Commit | Line | Data |
---|---|---|
355b0502 GKH |
1 | /* |
2 | * drivers/misc/logger.c | |
3 | * | |
4 | * A Logging Subsystem | |
5 | * | |
6 | * Copyright (C) 2007-2008 Google, Inc. | |
7 | * | |
8 | * Robert Love <rlove@google.com> | |
9 | * | |
10 | * This software is licensed under the terms of the GNU General Public | |
11 | * License version 2, as published by the Free Software Foundation, and | |
12 | * may be copied, distributed, and modified under those terms. | |
13 | * | |
14 | * This program is distributed in the hope that it will be useful, | |
15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | * GNU General Public License for more details. | |
18 | */ | |
19 | ||
23687af9 | 20 | #include <linux/sched.h> |
355b0502 GKH |
21 | #include <linux/module.h> |
22 | #include <linux/fs.h> | |
23 | #include <linux/miscdevice.h> | |
24 | #include <linux/uaccess.h> | |
25 | #include <linux/poll.h> | |
c11a166c | 26 | #include <linux/slab.h> |
355b0502 GKH |
27 | #include <linux/time.h> |
28 | #include "logger.h" | |
29 | ||
30 | #include <asm/ioctls.h> | |
31 | ||
32 | /* | |
33 | * struct logger_log - represents a specific log, such as 'main' or 'radio' | |
34 | * | |
35 | * This structure lives from module insertion until module removal, so it does | |
36 | * not need additional reference counting. The structure is protected by the | |
37 | * mutex 'mutex'. | |
38 | */ | |
39 | struct logger_log { | |
277cdd01 | 40 | unsigned char *buffer;/* the ring buffer itself */ |
355b0502 GKH |
41 | struct miscdevice misc; /* misc device representing the log */ |
42 | wait_queue_head_t wq; /* wait queue for readers */ | |
43 | struct list_head readers; /* this log's readers */ | |
44 | struct mutex mutex; /* mutex protecting buffer */ | |
45 | size_t w_off; /* current write head offset */ | |
46 | size_t head; /* new readers start here */ | |
47 | size_t size; /* size of the log */ | |
48 | }; | |
49 | ||
50 | /* | |
51 | * struct logger_reader - a logging device open for reading | |
52 | * | |
53 | * This object lives from open to release, so we don't need additional | |
54 | * reference counting. The structure is protected by log->mutex. | |
55 | */ | |
56 | struct logger_reader { | |
57 | struct logger_log *log; /* associated log */ | |
58 | struct list_head list; /* entry in logger_log's list */ | |
59 | size_t r_off; /* current read head offset */ | |
60 | }; | |
61 | ||
62 | /* logger_offset - returns index 'n' into the log via (optimized) modulus */ | |
c626224d TB |
63 | size_t logger_offset(struct logger_log *log, size_t n) |
64 | { | |
65 | return n & (log->size-1); | |
66 | } | |
67 | ||
355b0502 GKH |
68 | |
69 | /* | |
70 | * file_get_log - Given a file structure, return the associated log | |
71 | * | |
72 | * This isn't aesthetic. We have several goals: | |
73 | * | |
277cdd01 MN |
74 | * 1) Need to quickly obtain the associated log during an I/O operation |
75 | * 2) Readers need to maintain state (logger_reader) | |
76 | * 3) Writers need to be very fast (open() should be a near no-op) | |
355b0502 GKH |
77 | * |
78 | * In the reader case, we can trivially go file->logger_reader->logger_log. | |
79 | * For a writer, we don't want to maintain a logger_reader, so we just go | |
80 | * file->logger_log. Thus what file->private_data points at depends on whether | |
81 | * or not the file was opened for reading. This function hides that dirtiness. | |
82 | */ | |
83 | static inline struct logger_log *file_get_log(struct file *file) | |
84 | { | |
85 | if (file->f_mode & FMODE_READ) { | |
86 | struct logger_reader *reader = file->private_data; | |
87 | return reader->log; | |
88 | } else | |
89 | return file->private_data; | |
90 | } | |
91 | ||
92 | /* | |
93 | * get_entry_len - Grabs the length of the payload of the next entry starting | |
94 | * from 'off'. | |
95 | * | |
96 | * Caller needs to hold log->mutex. | |
97 | */ | |
98 | static __u32 get_entry_len(struct logger_log *log, size_t off) | |
99 | { | |
100 | __u16 val; | |
101 | ||
102 | switch (log->size - off) { | |
103 | case 1: | |
104 | memcpy(&val, log->buffer + off, 1); | |
105 | memcpy(((char *) &val) + 1, log->buffer, 1); | |
106 | break; | |
107 | default: | |
108 | memcpy(&val, log->buffer + off, 2); | |
109 | } | |
110 | ||
111 | return sizeof(struct logger_entry) + val; | |
112 | } | |
113 | ||
114 | /* | |
115 | * do_read_log_to_user - reads exactly 'count' bytes from 'log' into the | |
116 | * user-space buffer 'buf'. Returns 'count' on success. | |
117 | * | |
118 | * Caller must hold log->mutex. | |
119 | */ | |
120 | static ssize_t do_read_log_to_user(struct logger_log *log, | |
121 | struct logger_reader *reader, | |
122 | char __user *buf, | |
123 | size_t count) | |
124 | { | |
125 | size_t len; | |
126 | ||
127 | /* | |
128 | * We read from the log in two disjoint operations. First, we read from | |
129 | * the current read head offset up to 'count' bytes or to the end of | |
130 | * the log, whichever comes first. | |
131 | */ | |
132 | len = min(count, log->size - reader->r_off); | |
133 | if (copy_to_user(buf, log->buffer + reader->r_off, len)) | |
134 | return -EFAULT; | |
135 | ||
136 | /* | |
137 | * Second, we read any remaining bytes, starting back at the head of | |
138 | * the log. | |
139 | */ | |
140 | if (count != len) | |
141 | if (copy_to_user(buf + len, log->buffer, count - len)) | |
142 | return -EFAULT; | |
143 | ||
c626224d | 144 | reader->r_off = logger_offset(log, reader->r_off + count); |
355b0502 GKH |
145 | |
146 | return count; | |
147 | } | |
148 | ||
149 | /* | |
150 | * logger_read - our log's read() method | |
151 | * | |
152 | * Behavior: | |
153 | * | |
277cdd01 MN |
154 | * - O_NONBLOCK works |
155 | * - If there are no log entries to read, blocks until log is written to | |
156 | * - Atomically reads exactly one log entry | |
355b0502 GKH |
157 | * |
158 | * Optimal read size is LOGGER_ENTRY_MAX_LEN. Will set errno to EINVAL if read | |
159 | * buffer is insufficient to hold next entry. | |
160 | */ | |
161 | static ssize_t logger_read(struct file *file, char __user *buf, | |
162 | size_t count, loff_t *pos) | |
163 | { | |
164 | struct logger_reader *reader = file->private_data; | |
165 | struct logger_log *log = reader->log; | |
166 | ssize_t ret; | |
167 | DEFINE_WAIT(wait); | |
168 | ||
169 | start: | |
170 | while (1) { | |
171 | prepare_to_wait(&log->wq, &wait, TASK_INTERRUPTIBLE); | |
172 | ||
173 | mutex_lock(&log->mutex); | |
174 | ret = (log->w_off == reader->r_off); | |
175 | mutex_unlock(&log->mutex); | |
176 | if (!ret) | |
177 | break; | |
178 | ||
179 | if (file->f_flags & O_NONBLOCK) { | |
180 | ret = -EAGAIN; | |
181 | break; | |
182 | } | |
183 | ||
184 | if (signal_pending(current)) { | |
185 | ret = -EINTR; | |
186 | break; | |
187 | } | |
188 | ||
189 | schedule(); | |
190 | } | |
191 | ||
192 | finish_wait(&log->wq, &wait); | |
193 | if (ret) | |
194 | return ret; | |
195 | ||
196 | mutex_lock(&log->mutex); | |
197 | ||
198 | /* is there still something to read or did we race? */ | |
199 | if (unlikely(log->w_off == reader->r_off)) { | |
200 | mutex_unlock(&log->mutex); | |
201 | goto start; | |
202 | } | |
203 | ||
204 | /* get the size of the next entry */ | |
205 | ret = get_entry_len(log, reader->r_off); | |
206 | if (count < ret) { | |
207 | ret = -EINVAL; | |
208 | goto out; | |
209 | } | |
210 | ||
211 | /* get exactly one entry from the log */ | |
212 | ret = do_read_log_to_user(log, reader, buf, ret); | |
213 | ||
214 | out: | |
215 | mutex_unlock(&log->mutex); | |
216 | ||
217 | return ret; | |
218 | } | |
219 | ||
220 | /* | |
221 | * get_next_entry - return the offset of the first valid entry at least 'len' | |
222 | * bytes after 'off'. | |
223 | * | |
224 | * Caller must hold log->mutex. | |
225 | */ | |
226 | static size_t get_next_entry(struct logger_log *log, size_t off, size_t len) | |
227 | { | |
228 | size_t count = 0; | |
229 | ||
230 | do { | |
231 | size_t nr = get_entry_len(log, off); | |
c626224d | 232 | off = logger_offset(log, off + nr); |
355b0502 GKH |
233 | count += nr; |
234 | } while (count < len); | |
235 | ||
236 | return off; | |
237 | } | |
238 | ||
239 | /* | |
240 | * clock_interval - is a < c < b in mod-space? Put another way, does the line | |
241 | * from a to b cross c? | |
242 | */ | |
243 | static inline int clock_interval(size_t a, size_t b, size_t c) | |
244 | { | |
245 | if (b < a) { | |
246 | if (a < c || b >= c) | |
247 | return 1; | |
248 | } else { | |
249 | if (a < c && b >= c) | |
250 | return 1; | |
251 | } | |
252 | ||
253 | return 0; | |
254 | } | |
255 | ||
256 | /* | |
257 | * fix_up_readers - walk the list of all readers and "fix up" any who were | |
258 | * lapped by the writer; also do the same for the default "start head". | |
259 | * We do this by "pulling forward" the readers and start head to the first | |
260 | * entry after the new write head. | |
261 | * | |
262 | * The caller needs to hold log->mutex. | |
263 | */ | |
264 | static void fix_up_readers(struct logger_log *log, size_t len) | |
265 | { | |
266 | size_t old = log->w_off; | |
c626224d | 267 | size_t new = logger_offset(log, old + len); |
355b0502 GKH |
268 | struct logger_reader *reader; |
269 | ||
270 | if (clock_interval(old, new, log->head)) | |
271 | log->head = get_next_entry(log, log->head, len); | |
272 | ||
273 | list_for_each_entry(reader, &log->readers, list) | |
274 | if (clock_interval(old, new, reader->r_off)) | |
275 | reader->r_off = get_next_entry(log, reader->r_off, len); | |
276 | } | |
277 | ||
278 | /* | |
279 | * do_write_log - writes 'len' bytes from 'buf' to 'log' | |
280 | * | |
281 | * The caller needs to hold log->mutex. | |
282 | */ | |
283 | static void do_write_log(struct logger_log *log, const void *buf, size_t count) | |
284 | { | |
285 | size_t len; | |
286 | ||
287 | len = min(count, log->size - log->w_off); | |
288 | memcpy(log->buffer + log->w_off, buf, len); | |
289 | ||
290 | if (count != len) | |
291 | memcpy(log->buffer, buf + len, count - len); | |
292 | ||
c626224d | 293 | log->w_off = logger_offset(log, log->w_off + count); |
355b0502 GKH |
294 | |
295 | } | |
296 | ||
297 | /* | |
298 | * do_write_log_user - writes 'len' bytes from the user-space buffer 'buf' to | |
299 | * the log 'log' | |
300 | * | |
301 | * The caller needs to hold log->mutex. | |
302 | * | |
303 | * Returns 'count' on success, negative error code on failure. | |
304 | */ | |
305 | static ssize_t do_write_log_from_user(struct logger_log *log, | |
306 | const void __user *buf, size_t count) | |
307 | { | |
308 | size_t len; | |
309 | ||
310 | len = min(count, log->size - log->w_off); | |
311 | if (len && copy_from_user(log->buffer + log->w_off, buf, len)) | |
312 | return -EFAULT; | |
313 | ||
314 | if (count != len) | |
315 | if (copy_from_user(log->buffer, buf + len, count - len)) | |
316 | return -EFAULT; | |
317 | ||
c626224d | 318 | log->w_off = logger_offset(log, log->w_off + count); |
355b0502 GKH |
319 | |
320 | return count; | |
321 | } | |
322 | ||
323 | /* | |
324 | * logger_aio_write - our write method, implementing support for write(), | |
325 | * writev(), and aio_write(). Writes are our fast path, and we try to optimize | |
326 | * them above all else. | |
327 | */ | |
328 | ssize_t logger_aio_write(struct kiocb *iocb, const struct iovec *iov, | |
329 | unsigned long nr_segs, loff_t ppos) | |
330 | { | |
331 | struct logger_log *log = file_get_log(iocb->ki_filp); | |
332 | size_t orig = log->w_off; | |
333 | struct logger_entry header; | |
334 | struct timespec now; | |
335 | ssize_t ret = 0; | |
336 | ||
337 | now = current_kernel_time(); | |
338 | ||
339 | header.pid = current->tgid; | |
340 | header.tid = current->pid; | |
341 | header.sec = now.tv_sec; | |
342 | header.nsec = now.tv_nsec; | |
343 | header.len = min_t(size_t, iocb->ki_left, LOGGER_ENTRY_MAX_PAYLOAD); | |
344 | ||
345 | /* null writes succeed, return zero */ | |
346 | if (unlikely(!header.len)) | |
347 | return 0; | |
348 | ||
349 | mutex_lock(&log->mutex); | |
350 | ||
351 | /* | |
352 | * Fix up any readers, pulling them forward to the first readable | |
353 | * entry after (what will be) the new write offset. We do this now | |
354 | * because if we partially fail, we can end up with clobbered log | |
355 | * entries that encroach on readable buffer. | |
356 | */ | |
357 | fix_up_readers(log, sizeof(struct logger_entry) + header.len); | |
358 | ||
359 | do_write_log(log, &header, sizeof(struct logger_entry)); | |
360 | ||
361 | while (nr_segs-- > 0) { | |
362 | size_t len; | |
363 | ssize_t nr; | |
364 | ||
365 | /* figure out how much of this vector we can keep */ | |
366 | len = min_t(size_t, iov->iov_len, header.len - ret); | |
367 | ||
368 | /* write out this segment's payload */ | |
369 | nr = do_write_log_from_user(log, iov->iov_base, len); | |
370 | if (unlikely(nr < 0)) { | |
371 | log->w_off = orig; | |
372 | mutex_unlock(&log->mutex); | |
373 | return nr; | |
374 | } | |
375 | ||
376 | iov++; | |
377 | ret += nr; | |
378 | } | |
379 | ||
380 | mutex_unlock(&log->mutex); | |
381 | ||
382 | /* wake up any blocked readers */ | |
383 | wake_up_interruptible(&log->wq); | |
384 | ||
385 | return ret; | |
386 | } | |
387 | ||
388 | static struct logger_log *get_log_from_minor(int); | |
389 | ||
390 | /* | |
391 | * logger_open - the log's open() file operation | |
392 | * | |
393 | * Note how near a no-op this is in the write-only case. Keep it that way! | |
394 | */ | |
395 | static int logger_open(struct inode *inode, struct file *file) | |
396 | { | |
397 | struct logger_log *log; | |
398 | int ret; | |
399 | ||
400 | ret = nonseekable_open(inode, file); | |
401 | if (ret) | |
402 | return ret; | |
403 | ||
404 | log = get_log_from_minor(MINOR(inode->i_rdev)); | |
405 | if (!log) | |
406 | return -ENODEV; | |
407 | ||
408 | if (file->f_mode & FMODE_READ) { | |
409 | struct logger_reader *reader; | |
410 | ||
411 | reader = kmalloc(sizeof(struct logger_reader), GFP_KERNEL); | |
412 | if (!reader) | |
413 | return -ENOMEM; | |
414 | ||
415 | reader->log = log; | |
416 | INIT_LIST_HEAD(&reader->list); | |
417 | ||
418 | mutex_lock(&log->mutex); | |
419 | reader->r_off = log->head; | |
420 | list_add_tail(&reader->list, &log->readers); | |
421 | mutex_unlock(&log->mutex); | |
422 | ||
423 | file->private_data = reader; | |
424 | } else | |
425 | file->private_data = log; | |
426 | ||
427 | return 0; | |
428 | } | |
429 | ||
430 | /* | |
431 | * logger_release - the log's release file operation | |
432 | * | |
433 | * Note this is a total no-op in the write-only case. Keep it that way! | |
434 | */ | |
435 | static int logger_release(struct inode *ignored, struct file *file) | |
436 | { | |
437 | if (file->f_mode & FMODE_READ) { | |
438 | struct logger_reader *reader = file->private_data; | |
439 | list_del(&reader->list); | |
440 | kfree(reader); | |
441 | } | |
442 | ||
443 | return 0; | |
444 | } | |
445 | ||
446 | /* | |
447 | * logger_poll - the log's poll file operation, for poll/select/epoll | |
448 | * | |
449 | * Note we always return POLLOUT, because you can always write() to the log. | |
450 | * Note also that, strictly speaking, a return value of POLLIN does not | |
451 | * guarantee that the log is readable without blocking, as there is a small | |
452 | * chance that the writer can lap the reader in the interim between poll() | |
453 | * returning and the read() request. | |
454 | */ | |
455 | static unsigned int logger_poll(struct file *file, poll_table *wait) | |
456 | { | |
457 | struct logger_reader *reader; | |
458 | struct logger_log *log; | |
459 | unsigned int ret = POLLOUT | POLLWRNORM; | |
460 | ||
461 | if (!(file->f_mode & FMODE_READ)) | |
462 | return ret; | |
463 | ||
464 | reader = file->private_data; | |
465 | log = reader->log; | |
466 | ||
467 | poll_wait(file, &log->wq, wait); | |
468 | ||
469 | mutex_lock(&log->mutex); | |
470 | if (log->w_off != reader->r_off) | |
471 | ret |= POLLIN | POLLRDNORM; | |
472 | mutex_unlock(&log->mutex); | |
473 | ||
474 | return ret; | |
475 | } | |
476 | ||
477 | static long logger_ioctl(struct file *file, unsigned int cmd, unsigned long arg) | |
478 | { | |
479 | struct logger_log *log = file_get_log(file); | |
480 | struct logger_reader *reader; | |
481 | long ret = -ENOTTY; | |
482 | ||
483 | mutex_lock(&log->mutex); | |
484 | ||
485 | switch (cmd) { | |
486 | case LOGGER_GET_LOG_BUF_SIZE: | |
487 | ret = log->size; | |
488 | break; | |
489 | case LOGGER_GET_LOG_LEN: | |
490 | if (!(file->f_mode & FMODE_READ)) { | |
491 | ret = -EBADF; | |
492 | break; | |
493 | } | |
494 | reader = file->private_data; | |
495 | if (log->w_off >= reader->r_off) | |
496 | ret = log->w_off - reader->r_off; | |
497 | else | |
498 | ret = (log->size - reader->r_off) + log->w_off; | |
499 | break; | |
500 | case LOGGER_GET_NEXT_ENTRY_LEN: | |
501 | if (!(file->f_mode & FMODE_READ)) { | |
502 | ret = -EBADF; | |
503 | break; | |
504 | } | |
505 | reader = file->private_data; | |
506 | if (log->w_off != reader->r_off) | |
507 | ret = get_entry_len(log, reader->r_off); | |
508 | else | |
509 | ret = 0; | |
510 | break; | |
511 | case LOGGER_FLUSH_LOG: | |
512 | if (!(file->f_mode & FMODE_WRITE)) { | |
513 | ret = -EBADF; | |
514 | break; | |
515 | } | |
516 | list_for_each_entry(reader, &log->readers, list) | |
517 | reader->r_off = log->w_off; | |
518 | log->head = log->w_off; | |
519 | ret = 0; | |
520 | break; | |
521 | } | |
522 | ||
523 | mutex_unlock(&log->mutex); | |
524 | ||
525 | return ret; | |
526 | } | |
527 | ||
528 | static const struct file_operations logger_fops = { | |
529 | .owner = THIS_MODULE, | |
530 | .read = logger_read, | |
531 | .aio_write = logger_aio_write, | |
532 | .poll = logger_poll, | |
533 | .unlocked_ioctl = logger_ioctl, | |
534 | .compat_ioctl = logger_ioctl, | |
535 | .open = logger_open, | |
536 | .release = logger_release, | |
537 | }; | |
538 | ||
539 | /* | |
540 | * Defines a log structure with name 'NAME' and a size of 'SIZE' bytes, which | |
541 | * must be a power of two, greater than LOGGER_ENTRY_MAX_LEN, and less than | |
542 | * LONG_MAX minus LOGGER_ENTRY_MAX_LEN. | |
543 | */ | |
544 | #define DEFINE_LOGGER_DEVICE(VAR, NAME, SIZE) \ | |
545 | static unsigned char _buf_ ## VAR[SIZE]; \ | |
546 | static struct logger_log VAR = { \ | |
547 | .buffer = _buf_ ## VAR, \ | |
548 | .misc = { \ | |
549 | .minor = MISC_DYNAMIC_MINOR, \ | |
550 | .name = NAME, \ | |
551 | .fops = &logger_fops, \ | |
552 | .parent = NULL, \ | |
553 | }, \ | |
554 | .wq = __WAIT_QUEUE_HEAD_INITIALIZER(VAR .wq), \ | |
555 | .readers = LIST_HEAD_INIT(VAR .readers), \ | |
556 | .mutex = __MUTEX_INITIALIZER(VAR .mutex), \ | |
557 | .w_off = 0, \ | |
558 | .head = 0, \ | |
559 | .size = SIZE, \ | |
560 | }; | |
561 | ||
2b374956 | 562 | DEFINE_LOGGER_DEVICE(log_main, LOGGER_LOG_MAIN, 256*1024) |
355b0502 | 563 | DEFINE_LOGGER_DEVICE(log_events, LOGGER_LOG_EVENTS, 256*1024) |
2b374956 JA |
564 | DEFINE_LOGGER_DEVICE(log_radio, LOGGER_LOG_RADIO, 256*1024) |
565 | DEFINE_LOGGER_DEVICE(log_system, LOGGER_LOG_SYSTEM, 256*1024) | |
355b0502 GKH |
566 | |
567 | static struct logger_log *get_log_from_minor(int minor) | |
568 | { | |
569 | if (log_main.misc.minor == minor) | |
570 | return &log_main; | |
571 | if (log_events.misc.minor == minor) | |
572 | return &log_events; | |
573 | if (log_radio.misc.minor == minor) | |
574 | return &log_radio; | |
3537cdaa SM |
575 | if (log_system.misc.minor == minor) |
576 | return &log_system; | |
355b0502 GKH |
577 | return NULL; |
578 | } | |
579 | ||
580 | static int __init init_log(struct logger_log *log) | |
581 | { | |
582 | int ret; | |
583 | ||
584 | ret = misc_register(&log->misc); | |
585 | if (unlikely(ret)) { | |
586 | printk(KERN_ERR "logger: failed to register misc " | |
587 | "device for log '%s'!\n", log->misc.name); | |
588 | return ret; | |
589 | } | |
590 | ||
591 | printk(KERN_INFO "logger: created %luK log '%s'\n", | |
592 | (unsigned long) log->size >> 10, log->misc.name); | |
593 | ||
594 | return 0; | |
595 | } | |
596 | ||
597 | static int __init logger_init(void) | |
598 | { | |
599 | int ret; | |
600 | ||
601 | ret = init_log(&log_main); | |
602 | if (unlikely(ret)) | |
603 | goto out; | |
604 | ||
605 | ret = init_log(&log_events); | |
606 | if (unlikely(ret)) | |
607 | goto out; | |
608 | ||
609 | ret = init_log(&log_radio); | |
610 | if (unlikely(ret)) | |
611 | goto out; | |
612 | ||
3537cdaa SM |
613 | ret = init_log(&log_system); |
614 | if (unlikely(ret)) | |
615 | goto out; | |
616 | ||
355b0502 GKH |
617 | out: |
618 | return ret; | |
619 | } | |
620 | device_initcall(logger_init); |