3 Copyright 2003 Jonathan Corbet <corbet@lwn.net>
4 This file is originally from the LWN.net Driver Porting series at
5 http://lwn.net/Articles/driver-porting/
8 There are numerous ways for a device driver (or other kernel component) to
9 provide information to the user or system administrator. One useful
10 technique is the creation of virtual files, in debugfs, /proc or elsewhere.
11 Virtual files can provide human-readable output that is easy to get at
12 without any special utility programs; they can also make life easier for
13 script writers. It is not surprising that the use of virtual files has
16 Creating those files correctly has always been a bit of a challenge,
17 however. It is not that hard to make a virtual file which returns a
18 string. But life gets trickier if the output is long - anything greater
19 than an application is likely to read in a single operation. Handling
20 multiple reads (and seeks) requires careful attention to the reader's
21 position within the virtual file - that position is, likely as not, in the
22 middle of a line of output. The kernel has traditionally had a number of
23 implementations that got this wrong.
25 The 2.6 kernel contains a set of functions (implemented by Alexander Viro)
26 which are designed to make it easy for virtual file creators to get it
29 The seq_file interface is available via <linux/seq_file.h>. There are
30 three aspects to seq_file:
32 * An iterator interface which lets a virtual file implementation
33 step through the objects it is presenting.
35 * Some utility functions for formatting objects for output without
36 needing to worry about things like output buffers.
38 * A set of canned file_operations which implement most operations on
41 We'll look at the seq_file interface via an extremely simple example: a
42 loadable module which creates a file called /proc/sequence. The file, when
43 read, simply produces a set of increasing integer values, one per line. The
44 sequence will continue until the user loses patience and finds something
45 better to do. The file is seekable, in that one can do something like the
48 dd if=/proc/sequence of=out1 count=1
49 dd if=/proc/sequence skip=1 of=out2 count=1
51 Then concatenate the output files out1 and out2 and get the right
52 result. Yes, it is a thoroughly useless module, but the point is to show
53 how the mechanism works without getting lost in other details. (Those
54 wanting to see the full source for this module can find it at
55 http://lwn.net/Articles/22359/).
57 Deprecated create_proc_entry
59 Note that the above article uses create_proc_entry which was removed in
60 kernel 3.10. Current versions require the following update
62 - entry = create_proc_entry("sequence", 0, NULL);
64 - entry->proc_fops = &ct_file_ops;
65 + entry = proc_create("sequence", 0, NULL, &ct_file_ops);
67 The iterator interface
69 Modules implementing a virtual file with seq_file must implement an
70 iterator object that allows stepping through the data of interest
71 during a "session" (roughly one read() system call). If the iterator
72 is able to move to a specific position - like the file they implement,
73 though with freedom to map the position number to a sequence location
74 in whatever way is convenient - the iterator need only exist
75 transiently during a session. If the iterator cannot easily find a
76 numerical position but works well with a first/next interface, the
77 iterator can be stored in the private data area and continue from one
80 A seq_file implementation that is formatting firewall rules from a
81 table, for example, could provide a simple iterator that interprets
82 position N as the Nth rule in the chain. A seq_file implementation
83 that presents the content of a, potentially volatile, linked list
84 might record a pointer into that list, providing that can be done
85 without risk of the current location being removed.
87 Positioning can thus be done in whatever way makes the most sense for
88 the generator of the data, which need not be aware of how a position
89 translates to an offset in the virtual file. The one obvious exception
90 is that a position of zero should indicate the beginning of the file.
92 The /proc/sequence iterator just uses the count of the next number it
93 will output as its position.
95 Four functions must be implemented to make the iterator work. The
96 first, called start(), starts a session and takes a position as an
97 argument, returning an iterator which will start reading at that
98 position. The pos passed to start() will always be either zero, or
99 the most recent pos used in the previous session.
101 For our simple sequence example,
102 the start() function looks like:
104 static void *ct_seq_start(struct seq_file *s, loff_t *pos)
106 loff_t *spos = kmalloc(sizeof(loff_t), GFP_KERNEL);
113 The entire data structure for this iterator is a single loff_t value
114 holding the current position. There is no upper bound for the sequence
115 iterator, but that will not be the case for most other seq_file
116 implementations; in most cases the start() function should check for a
117 "past end of file" condition and return NULL if need be.
119 For more complicated applications, the private field of the seq_file
120 structure can be used to hold state from session to session. There is
121 also a special value which can be returned by the start() function
122 called SEQ_START_TOKEN; it can be used if you wish to instruct your
123 show() function (described below) to print a header at the top of the
124 output. SEQ_START_TOKEN should only be used if the offset is zero,
127 The next function to implement is called, amazingly, next(); its job is to
128 move the iterator forward to the next position in the sequence. The
129 example module can simply increment the position by one; more useful
130 modules will do what is needed to step through some data structure. The
131 next() function returns a new iterator, or NULL if the sequence is
132 complete. Here's the example version:
134 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
141 The stop() function closes a session; its job, of course, is to clean
142 up. If dynamic memory is allocated for the iterator, stop() is the
143 place to free it; if a lock was taken by start(), stop() must release
144 that lock. The value that *pos was set to by the last next() call
145 before stop() is remembered, and used for the first start() call of
146 the next session unless lseek() has been called on the file; in that
147 case next start() will be asked to start at position zero.
149 static void ct_seq_stop(struct seq_file *s, void *v)
154 Finally, the show() function should format the object currently pointed to
155 by the iterator for output. The example module's show() function is:
157 static int ct_seq_show(struct seq_file *s, void *v)
160 seq_printf(s, "%lld\n", (long long)*spos);
164 If all is well, the show() function should return zero. A negative error
165 code in the usual manner indicates that something went wrong; it will be
166 passed back to user space. This function can also return SEQ_SKIP, which
167 causes the current item to be skipped; if the show() function has already
168 generated output before returning SEQ_SKIP, that output will be dropped.
170 We will look at seq_printf() in a moment. But first, the definition of the
171 seq_file iterator is finished by creating a seq_operations structure with
172 the four functions we have just defined:
174 static const struct seq_operations ct_seq_ops = {
175 .start = ct_seq_start,
181 This structure will be needed to tie our iterator to the /proc file in
184 It's worth noting that the iterator value returned by start() and
185 manipulated by the other functions is considered to be completely opaque by
186 the seq_file code. It can thus be anything that is useful in stepping
187 through the data to be output. Counters can be useful, but it could also be
188 a direct pointer into an array or linked list. Anything goes, as long as
189 the programmer is aware that things can happen between calls to the
190 iterator function. However, the seq_file code (by design) will not sleep
191 between the calls to start() and stop(), so holding a lock during that time
192 is a reasonable thing to do. The seq_file code will also avoid taking any
193 other locks while the iterator is active.
195 The iterater value returned by start() or next() is guaranteed to be
196 passed to a subsequent next() or stop() call. This allows resources
197 such as locks that were taken to be reliably released. There is *no*
198 guarantee that the iterator will be passed to show(), though in practice
204 The seq_file code manages positioning within the output created by the
205 iterator and getting it into the user's buffer. But, for that to work, that
206 output must be passed to the seq_file code. Some utility functions have
207 been defined which make this task easy.
209 Most code will simply use seq_printf(), which works pretty much like
210 printk(), but which requires the seq_file pointer as an argument.
212 For straight character output, the following functions may be used:
214 seq_putc(struct seq_file *m, char c);
215 seq_puts(struct seq_file *m, const char *s);
216 seq_escape(struct seq_file *m, const char *s, const char *esc);
218 The first two output a single character and a string, just like one would
219 expect. seq_escape() is like seq_puts(), except that any character in s
220 which is in the string esc will be represented in octal form in the output.
222 There are also a pair of functions for printing filenames:
224 int seq_path(struct seq_file *m, const struct path *path,
226 int seq_path_root(struct seq_file *m, const struct path *path,
227 const struct path *root, const char *esc)
229 Here, path indicates the file of interest, and esc is a set of characters
230 which should be escaped in the output. A call to seq_path() will output
231 the path relative to the current process's filesystem root. If a different
232 root is desired, it can be used with seq_path_root(). If it turns out that
233 path cannot be reached from root, seq_path_root() returns SEQ_SKIP.
235 A function producing complicated output may want to check
236 bool seq_has_overflowed(struct seq_file *m);
237 and avoid further seq_<output> calls if true is returned.
239 A true return from seq_has_overflowed means that the seq_file buffer will
240 be discarded and the seq_show function will attempt to allocate a larger
241 buffer and retry printing.
246 So far, we have a nice set of functions which can produce output within the
247 seq_file system, but we have not yet turned them into a file that a user
248 can see. Creating a file within the kernel requires, of course, the
249 creation of a set of file_operations which implement the operations on that
250 file. The seq_file interface provides a set of canned operations which do
251 most of the work. The virtual file author still must implement the open()
252 method, however, to hook everything up. The open function is often a single
253 line, as in the example module:
255 static int ct_open(struct inode *inode, struct file *file)
257 return seq_open(file, &ct_seq_ops);
260 Here, the call to seq_open() takes the seq_operations structure we created
261 before, and gets set up to iterate through the virtual file.
263 On a successful open, seq_open() stores the struct seq_file pointer in
264 file->private_data. If you have an application where the same iterator can
265 be used for more than one file, you can store an arbitrary pointer in the
266 private field of the seq_file structure; that value can then be retrieved
267 by the iterator functions.
269 There is also a wrapper function to seq_open() called seq_open_private(). It
270 kmallocs a zero filled block of memory and stores a pointer to it in the
271 private field of the seq_file structure, returning 0 on success. The
272 block size is specified in a third parameter to the function, e.g.:
274 static int ct_open(struct inode *inode, struct file *file)
276 return seq_open_private(file, &ct_seq_ops,
277 sizeof(struct mystruct));
280 There is also a variant function, __seq_open_private(), which is functionally
281 identical except that, if successful, it returns the pointer to the allocated
282 memory block, allowing further initialisation e.g.:
284 static int ct_open(struct inode *inode, struct file *file)
287 __seq_open_private(file, &ct_seq_ops, sizeof(*p));
292 p->foo = bar; /* initialize my stuff */
299 A corresponding close function, seq_release_private() is available which
300 frees the memory allocated in the corresponding open.
302 The other operations of interest - read(), llseek(), and release() - are
303 all implemented by the seq_file code itself. So a virtual file's
304 file_operations structure will look like:
306 static const struct file_operations ct_file_ops = {
307 .owner = THIS_MODULE,
311 .release = seq_release
314 There is also a seq_release_private() which passes the contents of the
315 seq_file private field to kfree() before releasing the structure.
317 The final step is the creation of the /proc file itself. In the example
318 code, that is done in the initialization code in the usual way:
320 static int ct_init(void)
322 struct proc_dir_entry *entry;
324 proc_create("sequence", 0, NULL, &ct_file_ops);
328 module_init(ct_init);
330 And that is pretty much it.
335 If your file will be iterating through a linked list, you may find these
338 struct list_head *seq_list_start(struct list_head *head,
340 struct list_head *seq_list_start_head(struct list_head *head,
342 struct list_head *seq_list_next(void *v, struct list_head *head,
345 These helpers will interpret pos as a position within the list and iterate
346 accordingly. Your start() and next() functions need only invoke the
347 seq_list_* helpers with a pointer to the appropriate list_head structure.
350 The extra-simple version
352 For extremely simple virtual files, there is an even easier interface. A
353 module can define only the show() function, which should create all the
354 output that the virtual file will contain. The file's open() method then
357 int single_open(struct file *file,
358 int (*show)(struct seq_file *m, void *p),
361 When output time comes, the show() function will be called once. The data
362 value given to single_open() can be found in the private field of the
363 seq_file structure. When using single_open(), the programmer should use
364 single_release() instead of seq_release() in the file_operations structure
365 to avoid a memory leak.