 kkkkkkkkkkkkkkk. |
| 248 | backtrace: |
| 249 | [<00000000e0a73ec7>] 0xffffffffc01d2036 |
| 250 | [<000000000c5d2a46>] do_one_initcall+0x41/0x1df |
| 251 | [<0000000046db7e0a>] do_init_module+0x55/0x200 |
| 252 | [<00000000542b9814>] load_module+0x203c/0x2480 |
| 253 | [<00000000c2850256>] __do_sys_finit_module+0xba/0xe0 |
| 254 | [<000000006564e7ef>] do_syscall_64+0x43/0x110 |
| 255 | [<000000007c873fa6>] entry_SYSCALL_64_after_hwframe+0x44/0xa9 |
| 256 | ... |
| 257 | |
| 258 | Removing the module with ``rmmod kmemleak_test`` should also trigger some |
| 259 | kmemleak results. |