1 .. SPDX-License-Identifier: GPL-2.0
2
3 ==================================
4 relay interface (formerly relayfs)
5 ==================================
6
7 The relay interface provides a means for kernel applications to
8 efficiently log and transfer large quantities of data from the kernel
9 to userspace via user-defined 'relay channels'.
10
11 A 'relay channel' is a kernel->user data relay mechanism implemented
12 as a set of per-cpu kernel buffers ('channel buffers'), each
13 represented as a regular file ('relay file') in user space. Kernel
14 clients write into the channel buffers using efficient write
15 functions; these automatically log into the current cpu's channel
16 buffer. User space applications mmap() or read() from the relay files
17 and retrieve the data as it becomes available. The relay files
18 themselves are files created in a host filesystem, e.g. debugfs, and
19 are associated with the channel buffers using the API described below.
20
21 The format of the data logged into the channel buffers is completely
22 up to the kernel client; the relay interface does however provide
23 hooks which allow kernel clients to impose some structure on the
24 buffer data. The relay interface doesn't implement any form of data
25 filtering - this also is left to the kernel client. The purpose is to
26 keep things as simple as possible.
27
28 This document provides an overview of the relay interface API. The
29 details of the function parameters are documented along with the
30 functions in the relay interface code - please see that for details.
31
32 Semantics
33 =========
34
35 Each relay channel has one buffer per CPU, each buffer has one or more
36 sub-buffers. Messages are written to the first sub-buffer until it is
37 too full to contain a new message, in which case it is written to
38 the next (if available). Messages are never split across sub-buffers.
39 At this point, userspace can be notified so it empties the first
40 sub-buffer, while the kernel continues writing to the next.
41
42 When notified that a sub-buffer is full, the kernel knows how many
43 bytes of it are padding i.e. unused space occurring because a complete
44 message couldn't fit into a sub-buffer. Userspace can use this
45 knowledge to copy only valid data.
46
47 After copying it, userspace can notify the kernel that a sub-buffer
48 has been consumed.
49
50 A relay channel can operate in a mode where it will overwrite data not
51 yet collected by userspace, and not wait for it to be consumed.
52
53 The relay channel itself does not provide for communication of such
54 data between userspace and kernel, allowing the kernel side to remain
55 simple and not impose a single interface on userspace. It does
56 provide a set of examples and a separate helper though, described
57 below.
58
59 The read() interface both removes padding and internally consumes the
60 read sub-buffers; thus in cases where read(2) is being used to drain
61 the channel buffers, special-purpose communication between kernel and
62 user isn't necessary for basic operation.
63
64 One of the major goals of the relay interface is to provide a low
65 overhead mechanism for conveying kernel data to userspace. While the
66 read() interface is easy to use, it's not as efficient as the mmap()
67 approach; the example code attempts to make the tradeoff between the
68 two approaches as small as possible.
69
70 klog and relay-apps example code
71 ================================
72
73 The relay interface itself is ready to use, but to make things easier,
74 a couple simple utility functions and a set of examples are provided.
75
76 The relay-apps example tarball, available on the relay sourceforge
77 site, contains a set of self-contained examples, each consisting of a
78 pair of .c files containing boilerplate code for each of the user and
79 kernel sides of a relay application. When combined these two sets of
80 boilerplate code provide glue to easily stream data to disk, without
81 having to bother with mundane housekeeping chores.
82
83 The 'klog debugging functions' patch (klog.patch in the relay-apps
84 tarball) provides a couple of high-level logging functions to the
85 kernel which allow writing formatted text or raw data to a channel,
86 regardless of whether a channel to write into exists or not, or even
87 whether the relay interface is compiled into the kernel or not. These
88 functions allow you to put unconditional 'trace' statements anywhere
89 in the kernel or kernel modules; only when there is a 'klog handler'
90 registered will data actually be logged (see the klog and kleak
91 examples for details).
92
93 It is of course possible to use the relay interface from scratch,
94 i.e. without using any of the relay-apps example code or klog, but
95 you'll have to implement communication between userspace and kernel,
96 allowing both to convey the state of buffers (full, empty, amount of
97 padding). The read() interface both removes padding and internally
98 consumes the read sub-buffers; thus in cases where read(2) is being
99 used to drain the channel buffers, special-purpose communication
100 between kernel and user isn't necessary for basic operation. Things
101 such as buffer-full conditions would still need to be communicated via
102 some channel though.
103
104 klog and the relay-apps examples can be found in the relay-apps
105 tarball on http://relayfs.sourceforge.net
106
107 The relay interface user space API
108 ==================================
109
110 The relay interface implements basic file operations for user space
111 access to relay channel buffer data. Here are the file operations
112 that are available and some comments regarding their behavior:
113
114 =========== ============================================================
115 open() enables user to open an _existing_ channel buffer.
116
117 mmap() results in channel buffer being mapped into the caller's
118 memory space. Note that you can't do a partial mmap - you
119 must map the entire file, which is NRBUF * SUBBUFSIZE.
120
121 read() read the contents of a channel buffer. The bytes read are
122 'consumed' by the reader, i.e. they won't be available
123 again to subsequent reads. If the channel is being used
124 in no-overwrite mode (the default), it can be read at any
125 time even if there's an active kernel writer. If the
126 channel is being used in overwrite mode and there are
127 active channel writers, results may be unpredictable -
128 users should make sure that all logging to the channel has
129 ended before using read() with overwrite mode. Sub-buffer
130 padding is automatically removed and will not be seen by
131 the reader.
132
133 sendfile() transfer data from a channel buffer to an output file
134 descriptor. Sub-buffer padding is automatically removed
135 and will not be seen by the reader.
136
137 poll() POLLIN/POLLRDNORM/POLLERR supported. User applications are
138 notified when sub-buffer boundaries are crossed.
139
140 close() decrements the channel buffer's refcount. When the refcount
141 reaches 0, i.e. when no process or kernel client has the
142 buffer open, the channel buffer is freed.
143 =========== ============================================================
144
145 In order for a user application to make use of relay files, the
146 host filesystem must be mounted. For example::
147
148 mount -t debugfs debugfs /sys/kernel/debug
149
150 .. Note::
151
152 the host filesystem doesn't need to be mounted for kernel
153 clients to create or use channels - it only needs to be
154 mounted when user space applications need access to the buffer
155 data.
156
157
158 The relay interface kernel API
159 ==============================
160
161 Here's a summary of the API the relay interface provides to in-kernel clients:
162
163 TBD(curr. line MT:/API/)
164 channel management functions::
165
166 relay_open(base_filename, parent, subbuf_size, n_subbufs,
167 callbacks, private_data)
168 relay_close(chan)
169 relay_flush(chan)
170 relay_reset(chan)
171
172 channel management typically called on instigation of userspace::
173
174 relay_subbufs_consumed(chan, cpu, subbufs_consumed)
175
176 write functions::
177
178 relay_write(chan, data, length)
179 __relay_write(chan, data, length)
180 relay_reserve(chan, length)
181
182 callbacks::
183
184 subbuf_start(buf, subbuf, prev_subbuf, prev_padding)
185 buf_mapped(buf, filp)
186 buf_unmapped(buf, filp)
187 create_buf_file(filename, parent, mode, buf, is_global)
188 remove_buf_file(dentry)
189
190 helper functions::
191
192 relay_buf_full(buf)
193 subbuf_start_reserve(buf, length)
194
195
196 Creating a channel
197 ------------------
198
199 relay_open() is used to create a channel, along with its per-cpu
200 channel buffers. Each channel buffer will have an associated file
201 created for it in the host filesystem, which can be and mmapped or
202 read from in user space. The files are named basename0...basenameN-1
203 where N is the number of online cpus, and by default will be created
204 in the root of the filesystem (if the parent param is NULL). If you
205 want a directory structure to contain your relay files, you should
206 create it using the host filesystem's directory creation function,
207 e.g. debugfs_create_dir(), and pass the parent directory to
208 relay_open(). Users are responsible for cleaning up any directory
209 structure they create, when the channel is closed - again the host
210 filesystem's directory removal functions should be used for that,
211 e.g. debugfs_remove().
212
213 In order for a channel to be created and the host filesystem's files
214 associated with its channel buffers, the user must provide definitions
215 for two callback functions, create_buf_file() and remove_buf_file().
216 create_buf_file() is called once for each per-cpu buffer from
217 relay_open() and allows the user to create the file which will be used
218 to represent the corresponding channel buffer. The callback should
219 return the dentry of the file created to represent the channel buffer.
220 remove_buf_file() must also be defined; it's responsible for deleting
221 the file(s) created in create_buf_file() and is called during
222 relay_close().
223
224 Here are some typical definitions for these callbacks, in this case
225 using debugfs::
226
227 /*
228 * create_buf_file() callback. Creates relay file in debugfs.
229 */
230 static struct dentry *create_buf_file_handler(const char *filename,
231 struct dentry *parent,
232 umode_t mode,
233 struct rchan_buf *buf,
234 int *is_global)
235 {
236 return debugfs_create_file(filename, mode, parent, buf,
237 &relay_file_operations);
238 }
239
240 /*
241 * remove_buf_file() callback. Removes relay file from debugfs.
242 */
243 static int remove_buf_file_handler(struct dentry *dentry)
244 {
245 debugfs_remove(dentry);
246
247 return 0;
248 }
249
250 /*
251 * relay interface callbacks
252 */
253 static struct rchan_callbacks relay_callbacks =
254 {
255 .create_buf_file = create_buf_file_handler,
256 .remove_buf_file = remove_buf_file_handler,
257 };
258
259 And an example relay_open() invocation using them::
260
261 chan = relay_open("cpu", NULL, SUBBUF_SIZE, N_SUBBUFS, &relay_callbacks, NULL);
262
263 If the create_buf_file() callback fails, or isn't defined, channel
264 creation and thus relay_open() will fail.
265
266 The total size of each per-cpu buffer is calculated by multiplying the
267 number of sub-buffers by the sub-buffer size passed into relay_open().
268 The idea behind sub-buffers is that they're basically an extension of
269 double-buffering to N buffers, and they also allow applications to
270 easily implement random-access-on-buffer-boundary schemes, which can
271 be important for some high-volume applications. The number and size
272 of sub-buffers is completely dependent on the application and even for
273 the same application, different conditions will warrant different
274 values for these parameters at different times. Typically, the right
275 values to use are best decided after some experimentation; in general,
276 though, it's safe to assume that having only 1 sub-buffer is a bad
277 idea - you're guaranteed to either overwrite data or lose events
278 depending on the channel mode being used.
279
280 The create_buf_file() implementation can also be defined in such a way
281 as to allow the creation of a single 'global' buffer instead of the
282 default per-cpu set. This can be useful for applications interested
283 mainly in seeing the relative ordering of system-wide events without
284 the need to bother with saving explicit timestamps for the purpose of
285 merging/sorting per-cpu files in a postprocessing step.
286
287 To have relay_open() create a global buffer, the create_buf_file()
288 implementation should set the value of the is_global outparam to a
289 non-zero value in addition to creating the file that will be used to
290 represent the single buffer. In the case of a global buffer,
291 create_buf_file() and remove_buf_file() will be called only once. The
292 normal channel-writing functions, e.g. relay_write(), can still be
293 used - writes from any cpu will transparently end up in the global
294 buffer - but since it is a global buffer, callers should make sure
295 they use the proper locking for such a buffer, either by wrapping
296 writes in a spinlock, or by copying a write function from relay.h and
297 creating a local version that internally does the proper locking.
298
299 The private_data passed into relay_open() allows clients to associate
300 user-defined data with a channel, and is immediately available
301 (including in create_buf_file()) via chan->private_data or
302 buf->chan->private_data.
303
304 Buffer-only channels
305 --------------------
306
307 These channels have no files associated and can be created with
308 relay_open(NULL, NULL, ...). Such channels are useful in scenarios such
309 as when doing early tracing in the kernel, before the VFS is up. In these
310 cases, one may open a buffer-only channel and then call
311 relay_late_setup_files() when the kernel is ready to handle files,
312 to expose the buffered data to the userspace.
313
314 Channel 'modes'
315 ---------------
316
317 relay channels can be used in either of two modes - 'overwrite' or
318 'no-overwrite'. The mode is entirely determined by the implementation
319 of the subbuf_start() callback, as described below. The default if no
320 subbuf_start() callback is defined is 'no-overwrite' mode. If the
321 default mode suits your needs, and you plan to use the read()
322 interface to retrieve channel data, you can ignore the details of this
323 section, as it pertains mainly to mmap() implementations.
324
325 In 'overwrite' mode, also known as 'flight recorder' mode, writes
326 continuously cycle around the buffer and will never fail, but will
327 unconditionally overwrite old data regardless of whether it's actually
328 been consumed. In no-overwrite mode, writes will fail, i.e. data will
329 be lost, if the number of unconsumed sub-buffers equals the total
330 number of sub-buffers in the channel. It should be clear that if
331 there is no consumer or if the consumer can't consume sub-buffers fast
332 enough, data will be lost in either case; the only difference is
333 whether data is lost from the beginning or the end of a buffer.
334
335 As explained above, a relay channel is made of up one or more
336 per-cpu channel buffers, each implemented as a circular buffer
337 subdivided into one or more sub-buffers. Messages are written into
338 the current sub-buffer of the channel's current per-cpu buffer via the
339 write functions described below. Whenever a message can't fit into
340 the current sub-buffer, because there's no room left for it, the
341 client is notified via the subbuf_start() callback that a switch to a
342 new sub-buffer is about to occur. The client uses this callback to 1)
343 initialize the next sub-buffer if appropriate 2) finalize the previous
344 sub-buffer if appropriate and 3) return a boolean value indicating
345 whether or not to actually move on to the next sub-buffer.
346
347 To implement 'no-overwrite' mode, the userspace client would provide
348 an implementation of the subbuf_start() callback something like the
349 following::
350
351 static int subbuf_start(struct rchan_buf *buf,
352 void *subbuf,
353 void *prev_subbuf,
354 unsigned int prev_padding)
355 {
356 if (prev_subbuf)
357 *((unsigned *)prev_subbuf) = prev_padding;
358
359 if (relay_buf_full(buf))
360 return 0;
361
362 subbuf_start_reserve(buf, sizeof(unsigned int));
363
364 return 1;
365 }
366
367 If the current buffer is full, i.e. all sub-buffers remain unconsumed,
368 the callback returns 0 to indicate that the buffer switch should not
369 occur yet, i.e. until the consumer has had a chance to read the
370 current set of ready sub-buffers. For the relay_buf_full() function
371 to make sense, the consumer is responsible for notifying the relay
372 interface when sub-buffers have been consumed via
373 relay_subbufs_consumed(). Any subsequent attempts to write into the
374 buffer will again invoke the subbuf_start() callback with the same
375 parameters; only when the consumer has consumed one or more of the
376 ready sub-buffers will relay_buf_full() return 0, in which case the
377 buffer switch can continue.
378
379 The implementation of the subbuf_start() callback for 'overwrite' mode
380 would be very similar::
381
382 static int subbuf_start(struct rchan_buf *buf,
383 void *subbuf,
384 void *prev_subbuf,
385 size_t prev_padding)
386 {
387 if (prev_subbuf)
388 *((unsigned *)prev_subbuf) = prev_padding;
389
390 subbuf_start_reserve(buf, sizeof(unsigned int));
391
392 return 1;
393 }
394
395 In this case, the relay_buf_full() check is meaningless and the
396 callback always returns 1, causing the buffer switch to occur
397 unconditionally. It's also meaningless for the client to use the
398 relay_subbufs_consumed() function in this mode, as it's never
399 consulted.
400
401 The default subbuf_start() implementation, used if the client doesn't
402 define any callbacks, or doesn't define the subbuf_start() callback,
403 implements the simplest possible 'no-overwrite' mode, i.e. it does
404 nothing but return 0.
405
406 Header information can be reserved at the beginning of each sub-buffer
407 by calling the subbuf_start_reserve() helper function from within the
408 subbuf_start() callback. This reserved area can be used to store
409 whatever information the client wants. In the example above, room is
410 reserved in each sub-buffer to store the padding count for that
411 sub-buffer. This is filled in for the previous sub-buffer in the
412 subbuf_start() implementation; the padding value for the previous
413 sub-buffer is passed into the subbuf_start() callback along with a
414 pointer to the previous sub-buffer, since the padding value isn't
415 known until a sub-buffer is filled. The subbuf_start() callback is
416 also called for the first sub-buffer when the channel is opened, to
417 give the client a chance to reserve space in it. In this case the
418 previous sub-buffer pointer passed into the callback will be NULL, so
419 the client should check the value of the prev_subbuf pointer before
420 writing into the previous sub-buffer.
421
422 Writing to a channel
423 --------------------
424
425 Kernel clients write data into the current cpu's channel buffer using
426 relay_write() or __relay_write(). relay_write() is the main logging
427 function - it uses local_irqsave() to protect the buffer and should be
428 used if you might be logging from interrupt context. If you know
429 you'll never be logging from interrupt context, you can use
430 __relay_write(), which only disables preemption. These functions
431 don't return a value, so you can't determine whether or not they
432 failed - the assumption is that you wouldn't want to check a return
433 value in the fast logging path anyway, and that they'll always succeed
434 unless the buffer is full and no-overwrite mode is being used, in
435 which case you can detect a failed write in the subbuf_start()
436 callback by calling the relay_buf_full() helper function.
437
438 relay_reserve() is used to reserve a slot in a channel buffer which
439 can be written to later. This would typically be used in applications
440 that need to write directly into a channel buffer without having to
441 stage data in a temporary buffer beforehand. Because the actual write
442 may not happen immediately after the slot is reserved, applications
443 using relay_reserve() can keep a count of the number of bytes actually
444 written, either in space reserved in the sub-buffers themselves or as
445 a separate array. See the 'reserve' example in the relay-apps tarball
446 at http://relayfs.sourceforge.net for an example of how this can be
447 done. Because the write is under control of the client and is
448 separated from the reserve, relay_reserve() doesn't protect the buffer
449 at all - it's up to the client to provide the appropriate
450 synchronization when using relay_reserve().
451
452 Closing a channel
453 -----------------
454
455 The client calls relay_close() when it's finished using the channel.
456 The channel and its associated buffers are destroyed when there are no
457 longer any references to any of the channel buffers. relay_flush()
458 forces a sub-buffer switch on all the channel buffers, and can be used
459 to finalize and process the last sub-buffers before the channel is
460 closed.
461
462 Misc
463 ----
464
465 Some applications may want to keep a channel around and re-use it
466 rather than open and close a new channel for each use. relay_reset()
467 can be used for this purpose - it resets a channel to its initial
468 state without reallocating channel buffer memory or destroying
469 existing mappings. It should however only be called when it's safe to
470 do so, i.e. when the channel isn't currently being written to.
471
472 Finally, there are a couple of utility callbacks that can be used for
473 different purposes. buf_mapped() is called whenever a channel buffer
474 is mmapped from user space and buf_unmapped() is called when it's
475 unmapped. The client can use this notification to trigger actions
476 within the kernel application, such as enabling/disabling logging to
477 the channel.
478
479
480 Resources
481 =========
482
483 For news, example code, mailing list, etc. see the relay interface homepage:
484
485 http://relayfs.sourceforge.net
486
487
488 Credits
489 =======
490
491 The ideas and specs for the relay interface came about as a result of
492 discussions on tracing involving the following:
493
494 Michel Dagenais <michel.dagenais@polymtl.ca>
495 Richard Moore <richardj_moore@uk.ibm.com>
496 Bob Wisniewski <bob@watson.ibm.com>
497 Karim Yaghmour <karim@opersys.com>
498 Tom Zanussi <zanussi@us.ibm.com>
499
500 Also thanks to Hubertus Franke for a lot of useful suggestions and bug
501 reports.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.