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