kprobetrace.rst 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257
  1. ==========================
  2. Kprobe-based Event Tracing
  3. ==========================
  4. :Author: Masami Hiramatsu
  5. Overview
  6. --------
  7. These events are similar to tracepoint based events. Instead of Tracepoint,
  8. this is based on kprobes (kprobe and kretprobe). So it can probe wherever
  9. kprobes can probe (this means, all functions except those with
  10. __kprobes/nokprobe_inline annotation and those marked NOKPROBE_SYMBOL).
  11. Unlike the Tracepoint based event, this can be added and removed
  12. dynamically, on the fly.
  13. To enable this feature, build your kernel with CONFIG_KPROBE_EVENTS=y.
  14. Similar to the events tracer, this doesn't need to be activated via
  15. current_tracer. Instead of that, add probe points via
  16. /sys/kernel/debug/tracing/kprobe_events, and enable it via
  17. /sys/kernel/debug/tracing/events/kprobes/<EVENT>/enable.
  18. You can also use /sys/kernel/debug/tracing/dynamic_events instead of
  19. kprobe_events. That interface will provide unified access to other
  20. dynamic events too.
  21. Synopsis of kprobe_events
  22. -------------------------
  23. ::
  24. p[:[GRP/]EVENT] [MOD:]SYM[+offs]|MEMADDR [FETCHARGS] : Set a probe
  25. r[MAXACTIVE][:[GRP/]EVENT] [MOD:]SYM[+0] [FETCHARGS] : Set a return probe
  26. p:[GRP/]EVENT] [MOD:]SYM[+0]%return [FETCHARGS] : Set a return probe
  27. -:[GRP/]EVENT : Clear a probe
  28. GRP : Group name. If omitted, use "kprobes" for it.
  29. EVENT : Event name. If omitted, the event name is generated
  30. based on SYM+offs or MEMADDR.
  31. MOD : Module name which has given SYM.
  32. SYM[+offs] : Symbol+offset where the probe is inserted.
  33. SYM%return : Return address of the symbol
  34. MEMADDR : Address where the probe is inserted.
  35. MAXACTIVE : Maximum number of instances of the specified function that
  36. can be probed simultaneously, or 0 for the default value
  37. as defined in Documentation/trace/kprobes.rst section 1.3.1.
  38. FETCHARGS : Arguments. Each probe can have up to 128 args.
  39. %REG : Fetch register REG
  40. @ADDR : Fetch memory at ADDR (ADDR should be in kernel)
  41. @SYM[+|-offs] : Fetch memory at SYM +|- offs (SYM should be a data symbol)
  42. $stackN : Fetch Nth entry of stack (N >= 0)
  43. $stack : Fetch stack address.
  44. $argN : Fetch the Nth function argument. (N >= 1) (\*1)
  45. $retval : Fetch return value.(\*2)
  46. $comm : Fetch current task comm.
  47. +|-[u]OFFS(FETCHARG) : Fetch memory at FETCHARG +|- OFFS address.(\*3)(\*4)
  48. \IMM : Store an immediate value to the argument.
  49. NAME=FETCHARG : Set NAME as the argument name of FETCHARG.
  50. FETCHARG:TYPE : Set TYPE as the type of FETCHARG. Currently, basic types
  51. (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal types
  52. (x8/x16/x32/x64), "string", "ustring" and bitfield
  53. are supported.
  54. (\*1) only for the probe on function entry (offs == 0).
  55. (\*2) only for return probe.
  56. (\*3) this is useful for fetching a field of data structures.
  57. (\*4) "u" means user-space dereference. See :ref:`user_mem_access`.
  58. Types
  59. -----
  60. Several types are supported for fetch-args. Kprobe tracer will access memory
  61. by given type. Prefix 's' and 'u' means those types are signed and unsigned
  62. respectively. 'x' prefix implies it is unsigned. Traced arguments are shown
  63. in decimal ('s' and 'u') or hexadecimal ('x'). Without type casting, 'x32'
  64. or 'x64' is used depends on the architecture (e.g. x86-32 uses x32, and
  65. x86-64 uses x64).
  66. These value types can be an array. To record array data, you can add '[N]'
  67. (where N is a fixed number, less than 64) to the base type.
  68. E.g. 'x16[4]' means an array of x16 (2bytes hex) with 4 elements.
  69. Note that the array can be applied to memory type fetchargs, you can not
  70. apply it to registers/stack-entries etc. (for example, '$stack1:x8[8]' is
  71. wrong, but '+8($stack):x8[8]' is OK.)
  72. String type is a special type, which fetches a "null-terminated" string from
  73. kernel space. This means it will fail and store NULL if the string container
  74. has been paged out. "ustring" type is an alternative of string for user-space.
  75. See :ref:`user_mem_access` for more info..
  76. The string array type is a bit different from other types. For other base
  77. types, <base-type>[1] is equal to <base-type> (e.g. +0(%di):x32[1] is same
  78. as +0(%di):x32.) But string[1] is not equal to string. The string type itself
  79. represents "char array", but string array type represents "char * array".
  80. So, for example, +0(%di):string[1] is equal to +0(+0(%di)):string.
  81. Bitfield is another special type, which takes 3 parameters, bit-width, bit-
  82. offset, and container-size (usually 32). The syntax is::
  83. b<bit-width>@<bit-offset>/<container-size>
  84. Symbol type('symbol') is an alias of u32 or u64 type (depends on BITS_PER_LONG)
  85. which shows given pointer in "symbol+offset" style.
  86. For $comm, the default type is "string"; any other type is invalid.
  87. .. _user_mem_access:
  88. User Memory Access
  89. ------------------
  90. Kprobe events supports user-space memory access. For that purpose, you can use
  91. either user-space dereference syntax or 'ustring' type.
  92. The user-space dereference syntax allows you to access a field of a data
  93. structure in user-space. This is done by adding the "u" prefix to the
  94. dereference syntax. For example, +u4(%si) means it will read memory from the
  95. address in the register %si offset by 4, and the memory is expected to be in
  96. user-space. You can use this for strings too, e.g. +u0(%si):string will read
  97. a string from the address in the register %si that is expected to be in user-
  98. space. 'ustring' is a shortcut way of performing the same task. That is,
  99. +0(%si):ustring is equivalent to +u0(%si):string.
  100. Note that kprobe-event provides the user-memory access syntax but it doesn't
  101. use it transparently. This means if you use normal dereference or string type
  102. for user memory, it might fail, and may always fail on some archs. The user
  103. has to carefully check if the target data is in kernel or user space.
  104. Per-Probe Event Filtering
  105. -------------------------
  106. Per-probe event filtering feature allows you to set different filter on each
  107. probe and gives you what arguments will be shown in trace buffer. If an event
  108. name is specified right after 'p:' or 'r:' in kprobe_events, it adds an event
  109. under tracing/events/kprobes/<EVENT>, at the directory you can see 'id',
  110. 'enable', 'format', 'filter' and 'trigger'.
  111. enable:
  112. You can enable/disable the probe by writing 1 or 0 on it.
  113. format:
  114. This shows the format of this probe event.
  115. filter:
  116. You can write filtering rules of this event.
  117. id:
  118. This shows the id of this probe event.
  119. trigger:
  120. This allows to install trigger commands which are executed when the event is
  121. hit (for details, see Documentation/trace/events.rst, section 6).
  122. Event Profiling
  123. ---------------
  124. You can check the total number of probe hits and probe miss-hits via
  125. /sys/kernel/debug/tracing/kprobe_profile.
  126. The first column is event name, the second is the number of probe hits,
  127. the third is the number of probe miss-hits.
  128. Kernel Boot Parameter
  129. ---------------------
  130. You can add and enable new kprobe events when booting up the kernel by
  131. "kprobe_event=" parameter. The parameter accepts a semicolon-delimited
  132. kprobe events, which format is similar to the kprobe_events.
  133. The difference is that the probe definition parameters are comma-delimited
  134. instead of space. For example, adding myprobe event on do_sys_open like below
  135. p:myprobe do_sys_open dfd=%ax filename=%dx flags=%cx mode=+4($stack)
  136. should be below for kernel boot parameter (just replace spaces with comma)
  137. p:myprobe,do_sys_open,dfd=%ax,filename=%dx,flags=%cx,mode=+4($stack)
  138. Usage examples
  139. --------------
  140. To add a probe as a new event, write a new definition to kprobe_events
  141. as below::
  142. echo 'p:myprobe do_sys_open dfd=%ax filename=%dx flags=%cx mode=+4($stack)' > /sys/kernel/debug/tracing/kprobe_events
  143. This sets a kprobe on the top of do_sys_open() function with recording
  144. 1st to 4th arguments as "myprobe" event. Note, which register/stack entry is
  145. assigned to each function argument depends on arch-specific ABI. If you unsure
  146. the ABI, please try to use probe subcommand of perf-tools (you can find it
  147. under tools/perf/).
  148. As this example shows, users can choose more familiar names for each arguments.
  149. ::
  150. echo 'r:myretprobe do_sys_open $retval' >> /sys/kernel/debug/tracing/kprobe_events
  151. This sets a kretprobe on the return point of do_sys_open() function with
  152. recording return value as "myretprobe" event.
  153. You can see the format of these events via
  154. /sys/kernel/debug/tracing/events/kprobes/<EVENT>/format.
  155. ::
  156. cat /sys/kernel/debug/tracing/events/kprobes/myprobe/format
  157. name: myprobe
  158. ID: 780
  159. format:
  160. field:unsigned short common_type; offset:0; size:2; signed:0;
  161. field:unsigned char common_flags; offset:2; size:1; signed:0;
  162. field:unsigned char common_preempt_count; offset:3; size:1;signed:0;
  163. field:int common_pid; offset:4; size:4; signed:1;
  164. field:unsigned long __probe_ip; offset:12; size:4; signed:0;
  165. field:int __probe_nargs; offset:16; size:4; signed:1;
  166. field:unsigned long dfd; offset:20; size:4; signed:0;
  167. field:unsigned long filename; offset:24; size:4; signed:0;
  168. field:unsigned long flags; offset:28; size:4; signed:0;
  169. field:unsigned long mode; offset:32; size:4; signed:0;
  170. print fmt: "(%lx) dfd=%lx filename=%lx flags=%lx mode=%lx", REC->__probe_ip,
  171. REC->dfd, REC->filename, REC->flags, REC->mode
  172. You can see that the event has 4 arguments as in the expressions you specified.
  173. ::
  174. echo > /sys/kernel/debug/tracing/kprobe_events
  175. This clears all probe points.
  176. Or,
  177. ::
  178. echo -:myprobe >> kprobe_events
  179. This clears probe points selectively.
  180. Right after definition, each event is disabled by default. For tracing these
  181. events, you need to enable it.
  182. ::
  183. echo 1 > /sys/kernel/debug/tracing/events/kprobes/myprobe/enable
  184. echo 1 > /sys/kernel/debug/tracing/events/kprobes/myretprobe/enable
  185. Use the following command to start tracing in an interval.
  186. ::
  187. # echo 1 > tracing_on
  188. Open something...
  189. # echo 0 > tracing_on
  190. And you can see the traced information via /sys/kernel/debug/tracing/trace.
  191. ::
  192. cat /sys/kernel/debug/tracing/trace
  193. # tracer: nop
  194. #
  195. # TASK-PID CPU# TIMESTAMP FUNCTION
  196. # | | | | |
  197. <...>-1447 [001] 1038282.286875: myprobe: (do_sys_open+0x0/0xd6) dfd=3 filename=7fffd1ec4440 flags=8000 mode=0
  198. <...>-1447 [001] 1038282.286878: myretprobe: (sys_openat+0xc/0xe <- do_sys_open) $retval=fffffffffffffffe
  199. <...>-1447 [001] 1038282.286885: myprobe: (do_sys_open+0x0/0xd6) dfd=ffffff9c filename=40413c flags=8000 mode=1b6
  200. <...>-1447 [001] 1038282.286915: myretprobe: (sys_open+0x1b/0x1d <- do_sys_open) $retval=3
  201. <...>-1447 [001] 1038282.286969: myprobe: (do_sys_open+0x0/0xd6) dfd=ffffff9c filename=4041c6 flags=98800 mode=10
  202. <...>-1447 [001] 1038282.286976: myretprobe: (sys_open+0x1b/0x1d <- do_sys_open) $retval=3
  203. Each line shows when the kernel hits an event, and <- SYMBOL means kernel
  204. returns from SYMBOL(e.g. "sys_open+0x1b/0x1d <- do_sys_open" means kernel
  205. returns from do_sys_open to sys_open+0x1b).