Documentation/kmemleak.txt

   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 Please check DEBUG_KMEMLEAK dependencies in lib/Kconfig.debug for supported
  16 architectures.
  17
  18 Usage
  19 -----
  20
  21 CONFIG_DEBUG_KMEMLEAK in "Kernel hacking" has to be enabled. A kernel
  22 thread scans the memory every 10 minutes (by default) and prints the
  23 number of new unreferenced objects found. To display the details of all
  24 the possible memory leaks:
  25
  26   # mount -t debugfs nodev /sys/kernel/debug/
  27   # cat /sys/kernel/debug/kmemleak
  28
  29 To trigger an intermediate memory scan:
  30
  31   # echo scan > /sys/kernel/debug/kmemleak
  32
  33 To clear the list of all current possible memory leaks:
  34
  35   # echo clear > /sys/kernel/debug/kmemleak
  36
  37 New leaks will then come up upon reading /sys/kernel/debug/kmemleak
  38 again.
  39
  40 Note that the orphan objects are listed in the order they were allocated
  41 and one object at the beginning of the list may cause other subsequent
  42 objects to be reported as orphan.
  43
  44 Memory scanning parameters can be modified at run-time by writing to the
  45 /sys/kernel/debug/kmemleak file. The following parameters are supported:
  46
  47   off           - disable kmemleak (irreversible)
  48   stack=on      - enable the task stacks scanning (default)
  49   stack=off     - disable the tasks stacks scanning
  50   scan=on       - start the automatic memory scanning thread (default)
  51   scan=off      - stop the automatic memory scanning thread
  52   scan=<secs>   - set the automatic memory scanning period in seconds
  53                   (default 600, 0 to stop the automatic scanning)
  54   scan          - trigger a memory scan
  55   clear         - clear list of current memory leak suspects, done by
  56                   marking all current reported unreferenced objects grey,
  57                   or free all kmemleak objects if kmemleak has been disabled.
  58   dump=<addr>   - dump information about the object found at <addr>
  59
  60 Kmemleak can also be disabled at boot-time by passing "kmemleak=off" on
  61 the kernel command line.
  62
  63 Memory may be allocated or freed before kmemleak is initialised and
  64 these actions are stored in an early log buffer. The size of this buffer
  65 is configured via the CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE option.
  66
  67 Basic Algorithm
  68 ---------------
  69
  70 The memory allocations via kmalloc, vmalloc, kmem_cache_alloc and
  71 friends are traced and the pointers, together with additional
  72 information like size and stack trace, are stored in a prio search tree.
  73 The corresponding freeing function calls are tracked and the pointers
  74 removed from the kmemleak data structures.
  75
  76 An allocated block of memory is considered orphan if no pointer to its
  77 start address or to any location inside the block can be found by
  78 scanning the memory (including saved registers). This means that there
  79 might be no way for the kernel to pass the address of the allocated
  80 block to a freeing function and therefore the block is considered a
  81 memory leak.
  82
  83 The scanning algorithm steps:
  84
  85   1. mark all objects as white (remaining white objects will later be
  86      considered orphan)
  87   2. scan the memory starting with the data section and stacks, checking
  88      the values against the addresses stored in the prio search tree. If
  89      a pointer to a white object is found, the object is added to the
  90      gray list
  91   3. scan the gray objects for matching addresses (some white objects
  92      can become gray and added at the end of the gray list) until the
  93      gray set is finished
  94   4. the remaining white objects are considered orphan and reported via
  95      /sys/kernel/debug/kmemleak
  96
  97 Some allocated memory blocks have pointers stored in the kernel's
  98 internal data structures and they cannot be detected as orphans. To
  99 avoid this, kmemleak can also store the number of values pointing to an
 100 address inside the block address range that need to be found so that the
 101 block is not considered a leak. One example is __vmalloc().
 102
 103 Testing specific sections with kmemleak
 104 ---------------------------------------
 105
 106 Upon initial bootup your /sys/kernel/debug/kmemleak output page may be
 107 quite extensive. This can also be the case if you have very buggy code
 108 when doing development. To work around these situations you can use the
 109 'clear' command to clear all reported unreferenced objects from the
 110 /sys/kernel/debug/kmemleak output. By issuing a 'scan' after a 'clear'
 111 you can find new unreferenced objects; this should help with testing
 112 specific sections of code.
 113
 114 To test a critical section on demand with a clean kmemleak do:
 115
 116   # echo clear > /sys/kernel/debug/kmemleak
 117   ... test your kernel or modules ...
 118   # echo scan > /sys/kernel/debug/kmemleak
 119
 120 Then as usual to get your report with:
 121
 122   # cat /sys/kernel/debug/kmemleak
 123
 124 Freeing kmemleak internal objects
 125 ---------------------------------
 126
 127 To allow access to previosuly found memory leaks after kmemleak has been
 128 disabled by the user or due to an fatal error, internal kmemleak objects
 129 won't be freed when kmemleak is disabled, and those objects may occupy
 130 a large part of physical memory.
 131
 132 In this situation, you may reclaim memory with:
 133
 134   # echo clear > /sys/kernel/debug/kmemleak
 135
 136 Kmemleak API
 137 ------------
 138
 139 See the include/linux/kmemleak.h header for the functions prototype.
 140
 141 kmemleak_init            - initialize kmemleak
 142 kmemleak_alloc           - notify of a memory block allocation
 143 kmemleak_alloc_percpu    - notify of a percpu memory block allocation
 144 kmemleak_free            - notify of a memory block freeing
 145 kmemleak_free_part       - notify of a partial memory block freeing
 146 kmemleak_free_percpu     - notify of a percpu memory block freeing
 147 kmemleak_not_leak        - mark an object as not a leak
 148 kmemleak_ignore          - do not scan or report an object as leak
 149 kmemleak_scan_area       - add scan areas inside a memory block
 150 kmemleak_no_scan         - do not scan a memory block
 151 kmemleak_erase           - erase an old value in a pointer variable
 152 kmemleak_alloc_recursive - as kmemleak_alloc but checks the recursiveness
 153 kmemleak_free_recursive  - as kmemleak_free but checks the recursiveness
 154
 155 Dealing with false positives/negatives
 156 --------------------------------------
 157
 158 The false negatives are real memory leaks (orphan objects) but not
 159 reported by kmemleak because values found during the memory scanning
 160 point to such objects. To reduce the number of false negatives, kmemleak
 161 provides the kmemleak_ignore, kmemleak_scan_area, kmemleak_no_scan and
 162 kmemleak_erase functions (see above). The task stacks also increase the
 163 amount of false negatives and their scanning is not enabled by default.
 164
 165 The false positives are objects wrongly reported as being memory leaks
 166 (orphan). For objects known not to be leaks, kmemleak provides the
 167 kmemleak_not_leak function. The kmemleak_ignore could also be used if
 168 the memory block is known not to contain other pointers and it will no
 169 longer be scanned.
 170
 171 Some of the reported leaks are only transient, especially on SMP
 172 systems, because of pointers temporarily stored in CPU registers or
 173 stacks. Kmemleak defines MSECS_MIN_AGE (defaulting to 1000) representing
 174 the minimum age of an object to be reported as a memory leak.
 175
 176 Limitations and Drawbacks
 177 -------------------------
 178
 179 The main drawback is the reduced performance of memory allocation and
 180 freeing. To avoid other penalties, the memory scanning is only performed
 181 when the /sys/kernel/debug/kmemleak file is read. Anyway, this tool is
 182 intended for debugging purposes where the performance might not be the
 183 most important requirement.
 184
 185 To keep the algorithm simple, kmemleak scans for values pointing to any
 186 address inside a block's address range. This may lead to an increased
 187 number of false negatives. However, it is likely that a real memory leak
 188 will eventually become visible.
 189
 190 Another source of false negatives is the data stored in non-pointer
 191 values. In a future version, kmemleak could only scan the pointer
 192 members in the allocated structures. This feature would solve many of
 193 the false negative cases described above.
 194
 195 The tool can report false positives. These are cases where an allocated
 196 block doesn't need to be freed (some cases in the init_call functions),
 197 the pointer is calculated by other methods than the usual container_of
 198 macro or the pointer is stored in a location not scanned by kmemleak.
 199
 200 Page allocations and ioremap are not tracked.