seq_file.rst 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396
  1. .. SPDX-License-Identifier: GPL-2.0
  2. ======================
  3. The seq_file Interface
  4. ======================
  5. Copyright 2003 Jonathan Corbet <corbet@lwn.net>
  6. This file is originally from the LWN.net Driver Porting series at
  7. https://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
  14. grown over the years.
  15. Creating those files correctly has always been a bit of a challenge,
  16. however. It is not that hard to make a virtual file which returns a
  17. string. But life gets trickier if the output is long - anything greater
  18. than an application is likely to read in a single operation. Handling
  19. multiple reads (and seeks) requires careful attention to the reader's
  20. position within the virtual file - that position is, likely as not, in the
  21. middle of a line of output. The kernel has traditionally had a number of
  22. implementations that got this wrong.
  23. The 2.6 kernel contains a set of functions (implemented by Alexander Viro)
  24. which are designed to make it easy for virtual file creators to get it
  25. right.
  26. The seq_file interface is available via <linux/seq_file.h>. There are
  27. three aspects to seq_file:
  28. * An iterator interface which lets a virtual file implementation
  29. step through the objects it is presenting.
  30. * Some utility functions for formatting objects for output without
  31. needing to worry about things like output buffers.
  32. * A set of canned file_operations which implement most operations on
  33. the virtual file.
  34. We'll look at the seq_file interface via an extremely simple example: a
  35. loadable module which creates a file called /proc/sequence. The file, when
  36. read, simply produces a set of increasing integer values, one per line. The
  37. sequence will continue until the user loses patience and finds something
  38. better to do. The file is seekable, in that one can do something like the
  39. following::
  40. dd if=/proc/sequence of=out1 count=1
  41. dd if=/proc/sequence skip=1 of=out2 count=1
  42. Then concatenate the output files out1 and out2 and get the right
  43. result. Yes, it is a thoroughly useless module, but the point is to show
  44. how the mechanism works without getting lost in other details. (Those
  45. wanting to see the full source for this module can find it at
  46. https://lwn.net/Articles/22359/).
  47. Deprecated create_proc_entry
  48. ============================
  49. Note that the above article uses create_proc_entry which was removed in
  50. kernel 3.10. Current versions require the following update::
  51. - entry = create_proc_entry("sequence", 0, NULL);
  52. - if (entry)
  53. - entry->proc_fops = &ct_file_ops;
  54. + entry = proc_create("sequence", 0, NULL, &ct_file_ops);
  55. The iterator interface
  56. ======================
  57. Modules implementing a virtual file with seq_file must implement an
  58. iterator object that allows stepping through the data of interest
  59. during a "session" (roughly one read() system call). If the iterator
  60. is able to move to a specific position - like the file they implement,
  61. though with freedom to map the position number to a sequence location
  62. in whatever way is convenient - the iterator need only exist
  63. transiently during a session. If the iterator cannot easily find a
  64. numerical position but works well with a first/next interface, the
  65. iterator can be stored in the private data area and continue from one
  66. session to the next.
  67. A seq_file implementation that is formatting firewall rules from a
  68. table, for example, could provide a simple iterator that interprets
  69. position N as the Nth rule in the chain. A seq_file implementation
  70. that presents the content of a, potentially volatile, linked list
  71. might record a pointer into that list, providing that can be done
  72. without risk of the current location being removed.
  73. Positioning can thus be done in whatever way makes the most sense for
  74. the generator of the data, which need not be aware of how a position
  75. translates to an offset in the virtual file. The one obvious exception
  76. is that a position of zero should indicate the beginning of the file.
  77. The /proc/sequence iterator just uses the count of the next number it
  78. will output as its position.
  79. Four functions must be implemented to make the iterator work. The
  80. first, called start(), starts a session and takes a position as an
  81. argument, returning an iterator which will start reading at that
  82. position. The pos passed to start() will always be either zero, or
  83. the most recent pos used in the previous session.
  84. For our simple sequence example,
  85. the start() function looks like::
  86. static void *ct_seq_start(struct seq_file *s, loff_t *pos)
  87. {
  88. loff_t *spos = kmalloc(sizeof(loff_t), GFP_KERNEL);
  89. if (! spos)
  90. return NULL;
  91. *spos = *pos;
  92. return spos;
  93. }
  94. The entire data structure for this iterator is a single loff_t value
  95. holding the current position. There is no upper bound for the sequence
  96. iterator, but that will not be the case for most other seq_file
  97. implementations; in most cases the start() function should check for a
  98. "past end of file" condition and return NULL if need be.
  99. For more complicated applications, the private field of the seq_file
  100. structure can be used to hold state from session to session. There is
  101. also a special value which can be returned by the start() function
  102. called SEQ_START_TOKEN; it can be used if you wish to instruct your
  103. show() function (described below) to print a header at the top of the
  104. output. SEQ_START_TOKEN should only be used if the offset is zero,
  105. however. SEQ_START_TOKEN has no special meaning to the core seq_file
  106. code. It is provided as a convenience for a start() funciton to
  107. communicate with the next() and show() functions.
  108. The next function to implement is called, amazingly, next(); its job is to
  109. move the iterator forward to the next position in the sequence. The
  110. example module can simply increment the position by one; more useful
  111. modules will do what is needed to step through some data structure. The
  112. next() function returns a new iterator, or NULL if the sequence is
  113. complete. Here's the example version::
  114. static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
  115. {
  116. loff_t *spos = v;
  117. *pos = ++*spos;
  118. return spos;
  119. }
  120. The next() function should set ``*pos`` to a value that start() can use
  121. to find the new location in the sequence. When the iterator is being
  122. stored in the private data area, rather than being reinitialized on each
  123. start(), it might seem sufficient to simply set ``*pos`` to any non-zero
  124. value (zero always tells start() to restart the sequence). This is not
  125. sufficient due to historical problems.
  126. Historically, many next() functions have *not* updated ``*pos`` at
  127. end-of-file. If the value is then used by start() to initialise the
  128. iterator, this can result in corner cases where the last entry in the
  129. sequence is reported twice in the file. In order to discourage this bug
  130. from being resurrected, the core seq_file code now produces a warning if
  131. a next() function does not change the value of ``*pos``. Consequently a
  132. next() function *must* change the value of ``*pos``, and of course must
  133. set it to a non-zero value.
  134. The stop() function closes a session; its job, of course, is to clean
  135. up. If dynamic memory is allocated for the iterator, stop() is the
  136. place to free it; if a lock was taken by start(), stop() must release
  137. that lock. The value that ``*pos`` was set to by the last next() call
  138. before stop() is remembered, and used for the first start() call of
  139. the next session unless lseek() has been called on the file; in that
  140. case next start() will be asked to start at position zero::
  141. static void ct_seq_stop(struct seq_file *s, void *v)
  142. {
  143. kfree(v);
  144. }
  145. Finally, the show() function should format the object currently pointed to
  146. by the iterator for output. The example module's show() function is::
  147. static int ct_seq_show(struct seq_file *s, void *v)
  148. {
  149. loff_t *spos = v;
  150. seq_printf(s, "%lld\n", (long long)*spos);
  151. return 0;
  152. }
  153. If all is well, the show() function should return zero. A negative error
  154. code in the usual manner indicates that something went wrong; it will be
  155. passed back to user space. This function can also return SEQ_SKIP, which
  156. causes the current item to be skipped; if the show() function has already
  157. generated output before returning SEQ_SKIP, that output will be dropped.
  158. We will look at seq_printf() in a moment. But first, the definition of the
  159. seq_file iterator is finished by creating a seq_operations structure with
  160. the four functions we have just defined::
  161. static const struct seq_operations ct_seq_ops = {
  162. .start = ct_seq_start,
  163. .next = ct_seq_next,
  164. .stop = ct_seq_stop,
  165. .show = ct_seq_show
  166. };
  167. This structure will be needed to tie our iterator to the /proc file in
  168. a little bit.
  169. It's worth noting that the iterator value returned by start() and
  170. manipulated by the other functions is considered to be completely opaque by
  171. the seq_file code. It can thus be anything that is useful in stepping
  172. through the data to be output. Counters can be useful, but it could also be
  173. a direct pointer into an array or linked list. Anything goes, as long as
  174. the programmer is aware that things can happen between calls to the
  175. iterator function. However, the seq_file code (by design) will not sleep
  176. between the calls to start() and stop(), so holding a lock during that time
  177. is a reasonable thing to do. The seq_file code will also avoid taking any
  178. other locks while the iterator is active.
  179. The iterater value returned by start() or next() is guaranteed to be
  180. passed to a subsequent next() or stop() call. This allows resources
  181. such as locks that were taken to be reliably released. There is *no*
  182. guarantee that the iterator will be passed to show(), though in practice
  183. it often will be.
  184. Formatted output
  185. ================
  186. The seq_file code manages positioning within the output created by the
  187. iterator and getting it into the user's buffer. But, for that to work, that
  188. output must be passed to the seq_file code. Some utility functions have
  189. been defined which make this task easy.
  190. Most code will simply use seq_printf(), which works pretty much like
  191. printk(), but which requires the seq_file pointer as an argument.
  192. For straight character output, the following functions may be used::
  193. seq_putc(struct seq_file *m, char c);
  194. seq_puts(struct seq_file *m, const char *s);
  195. seq_escape(struct seq_file *m, const char *s, const char *esc);
  196. The first two output a single character and a string, just like one would
  197. expect. seq_escape() is like seq_puts(), except that any character in s
  198. which is in the string esc will be represented in octal form in the output.
  199. There are also a pair of functions for printing filenames::
  200. int seq_path(struct seq_file *m, const struct path *path,
  201. const char *esc);
  202. int seq_path_root(struct seq_file *m, const struct path *path,
  203. const struct path *root, const char *esc)
  204. Here, path indicates the file of interest, and esc is a set of characters
  205. which should be escaped in the output. A call to seq_path() will output
  206. the path relative to the current process's filesystem root. If a different
  207. root is desired, it can be used with seq_path_root(). If it turns out that
  208. path cannot be reached from root, seq_path_root() returns SEQ_SKIP.
  209. A function producing complicated output may want to check::
  210. bool seq_has_overflowed(struct seq_file *m);
  211. and avoid further seq_<output> calls if true is returned.
  212. A true return from seq_has_overflowed means that the seq_file buffer will
  213. be discarded and the seq_show function will attempt to allocate a larger
  214. buffer and retry printing.
  215. Making it all work
  216. ==================
  217. So far, we have a nice set of functions which can produce output within the
  218. seq_file system, but we have not yet turned them into a file that a user
  219. can see. Creating a file within the kernel requires, of course, the
  220. creation of a set of file_operations which implement the operations on that
  221. file. The seq_file interface provides a set of canned operations which do
  222. most of the work. The virtual file author still must implement the open()
  223. method, however, to hook everything up. The open function is often a single
  224. line, as in the example module::
  225. static int ct_open(struct inode *inode, struct file *file)
  226. {
  227. return seq_open(file, &ct_seq_ops);
  228. }
  229. Here, the call to seq_open() takes the seq_operations structure we created
  230. before, and gets set up to iterate through the virtual file.
  231. On a successful open, seq_open() stores the struct seq_file pointer in
  232. file->private_data. If you have an application where the same iterator can
  233. be used for more than one file, you can store an arbitrary pointer in the
  234. private field of the seq_file structure; that value can then be retrieved
  235. by the iterator functions.
  236. There is also a wrapper function to seq_open() called seq_open_private(). It
  237. kmallocs a zero filled block of memory and stores a pointer to it in the
  238. private field of the seq_file structure, returning 0 on success. The
  239. block size is specified in a third parameter to the function, e.g.::
  240. static int ct_open(struct inode *inode, struct file *file)
  241. {
  242. return seq_open_private(file, &ct_seq_ops,
  243. sizeof(struct mystruct));
  244. }
  245. There is also a variant function, __seq_open_private(), which is functionally
  246. identical except that, if successful, it returns the pointer to the allocated
  247. memory block, allowing further initialisation e.g.::
  248. static int ct_open(struct inode *inode, struct file *file)
  249. {
  250. struct mystruct *p =
  251. __seq_open_private(file, &ct_seq_ops, sizeof(*p));
  252. if (!p)
  253. return -ENOMEM;
  254. p->foo = bar; /* initialize my stuff */
  255. ...
  256. p->baz = true;
  257. return 0;
  258. }
  259. A corresponding close function, seq_release_private() is available which
  260. frees the memory allocated in the corresponding open.
  261. The other operations of interest - read(), llseek(), and release() - are
  262. all implemented by the seq_file code itself. So a virtual file's
  263. file_operations structure will look like::
  264. static const struct file_operations ct_file_ops = {
  265. .owner = THIS_MODULE,
  266. .open = ct_open,
  267. .read = seq_read,
  268. .llseek = seq_lseek,
  269. .release = seq_release
  270. };
  271. There is also a seq_release_private() which passes the contents of the
  272. seq_file private field to kfree() before releasing the structure.
  273. The final step is the creation of the /proc file itself. In the example
  274. code, that is done in the initialization code in the usual way::
  275. static int ct_init(void)
  276. {
  277. struct proc_dir_entry *entry;
  278. proc_create("sequence", 0, NULL, &ct_file_ops);
  279. return 0;
  280. }
  281. module_init(ct_init);
  282. And that is pretty much it.
  283. seq_list
  284. ========
  285. If your file will be iterating through a linked list, you may find these
  286. routines useful::
  287. struct list_head *seq_list_start(struct list_head *head,
  288. loff_t pos);
  289. struct list_head *seq_list_start_head(struct list_head *head,
  290. loff_t pos);
  291. struct list_head *seq_list_next(void *v, struct list_head *head,
  292. loff_t *ppos);
  293. These helpers will interpret pos as a position within the list and iterate
  294. accordingly. Your start() and next() functions need only invoke the
  295. ``seq_list_*`` helpers with a pointer to the appropriate list_head structure.
  296. The extra-simple version
  297. ========================
  298. For extremely simple virtual files, there is an even easier interface. A
  299. module can define only the show() function, which should create all the
  300. output that the virtual file will contain. The file's open() method then
  301. calls::
  302. int single_open(struct file *file,
  303. int (*show)(struct seq_file *m, void *p),
  304. void *data);
  305. When output time comes, the show() function will be called once. The data
  306. value given to single_open() can be found in the private field of the
  307. seq_file structure. When using single_open(), the programmer should use
  308. single_release() instead of seq_release() in the file_operations structure
  309. to avoid a memory leak.