1.. SPDX-License-Identifier: GPL-2.0
2.. include:: <isonum.txt>
3
4=======
5DebugFS
6=======
7
8Copyright |copy| 2009 Jonathan Corbet <corbet@lwn.net>
9
10Debugfs exists as a simple way for kernel developers to make information
11available to user space.  Unlike /proc, which is only meant for information
12about a process, or sysfs, which has strict one-value-per-file rules,
13debugfs has no rules at all.  Developers can put any information they want
14there.  The debugfs filesystem is also intended to not serve as a stable
15ABI to user space; in theory, there are no stability constraints placed on
16files exported there.  The real world is not always so simple, though [1]_;
17even debugfs interfaces are best designed with the idea that they will need
18to be maintained forever.
19
20Debugfs is typically mounted with a command like::
21
22    mount -t debugfs none /sys/kernel/debug
23
24(Or an equivalent /etc/fstab line).
25The debugfs root directory is accessible only to the root user by
26default. To change access to the tree the "uid", "gid" and "mode" mount
27options can be used.
28
29Note that the debugfs API is exported GPL-only to modules.
30
31Code using debugfs should include <linux/debugfs.h>.  Then, the first order
32of business will be to create at least one directory to hold a set of
33debugfs files::
34
35    struct dentry *debugfs_create_dir(const char *name, struct dentry *parent);
36
37This call, if successful, will make a directory called name underneath the
38indicated parent directory.  If parent is NULL, the directory will be
39created in the debugfs root.  On success, the return value is a struct
40dentry pointer which can be used to create files in the directory (and to
41clean it up at the end).  An ERR_PTR(-ERROR) return value indicates that
42something went wrong.  If ERR_PTR(-ENODEV) is returned, that is an
43indication that the kernel has been built without debugfs support and none
44of the functions described below will work.
45
46The most general way to create a file within a debugfs directory is with::
47
48    struct dentry *debugfs_create_file(const char *name, umode_t mode,
49				       struct dentry *parent, void *data,
50				       const struct file_operations *fops);
51
52Here, name is the name of the file to create, mode describes the access
53permissions the file should have, parent indicates the directory which
54should hold the file, data will be stored in the i_private field of the
55resulting inode structure, and fops is a set of file operations which
56implement the file's behavior.  At a minimum, the read() and/or write()
57operations should be provided; others can be included as needed.  Again,
58the return value will be a dentry pointer to the created file,
59ERR_PTR(-ERROR) on error, or ERR_PTR(-ENODEV) if debugfs support is
60missing.
61
62Create a file with an initial size, the following function can be used
63instead::
64
65    void debugfs_create_file_size(const char *name, umode_t mode,
66				  struct dentry *parent, void *data,
67				  const struct file_operations *fops,
68				  loff_t file_size);
69
70file_size is the initial file size. The other parameters are the same
71as the function debugfs_create_file.
72
73In a number of cases, the creation of a set of file operations is not
74actually necessary; the debugfs code provides a number of helper functions
75for simple situations.  Files containing a single integer value can be
76created with any of::
77
78    void debugfs_create_u8(const char *name, umode_t mode,
79			   struct dentry *parent, u8 *value);
80    void debugfs_create_u16(const char *name, umode_t mode,
81			    struct dentry *parent, u16 *value);
82    void debugfs_create_u32(const char *name, umode_t mode,
83			    struct dentry *parent, u32 *value);
84    void debugfs_create_u64(const char *name, umode_t mode,
85			    struct dentry *parent, u64 *value);
86
87These files support both reading and writing the given value; if a specific
88file should not be written to, simply set the mode bits accordingly.  The
89values in these files are in decimal; if hexadecimal is more appropriate,
90the following functions can be used instead::
91
92    void debugfs_create_x8(const char *name, umode_t mode,
93			   struct dentry *parent, u8 *value);
94    void debugfs_create_x16(const char *name, umode_t mode,
95			    struct dentry *parent, u16 *value);
96    void debugfs_create_x32(const char *name, umode_t mode,
97			    struct dentry *parent, u32 *value);
98    void debugfs_create_x64(const char *name, umode_t mode,
99			    struct dentry *parent, u64 *value);
100
101These functions are useful as long as the developer knows the size of the
102value to be exported.  Some types can have different widths on different
103architectures, though, complicating the situation somewhat.  There are
104functions meant to help out in such special cases::
105
106    void debugfs_create_size_t(const char *name, umode_t mode,
107			       struct dentry *parent, size_t *value);
108
109As might be expected, this function will create a debugfs file to represent
110a variable of type size_t.
111
112Similarly, there are helpers for variables of type unsigned long, in decimal
113and hexadecimal::
114
115    struct dentry *debugfs_create_ulong(const char *name, umode_t mode,
116					struct dentry *parent,
117					unsigned long *value);
118    void debugfs_create_xul(const char *name, umode_t mode,
119			    struct dentry *parent, unsigned long *value);
120
121Boolean values can be placed in debugfs with::
122
123    struct dentry *debugfs_create_bool(const char *name, umode_t mode,
124				       struct dentry *parent, bool *value);
125
126A read on the resulting file will yield either Y (for non-zero values) or
127N, followed by a newline.  If written to, it will accept either upper- or
128lower-case values, or 1 or 0.  Any other input will be silently ignored.
129
130Also, atomic_t values can be placed in debugfs with::
131
132    void debugfs_create_atomic_t(const char *name, umode_t mode,
133				 struct dentry *parent, atomic_t *value)
134
135A read of this file will get atomic_t values, and a write of this file
136will set atomic_t values.
137
138Another option is exporting a block of arbitrary binary data, with
139this structure and function::
140
141    struct debugfs_blob_wrapper {
142	void *data;
143	unsigned long size;
144    };
145
146    struct dentry *debugfs_create_blob(const char *name, umode_t mode,
147				       struct dentry *parent,
148				       struct debugfs_blob_wrapper *blob);
149
150A read of this file will return the data pointed to by the
151debugfs_blob_wrapper structure.  Some drivers use "blobs" as a simple way
152to return several lines of (static) formatted text output.  This function
153can be used to export binary information, but there does not appear to be
154any code which does so in the mainline.  Note that all files created with
155debugfs_create_blob() are read-only.
156
157If you want to dump a block of registers (something that happens quite
158often during development, even if little such code reaches mainline.
159Debugfs offers two functions: one to make a registers-only file, and
160another to insert a register block in the middle of another sequential
161file::
162
163    struct debugfs_reg32 {
164	char *name;
165	unsigned long offset;
166    };
167
168    struct debugfs_regset32 {
169	const struct debugfs_reg32 *regs;
170	int nregs;
171	void __iomem *base;
172	struct device *dev;     /* Optional device for Runtime PM */
173    };
174
175    debugfs_create_regset32(const char *name, umode_t mode,
176			    struct dentry *parent,
177			    struct debugfs_regset32 *regset);
178
179    void debugfs_print_regs32(struct seq_file *s, const struct debugfs_reg32 *regs,
180			 int nregs, void __iomem *base, char *prefix);
181
182The "base" argument may be 0, but you may want to build the reg32 array
183using __stringify, and a number of register names (macros) are actually
184byte offsets over a base for the register block.
185
186If you want to dump an u32 array in debugfs, you can create file with::
187
188    struct debugfs_u32_array {
189	u32 *array;
190	u32 n_elements;
191    };
192
193    void debugfs_create_u32_array(const char *name, umode_t mode,
194			struct dentry *parent,
195			struct debugfs_u32_array *array);
196
197The "array" argument wraps a pointer to the array's data and the number
198of its elements. Note: Once array is created its size can not be changed.
199
200There is a helper function to create device related seq_file::
201
202   struct dentry *debugfs_create_devm_seqfile(struct device *dev,
203				const char *name,
204				struct dentry *parent,
205				int (*read_fn)(struct seq_file *s,
206					void *data));
207
208The "dev" argument is the device related to this debugfs file, and
209the "read_fn" is a function pointer which to be called to print the
210seq_file content.
211
212There are a couple of other directory-oriented helper functions::
213
214    struct dentry *debugfs_rename(struct dentry *old_dir,
215    				  struct dentry *old_dentry,
216		                  struct dentry *new_dir,
217				  const char *new_name);
218
219    struct dentry *debugfs_create_symlink(const char *name,
220                                          struct dentry *parent,
221				      	  const char *target);
222
223A call to debugfs_rename() will give a new name to an existing debugfs
224file, possibly in a different directory.  The new_name must not exist prior
225to the call; the return value is old_dentry with updated information.
226Symbolic links can be created with debugfs_create_symlink().
227
228There is one important thing that all debugfs users must take into account:
229there is no automatic cleanup of any directories created in debugfs.  If a
230module is unloaded without explicitly removing debugfs entries, the result
231will be a lot of stale pointers and no end of highly antisocial behavior.
232So all debugfs users - at least those which can be built as modules - must
233be prepared to remove all files and directories they create there.  A file
234can be removed with::
235
236    void debugfs_remove(struct dentry *dentry);
237
238The dentry value can be NULL or an error value, in which case nothing will
239be removed.
240
241Once upon a time, debugfs users were required to remember the dentry
242pointer for every debugfs file they created so that all files could be
243cleaned up.  We live in more civilized times now, though, and debugfs users
244can call::
245
246    void debugfs_remove_recursive(struct dentry *dentry);
247
248If this function is passed a pointer for the dentry corresponding to the
249top-level directory, the entire hierarchy below that directory will be
250removed.
251
252.. [1] http://lwn.net/Articles/309298/
253