1.. SPDX-License-Identifier: GPL-2.0
2
3==================================
4Fprobe - Function entry/exit probe
5==================================
6
7.. Author: Masami Hiramatsu <mhiramat@kernel.org>
8
9Introduction
10============
11
12Fprobe is a function entry/exit probe mechanism based on ftrace.
13Instead of using ftrace full feature, if you only want to attach callbacks
14on function entry and exit, similar to the kprobes and kretprobes, you can
15use fprobe. Compared with kprobes and kretprobes, fprobe gives faster
16instrumentation for multiple functions with single handler. This document
17describes how to use fprobe.
18
19The usage of fprobe
20===================
21
22The fprobe is a wrapper of ftrace (+ kretprobe-like return callback) to
23attach callbacks to multiple function entry and exit. User needs to set up
24the `struct fprobe` and pass it to `register_fprobe()`.
25
26Typically, `fprobe` data structure is initialized with the `entry_handler`
27and/or `exit_handler` as below.
28
29.. code-block:: c
30
31 struct fprobe fp = {
32        .entry_handler  = my_entry_callback,
33        .exit_handler   = my_exit_callback,
34 };
35
36To enable the fprobe, call one of register_fprobe(), register_fprobe_ips(), and
37register_fprobe_syms(). These functions register the fprobe with different types
38of parameters.
39
40The register_fprobe() enables a fprobe by function-name filters.
41E.g. this enables @fp on "func*()" function except "func2()".::
42
43  register_fprobe(&fp, "func*", "func2");
44
45The register_fprobe_ips() enables a fprobe by ftrace-location addresses.
46E.g.
47
48.. code-block:: c
49
50  unsigned long ips[] = { 0x.... };
51
52  register_fprobe_ips(&fp, ips, ARRAY_SIZE(ips));
53
54And the register_fprobe_syms() enables a fprobe by symbol names.
55E.g.
56
57.. code-block:: c
58
59  char syms[] = {"func1", "func2", "func3"};
60
61  register_fprobe_syms(&fp, syms, ARRAY_SIZE(syms));
62
63To disable (remove from functions) this fprobe, call::
64
65  unregister_fprobe(&fp);
66
67You can temporally (soft) disable the fprobe by::
68
69  disable_fprobe(&fp);
70
71and resume by::
72
73  enable_fprobe(&fp);
74
75The above is defined by including the header::
76
77  #include <linux/fprobe.h>
78
79Same as ftrace, the registered callbacks will start being called some time
80after the register_fprobe() is called and before it returns. See
81:file:`Documentation/trace/ftrace.rst`.
82
83Also, the unregister_fprobe() will guarantee that the both enter and exit
84handlers are no longer being called by functions after unregister_fprobe()
85returns as same as unregister_ftrace_function().
86
87The fprobe entry/exit handler
88=============================
89
90The prototype of the entry/exit callback function are as follows:
91
92.. code-block:: c
93
94 int entry_callback(struct fprobe *fp, unsigned long entry_ip, struct pt_regs *regs, void *entry_data);
95
96 void exit_callback(struct fprobe *fp, unsigned long entry_ip, struct pt_regs *regs, void *entry_data);
97
98Note that the @entry_ip is saved at function entry and passed to exit handler.
99If the entry callback function returns !0, the corresponding exit callback will be cancelled.
100
101@fp
102        This is the address of `fprobe` data structure related to this handler.
103        You can embed the `fprobe` to your data structure and get it by
104        container_of() macro from @fp. The @fp must not be NULL.
105
106@entry_ip
107        This is the ftrace address of the traced function (both entry and exit).
108        Note that this may not be the actual entry address of the function but
109        the address where the ftrace is instrumented.
110
111@regs
112        This is the `pt_regs` data structure at the entry and exit. Note that
113        the instruction pointer of @regs may be different from the @entry_ip
114        in the entry_handler. If you need traced instruction pointer, you need
115        to use @entry_ip. On the other hand, in the exit_handler, the instruction
116        pointer of @regs is set to the currect return address.
117
118@entry_data
119        This is a local storage to share the data between entry and exit handlers.
120        This storage is NULL by default. If the user specify `exit_handler` field
121        and `entry_data_size` field when registering the fprobe, the storage is
122        allocated and passed to both `entry_handler` and `exit_handler`.
123
124Share the callbacks with kprobes
125================================
126
127Since the recursion safeness of the fprobe (and ftrace) is a bit different
128from the kprobes, this may cause an issue if user wants to run the same
129code from the fprobe and the kprobes.
130
131Kprobes has per-cpu 'current_kprobe' variable which protects the kprobe
132handler from recursion in all cases. On the other hand, fprobe uses
133only ftrace_test_recursion_trylock(). This allows interrupt context to
134call another (or same) fprobe while the fprobe user handler is running.
135
136This is not a matter if the common callback code has its own recursion
137detection, or it can handle the recursion in the different contexts
138(normal/interrupt/NMI.)
139But if it relies on the 'current_kprobe' recursion lock, it has to check
140kprobe_running() and use kprobe_busy_*() APIs.
141
142Fprobe has FPROBE_FL_KPROBE_SHARED flag to do this. If your common callback
143code will be shared with kprobes, please set FPROBE_FL_KPROBE_SHARED
144*before* registering the fprobe, like:
145
146.. code-block:: c
147
148 fprobe.flags = FPROBE_FL_KPROBE_SHARED;
149
150 register_fprobe(&fprobe, "func*", NULL);
151
152This will protect your common callback from the nested call.
153
154The missed counter
155==================
156
157The `fprobe` data structure has `fprobe::nmissed` counter field as same as
158kprobes.
159This counter counts up when;
160
161 - fprobe fails to take ftrace_recursion lock. This usually means that a function
162   which is traced by other ftrace users is called from the entry_handler.
163
164 - fprobe fails to setup the function exit because of the shortage of rethook
165   (the shadow stack for hooking the function return.)
166
167The `fprobe::nmissed` field counts up in both cases. Therefore, the former
168skips both of entry and exit callback and the latter skips the exit
169callback, but in both case the counter will increase by 1.
170
171Note that if you set the FTRACE_OPS_FL_RECURSION and/or FTRACE_OPS_FL_RCU to
172`fprobe::ops::flags` (ftrace_ops::flags) when registering the fprobe, this
173counter may not work correctly, because ftrace skips the fprobe function which
174increase the counter.
175
176
177Functions and structures
178========================
179
180.. kernel-doc:: include/linux/fprobe.h
181.. kernel-doc:: kernel/trace/fprobe.c
182
183