xref: /linux/Documentation/trace/kprobetrace.rst (revision 3a2c4d55e32ad65efebdb6de44eef3bfa08bb49d)
1==========================
2Kprobe-based Event Tracing
3==========================
4
5:Author: Masami Hiramatsu
6
7Overview
8--------
9These events are similar to tracepoint-based events. Instead of tracepoints,
10this is based on kprobes (kprobe and kretprobe). So it can probe wherever
11kprobes can probe (this means, all functions except those with
12__kprobes/nokprobe_inline annotation and those marked NOKPROBE_SYMBOL).
13Unlike the tracepoint-based event, this can be added and removed
14dynamically, on the fly.
15
16To enable this feature, build your kernel with CONFIG_KPROBE_EVENTS=y.
17
18Similar to the event tracer, this doesn't need to be activated via
19current_tracer. Instead of that, add probe points via
20/sys/kernel/tracing/kprobe_events, and enable it via
21/sys/kernel/tracing/events/kprobes/<EVENT>/enable.
22
23You can also use /sys/kernel/tracing/dynamic_events instead of
24kprobe_events. That interface will provide unified access to other
25dynamic events too.
26
27Synopsis of kprobe_events
28-------------------------
29::
30
31  p[:[GRP/][EVENT]] [MOD:]SYM[+offs]|MEMADDR [FETCHARGS]	: Set a probe
32  r[MAXACTIVE][:[GRP/][EVENT]] [MOD:]SYM[+0] [FETCHARGS]	: Set a return probe
33  p[:[GRP/][EVENT]] [MOD:]SYM[+0]%return [FETCHARGS]	: Set a return probe
34  -:[GRP/][EVENT]						: Clear a probe
35
36 GRP		: Group name. If omitted, use "kprobes" for it.
37 EVENT		: Event name. If omitted, the event name is generated
38		  based on SYM+offs or MEMADDR.
39 MOD		: Module name which has given SYM.
40 SYM[+offs]	: Symbol+offset where the probe is inserted.
41 SYM%return	: Return address of the symbol
42 MEMADDR	: Address where the probe is inserted.
43 MAXACTIVE	: Maximum number of instances of the specified function that
44		  can be probed simultaneously, or 0 for the default value
45		  as defined in Documentation/trace/kprobes.rst section 1.3.1.
46
47 FETCHARGS	: Arguments. Each probe can have up to 128 args.
48  %REG		: Fetch register REG
49  @ADDR		: Fetch memory at ADDR (ADDR should be in kernel)
50  @SYM[+|-offs]	: Fetch memory at SYM +|- offs (SYM should be a data symbol)
51  $stackN	: Fetch Nth entry of stack (N >= 0)
52  $stack	: Fetch stack address.
53  $argN		: Fetch the Nth function argument. (N >= 1) (\*1)
54  $retval	: Fetch return value.(\*2)
55  $comm		: Fetch current task comm.
56  $current      : Fetch the address of the current task_struct.
57  +|-[u]OFFS(FETCHARG) : Fetch memory at FETCHARG +|- OFFS address.(\*3)(\*4)
58  this_cpu_read(FETCHARG) : Read the value of the per-CPU variable FETCHARG on the current CPU.
59  this_cpu_ptr(FETCHARG) : Get the address of the per-CPU variable FETCHARG on the current CPU.
60  \IMM		: Store an immediate value to the argument.
61  NAME=FETCHARG : Set NAME as the argument name of FETCHARG.
62  FETCHARG:TYPE : Set TYPE as the type of FETCHARG. Currently, basic types
63		  (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal types
64		  (x8/x16/x32/x64), VFS layer common type(%pd/%pD), "char",
65                  "string", "ustring", "symbol", "symstr" and bitfield are
66                  supported.
67  (STRUCT[,ASGN])FIELD->MEMBER[->MEMBER] : If BTF is supported, typecast FIELD to
68                  a pointer to STRUCT and then derference the pointer defined by
69                  ->MEMBER. Note that this is available only when the probe is
70		   on function entry. ASGN can be specified optionally. If ASGN
71		   is specified, FIELD will be cast to the same offset position
72		   as the ASGN member, rather than to the beginning of the STRUCT.
73  (STRUCT[,ASGN])(FETCHARG)->MEMBER[->MEMBER] : typecast can nest, so the above can
74                 also be used with another FETCHARG instead of FIELD.
75
76  (\*1) only for the probe on function entry (offs == 0). Note, this argument access
77        is best effort, because depending on the argument type, it may be passed on
78        the stack. But this only support the arguments via registers.
79  (\*2) only for return probe. Note that this is also best effort. Depending on the
80        return value type, it might be passed via a pair of registers. But this only
81        accesses one register.
82  (\*3) this is useful for fetching a field of data structures.
83  (\*4) "u" means user-space dereference. See :ref:`user_mem_access`.
84
85Function arguments at kretprobe
86-------------------------------
87Function arguments can be accessed at kretprobe using $arg<N> fetcharg. This
88is useful to record the function parameter and return value at once, and
89trace the difference of structure fields (for debugging a function whether it
90correctly updates the given data structure or not).
91See the :ref:`sample<fprobetrace_exit_args_sample>` in fprobe event for how
92it works.
93
94.. _kprobetrace_types:
95
96Types
97-----
98Several types are supported for fetchargs. Kprobe tracer will access memory
99by given type. Prefix 's' and 'u' means those types are signed and unsigned
100respectively. 'x' prefix implies it is unsigned. Traced arguments are shown
101in decimal ('s' and 'u') or hexadecimal ('x'). Without type casting, 'x32'
102or 'x64' is used depends on the architecture (e.g. x86-32 uses x32, and
103x86-64 uses x64).
104
105These value types can be an array. To record array data, you can add '[N]'
106(where N is a fixed number, less than 64) to the base type.
107E.g. 'x16[4]' means an array of x16 (2-byte hex) with 4 elements.
108Note that the array can be applied to memory type fetchargs, you can not
109apply it to registers/stack-entries etc. (for example, '$stack1:x8[8]' is
110wrong, but '+8($stack):x8[8]' is OK.)
111
112Char type can be used to show the character value of traced arguments.
113
114String type is a special type, which fetches a "null-terminated" string from
115kernel space. This means it will fail and store NULL if the string container
116has been paged out. "ustring" type is an alternative of string for user-space.
117See :ref:`user_mem_access` for more info.
118
119The string array type is a bit different from other types. For other base
120types, <base-type>[1] is equal to <base-type> (e.g. +0(%di):x32[1] is same
121as +0(%di):x32.) But string[1] is not equal to string. The string type itself
122represents "char array", but string array type represents "char * array".
123So, for example, +0(%di):string[1] is equal to +0(+0(%di)):string.
124Bitfield is another special type, which takes 3 parameters, bit-width, bit-
125offset, and container-size (usually 32). The syntax is::
126
127 b<bit-width>@<bit-offset>/<container-size>
128
129Symbol type('symbol') is an alias of u32 or u64 type (depends on BITS_PER_LONG)
130which shows given pointer in "symbol+offset" style.
131On the other hand, symbol-string type ('symstr') converts the given address to
132"symbol+offset/symbolsize" style and stores it as a null-terminated string.
133With 'symstr' type, you can filter the event with wildcard pattern of the
134symbols, and you don't need to solve symbol name by yourself.
135For $comm, the default type is "string"; any other type is invalid.
136
137VFS layer common type(%pd/%pD) is a special type, which fetches dentry's or
138file's name from struct dentry's address or struct file's address.
139
140.. _user_mem_access:
141
142User Memory Access
143------------------
144Kprobe events supports user-space memory access. For that purpose, you can use
145either user-space dereference syntax or 'ustring' type.
146
147The user-space dereference syntax allows you to access a field of a data
148structure in user-space. This is done by adding the "u" prefix to the
149dereference syntax. For example, +u4(%si) means it will read memory from the
150address in the register %si offset by 4, and the memory is expected to be in
151user-space. You can use this for strings too, e.g. +u0(%si):string will read
152a string from the address in the register %si that is expected to be in user-
153space. 'ustring' is a shortcut way of performing the same task. That is,
154+0(%si):ustring is equivalent to +u0(%si):string.
155
156Note that kprobe-event provides the user-memory access syntax but it doesn't
157use it transparently. This means if you use normal dereference or string type
158for user memory, it might fail, and may always fail on some architectures. The
159user has to carefully check if the target data is in kernel or user space.
160
161Per-Probe Event Filtering
162-------------------------
163Per-probe event filtering feature allows you to set different filter on each
164probe and gives you what arguments will be shown in trace buffer. If an event
165name is specified right after 'p:' or 'r:' in kprobe_events, it adds an event
166under tracing/events/kprobes/<EVENT>, at the directory you can see 'id',
167'enable', 'format', 'filter' and 'trigger'.
168
169enable:
170  You can enable/disable the probe by writing 1 or 0 on it.
171
172format:
173  This shows the format of this probe event.
174
175filter:
176  You can write filtering rules of this event.
177
178id:
179  This shows the id of this probe event.
180
181trigger:
182  This allows to install trigger commands which are executed when the event is
183  hit (for details, see Documentation/trace/events.rst, section 6).
184
185Event Profiling
186---------------
187You can check the total number of probe hits and probe miss-hits via
188/sys/kernel/tracing/kprobe_profile.
189The first column is event name, the second is the number of probe hits,
190the third is the number of probe miss-hits.
191
192Kernel Boot Parameter
193---------------------
194You can add and enable new kprobe events when booting up the kernel by
195"kprobe_event=" parameter. The parameter accepts a semicolon-delimited
196kprobe events, which format is similar to the kprobe_events.
197The difference is that the probe definition parameters are comma-delimited
198instead of space. For example, adding myprobe event on do_sys_open like below::
199
200  p:myprobe do_sys_open dfd=%ax filename=%dx flags=%cx mode=+4($stack)
201
202should be below for kernel boot parameter (just replace spaces with comma)::
203
204  p:myprobe,do_sys_open,dfd=%ax,filename=%dx,flags=%cx,mode=+4($stack)
205
206
207Usage examples
208--------------
209To add a probe as a new event, write a new definition to kprobe_events
210as below::
211
212  echo 'p:myprobe do_sys_open dfd=%ax filename=%dx flags=%cx mode=+4($stack)' > /sys/kernel/tracing/kprobe_events
213
214This sets a kprobe on the top of do_sys_open() function with recording
2151st to 4th arguments as "myprobe" event. Note, which register/stack entry is
216assigned to each function argument depends on arch-specific ABI. If you unsure
217the ABI, please try to use probe subcommand of perf-tools (you can find it
218under tools/perf/).
219As this example shows, users can choose more familiar names for each arguments.
220::
221
222  echo 'r:myretprobe do_sys_open $retval' >> /sys/kernel/tracing/kprobe_events
223
224This sets a kretprobe on the return point of do_sys_open() function with
225recording return value as "myretprobe" event.
226You can see the format of these events via
227/sys/kernel/tracing/events/kprobes/<EVENT>/format.
228::
229
230  cat /sys/kernel/tracing/events/kprobes/myprobe/format
231  name: myprobe
232  ID: 780
233  format:
234          field:unsigned short common_type;       offset:0;       size:2; signed:0;
235          field:unsigned char common_flags;       offset:2;       size:1; signed:0;
236          field:unsigned char common_preempt_count;       offset:3; size:1;signed:0;
237          field:int common_pid;   offset:4;       size:4; signed:1;
238
239          field:unsigned long __probe_ip; offset:12;      size:4; signed:0;
240          field:int __probe_nargs;        offset:16;      size:4; signed:1;
241          field:unsigned long dfd;        offset:20;      size:4; signed:0;
242          field:unsigned long filename;   offset:24;      size:4; signed:0;
243          field:unsigned long flags;      offset:28;      size:4; signed:0;
244          field:unsigned long mode;       offset:32;      size:4; signed:0;
245
246
247  print fmt: "(%lx) dfd=%lx filename=%lx flags=%lx mode=%lx", REC->__probe_ip,
248  REC->dfd, REC->filename, REC->flags, REC->mode
249
250You can see that the event has 4 arguments as in the expressions you specified.
251::
252
253  echo > /sys/kernel/tracing/kprobe_events
254
255This clears all probe points.
256
257Or,
258::
259
260  echo -:myprobe >> kprobe_events
261
262This clears probe points selectively.
263
264Right after definition, each event is disabled by default. For tracing these
265events, you need to enable it.
266::
267
268  echo 1 > /sys/kernel/tracing/events/kprobes/myprobe/enable
269  echo 1 > /sys/kernel/tracing/events/kprobes/myretprobe/enable
270
271Use the following command to start tracing in an interval.
272::
273
274    # echo 1 > tracing_on
275    Open something...
276    # echo 0 > tracing_on
277
278And you can see the traced information via /sys/kernel/tracing/trace.
279::
280
281  cat /sys/kernel/tracing/trace
282  # tracer: nop
283  #
284  #           TASK-PID    CPU#    TIMESTAMP  FUNCTION
285  #              | |       |          |         |
286             <...>-1447  [001] 1038282.286875: myprobe: (do_sys_open+0x0/0xd6) dfd=3 filename=7fffd1ec4440 flags=8000 mode=0
287             <...>-1447  [001] 1038282.286878: myretprobe: (sys_openat+0xc/0xe <- do_sys_open) $retval=fffffffffffffffe
288             <...>-1447  [001] 1038282.286885: myprobe: (do_sys_open+0x0/0xd6) dfd=ffffff9c filename=40413c flags=8000 mode=1b6
289             <...>-1447  [001] 1038282.286915: myretprobe: (sys_open+0x1b/0x1d <- do_sys_open) $retval=3
290             <...>-1447  [001] 1038282.286969: myprobe: (do_sys_open+0x0/0xd6) dfd=ffffff9c filename=4041c6 flags=98800 mode=10
291             <...>-1447  [001] 1038282.286976: myretprobe: (sys_open+0x1b/0x1d <- do_sys_open) $retval=3
292
293
294Each line shows when the kernel hits an event, and <- SYMBOL means kernel
295returns from SYMBOL(e.g. "sys_open+0x1b/0x1d <- do_sys_open" means kernel
296returns from do_sys_open to sys_open+0x1b).
297