]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * linux/kernel/printk.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * Modified to make sys_syslog() more flexible: added commands to | |
7 | * return the last 4k of kernel messages, regardless of whether | |
8 | * they've been read or not. Added option to suppress kernel printk's | |
9 | * to the console. Added hook for sending the console messages | |
10 | * elsewhere, in preparation for a serial line console (someday). | |
11 | * Ted Ts'o, 2/11/93. | |
12 | * Modified for sysctl support, 1/8/97, Chris Horn. | |
13 | * Fixed SMP synchronization, 08/08/99, Manfred Spraul | |
14 | * manfred@colorfullife.com | |
15 | * Rewrote bits to get rid of console_lock | |
16 | * 01Mar01 Andrew Morton | |
17 | */ | |
18 | ||
19 | #include <linux/kernel.h> | |
20 | #include <linux/mm.h> | |
21 | #include <linux/tty.h> | |
22 | #include <linux/tty_driver.h> | |
23 | #include <linux/console.h> | |
24 | #include <linux/init.h> | |
25 | #include <linux/jiffies.h> | |
26 | #include <linux/nmi.h> | |
27 | #include <linux/module.h> | |
28 | #include <linux/moduleparam.h> | |
29 | #include <linux/interrupt.h> /* For in_interrupt() */ | |
30 | #include <linux/delay.h> | |
31 | #include <linux/smp.h> | |
32 | #include <linux/security.h> | |
33 | #include <linux/bootmem.h> | |
34 | #include <linux/memblock.h> | |
35 | #include <linux/syscalls.h> | |
36 | #include <linux/kexec.h> | |
37 | #include <linux/kdb.h> | |
38 | #include <linux/ratelimit.h> | |
39 | #include <linux/kmsg_dump.h> | |
40 | #include <linux/syslog.h> | |
41 | #include <linux/cpu.h> | |
42 | #include <linux/notifier.h> | |
43 | #include <linux/rculist.h> | |
44 | #include <linux/poll.h> | |
45 | ||
46 | #include <asm/uaccess.h> | |
47 | ||
48 | #define CREATE_TRACE_POINTS | |
49 | #include <trace/events/printk.h> | |
50 | ||
51 | /* | |
52 | * Architectures can override it: | |
53 | */ | |
54 | void asmlinkage __attribute__((weak)) early_printk(const char *fmt, ...) | |
55 | { | |
56 | } | |
57 | ||
58 | /* printk's without a loglevel use this.. */ | |
59 | #define DEFAULT_MESSAGE_LOGLEVEL CONFIG_DEFAULT_MESSAGE_LOGLEVEL | |
60 | ||
61 | /* We show everything that is MORE important than this.. */ | |
62 | #define MINIMUM_CONSOLE_LOGLEVEL 1 /* Minimum loglevel we let people use */ | |
63 | #define DEFAULT_CONSOLE_LOGLEVEL 7 /* anything MORE serious than KERN_DEBUG */ | |
64 | ||
65 | DECLARE_WAIT_QUEUE_HEAD(log_wait); | |
66 | ||
67 | int console_printk[4] = { | |
68 | DEFAULT_CONSOLE_LOGLEVEL, /* console_loglevel */ | |
69 | DEFAULT_MESSAGE_LOGLEVEL, /* default_message_loglevel */ | |
70 | MINIMUM_CONSOLE_LOGLEVEL, /* minimum_console_loglevel */ | |
71 | DEFAULT_CONSOLE_LOGLEVEL, /* default_console_loglevel */ | |
72 | }; | |
73 | ||
74 | /* | |
75 | * Low level drivers may need that to know if they can schedule in | |
76 | * their unblank() callback or not. So let's export it. | |
77 | */ | |
78 | int oops_in_progress; | |
79 | EXPORT_SYMBOL(oops_in_progress); | |
80 | ||
81 | /* | |
82 | * console_sem protects the console_drivers list, and also | |
83 | * provides serialisation for access to the entire console | |
84 | * driver system. | |
85 | */ | |
86 | static DEFINE_SEMAPHORE(console_sem); | |
87 | struct console *console_drivers; | |
88 | EXPORT_SYMBOL_GPL(console_drivers); | |
89 | ||
90 | /* | |
91 | * This is used for debugging the mess that is the VT code by | |
92 | * keeping track if we have the console semaphore held. It's | |
93 | * definitely not the perfect debug tool (we don't know if _WE_ | |
94 | * hold it are racing, but it helps tracking those weird code | |
95 | * path in the console code where we end up in places I want | |
96 | * locked without the console sempahore held | |
97 | */ | |
98 | static int console_locked, console_suspended; | |
99 | ||
100 | /* | |
101 | * If exclusive_console is non-NULL then only this console is to be printed to. | |
102 | */ | |
103 | static struct console *exclusive_console; | |
104 | ||
105 | /* | |
106 | * Array of consoles built from command line options (console=) | |
107 | */ | |
108 | struct console_cmdline | |
109 | { | |
110 | char name[8]; /* Name of the driver */ | |
111 | int index; /* Minor dev. to use */ | |
112 | char *options; /* Options for the driver */ | |
113 | #ifdef CONFIG_A11Y_BRAILLE_CONSOLE | |
114 | char *brl_options; /* Options for braille driver */ | |
115 | #endif | |
116 | }; | |
117 | ||
118 | #define MAX_CMDLINECONSOLES 8 | |
119 | ||
120 | static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES]; | |
121 | static int selected_console = -1; | |
122 | static int preferred_console = -1; | |
123 | int console_set_on_cmdline; | |
124 | EXPORT_SYMBOL(console_set_on_cmdline); | |
125 | ||
126 | /* Flag: console code may call schedule() */ | |
127 | static int console_may_schedule; | |
128 | ||
129 | /* | |
130 | * The printk log buffer consists of a chain of concatenated variable | |
131 | * length records. Every record starts with a record header, containing | |
132 | * the overall length of the record. | |
133 | * | |
134 | * The heads to the first and last entry in the buffer, as well as the | |
135 | * sequence numbers of these both entries are maintained when messages | |
136 | * are stored.. | |
137 | * | |
138 | * If the heads indicate available messages, the length in the header | |
139 | * tells the start next message. A length == 0 for the next message | |
140 | * indicates a wrap-around to the beginning of the buffer. | |
141 | * | |
142 | * Every record carries the monotonic timestamp in microseconds, as well as | |
143 | * the standard userspace syslog level and syslog facility. The usual | |
144 | * kernel messages use LOG_KERN; userspace-injected messages always carry | |
145 | * a matching syslog facility, by default LOG_USER. The origin of every | |
146 | * message can be reliably determined that way. | |
147 | * | |
148 | * The human readable log message directly follows the message header. The | |
149 | * length of the message text is stored in the header, the stored message | |
150 | * is not terminated. | |
151 | * | |
152 | * Optionally, a message can carry a dictionary of properties (key/value pairs), | |
153 | * to provide userspace with a machine-readable message context. | |
154 | * | |
155 | * Examples for well-defined, commonly used property names are: | |
156 | * DEVICE=b12:8 device identifier | |
157 | * b12:8 block dev_t | |
158 | * c127:3 char dev_t | |
159 | * n8 netdev ifindex | |
160 | * +sound:card0 subsystem:devname | |
161 | * SUBSYSTEM=pci driver-core subsystem name | |
162 | * | |
163 | * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value | |
164 | * follows directly after a '=' character. Every property is terminated by | |
165 | * a '\0' character. The last property is not terminated. | |
166 | * | |
167 | * Example of a message structure: | |
168 | * 0000 ff 8f 00 00 00 00 00 00 monotonic time in nsec | |
169 | * 0008 34 00 record is 52 bytes long | |
170 | * 000a 0b 00 text is 11 bytes long | |
171 | * 000c 1f 00 dictionary is 23 bytes long | |
172 | * 000e 03 00 LOG_KERN (facility) LOG_ERR (level) | |
173 | * 0010 69 74 27 73 20 61 20 6c "it's a l" | |
174 | * 69 6e 65 "ine" | |
175 | * 001b 44 45 56 49 43 "DEVIC" | |
176 | * 45 3d 62 38 3a 32 00 44 "E=b8:2\0D" | |
177 | * 52 49 56 45 52 3d 62 75 "RIVER=bu" | |
178 | * 67 "g" | |
179 | * 0032 00 00 00 padding to next message header | |
180 | * | |
181 | * The 'struct log' buffer header must never be directly exported to | |
182 | * userspace, it is a kernel-private implementation detail that might | |
183 | * need to be changed in the future, when the requirements change. | |
184 | * | |
185 | * /dev/kmsg exports the structured data in the following line format: | |
186 | * "level,sequnum,timestamp;<message text>\n" | |
187 | * | |
188 | * The optional key/value pairs are attached as continuation lines starting | |
189 | * with a space character and terminated by a newline. All possible | |
190 | * non-prinatable characters are escaped in the "\xff" notation. | |
191 | * | |
192 | * Users of the export format should ignore possible additional values | |
193 | * separated by ',', and find the message after the ';' character. | |
194 | */ | |
195 | ||
196 | enum log_flags { | |
197 | LOG_NOCONS = 1, /* already flushed, do not print to console */ | |
198 | LOG_NEWLINE = 2, /* text ended with a newline */ | |
199 | LOG_PREFIX = 4, /* text started with a prefix */ | |
200 | LOG_CONT = 8, /* text is a fragment of a continuation line */ | |
201 | }; | |
202 | ||
203 | struct log { | |
204 | u64 ts_nsec; /* timestamp in nanoseconds */ | |
205 | u16 len; /* length of entire record */ | |
206 | u16 text_len; /* length of text buffer */ | |
207 | u16 dict_len; /* length of dictionary buffer */ | |
208 | u8 facility; /* syslog facility */ | |
209 | u8 flags:5; /* internal record flags */ | |
210 | u8 level:3; /* syslog level */ | |
211 | }; | |
212 | ||
213 | /* | |
214 | * The logbuf_lock protects kmsg buffer, indices, counters. It is also | |
215 | * used in interesting ways to provide interlocking in console_unlock(); | |
216 | */ | |
217 | static DEFINE_RAW_SPINLOCK(logbuf_lock); | |
218 | ||
219 | /* the next printk record to read by syslog(READ) or /proc/kmsg */ | |
220 | static u64 syslog_seq; | |
221 | static u32 syslog_idx; | |
222 | static enum log_flags syslog_prev; | |
223 | static size_t syslog_partial; | |
224 | ||
225 | /* index and sequence number of the first record stored in the buffer */ | |
226 | static u64 log_first_seq; | |
227 | static u32 log_first_idx; | |
228 | ||
229 | /* index and sequence number of the next record to store in the buffer */ | |
230 | static u64 log_next_seq; | |
231 | #ifdef CONFIG_PRINTK | |
232 | static u32 log_next_idx; | |
233 | ||
234 | /* the next printk record to read after the last 'clear' command */ | |
235 | static u64 clear_seq; | |
236 | static u32 clear_idx; | |
237 | ||
238 | #define PREFIX_MAX 32 | |
239 | #define LOG_LINE_MAX 1024 - PREFIX_MAX | |
240 | ||
241 | /* record buffer */ | |
242 | #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) | |
243 | #define LOG_ALIGN 4 | |
244 | #else | |
245 | #define LOG_ALIGN __alignof__(struct log) | |
246 | #endif | |
247 | #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT) | |
248 | static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN); | |
249 | static char *log_buf = __log_buf; | |
250 | static u32 log_buf_len = __LOG_BUF_LEN; | |
251 | ||
252 | /* cpu currently holding logbuf_lock */ | |
253 | static volatile unsigned int logbuf_cpu = UINT_MAX; | |
254 | ||
255 | /* human readable text of the record */ | |
256 | static char *log_text(const struct log *msg) | |
257 | { | |
258 | return (char *)msg + sizeof(struct log); | |
259 | } | |
260 | ||
261 | /* optional key/value pair dictionary attached to the record */ | |
262 | static char *log_dict(const struct log *msg) | |
263 | { | |
264 | return (char *)msg + sizeof(struct log) + msg->text_len; | |
265 | } | |
266 | ||
267 | /* get record by index; idx must point to valid msg */ | |
268 | static struct log *log_from_idx(u32 idx) | |
269 | { | |
270 | struct log *msg = (struct log *)(log_buf + idx); | |
271 | ||
272 | /* | |
273 | * A length == 0 record is the end of buffer marker. Wrap around and | |
274 | * read the message at the start of the buffer. | |
275 | */ | |
276 | if (!msg->len) | |
277 | return (struct log *)log_buf; | |
278 | return msg; | |
279 | } | |
280 | ||
281 | /* get next record; idx must point to valid msg */ | |
282 | static u32 log_next(u32 idx) | |
283 | { | |
284 | struct log *msg = (struct log *)(log_buf + idx); | |
285 | ||
286 | /* length == 0 indicates the end of the buffer; wrap */ | |
287 | /* | |
288 | * A length == 0 record is the end of buffer marker. Wrap around and | |
289 | * read the message at the start of the buffer as *this* one, and | |
290 | * return the one after that. | |
291 | */ | |
292 | if (!msg->len) { | |
293 | msg = (struct log *)log_buf; | |
294 | return msg->len; | |
295 | } | |
296 | return idx + msg->len; | |
297 | } | |
298 | ||
299 | /* insert record into the buffer, discard old ones, update heads */ | |
300 | static void log_store(int facility, int level, | |
301 | enum log_flags flags, u64 ts_nsec, | |
302 | const char *dict, u16 dict_len, | |
303 | const char *text, u16 text_len) | |
304 | { | |
305 | struct log *msg; | |
306 | u32 size, pad_len; | |
307 | ||
308 | /* number of '\0' padding bytes to next message */ | |
309 | size = sizeof(struct log) + text_len + dict_len; | |
310 | pad_len = (-size) & (LOG_ALIGN - 1); | |
311 | size += pad_len; | |
312 | ||
313 | while (log_first_seq < log_next_seq) { | |
314 | u32 free; | |
315 | ||
316 | if (log_next_idx > log_first_idx) | |
317 | free = max(log_buf_len - log_next_idx, log_first_idx); | |
318 | else | |
319 | free = log_first_idx - log_next_idx; | |
320 | ||
321 | if (free > size + sizeof(struct log)) | |
322 | break; | |
323 | ||
324 | /* drop old messages until we have enough contiuous space */ | |
325 | log_first_idx = log_next(log_first_idx); | |
326 | log_first_seq++; | |
327 | } | |
328 | ||
329 | if (log_next_idx + size + sizeof(struct log) >= log_buf_len) { | |
330 | /* | |
331 | * This message + an additional empty header does not fit | |
332 | * at the end of the buffer. Add an empty header with len == 0 | |
333 | * to signify a wrap around. | |
334 | */ | |
335 | memset(log_buf + log_next_idx, 0, sizeof(struct log)); | |
336 | log_next_idx = 0; | |
337 | } | |
338 | ||
339 | /* fill message */ | |
340 | msg = (struct log *)(log_buf + log_next_idx); | |
341 | memcpy(log_text(msg), text, text_len); | |
342 | msg->text_len = text_len; | |
343 | memcpy(log_dict(msg), dict, dict_len); | |
344 | msg->dict_len = dict_len; | |
345 | msg->facility = facility; | |
346 | msg->level = level & 7; | |
347 | msg->flags = flags & 0x1f; | |
348 | if (ts_nsec > 0) | |
349 | msg->ts_nsec = ts_nsec; | |
350 | else | |
351 | msg->ts_nsec = local_clock(); | |
352 | memset(log_dict(msg) + dict_len, 0, pad_len); | |
353 | msg->len = sizeof(struct log) + text_len + dict_len + pad_len; | |
354 | ||
355 | /* insert message */ | |
356 | log_next_idx += msg->len; | |
357 | log_next_seq++; | |
358 | } | |
359 | ||
360 | /* /dev/kmsg - userspace message inject/listen interface */ | |
361 | struct devkmsg_user { | |
362 | u64 seq; | |
363 | u32 idx; | |
364 | struct mutex lock; | |
365 | char buf[8192]; | |
366 | }; | |
367 | ||
368 | static ssize_t devkmsg_writev(struct kiocb *iocb, const struct iovec *iv, | |
369 | unsigned long count, loff_t pos) | |
370 | { | |
371 | char *buf, *line; | |
372 | int i; | |
373 | int level = default_message_loglevel; | |
374 | int facility = 1; /* LOG_USER */ | |
375 | size_t len = iov_length(iv, count); | |
376 | ssize_t ret = len; | |
377 | ||
378 | if (len > LOG_LINE_MAX) | |
379 | return -EINVAL; | |
380 | buf = kmalloc(len+1, GFP_KERNEL); | |
381 | if (buf == NULL) | |
382 | return -ENOMEM; | |
383 | ||
384 | line = buf; | |
385 | for (i = 0; i < count; i++) { | |
386 | if (copy_from_user(line, iv[i].iov_base, iv[i].iov_len)) | |
387 | goto out; | |
388 | line += iv[i].iov_len; | |
389 | } | |
390 | ||
391 | /* | |
392 | * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace | |
393 | * the decimal value represents 32bit, the lower 3 bit are the log | |
394 | * level, the rest are the log facility. | |
395 | * | |
396 | * If no prefix or no userspace facility is specified, we | |
397 | * enforce LOG_USER, to be able to reliably distinguish | |
398 | * kernel-generated messages from userspace-injected ones. | |
399 | */ | |
400 | line = buf; | |
401 | if (line[0] == '<') { | |
402 | char *endp = NULL; | |
403 | ||
404 | i = simple_strtoul(line+1, &endp, 10); | |
405 | if (endp && endp[0] == '>') { | |
406 | level = i & 7; | |
407 | if (i >> 3) | |
408 | facility = i >> 3; | |
409 | endp++; | |
410 | len -= endp - line; | |
411 | line = endp; | |
412 | } | |
413 | } | |
414 | line[len] = '\0'; | |
415 | ||
416 | printk_emit(facility, level, NULL, 0, "%s", line); | |
417 | out: | |
418 | kfree(buf); | |
419 | return ret; | |
420 | } | |
421 | ||
422 | static ssize_t devkmsg_read(struct file *file, char __user *buf, | |
423 | size_t count, loff_t *ppos) | |
424 | { | |
425 | struct devkmsg_user *user = file->private_data; | |
426 | struct log *msg; | |
427 | u64 ts_usec; | |
428 | size_t i; | |
429 | size_t len; | |
430 | ssize_t ret; | |
431 | ||
432 | if (!user) | |
433 | return -EBADF; | |
434 | ||
435 | ret = mutex_lock_interruptible(&user->lock); | |
436 | if (ret) | |
437 | return ret; | |
438 | raw_spin_lock_irq(&logbuf_lock); | |
439 | while (user->seq == log_next_seq) { | |
440 | if (file->f_flags & O_NONBLOCK) { | |
441 | ret = -EAGAIN; | |
442 | raw_spin_unlock_irq(&logbuf_lock); | |
443 | goto out; | |
444 | } | |
445 | ||
446 | raw_spin_unlock_irq(&logbuf_lock); | |
447 | ret = wait_event_interruptible(log_wait, | |
448 | user->seq != log_next_seq); | |
449 | if (ret) | |
450 | goto out; | |
451 | raw_spin_lock_irq(&logbuf_lock); | |
452 | } | |
453 | ||
454 | if (user->seq < log_first_seq) { | |
455 | /* our last seen message is gone, return error and reset */ | |
456 | user->idx = log_first_idx; | |
457 | user->seq = log_first_seq; | |
458 | ret = -EPIPE; | |
459 | raw_spin_unlock_irq(&logbuf_lock); | |
460 | goto out; | |
461 | } | |
462 | ||
463 | msg = log_from_idx(user->idx); | |
464 | ts_usec = msg->ts_nsec; | |
465 | do_div(ts_usec, 1000); | |
466 | len = sprintf(user->buf, "%u,%llu,%llu;", | |
467 | (msg->facility << 3) | msg->level, user->seq, ts_usec); | |
468 | ||
469 | /* escape non-printable characters */ | |
470 | for (i = 0; i < msg->text_len; i++) { | |
471 | unsigned char c = log_text(msg)[i]; | |
472 | ||
473 | if (c < ' ' || c >= 127 || c == '\\') | |
474 | len += sprintf(user->buf + len, "\\x%02x", c); | |
475 | else | |
476 | user->buf[len++] = c; | |
477 | } | |
478 | user->buf[len++] = '\n'; | |
479 | ||
480 | if (msg->dict_len) { | |
481 | bool line = true; | |
482 | ||
483 | for (i = 0; i < msg->dict_len; i++) { | |
484 | unsigned char c = log_dict(msg)[i]; | |
485 | ||
486 | if (line) { | |
487 | user->buf[len++] = ' '; | |
488 | line = false; | |
489 | } | |
490 | ||
491 | if (c == '\0') { | |
492 | user->buf[len++] = '\n'; | |
493 | line = true; | |
494 | continue; | |
495 | } | |
496 | ||
497 | if (c < ' ' || c >= 127 || c == '\\') { | |
498 | len += sprintf(user->buf + len, "\\x%02x", c); | |
499 | continue; | |
500 | } | |
501 | ||
502 | user->buf[len++] = c; | |
503 | } | |
504 | user->buf[len++] = '\n'; | |
505 | } | |
506 | ||
507 | user->idx = log_next(user->idx); | |
508 | user->seq++; | |
509 | raw_spin_unlock_irq(&logbuf_lock); | |
510 | ||
511 | if (len > count) { | |
512 | ret = -EINVAL; | |
513 | goto out; | |
514 | } | |
515 | ||
516 | if (copy_to_user(buf, user->buf, len)) { | |
517 | ret = -EFAULT; | |
518 | goto out; | |
519 | } | |
520 | ret = len; | |
521 | out: | |
522 | mutex_unlock(&user->lock); | |
523 | return ret; | |
524 | } | |
525 | ||
526 | static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence) | |
527 | { | |
528 | struct devkmsg_user *user = file->private_data; | |
529 | loff_t ret = 0; | |
530 | ||
531 | if (!user) | |
532 | return -EBADF; | |
533 | if (offset) | |
534 | return -ESPIPE; | |
535 | ||
536 | raw_spin_lock_irq(&logbuf_lock); | |
537 | switch (whence) { | |
538 | case SEEK_SET: | |
539 | /* the first record */ | |
540 | user->idx = log_first_idx; | |
541 | user->seq = log_first_seq; | |
542 | break; | |
543 | case SEEK_DATA: | |
544 | /* | |
545 | * The first record after the last SYSLOG_ACTION_CLEAR, | |
546 | * like issued by 'dmesg -c'. Reading /dev/kmsg itself | |
547 | * changes no global state, and does not clear anything. | |
548 | */ | |
549 | user->idx = clear_idx; | |
550 | user->seq = clear_seq; | |
551 | break; | |
552 | case SEEK_END: | |
553 | /* after the last record */ | |
554 | user->idx = log_next_idx; | |
555 | user->seq = log_next_seq; | |
556 | break; | |
557 | default: | |
558 | ret = -EINVAL; | |
559 | } | |
560 | raw_spin_unlock_irq(&logbuf_lock); | |
561 | return ret; | |
562 | } | |
563 | ||
564 | static unsigned int devkmsg_poll(struct file *file, poll_table *wait) | |
565 | { | |
566 | struct devkmsg_user *user = file->private_data; | |
567 | int ret = 0; | |
568 | ||
569 | if (!user) | |
570 | return POLLERR|POLLNVAL; | |
571 | ||
572 | poll_wait(file, &log_wait, wait); | |
573 | ||
574 | raw_spin_lock_irq(&logbuf_lock); | |
575 | if (user->seq < log_next_seq) { | |
576 | /* return error when data has vanished underneath us */ | |
577 | if (user->seq < log_first_seq) | |
578 | ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI; | |
579 | ret = POLLIN|POLLRDNORM; | |
580 | } | |
581 | raw_spin_unlock_irq(&logbuf_lock); | |
582 | ||
583 | return ret; | |
584 | } | |
585 | ||
586 | static int devkmsg_open(struct inode *inode, struct file *file) | |
587 | { | |
588 | struct devkmsg_user *user; | |
589 | int err; | |
590 | ||
591 | /* write-only does not need any file context */ | |
592 | if ((file->f_flags & O_ACCMODE) == O_WRONLY) | |
593 | return 0; | |
594 | ||
595 | err = security_syslog(SYSLOG_ACTION_READ_ALL); | |
596 | if (err) | |
597 | return err; | |
598 | ||
599 | user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL); | |
600 | if (!user) | |
601 | return -ENOMEM; | |
602 | ||
603 | mutex_init(&user->lock); | |
604 | ||
605 | raw_spin_lock_irq(&logbuf_lock); | |
606 | user->idx = log_first_idx; | |
607 | user->seq = log_first_seq; | |
608 | raw_spin_unlock_irq(&logbuf_lock); | |
609 | ||
610 | file->private_data = user; | |
611 | return 0; | |
612 | } | |
613 | ||
614 | static int devkmsg_release(struct inode *inode, struct file *file) | |
615 | { | |
616 | struct devkmsg_user *user = file->private_data; | |
617 | ||
618 | if (!user) | |
619 | return 0; | |
620 | ||
621 | mutex_destroy(&user->lock); | |
622 | kfree(user); | |
623 | return 0; | |
624 | } | |
625 | ||
626 | const struct file_operations kmsg_fops = { | |
627 | .open = devkmsg_open, | |
628 | .read = devkmsg_read, | |
629 | .aio_write = devkmsg_writev, | |
630 | .llseek = devkmsg_llseek, | |
631 | .poll = devkmsg_poll, | |
632 | .release = devkmsg_release, | |
633 | }; | |
634 | ||
635 | #ifdef CONFIG_KEXEC | |
636 | /* | |
637 | * This appends the listed symbols to /proc/vmcoreinfo | |
638 | * | |
639 | * /proc/vmcoreinfo is used by various utiilties, like crash and makedumpfile to | |
640 | * obtain access to symbols that are otherwise very difficult to locate. These | |
641 | * symbols are specifically used so that utilities can access and extract the | |
642 | * dmesg log from a vmcore file after a crash. | |
643 | */ | |
644 | void log_buf_kexec_setup(void) | |
645 | { | |
646 | VMCOREINFO_SYMBOL(log_buf); | |
647 | VMCOREINFO_SYMBOL(log_buf_len); | |
648 | VMCOREINFO_SYMBOL(log_first_idx); | |
649 | VMCOREINFO_SYMBOL(log_next_idx); | |
650 | } | |
651 | #endif | |
652 | ||
653 | /* requested log_buf_len from kernel cmdline */ | |
654 | static unsigned long __initdata new_log_buf_len; | |
655 | ||
656 | /* save requested log_buf_len since it's too early to process it */ | |
657 | static int __init log_buf_len_setup(char *str) | |
658 | { | |
659 | unsigned size = memparse(str, &str); | |
660 | ||
661 | if (size) | |
662 | size = roundup_pow_of_two(size); | |
663 | if (size > log_buf_len) | |
664 | new_log_buf_len = size; | |
665 | ||
666 | return 0; | |
667 | } | |
668 | early_param("log_buf_len", log_buf_len_setup); | |
669 | ||
670 | void __init setup_log_buf(int early) | |
671 | { | |
672 | unsigned long flags; | |
673 | char *new_log_buf; | |
674 | int free; | |
675 | ||
676 | if (!new_log_buf_len) | |
677 | return; | |
678 | ||
679 | if (early) { | |
680 | unsigned long mem; | |
681 | ||
682 | mem = memblock_alloc(new_log_buf_len, PAGE_SIZE); | |
683 | if (!mem) | |
684 | return; | |
685 | new_log_buf = __va(mem); | |
686 | } else { | |
687 | new_log_buf = alloc_bootmem_nopanic(new_log_buf_len); | |
688 | } | |
689 | ||
690 | if (unlikely(!new_log_buf)) { | |
691 | pr_err("log_buf_len: %ld bytes not available\n", | |
692 | new_log_buf_len); | |
693 | return; | |
694 | } | |
695 | ||
696 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
697 | log_buf_len = new_log_buf_len; | |
698 | log_buf = new_log_buf; | |
699 | new_log_buf_len = 0; | |
700 | free = __LOG_BUF_LEN - log_next_idx; | |
701 | memcpy(log_buf, __log_buf, __LOG_BUF_LEN); | |
702 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
703 | ||
704 | pr_info("log_buf_len: %d\n", log_buf_len); | |
705 | pr_info("early log buf free: %d(%d%%)\n", | |
706 | free, (free * 100) / __LOG_BUF_LEN); | |
707 | } | |
708 | ||
709 | #ifdef CONFIG_BOOT_PRINTK_DELAY | |
710 | ||
711 | static int boot_delay; /* msecs delay after each printk during bootup */ | |
712 | static unsigned long long loops_per_msec; /* based on boot_delay */ | |
713 | ||
714 | static int __init boot_delay_setup(char *str) | |
715 | { | |
716 | unsigned long lpj; | |
717 | ||
718 | lpj = preset_lpj ? preset_lpj : 1000000; /* some guess */ | |
719 | loops_per_msec = (unsigned long long)lpj / 1000 * HZ; | |
720 | ||
721 | get_option(&str, &boot_delay); | |
722 | if (boot_delay > 10 * 1000) | |
723 | boot_delay = 0; | |
724 | ||
725 | pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, " | |
726 | "HZ: %d, loops_per_msec: %llu\n", | |
727 | boot_delay, preset_lpj, lpj, HZ, loops_per_msec); | |
728 | return 1; | |
729 | } | |
730 | __setup("boot_delay=", boot_delay_setup); | |
731 | ||
732 | static void boot_delay_msec(void) | |
733 | { | |
734 | unsigned long long k; | |
735 | unsigned long timeout; | |
736 | ||
737 | if (boot_delay == 0 || system_state != SYSTEM_BOOTING) | |
738 | return; | |
739 | ||
740 | k = (unsigned long long)loops_per_msec * boot_delay; | |
741 | ||
742 | timeout = jiffies + msecs_to_jiffies(boot_delay); | |
743 | while (k) { | |
744 | k--; | |
745 | cpu_relax(); | |
746 | /* | |
747 | * use (volatile) jiffies to prevent | |
748 | * compiler reduction; loop termination via jiffies | |
749 | * is secondary and may or may not happen. | |
750 | */ | |
751 | if (time_after(jiffies, timeout)) | |
752 | break; | |
753 | touch_nmi_watchdog(); | |
754 | } | |
755 | } | |
756 | #else | |
757 | static inline void boot_delay_msec(void) | |
758 | { | |
759 | } | |
760 | #endif | |
761 | ||
762 | #ifdef CONFIG_SECURITY_DMESG_RESTRICT | |
763 | int dmesg_restrict = 1; | |
764 | #else | |
765 | int dmesg_restrict; | |
766 | #endif | |
767 | ||
768 | static int syslog_action_restricted(int type) | |
769 | { | |
770 | if (dmesg_restrict) | |
771 | return 1; | |
772 | /* Unless restricted, we allow "read all" and "get buffer size" for everybody */ | |
773 | return type != SYSLOG_ACTION_READ_ALL && type != SYSLOG_ACTION_SIZE_BUFFER; | |
774 | } | |
775 | ||
776 | static int check_syslog_permissions(int type, bool from_file) | |
777 | { | |
778 | /* | |
779 | * If this is from /proc/kmsg and we've already opened it, then we've | |
780 | * already done the capabilities checks at open time. | |
781 | */ | |
782 | if (from_file && type != SYSLOG_ACTION_OPEN) | |
783 | return 0; | |
784 | ||
785 | if (syslog_action_restricted(type)) { | |
786 | if (capable(CAP_SYSLOG)) | |
787 | return 0; | |
788 | /* For historical reasons, accept CAP_SYS_ADMIN too, with a warning */ | |
789 | if (capable(CAP_SYS_ADMIN)) { | |
790 | printk_once(KERN_WARNING "%s (%d): " | |
791 | "Attempt to access syslog with CAP_SYS_ADMIN " | |
792 | "but no CAP_SYSLOG (deprecated).\n", | |
793 | current->comm, task_pid_nr(current)); | |
794 | return 0; | |
795 | } | |
796 | return -EPERM; | |
797 | } | |
798 | return 0; | |
799 | } | |
800 | ||
801 | #if defined(CONFIG_PRINTK_TIME) | |
802 | static bool printk_time = 1; | |
803 | #else | |
804 | static bool printk_time; | |
805 | #endif | |
806 | module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR); | |
807 | ||
808 | static size_t print_time(u64 ts, char *buf) | |
809 | { | |
810 | unsigned long rem_nsec; | |
811 | ||
812 | if (!printk_time) | |
813 | return 0; | |
814 | ||
815 | if (!buf) | |
816 | return 15; | |
817 | ||
818 | rem_nsec = do_div(ts, 1000000000); | |
819 | return sprintf(buf, "[%5lu.%06lu] ", | |
820 | (unsigned long)ts, rem_nsec / 1000); | |
821 | } | |
822 | ||
823 | static size_t print_prefix(const struct log *msg, bool syslog, char *buf) | |
824 | { | |
825 | size_t len = 0; | |
826 | unsigned int prefix = (msg->facility << 3) | msg->level; | |
827 | ||
828 | if (syslog) { | |
829 | if (buf) { | |
830 | len += sprintf(buf, "<%u>", prefix); | |
831 | } else { | |
832 | len += 3; | |
833 | if (prefix > 999) | |
834 | len += 3; | |
835 | else if (prefix > 99) | |
836 | len += 2; | |
837 | else if (prefix > 9) | |
838 | len++; | |
839 | } | |
840 | } | |
841 | ||
842 | len += print_time(msg->ts_nsec, buf ? buf + len : NULL); | |
843 | return len; | |
844 | } | |
845 | ||
846 | static size_t msg_print_text(const struct log *msg, enum log_flags prev, | |
847 | bool syslog, char *buf, size_t size) | |
848 | { | |
849 | const char *text = log_text(msg); | |
850 | size_t text_size = msg->text_len; | |
851 | bool prefix = true; | |
852 | bool newline = true; | |
853 | size_t len = 0; | |
854 | ||
855 | if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX)) | |
856 | prefix = false; | |
857 | ||
858 | if (msg->flags & LOG_CONT) { | |
859 | if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE)) | |
860 | prefix = false; | |
861 | ||
862 | if (!(msg->flags & LOG_NEWLINE)) | |
863 | newline = false; | |
864 | } | |
865 | ||
866 | do { | |
867 | const char *next = memchr(text, '\n', text_size); | |
868 | size_t text_len; | |
869 | ||
870 | if (next) { | |
871 | text_len = next - text; | |
872 | next++; | |
873 | text_size -= next - text; | |
874 | } else { | |
875 | text_len = text_size; | |
876 | } | |
877 | ||
878 | if (buf) { | |
879 | if (print_prefix(msg, syslog, NULL) + | |
880 | text_len + 1 >= size - len) | |
881 | break; | |
882 | ||
883 | if (prefix) | |
884 | len += print_prefix(msg, syslog, buf + len); | |
885 | memcpy(buf + len, text, text_len); | |
886 | len += text_len; | |
887 | if (next || newline) | |
888 | buf[len++] = '\n'; | |
889 | } else { | |
890 | /* SYSLOG_ACTION_* buffer size only calculation */ | |
891 | if (prefix) | |
892 | len += print_prefix(msg, syslog, NULL); | |
893 | len += text_len; | |
894 | if (next || newline) | |
895 | len++; | |
896 | } | |
897 | ||
898 | prefix = true; | |
899 | text = next; | |
900 | } while (text); | |
901 | ||
902 | return len; | |
903 | } | |
904 | ||
905 | static int syslog_print(char __user *buf, int size) | |
906 | { | |
907 | char *text; | |
908 | struct log *msg; | |
909 | int len = 0; | |
910 | ||
911 | text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL); | |
912 | if (!text) | |
913 | return -ENOMEM; | |
914 | ||
915 | while (size > 0) { | |
916 | size_t n; | |
917 | size_t skip; | |
918 | ||
919 | raw_spin_lock_irq(&logbuf_lock); | |
920 | if (syslog_seq < log_first_seq) { | |
921 | /* messages are gone, move to first one */ | |
922 | syslog_seq = log_first_seq; | |
923 | syslog_idx = log_first_idx; | |
924 | syslog_prev = 0; | |
925 | syslog_partial = 0; | |
926 | } | |
927 | if (syslog_seq == log_next_seq) { | |
928 | raw_spin_unlock_irq(&logbuf_lock); | |
929 | break; | |
930 | } | |
931 | ||
932 | skip = syslog_partial; | |
933 | msg = log_from_idx(syslog_idx); | |
934 | n = msg_print_text(msg, syslog_prev, true, text, | |
935 | LOG_LINE_MAX + PREFIX_MAX); | |
936 | if (n - syslog_partial <= size) { | |
937 | /* message fits into buffer, move forward */ | |
938 | syslog_idx = log_next(syslog_idx); | |
939 | syslog_seq++; | |
940 | syslog_prev = msg->flags; | |
941 | n -= syslog_partial; | |
942 | syslog_partial = 0; | |
943 | } else if (!len){ | |
944 | /* partial read(), remember position */ | |
945 | n = size; | |
946 | syslog_partial += n; | |
947 | } else | |
948 | n = 0; | |
949 | raw_spin_unlock_irq(&logbuf_lock); | |
950 | ||
951 | if (!n) | |
952 | break; | |
953 | ||
954 | if (copy_to_user(buf, text + skip, n)) { | |
955 | if (!len) | |
956 | len = -EFAULT; | |
957 | break; | |
958 | } | |
959 | ||
960 | len += n; | |
961 | size -= n; | |
962 | buf += n; | |
963 | } | |
964 | ||
965 | kfree(text); | |
966 | return len; | |
967 | } | |
968 | ||
969 | static int syslog_print_all(char __user *buf, int size, bool clear) | |
970 | { | |
971 | char *text; | |
972 | int len = 0; | |
973 | ||
974 | text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL); | |
975 | if (!text) | |
976 | return -ENOMEM; | |
977 | ||
978 | raw_spin_lock_irq(&logbuf_lock); | |
979 | if (buf) { | |
980 | u64 next_seq; | |
981 | u64 seq; | |
982 | u32 idx; | |
983 | enum log_flags prev; | |
984 | ||
985 | if (clear_seq < log_first_seq) { | |
986 | /* messages are gone, move to first available one */ | |
987 | clear_seq = log_first_seq; | |
988 | clear_idx = log_first_idx; | |
989 | } | |
990 | ||
991 | /* | |
992 | * Find first record that fits, including all following records, | |
993 | * into the user-provided buffer for this dump. | |
994 | */ | |
995 | seq = clear_seq; | |
996 | idx = clear_idx; | |
997 | prev = 0; | |
998 | while (seq < log_next_seq) { | |
999 | struct log *msg = log_from_idx(idx); | |
1000 | ||
1001 | len += msg_print_text(msg, prev, true, NULL, 0); | |
1002 | idx = log_next(idx); | |
1003 | seq++; | |
1004 | } | |
1005 | ||
1006 | /* move first record forward until length fits into the buffer */ | |
1007 | seq = clear_seq; | |
1008 | idx = clear_idx; | |
1009 | prev = 0; | |
1010 | while (len > size && seq < log_next_seq) { | |
1011 | struct log *msg = log_from_idx(idx); | |
1012 | ||
1013 | len -= msg_print_text(msg, prev, true, NULL, 0); | |
1014 | idx = log_next(idx); | |
1015 | seq++; | |
1016 | } | |
1017 | ||
1018 | /* last message fitting into this dump */ | |
1019 | next_seq = log_next_seq; | |
1020 | ||
1021 | len = 0; | |
1022 | prev = 0; | |
1023 | while (len >= 0 && seq < next_seq) { | |
1024 | struct log *msg = log_from_idx(idx); | |
1025 | int textlen; | |
1026 | ||
1027 | textlen = msg_print_text(msg, prev, true, text, | |
1028 | LOG_LINE_MAX + PREFIX_MAX); | |
1029 | if (textlen < 0) { | |
1030 | len = textlen; | |
1031 | break; | |
1032 | } | |
1033 | idx = log_next(idx); | |
1034 | seq++; | |
1035 | prev = msg->flags; | |
1036 | ||
1037 | raw_spin_unlock_irq(&logbuf_lock); | |
1038 | if (copy_to_user(buf + len, text, textlen)) | |
1039 | len = -EFAULT; | |
1040 | else | |
1041 | len += textlen; | |
1042 | raw_spin_lock_irq(&logbuf_lock); | |
1043 | ||
1044 | if (seq < log_first_seq) { | |
1045 | /* messages are gone, move to next one */ | |
1046 | seq = log_first_seq; | |
1047 | idx = log_first_idx; | |
1048 | prev = 0; | |
1049 | } | |
1050 | } | |
1051 | } | |
1052 | ||
1053 | if (clear) { | |
1054 | clear_seq = log_next_seq; | |
1055 | clear_idx = log_next_idx; | |
1056 | } | |
1057 | raw_spin_unlock_irq(&logbuf_lock); | |
1058 | ||
1059 | kfree(text); | |
1060 | return len; | |
1061 | } | |
1062 | ||
1063 | int do_syslog(int type, char __user *buf, int len, bool from_file) | |
1064 | { | |
1065 | bool clear = false; | |
1066 | static int saved_console_loglevel = -1; | |
1067 | int error; | |
1068 | ||
1069 | error = check_syslog_permissions(type, from_file); | |
1070 | if (error) | |
1071 | goto out; | |
1072 | ||
1073 | error = security_syslog(type); | |
1074 | if (error) | |
1075 | return error; | |
1076 | ||
1077 | switch (type) { | |
1078 | case SYSLOG_ACTION_CLOSE: /* Close log */ | |
1079 | break; | |
1080 | case SYSLOG_ACTION_OPEN: /* Open log */ | |
1081 | break; | |
1082 | case SYSLOG_ACTION_READ: /* Read from log */ | |
1083 | error = -EINVAL; | |
1084 | if (!buf || len < 0) | |
1085 | goto out; | |
1086 | error = 0; | |
1087 | if (!len) | |
1088 | goto out; | |
1089 | if (!access_ok(VERIFY_WRITE, buf, len)) { | |
1090 | error = -EFAULT; | |
1091 | goto out; | |
1092 | } | |
1093 | error = wait_event_interruptible(log_wait, | |
1094 | syslog_seq != log_next_seq); | |
1095 | if (error) | |
1096 | goto out; | |
1097 | error = syslog_print(buf, len); | |
1098 | break; | |
1099 | /* Read/clear last kernel messages */ | |
1100 | case SYSLOG_ACTION_READ_CLEAR: | |
1101 | clear = true; | |
1102 | /* FALL THRU */ | |
1103 | /* Read last kernel messages */ | |
1104 | case SYSLOG_ACTION_READ_ALL: | |
1105 | error = -EINVAL; | |
1106 | if (!buf || len < 0) | |
1107 | goto out; | |
1108 | error = 0; | |
1109 | if (!len) | |
1110 | goto out; | |
1111 | if (!access_ok(VERIFY_WRITE, buf, len)) { | |
1112 | error = -EFAULT; | |
1113 | goto out; | |
1114 | } | |
1115 | error = syslog_print_all(buf, len, clear); | |
1116 | break; | |
1117 | /* Clear ring buffer */ | |
1118 | case SYSLOG_ACTION_CLEAR: | |
1119 | syslog_print_all(NULL, 0, true); | |
1120 | break; | |
1121 | /* Disable logging to console */ | |
1122 | case SYSLOG_ACTION_CONSOLE_OFF: | |
1123 | if (saved_console_loglevel == -1) | |
1124 | saved_console_loglevel = console_loglevel; | |
1125 | console_loglevel = minimum_console_loglevel; | |
1126 | break; | |
1127 | /* Enable logging to console */ | |
1128 | case SYSLOG_ACTION_CONSOLE_ON: | |
1129 | if (saved_console_loglevel != -1) { | |
1130 | console_loglevel = saved_console_loglevel; | |
1131 | saved_console_loglevel = -1; | |
1132 | } | |
1133 | break; | |
1134 | /* Set level of messages printed to console */ | |
1135 | case SYSLOG_ACTION_CONSOLE_LEVEL: | |
1136 | error = -EINVAL; | |
1137 | if (len < 1 || len > 8) | |
1138 | goto out; | |
1139 | if (len < minimum_console_loglevel) | |
1140 | len = minimum_console_loglevel; | |
1141 | console_loglevel = len; | |
1142 | /* Implicitly re-enable logging to console */ | |
1143 | saved_console_loglevel = -1; | |
1144 | error = 0; | |
1145 | break; | |
1146 | /* Number of chars in the log buffer */ | |
1147 | case SYSLOG_ACTION_SIZE_UNREAD: | |
1148 | raw_spin_lock_irq(&logbuf_lock); | |
1149 | if (syslog_seq < log_first_seq) { | |
1150 | /* messages are gone, move to first one */ | |
1151 | syslog_seq = log_first_seq; | |
1152 | syslog_idx = log_first_idx; | |
1153 | syslog_prev = 0; | |
1154 | syslog_partial = 0; | |
1155 | } | |
1156 | if (from_file) { | |
1157 | /* | |
1158 | * Short-cut for poll(/"proc/kmsg") which simply checks | |
1159 | * for pending data, not the size; return the count of | |
1160 | * records, not the length. | |
1161 | */ | |
1162 | error = log_next_idx - syslog_idx; | |
1163 | } else { | |
1164 | u64 seq = syslog_seq; | |
1165 | u32 idx = syslog_idx; | |
1166 | enum log_flags prev = syslog_prev; | |
1167 | ||
1168 | error = 0; | |
1169 | while (seq < log_next_seq) { | |
1170 | struct log *msg = log_from_idx(idx); | |
1171 | ||
1172 | error += msg_print_text(msg, prev, true, NULL, 0); | |
1173 | idx = log_next(idx); | |
1174 | seq++; | |
1175 | prev = msg->flags; | |
1176 | } | |
1177 | error -= syslog_partial; | |
1178 | } | |
1179 | raw_spin_unlock_irq(&logbuf_lock); | |
1180 | break; | |
1181 | /* Size of the log buffer */ | |
1182 | case SYSLOG_ACTION_SIZE_BUFFER: | |
1183 | error = log_buf_len; | |
1184 | break; | |
1185 | default: | |
1186 | error = -EINVAL; | |
1187 | break; | |
1188 | } | |
1189 | out: | |
1190 | return error; | |
1191 | } | |
1192 | ||
1193 | SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len) | |
1194 | { | |
1195 | return do_syslog(type, buf, len, SYSLOG_FROM_CALL); | |
1196 | } | |
1197 | ||
1198 | #ifdef CONFIG_KGDB_KDB | |
1199 | /* kdb dmesg command needs access to the syslog buffer. do_syslog() | |
1200 | * uses locks so it cannot be used during debugging. Just tell kdb | |
1201 | * where the start and end of the physical and logical logs are. This | |
1202 | * is equivalent to do_syslog(3). | |
1203 | */ | |
1204 | void kdb_syslog_data(char *syslog_data[4]) | |
1205 | { | |
1206 | syslog_data[0] = log_buf; | |
1207 | syslog_data[1] = log_buf + log_buf_len; | |
1208 | syslog_data[2] = log_buf + log_first_idx; | |
1209 | syslog_data[3] = log_buf + log_next_idx; | |
1210 | } | |
1211 | #endif /* CONFIG_KGDB_KDB */ | |
1212 | ||
1213 | static bool __read_mostly ignore_loglevel; | |
1214 | ||
1215 | static int __init ignore_loglevel_setup(char *str) | |
1216 | { | |
1217 | ignore_loglevel = 1; | |
1218 | printk(KERN_INFO "debug: ignoring loglevel setting.\n"); | |
1219 | ||
1220 | return 0; | |
1221 | } | |
1222 | ||
1223 | early_param("ignore_loglevel", ignore_loglevel_setup); | |
1224 | module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR); | |
1225 | MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to" | |
1226 | "print all kernel messages to the console."); | |
1227 | ||
1228 | /* | |
1229 | * Call the console drivers, asking them to write out | |
1230 | * log_buf[start] to log_buf[end - 1]. | |
1231 | * The console_lock must be held. | |
1232 | */ | |
1233 | static void call_console_drivers(int level, const char *text, size_t len) | |
1234 | { | |
1235 | struct console *con; | |
1236 | ||
1237 | trace_console(text, 0, len, len); | |
1238 | ||
1239 | if (level >= console_loglevel && !ignore_loglevel) | |
1240 | return; | |
1241 | if (!console_drivers) | |
1242 | return; | |
1243 | ||
1244 | for_each_console(con) { | |
1245 | if (exclusive_console && con != exclusive_console) | |
1246 | continue; | |
1247 | if (!(con->flags & CON_ENABLED)) | |
1248 | continue; | |
1249 | if (!con->write) | |
1250 | continue; | |
1251 | if (!cpu_online(smp_processor_id()) && | |
1252 | !(con->flags & CON_ANYTIME)) | |
1253 | continue; | |
1254 | con->write(con, text, len); | |
1255 | } | |
1256 | } | |
1257 | ||
1258 | /* | |
1259 | * Zap console related locks when oopsing. Only zap at most once | |
1260 | * every 10 seconds, to leave time for slow consoles to print a | |
1261 | * full oops. | |
1262 | */ | |
1263 | static void zap_locks(void) | |
1264 | { | |
1265 | static unsigned long oops_timestamp; | |
1266 | ||
1267 | if (time_after_eq(jiffies, oops_timestamp) && | |
1268 | !time_after(jiffies, oops_timestamp + 30 * HZ)) | |
1269 | return; | |
1270 | ||
1271 | oops_timestamp = jiffies; | |
1272 | ||
1273 | debug_locks_off(); | |
1274 | /* If a crash is occurring, make sure we can't deadlock */ | |
1275 | raw_spin_lock_init(&logbuf_lock); | |
1276 | /* And make sure that we print immediately */ | |
1277 | sema_init(&console_sem, 1); | |
1278 | } | |
1279 | ||
1280 | /* Check if we have any console registered that can be called early in boot. */ | |
1281 | static int have_callable_console(void) | |
1282 | { | |
1283 | struct console *con; | |
1284 | ||
1285 | for_each_console(con) | |
1286 | if (con->flags & CON_ANYTIME) | |
1287 | return 1; | |
1288 | ||
1289 | return 0; | |
1290 | } | |
1291 | ||
1292 | /* | |
1293 | * Can we actually use the console at this time on this cpu? | |
1294 | * | |
1295 | * Console drivers may assume that per-cpu resources have | |
1296 | * been allocated. So unless they're explicitly marked as | |
1297 | * being able to cope (CON_ANYTIME) don't call them until | |
1298 | * this CPU is officially up. | |
1299 | */ | |
1300 | static inline int can_use_console(unsigned int cpu) | |
1301 | { | |
1302 | return cpu_online(cpu) || have_callable_console(); | |
1303 | } | |
1304 | ||
1305 | /* | |
1306 | * Try to get console ownership to actually show the kernel | |
1307 | * messages from a 'printk'. Return true (and with the | |
1308 | * console_lock held, and 'console_locked' set) if it | |
1309 | * is successful, false otherwise. | |
1310 | * | |
1311 | * This gets called with the 'logbuf_lock' spinlock held and | |
1312 | * interrupts disabled. It should return with 'lockbuf_lock' | |
1313 | * released but interrupts still disabled. | |
1314 | */ | |
1315 | static int console_trylock_for_printk(unsigned int cpu) | |
1316 | __releases(&logbuf_lock) | |
1317 | { | |
1318 | int retval = 0, wake = 0; | |
1319 | ||
1320 | if (console_trylock()) { | |
1321 | retval = 1; | |
1322 | ||
1323 | /* | |
1324 | * If we can't use the console, we need to release | |
1325 | * the console semaphore by hand to avoid flushing | |
1326 | * the buffer. We need to hold the console semaphore | |
1327 | * in order to do this test safely. | |
1328 | */ | |
1329 | if (!can_use_console(cpu)) { | |
1330 | console_locked = 0; | |
1331 | wake = 1; | |
1332 | retval = 0; | |
1333 | } | |
1334 | } | |
1335 | logbuf_cpu = UINT_MAX; | |
1336 | if (wake) | |
1337 | up(&console_sem); | |
1338 | raw_spin_unlock(&logbuf_lock); | |
1339 | return retval; | |
1340 | } | |
1341 | ||
1342 | int printk_delay_msec __read_mostly; | |
1343 | ||
1344 | static inline void printk_delay(void) | |
1345 | { | |
1346 | if (unlikely(printk_delay_msec)) { | |
1347 | int m = printk_delay_msec; | |
1348 | ||
1349 | while (m--) { | |
1350 | mdelay(1); | |
1351 | touch_nmi_watchdog(); | |
1352 | } | |
1353 | } | |
1354 | } | |
1355 | ||
1356 | /* | |
1357 | * Continuation lines are buffered, and not committed to the record buffer | |
1358 | * until the line is complete, or a race forces it. The line fragments | |
1359 | * though, are printed immediately to the consoles to ensure everything has | |
1360 | * reached the console in case of a kernel crash. | |
1361 | */ | |
1362 | static struct cont { | |
1363 | char buf[LOG_LINE_MAX]; | |
1364 | size_t len; /* length == 0 means unused buffer */ | |
1365 | size_t cons; /* bytes written to console */ | |
1366 | struct task_struct *owner; /* task of first print*/ | |
1367 | u64 ts_nsec; /* time of first print */ | |
1368 | u8 level; /* log level of first message */ | |
1369 | u8 facility; /* log level of first message */ | |
1370 | bool flushed:1; /* buffer sealed and committed */ | |
1371 | } cont; | |
1372 | ||
1373 | static void cont_flush(enum log_flags flags) | |
1374 | { | |
1375 | if (cont.flushed) | |
1376 | return; | |
1377 | if (cont.len == 0) | |
1378 | return; | |
1379 | ||
1380 | log_store(cont.facility, cont.level, LOG_NOCONS | flags, | |
1381 | cont.ts_nsec, NULL, 0, cont.buf, cont.len); | |
1382 | ||
1383 | cont.flushed = true; | |
1384 | } | |
1385 | ||
1386 | static bool cont_add(int facility, int level, const char *text, size_t len) | |
1387 | { | |
1388 | if (cont.len && cont.flushed) | |
1389 | return false; | |
1390 | ||
1391 | if (cont.len + len > sizeof(cont.buf)) { | |
1392 | /* the line gets too long, split it up in separate records */ | |
1393 | cont_flush(LOG_CONT); | |
1394 | return false; | |
1395 | } | |
1396 | ||
1397 | if (!cont.len) { | |
1398 | cont.facility = facility; | |
1399 | cont.level = level; | |
1400 | cont.owner = current; | |
1401 | cont.ts_nsec = local_clock(); | |
1402 | cont.cons = 0; | |
1403 | cont.flushed = false; | |
1404 | } | |
1405 | ||
1406 | memcpy(cont.buf + cont.len, text, len); | |
1407 | cont.len += len; | |
1408 | return true; | |
1409 | } | |
1410 | ||
1411 | static size_t cont_print_text(char *text, size_t size) | |
1412 | { | |
1413 | size_t textlen = 0; | |
1414 | size_t len; | |
1415 | ||
1416 | if (cont.cons == 0) { | |
1417 | textlen += print_time(cont.ts_nsec, text); | |
1418 | size -= textlen; | |
1419 | } | |
1420 | ||
1421 | len = cont.len - cont.cons; | |
1422 | if (len > 0) { | |
1423 | if (len+1 > size) | |
1424 | len = size-1; | |
1425 | memcpy(text + textlen, cont.buf + cont.cons, len); | |
1426 | textlen += len; | |
1427 | cont.cons = cont.len; | |
1428 | } | |
1429 | ||
1430 | if (cont.flushed) { | |
1431 | text[textlen++] = '\n'; | |
1432 | /* got everything, release buffer */ | |
1433 | cont.len = 0; | |
1434 | } | |
1435 | return textlen; | |
1436 | } | |
1437 | ||
1438 | asmlinkage int vprintk_emit(int facility, int level, | |
1439 | const char *dict, size_t dictlen, | |
1440 | const char *fmt, va_list args) | |
1441 | { | |
1442 | static int recursion_bug; | |
1443 | static char textbuf[LOG_LINE_MAX]; | |
1444 | char *text = textbuf; | |
1445 | size_t text_len; | |
1446 | enum log_flags lflags = 0; | |
1447 | unsigned long flags; | |
1448 | int this_cpu; | |
1449 | int printed_len = 0; | |
1450 | ||
1451 | boot_delay_msec(); | |
1452 | printk_delay(); | |
1453 | ||
1454 | /* This stops the holder of console_sem just where we want him */ | |
1455 | local_irq_save(flags); | |
1456 | this_cpu = smp_processor_id(); | |
1457 | ||
1458 | /* | |
1459 | * Ouch, printk recursed into itself! | |
1460 | */ | |
1461 | if (unlikely(logbuf_cpu == this_cpu)) { | |
1462 | /* | |
1463 | * If a crash is occurring during printk() on this CPU, | |
1464 | * then try to get the crash message out but make sure | |
1465 | * we can't deadlock. Otherwise just return to avoid the | |
1466 | * recursion and return - but flag the recursion so that | |
1467 | * it can be printed at the next appropriate moment: | |
1468 | */ | |
1469 | if (!oops_in_progress && !lockdep_recursing(current)) { | |
1470 | recursion_bug = 1; | |
1471 | goto out_restore_irqs; | |
1472 | } | |
1473 | zap_locks(); | |
1474 | } | |
1475 | ||
1476 | lockdep_off(); | |
1477 | raw_spin_lock(&logbuf_lock); | |
1478 | logbuf_cpu = this_cpu; | |
1479 | ||
1480 | if (recursion_bug) { | |
1481 | static const char recursion_msg[] = | |
1482 | "BUG: recent printk recursion!"; | |
1483 | ||
1484 | recursion_bug = 0; | |
1485 | printed_len += strlen(recursion_msg); | |
1486 | /* emit KERN_CRIT message */ | |
1487 | log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0, | |
1488 | NULL, 0, recursion_msg, printed_len); | |
1489 | } | |
1490 | ||
1491 | /* | |
1492 | * The printf needs to come first; we need the syslog | |
1493 | * prefix which might be passed-in as a parameter. | |
1494 | */ | |
1495 | text_len = vscnprintf(text, sizeof(textbuf), fmt, args); | |
1496 | ||
1497 | /* mark and strip a trailing newline */ | |
1498 | if (text_len && text[text_len-1] == '\n') { | |
1499 | text_len--; | |
1500 | lflags |= LOG_NEWLINE; | |
1501 | } | |
1502 | ||
1503 | /* strip syslog prefix and extract log level or control flags */ | |
1504 | if (text[0] == '<' && text[1] && text[2] == '>') { | |
1505 | switch (text[1]) { | |
1506 | case '0' ... '7': | |
1507 | if (level == -1) | |
1508 | level = text[1] - '0'; | |
1509 | case 'd': /* KERN_DEFAULT */ | |
1510 | lflags |= LOG_PREFIX; | |
1511 | case 'c': /* KERN_CONT */ | |
1512 | text += 3; | |
1513 | text_len -= 3; | |
1514 | } | |
1515 | } | |
1516 | ||
1517 | if (level == -1) | |
1518 | level = default_message_loglevel; | |
1519 | ||
1520 | if (dict) | |
1521 | lflags |= LOG_PREFIX|LOG_NEWLINE; | |
1522 | ||
1523 | if (!(lflags & LOG_NEWLINE)) { | |
1524 | /* | |
1525 | * Flush the conflicting buffer. An earlier newline was missing, | |
1526 | * or another task also prints continuation lines. | |
1527 | */ | |
1528 | if (cont.len && (lflags & LOG_PREFIX || cont.owner != current)) | |
1529 | cont_flush(0); | |
1530 | ||
1531 | /* buffer line if possible, otherwise store it right away */ | |
1532 | if (!cont_add(facility, level, text, text_len)) | |
1533 | log_store(facility, level, lflags | LOG_CONT, 0, | |
1534 | dict, dictlen, text, text_len); | |
1535 | } else { | |
1536 | bool stored = false; | |
1537 | ||
1538 | /* | |
1539 | * If an earlier newline was missing and it was the same task, | |
1540 | * either merge it with the current buffer and flush, or if | |
1541 | * there was a race with interrupts (prefix == true) then just | |
1542 | * flush it out and store this line separately. | |
1543 | */ | |
1544 | if (cont.len && cont.owner == current) { | |
1545 | if (!(lflags & LOG_PREFIX)) | |
1546 | stored = cont_add(facility, level, text, text_len); | |
1547 | cont_flush(0); | |
1548 | } | |
1549 | ||
1550 | if (!stored) | |
1551 | log_store(facility, level, lflags, 0, | |
1552 | dict, dictlen, text, text_len); | |
1553 | } | |
1554 | printed_len += text_len; | |
1555 | ||
1556 | /* | |
1557 | * Try to acquire and then immediately release the console semaphore. | |
1558 | * The release will print out buffers and wake up /dev/kmsg and syslog() | |
1559 | * users. | |
1560 | * | |
1561 | * The console_trylock_for_printk() function will release 'logbuf_lock' | |
1562 | * regardless of whether it actually gets the console semaphore or not. | |
1563 | */ | |
1564 | if (console_trylock_for_printk(this_cpu)) | |
1565 | console_unlock(); | |
1566 | ||
1567 | lockdep_on(); | |
1568 | out_restore_irqs: | |
1569 | local_irq_restore(flags); | |
1570 | ||
1571 | return printed_len; | |
1572 | } | |
1573 | EXPORT_SYMBOL(vprintk_emit); | |
1574 | ||
1575 | asmlinkage int vprintk(const char *fmt, va_list args) | |
1576 | { | |
1577 | return vprintk_emit(0, -1, NULL, 0, fmt, args); | |
1578 | } | |
1579 | EXPORT_SYMBOL(vprintk); | |
1580 | ||
1581 | asmlinkage int printk_emit(int facility, int level, | |
1582 | const char *dict, size_t dictlen, | |
1583 | const char *fmt, ...) | |
1584 | { | |
1585 | va_list args; | |
1586 | int r; | |
1587 | ||
1588 | va_start(args, fmt); | |
1589 | r = vprintk_emit(facility, level, dict, dictlen, fmt, args); | |
1590 | va_end(args); | |
1591 | ||
1592 | return r; | |
1593 | } | |
1594 | EXPORT_SYMBOL(printk_emit); | |
1595 | ||
1596 | /** | |
1597 | * printk - print a kernel message | |
1598 | * @fmt: format string | |
1599 | * | |
1600 | * This is printk(). It can be called from any context. We want it to work. | |
1601 | * | |
1602 | * We try to grab the console_lock. If we succeed, it's easy - we log the | |
1603 | * output and call the console drivers. If we fail to get the semaphore, we | |
1604 | * place the output into the log buffer and return. The current holder of | |
1605 | * the console_sem will notice the new output in console_unlock(); and will | |
1606 | * send it to the consoles before releasing the lock. | |
1607 | * | |
1608 | * One effect of this deferred printing is that code which calls printk() and | |
1609 | * then changes console_loglevel may break. This is because console_loglevel | |
1610 | * is inspected when the actual printing occurs. | |
1611 | * | |
1612 | * See also: | |
1613 | * printf(3) | |
1614 | * | |
1615 | * See the vsnprintf() documentation for format string extensions over C99. | |
1616 | */ | |
1617 | asmlinkage int printk(const char *fmt, ...) | |
1618 | { | |
1619 | va_list args; | |
1620 | int r; | |
1621 | ||
1622 | #ifdef CONFIG_KGDB_KDB | |
1623 | if (unlikely(kdb_trap_printk)) { | |
1624 | va_start(args, fmt); | |
1625 | r = vkdb_printf(fmt, args); | |
1626 | va_end(args); | |
1627 | return r; | |
1628 | } | |
1629 | #endif | |
1630 | va_start(args, fmt); | |
1631 | r = vprintk_emit(0, -1, NULL, 0, fmt, args); | |
1632 | va_end(args); | |
1633 | ||
1634 | return r; | |
1635 | } | |
1636 | EXPORT_SYMBOL(printk); | |
1637 | ||
1638 | #else | |
1639 | ||
1640 | #define LOG_LINE_MAX 0 | |
1641 | #define PREFIX_MAX 0 | |
1642 | static struct cont { | |
1643 | size_t len; | |
1644 | size_t cons; | |
1645 | u8 level; | |
1646 | bool flushed:1; | |
1647 | } cont; | |
1648 | static struct log *log_from_idx(u32 idx) { return NULL; } | |
1649 | static u32 log_next(u32 idx) { return 0; } | |
1650 | static void call_console_drivers(int level, const char *text, size_t len) {} | |
1651 | static size_t msg_print_text(const struct log *msg, enum log_flags prev, | |
1652 | bool syslog, char *buf, size_t size) { return 0; } | |
1653 | static size_t cont_print_text(char *text, size_t size) { return 0; } | |
1654 | ||
1655 | #endif /* CONFIG_PRINTK */ | |
1656 | ||
1657 | static int __add_preferred_console(char *name, int idx, char *options, | |
1658 | char *brl_options) | |
1659 | { | |
1660 | struct console_cmdline *c; | |
1661 | int i; | |
1662 | ||
1663 | /* | |
1664 | * See if this tty is not yet registered, and | |
1665 | * if we have a slot free. | |
1666 | */ | |
1667 | for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++) | |
1668 | if (strcmp(console_cmdline[i].name, name) == 0 && | |
1669 | console_cmdline[i].index == idx) { | |
1670 | if (!brl_options) | |
1671 | selected_console = i; | |
1672 | return 0; | |
1673 | } | |
1674 | if (i == MAX_CMDLINECONSOLES) | |
1675 | return -E2BIG; | |
1676 | if (!brl_options) | |
1677 | selected_console = i; | |
1678 | c = &console_cmdline[i]; | |
1679 | strlcpy(c->name, name, sizeof(c->name)); | |
1680 | c->options = options; | |
1681 | #ifdef CONFIG_A11Y_BRAILLE_CONSOLE | |
1682 | c->brl_options = brl_options; | |
1683 | #endif | |
1684 | c->index = idx; | |
1685 | return 0; | |
1686 | } | |
1687 | /* | |
1688 | * Set up a list of consoles. Called from init/main.c | |
1689 | */ | |
1690 | static int __init console_setup(char *str) | |
1691 | { | |
1692 | char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */ | |
1693 | char *s, *options, *brl_options = NULL; | |
1694 | int idx; | |
1695 | ||
1696 | #ifdef CONFIG_A11Y_BRAILLE_CONSOLE | |
1697 | if (!memcmp(str, "brl,", 4)) { | |
1698 | brl_options = ""; | |
1699 | str += 4; | |
1700 | } else if (!memcmp(str, "brl=", 4)) { | |
1701 | brl_options = str + 4; | |
1702 | str = strchr(brl_options, ','); | |
1703 | if (!str) { | |
1704 | printk(KERN_ERR "need port name after brl=\n"); | |
1705 | return 1; | |
1706 | } | |
1707 | *(str++) = 0; | |
1708 | } | |
1709 | #endif | |
1710 | ||
1711 | /* | |
1712 | * Decode str into name, index, options. | |
1713 | */ | |
1714 | if (str[0] >= '0' && str[0] <= '9') { | |
1715 | strcpy(buf, "ttyS"); | |
1716 | strncpy(buf + 4, str, sizeof(buf) - 5); | |
1717 | } else { | |
1718 | strncpy(buf, str, sizeof(buf) - 1); | |
1719 | } | |
1720 | buf[sizeof(buf) - 1] = 0; | |
1721 | if ((options = strchr(str, ',')) != NULL) | |
1722 | *(options++) = 0; | |
1723 | #ifdef __sparc__ | |
1724 | if (!strcmp(str, "ttya")) | |
1725 | strcpy(buf, "ttyS0"); | |
1726 | if (!strcmp(str, "ttyb")) | |
1727 | strcpy(buf, "ttyS1"); | |
1728 | #endif | |
1729 | for (s = buf; *s; s++) | |
1730 | if ((*s >= '0' && *s <= '9') || *s == ',') | |
1731 | break; | |
1732 | idx = simple_strtoul(s, NULL, 10); | |
1733 | *s = 0; | |
1734 | ||
1735 | __add_preferred_console(buf, idx, options, brl_options); | |
1736 | console_set_on_cmdline = 1; | |
1737 | return 1; | |
1738 | } | |
1739 | __setup("console=", console_setup); | |
1740 | ||
1741 | /** | |
1742 | * add_preferred_console - add a device to the list of preferred consoles. | |
1743 | * @name: device name | |
1744 | * @idx: device index | |
1745 | * @options: options for this console | |
1746 | * | |
1747 | * The last preferred console added will be used for kernel messages | |
1748 | * and stdin/out/err for init. Normally this is used by console_setup | |
1749 | * above to handle user-supplied console arguments; however it can also | |
1750 | * be used by arch-specific code either to override the user or more | |
1751 | * commonly to provide a default console (ie from PROM variables) when | |
1752 | * the user has not supplied one. | |
1753 | */ | |
1754 | int add_preferred_console(char *name, int idx, char *options) | |
1755 | { | |
1756 | return __add_preferred_console(name, idx, options, NULL); | |
1757 | } | |
1758 | ||
1759 | int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options) | |
1760 | { | |
1761 | struct console_cmdline *c; | |
1762 | int i; | |
1763 | ||
1764 | for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++) | |
1765 | if (strcmp(console_cmdline[i].name, name) == 0 && | |
1766 | console_cmdline[i].index == idx) { | |
1767 | c = &console_cmdline[i]; | |
1768 | strlcpy(c->name, name_new, sizeof(c->name)); | |
1769 | c->name[sizeof(c->name) - 1] = 0; | |
1770 | c->options = options; | |
1771 | c->index = idx_new; | |
1772 | return i; | |
1773 | } | |
1774 | /* not found */ | |
1775 | return -1; | |
1776 | } | |
1777 | ||
1778 | bool console_suspend_enabled = 1; | |
1779 | EXPORT_SYMBOL(console_suspend_enabled); | |
1780 | ||
1781 | static int __init console_suspend_disable(char *str) | |
1782 | { | |
1783 | console_suspend_enabled = 0; | |
1784 | return 1; | |
1785 | } | |
1786 | __setup("no_console_suspend", console_suspend_disable); | |
1787 | module_param_named(console_suspend, console_suspend_enabled, | |
1788 | bool, S_IRUGO | S_IWUSR); | |
1789 | MODULE_PARM_DESC(console_suspend, "suspend console during suspend" | |
1790 | " and hibernate operations"); | |
1791 | ||
1792 | /** | |
1793 | * suspend_console - suspend the console subsystem | |
1794 | * | |
1795 | * This disables printk() while we go into suspend states | |
1796 | */ | |
1797 | void suspend_console(void) | |
1798 | { | |
1799 | if (!console_suspend_enabled) | |
1800 | return; | |
1801 | printk("Suspending console(s) (use no_console_suspend to debug)\n"); | |
1802 | console_lock(); | |
1803 | console_suspended = 1; | |
1804 | up(&console_sem); | |
1805 | } | |
1806 | ||
1807 | void resume_console(void) | |
1808 | { | |
1809 | if (!console_suspend_enabled) | |
1810 | return; | |
1811 | down(&console_sem); | |
1812 | console_suspended = 0; | |
1813 | console_unlock(); | |
1814 | } | |
1815 | ||
1816 | /** | |
1817 | * console_cpu_notify - print deferred console messages after CPU hotplug | |
1818 | * @self: notifier struct | |
1819 | * @action: CPU hotplug event | |
1820 | * @hcpu: unused | |
1821 | * | |
1822 | * If printk() is called from a CPU that is not online yet, the messages | |
1823 | * will be spooled but will not show up on the console. This function is | |
1824 | * called when a new CPU comes online (or fails to come up), and ensures | |
1825 | * that any such output gets printed. | |
1826 | */ | |
1827 | static int __cpuinit console_cpu_notify(struct notifier_block *self, | |
1828 | unsigned long action, void *hcpu) | |
1829 | { | |
1830 | switch (action) { | |
1831 | case CPU_ONLINE: | |
1832 | case CPU_DEAD: | |
1833 | case CPU_DYING: | |
1834 | case CPU_DOWN_FAILED: | |
1835 | case CPU_UP_CANCELED: | |
1836 | console_lock(); | |
1837 | console_unlock(); | |
1838 | } | |
1839 | return NOTIFY_OK; | |
1840 | } | |
1841 | ||
1842 | /** | |
1843 | * console_lock - lock the console system for exclusive use. | |
1844 | * | |
1845 | * Acquires a lock which guarantees that the caller has | |
1846 | * exclusive access to the console system and the console_drivers list. | |
1847 | * | |
1848 | * Can sleep, returns nothing. | |
1849 | */ | |
1850 | void console_lock(void) | |
1851 | { | |
1852 | BUG_ON(in_interrupt()); | |
1853 | down(&console_sem); | |
1854 | if (console_suspended) | |
1855 | return; | |
1856 | console_locked = 1; | |
1857 | console_may_schedule = 1; | |
1858 | } | |
1859 | EXPORT_SYMBOL(console_lock); | |
1860 | ||
1861 | /** | |
1862 | * console_trylock - try to lock the console system for exclusive use. | |
1863 | * | |
1864 | * Tried to acquire a lock which guarantees that the caller has | |
1865 | * exclusive access to the console system and the console_drivers list. | |
1866 | * | |
1867 | * returns 1 on success, and 0 on failure to acquire the lock. | |
1868 | */ | |
1869 | int console_trylock(void) | |
1870 | { | |
1871 | if (down_trylock(&console_sem)) | |
1872 | return 0; | |
1873 | if (console_suspended) { | |
1874 | up(&console_sem); | |
1875 | return 0; | |
1876 | } | |
1877 | console_locked = 1; | |
1878 | console_may_schedule = 0; | |
1879 | return 1; | |
1880 | } | |
1881 | EXPORT_SYMBOL(console_trylock); | |
1882 | ||
1883 | int is_console_locked(void) | |
1884 | { | |
1885 | return console_locked; | |
1886 | } | |
1887 | ||
1888 | /* | |
1889 | * Delayed printk version, for scheduler-internal messages: | |
1890 | */ | |
1891 | #define PRINTK_BUF_SIZE 512 | |
1892 | ||
1893 | #define PRINTK_PENDING_WAKEUP 0x01 | |
1894 | #define PRINTK_PENDING_SCHED 0x02 | |
1895 | ||
1896 | static DEFINE_PER_CPU(int, printk_pending); | |
1897 | static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf); | |
1898 | ||
1899 | void printk_tick(void) | |
1900 | { | |
1901 | if (__this_cpu_read(printk_pending)) { | |
1902 | int pending = __this_cpu_xchg(printk_pending, 0); | |
1903 | if (pending & PRINTK_PENDING_SCHED) { | |
1904 | char *buf = __get_cpu_var(printk_sched_buf); | |
1905 | printk(KERN_WARNING "[sched_delayed] %s", buf); | |
1906 | } | |
1907 | if (pending & PRINTK_PENDING_WAKEUP) | |
1908 | wake_up_interruptible(&log_wait); | |
1909 | } | |
1910 | } | |
1911 | ||
1912 | int printk_needs_cpu(int cpu) | |
1913 | { | |
1914 | if (cpu_is_offline(cpu)) | |
1915 | printk_tick(); | |
1916 | return __this_cpu_read(printk_pending); | |
1917 | } | |
1918 | ||
1919 | void wake_up_klogd(void) | |
1920 | { | |
1921 | if (waitqueue_active(&log_wait)) | |
1922 | this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP); | |
1923 | } | |
1924 | ||
1925 | /* the next printk record to write to the console */ | |
1926 | static u64 console_seq; | |
1927 | static u32 console_idx; | |
1928 | static enum log_flags console_prev; | |
1929 | ||
1930 | /** | |
1931 | * console_unlock - unlock the console system | |
1932 | * | |
1933 | * Releases the console_lock which the caller holds on the console system | |
1934 | * and the console driver list. | |
1935 | * | |
1936 | * While the console_lock was held, console output may have been buffered | |
1937 | * by printk(). If this is the case, console_unlock(); emits | |
1938 | * the output prior to releasing the lock. | |
1939 | * | |
1940 | * If there is output waiting, we wake /dev/kmsg and syslog() users. | |
1941 | * | |
1942 | * console_unlock(); may be called from any context. | |
1943 | */ | |
1944 | void console_unlock(void) | |
1945 | { | |
1946 | static char text[LOG_LINE_MAX + PREFIX_MAX]; | |
1947 | static u64 seen_seq; | |
1948 | unsigned long flags; | |
1949 | bool wake_klogd = false; | |
1950 | bool retry; | |
1951 | ||
1952 | if (console_suspended) { | |
1953 | up(&console_sem); | |
1954 | return; | |
1955 | } | |
1956 | ||
1957 | console_may_schedule = 0; | |
1958 | ||
1959 | /* flush buffered message fragment immediately to console */ | |
1960 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
1961 | if (cont.len && (cont.cons < cont.len || cont.flushed)) { | |
1962 | size_t len; | |
1963 | ||
1964 | len = cont_print_text(text, sizeof(text)); | |
1965 | raw_spin_unlock(&logbuf_lock); | |
1966 | stop_critical_timings(); | |
1967 | call_console_drivers(cont.level, text, len); | |
1968 | start_critical_timings(); | |
1969 | local_irq_restore(flags); | |
1970 | } else | |
1971 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
1972 | ||
1973 | again: | |
1974 | for (;;) { | |
1975 | struct log *msg; | |
1976 | size_t len; | |
1977 | int level; | |
1978 | ||
1979 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
1980 | if (seen_seq != log_next_seq) { | |
1981 | wake_klogd = true; | |
1982 | seen_seq = log_next_seq; | |
1983 | } | |
1984 | ||
1985 | if (console_seq < log_first_seq) { | |
1986 | /* messages are gone, move to first one */ | |
1987 | console_seq = log_first_seq; | |
1988 | console_idx = log_first_idx; | |
1989 | console_prev = 0; | |
1990 | } | |
1991 | skip: | |
1992 | if (console_seq == log_next_seq) | |
1993 | break; | |
1994 | ||
1995 | msg = log_from_idx(console_idx); | |
1996 | if (msg->flags & LOG_NOCONS) { | |
1997 | /* | |
1998 | * Skip record we have buffered and already printed | |
1999 | * directly to the console when we received it. | |
2000 | */ | |
2001 | console_idx = log_next(console_idx); | |
2002 | console_seq++; | |
2003 | /* | |
2004 | * We will get here again when we register a new | |
2005 | * CON_PRINTBUFFER console. Clear the flag so we | |
2006 | * will properly dump everything later. | |
2007 | */ | |
2008 | msg->flags &= ~LOG_NOCONS; | |
2009 | goto skip; | |
2010 | } | |
2011 | ||
2012 | level = msg->level; | |
2013 | len = msg_print_text(msg, console_prev, false, | |
2014 | text, sizeof(text)); | |
2015 | console_idx = log_next(console_idx); | |
2016 | console_seq++; | |
2017 | console_prev = msg->flags; | |
2018 | raw_spin_unlock(&logbuf_lock); | |
2019 | ||
2020 | stop_critical_timings(); /* don't trace print latency */ | |
2021 | call_console_drivers(level, text, len); | |
2022 | start_critical_timings(); | |
2023 | local_irq_restore(flags); | |
2024 | } | |
2025 | console_locked = 0; | |
2026 | ||
2027 | /* Release the exclusive_console once it is used */ | |
2028 | if (unlikely(exclusive_console)) | |
2029 | exclusive_console = NULL; | |
2030 | ||
2031 | raw_spin_unlock(&logbuf_lock); | |
2032 | ||
2033 | up(&console_sem); | |
2034 | ||
2035 | /* | |
2036 | * Someone could have filled up the buffer again, so re-check if there's | |
2037 | * something to flush. In case we cannot trylock the console_sem again, | |
2038 | * there's a new owner and the console_unlock() from them will do the | |
2039 | * flush, no worries. | |
2040 | */ | |
2041 | raw_spin_lock(&logbuf_lock); | |
2042 | retry = console_seq != log_next_seq; | |
2043 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2044 | ||
2045 | if (retry && console_trylock()) | |
2046 | goto again; | |
2047 | ||
2048 | if (wake_klogd) | |
2049 | wake_up_klogd(); | |
2050 | } | |
2051 | EXPORT_SYMBOL(console_unlock); | |
2052 | ||
2053 | /** | |
2054 | * console_conditional_schedule - yield the CPU if required | |
2055 | * | |
2056 | * If the console code is currently allowed to sleep, and | |
2057 | * if this CPU should yield the CPU to another task, do | |
2058 | * so here. | |
2059 | * | |
2060 | * Must be called within console_lock();. | |
2061 | */ | |
2062 | void __sched console_conditional_schedule(void) | |
2063 | { | |
2064 | if (console_may_schedule) | |
2065 | cond_resched(); | |
2066 | } | |
2067 | EXPORT_SYMBOL(console_conditional_schedule); | |
2068 | ||
2069 | void console_unblank(void) | |
2070 | { | |
2071 | struct console *c; | |
2072 | ||
2073 | /* | |
2074 | * console_unblank can no longer be called in interrupt context unless | |
2075 | * oops_in_progress is set to 1.. | |
2076 | */ | |
2077 | if (oops_in_progress) { | |
2078 | if (down_trylock(&console_sem) != 0) | |
2079 | return; | |
2080 | } else | |
2081 | console_lock(); | |
2082 | ||
2083 | console_locked = 1; | |
2084 | console_may_schedule = 0; | |
2085 | for_each_console(c) | |
2086 | if ((c->flags & CON_ENABLED) && c->unblank) | |
2087 | c->unblank(); | |
2088 | console_unlock(); | |
2089 | } | |
2090 | ||
2091 | /* | |
2092 | * Return the console tty driver structure and its associated index | |
2093 | */ | |
2094 | struct tty_driver *console_device(int *index) | |
2095 | { | |
2096 | struct console *c; | |
2097 | struct tty_driver *driver = NULL; | |
2098 | ||
2099 | console_lock(); | |
2100 | for_each_console(c) { | |
2101 | if (!c->device) | |
2102 | continue; | |
2103 | driver = c->device(c, index); | |
2104 | if (driver) | |
2105 | break; | |
2106 | } | |
2107 | console_unlock(); | |
2108 | return driver; | |
2109 | } | |
2110 | ||
2111 | /* | |
2112 | * Prevent further output on the passed console device so that (for example) | |
2113 | * serial drivers can disable console output before suspending a port, and can | |
2114 | * re-enable output afterwards. | |
2115 | */ | |
2116 | void console_stop(struct console *console) | |
2117 | { | |
2118 | console_lock(); | |
2119 | console->flags &= ~CON_ENABLED; | |
2120 | console_unlock(); | |
2121 | } | |
2122 | EXPORT_SYMBOL(console_stop); | |
2123 | ||
2124 | void console_start(struct console *console) | |
2125 | { | |
2126 | console_lock(); | |
2127 | console->flags |= CON_ENABLED; | |
2128 | console_unlock(); | |
2129 | } | |
2130 | EXPORT_SYMBOL(console_start); | |
2131 | ||
2132 | static int __read_mostly keep_bootcon; | |
2133 | ||
2134 | static int __init keep_bootcon_setup(char *str) | |
2135 | { | |
2136 | keep_bootcon = 1; | |
2137 | printk(KERN_INFO "debug: skip boot console de-registration.\n"); | |
2138 | ||
2139 | return 0; | |
2140 | } | |
2141 | ||
2142 | early_param("keep_bootcon", keep_bootcon_setup); | |
2143 | ||
2144 | /* | |
2145 | * The console driver calls this routine during kernel initialization | |
2146 | * to register the console printing procedure with printk() and to | |
2147 | * print any messages that were printed by the kernel before the | |
2148 | * console driver was initialized. | |
2149 | * | |
2150 | * This can happen pretty early during the boot process (because of | |
2151 | * early_printk) - sometimes before setup_arch() completes - be careful | |
2152 | * of what kernel features are used - they may not be initialised yet. | |
2153 | * | |
2154 | * There are two types of consoles - bootconsoles (early_printk) and | |
2155 | * "real" consoles (everything which is not a bootconsole) which are | |
2156 | * handled differently. | |
2157 | * - Any number of bootconsoles can be registered at any time. | |
2158 | * - As soon as a "real" console is registered, all bootconsoles | |
2159 | * will be unregistered automatically. | |
2160 | * - Once a "real" console is registered, any attempt to register a | |
2161 | * bootconsoles will be rejected | |
2162 | */ | |
2163 | void register_console(struct console *newcon) | |
2164 | { | |
2165 | int i; | |
2166 | unsigned long flags; | |
2167 | struct console *bcon = NULL; | |
2168 | ||
2169 | /* | |
2170 | * before we register a new CON_BOOT console, make sure we don't | |
2171 | * already have a valid console | |
2172 | */ | |
2173 | if (console_drivers && newcon->flags & CON_BOOT) { | |
2174 | /* find the last or real console */ | |
2175 | for_each_console(bcon) { | |
2176 | if (!(bcon->flags & CON_BOOT)) { | |
2177 | printk(KERN_INFO "Too late to register bootconsole %s%d\n", | |
2178 | newcon->name, newcon->index); | |
2179 | return; | |
2180 | } | |
2181 | } | |
2182 | } | |
2183 | ||
2184 | if (console_drivers && console_drivers->flags & CON_BOOT) | |
2185 | bcon = console_drivers; | |
2186 | ||
2187 | if (preferred_console < 0 || bcon || !console_drivers) | |
2188 | preferred_console = selected_console; | |
2189 | ||
2190 | if (newcon->early_setup) | |
2191 | newcon->early_setup(); | |
2192 | ||
2193 | /* | |
2194 | * See if we want to use this console driver. If we | |
2195 | * didn't select a console we take the first one | |
2196 | * that registers here. | |
2197 | */ | |
2198 | if (preferred_console < 0) { | |
2199 | if (newcon->index < 0) | |
2200 | newcon->index = 0; | |
2201 | if (newcon->setup == NULL || | |
2202 | newcon->setup(newcon, NULL) == 0) { | |
2203 | newcon->flags |= CON_ENABLED; | |
2204 | if (newcon->device) { | |
2205 | newcon->flags |= CON_CONSDEV; | |
2206 | preferred_console = 0; | |
2207 | } | |
2208 | } | |
2209 | } | |
2210 | ||
2211 | /* | |
2212 | * See if this console matches one we selected on | |
2213 | * the command line. | |
2214 | */ | |
2215 | for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; | |
2216 | i++) { | |
2217 | if (strcmp(console_cmdline[i].name, newcon->name) != 0) | |
2218 | continue; | |
2219 | if (newcon->index >= 0 && | |
2220 | newcon->index != console_cmdline[i].index) | |
2221 | continue; | |
2222 | if (newcon->index < 0) | |
2223 | newcon->index = console_cmdline[i].index; | |
2224 | #ifdef CONFIG_A11Y_BRAILLE_CONSOLE | |
2225 | if (console_cmdline[i].brl_options) { | |
2226 | newcon->flags |= CON_BRL; | |
2227 | braille_register_console(newcon, | |
2228 | console_cmdline[i].index, | |
2229 | console_cmdline[i].options, | |
2230 | console_cmdline[i].brl_options); | |
2231 | return; | |
2232 | } | |
2233 | #endif | |
2234 | if (newcon->setup && | |
2235 | newcon->setup(newcon, console_cmdline[i].options) != 0) | |
2236 | break; | |
2237 | newcon->flags |= CON_ENABLED; | |
2238 | newcon->index = console_cmdline[i].index; | |
2239 | if (i == selected_console) { | |
2240 | newcon->flags |= CON_CONSDEV; | |
2241 | preferred_console = selected_console; | |
2242 | } | |
2243 | break; | |
2244 | } | |
2245 | ||
2246 | if (!(newcon->flags & CON_ENABLED)) | |
2247 | return; | |
2248 | ||
2249 | /* | |
2250 | * If we have a bootconsole, and are switching to a real console, | |
2251 | * don't print everything out again, since when the boot console, and | |
2252 | * the real console are the same physical device, it's annoying to | |
2253 | * see the beginning boot messages twice | |
2254 | */ | |
2255 | if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV)) | |
2256 | newcon->flags &= ~CON_PRINTBUFFER; | |
2257 | ||
2258 | /* | |
2259 | * Put this console in the list - keep the | |
2260 | * preferred driver at the head of the list. | |
2261 | */ | |
2262 | console_lock(); | |
2263 | if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) { | |
2264 | newcon->next = console_drivers; | |
2265 | console_drivers = newcon; | |
2266 | if (newcon->next) | |
2267 | newcon->next->flags &= ~CON_CONSDEV; | |
2268 | } else { | |
2269 | newcon->next = console_drivers->next; | |
2270 | console_drivers->next = newcon; | |
2271 | } | |
2272 | if (newcon->flags & CON_PRINTBUFFER) { | |
2273 | /* | |
2274 | * console_unlock(); will print out the buffered messages | |
2275 | * for us. | |
2276 | */ | |
2277 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
2278 | console_seq = syslog_seq; | |
2279 | console_idx = syslog_idx; | |
2280 | console_prev = syslog_prev; | |
2281 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2282 | /* | |
2283 | * We're about to replay the log buffer. Only do this to the | |
2284 | * just-registered console to avoid excessive message spam to | |
2285 | * the already-registered consoles. | |
2286 | */ | |
2287 | exclusive_console = newcon; | |
2288 | } | |
2289 | console_unlock(); | |
2290 | console_sysfs_notify(); | |
2291 | ||
2292 | /* | |
2293 | * By unregistering the bootconsoles after we enable the real console | |
2294 | * we get the "console xxx enabled" message on all the consoles - | |
2295 | * boot consoles, real consoles, etc - this is to ensure that end | |
2296 | * users know there might be something in the kernel's log buffer that | |
2297 | * went to the bootconsole (that they do not see on the real console) | |
2298 | */ | |
2299 | if (bcon && | |
2300 | ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) && | |
2301 | !keep_bootcon) { | |
2302 | /* we need to iterate through twice, to make sure we print | |
2303 | * everything out, before we unregister the console(s) | |
2304 | */ | |
2305 | printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n", | |
2306 | newcon->name, newcon->index); | |
2307 | for_each_console(bcon) | |
2308 | if (bcon->flags & CON_BOOT) | |
2309 | unregister_console(bcon); | |
2310 | } else { | |
2311 | printk(KERN_INFO "%sconsole [%s%d] enabled\n", | |
2312 | (newcon->flags & CON_BOOT) ? "boot" : "" , | |
2313 | newcon->name, newcon->index); | |
2314 | } | |
2315 | } | |
2316 | EXPORT_SYMBOL(register_console); | |
2317 | ||
2318 | int unregister_console(struct console *console) | |
2319 | { | |
2320 | struct console *a, *b; | |
2321 | int res = 1; | |
2322 | ||
2323 | #ifdef CONFIG_A11Y_BRAILLE_CONSOLE | |
2324 | if (console->flags & CON_BRL) | |
2325 | return braille_unregister_console(console); | |
2326 | #endif | |
2327 | ||
2328 | console_lock(); | |
2329 | if (console_drivers == console) { | |
2330 | console_drivers=console->next; | |
2331 | res = 0; | |
2332 | } else if (console_drivers) { | |
2333 | for (a=console_drivers->next, b=console_drivers ; | |
2334 | a; b=a, a=b->next) { | |
2335 | if (a == console) { | |
2336 | b->next = a->next; | |
2337 | res = 0; | |
2338 | break; | |
2339 | } | |
2340 | } | |
2341 | } | |
2342 | ||
2343 | /* | |
2344 | * If this isn't the last console and it has CON_CONSDEV set, we | |
2345 | * need to set it on the next preferred console. | |
2346 | */ | |
2347 | if (console_drivers != NULL && console->flags & CON_CONSDEV) | |
2348 | console_drivers->flags |= CON_CONSDEV; | |
2349 | ||
2350 | console_unlock(); | |
2351 | console_sysfs_notify(); | |
2352 | return res; | |
2353 | } | |
2354 | EXPORT_SYMBOL(unregister_console); | |
2355 | ||
2356 | static int __init printk_late_init(void) | |
2357 | { | |
2358 | struct console *con; | |
2359 | ||
2360 | for_each_console(con) { | |
2361 | if (!keep_bootcon && con->flags & CON_BOOT) { | |
2362 | printk(KERN_INFO "turn off boot console %s%d\n", | |
2363 | con->name, con->index); | |
2364 | unregister_console(con); | |
2365 | } | |
2366 | } | |
2367 | hotcpu_notifier(console_cpu_notify, 0); | |
2368 | return 0; | |
2369 | } | |
2370 | late_initcall(printk_late_init); | |
2371 | ||
2372 | #if defined CONFIG_PRINTK | |
2373 | ||
2374 | int printk_sched(const char *fmt, ...) | |
2375 | { | |
2376 | unsigned long flags; | |
2377 | va_list args; | |
2378 | char *buf; | |
2379 | int r; | |
2380 | ||
2381 | local_irq_save(flags); | |
2382 | buf = __get_cpu_var(printk_sched_buf); | |
2383 | ||
2384 | va_start(args, fmt); | |
2385 | r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args); | |
2386 | va_end(args); | |
2387 | ||
2388 | __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED); | |
2389 | local_irq_restore(flags); | |
2390 | ||
2391 | return r; | |
2392 | } | |
2393 | ||
2394 | /* | |
2395 | * printk rate limiting, lifted from the networking subsystem. | |
2396 | * | |
2397 | * This enforces a rate limit: not more than 10 kernel messages | |
2398 | * every 5s to make a denial-of-service attack impossible. | |
2399 | */ | |
2400 | DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10); | |
2401 | ||
2402 | int __printk_ratelimit(const char *func) | |
2403 | { | |
2404 | return ___ratelimit(&printk_ratelimit_state, func); | |
2405 | } | |
2406 | EXPORT_SYMBOL(__printk_ratelimit); | |
2407 | ||
2408 | /** | |
2409 | * printk_timed_ratelimit - caller-controlled printk ratelimiting | |
2410 | * @caller_jiffies: pointer to caller's state | |
2411 | * @interval_msecs: minimum interval between prints | |
2412 | * | |
2413 | * printk_timed_ratelimit() returns true if more than @interval_msecs | |
2414 | * milliseconds have elapsed since the last time printk_timed_ratelimit() | |
2415 | * returned true. | |
2416 | */ | |
2417 | bool printk_timed_ratelimit(unsigned long *caller_jiffies, | |
2418 | unsigned int interval_msecs) | |
2419 | { | |
2420 | if (*caller_jiffies == 0 | |
2421 | || !time_in_range(jiffies, *caller_jiffies, | |
2422 | *caller_jiffies | |
2423 | + msecs_to_jiffies(interval_msecs))) { | |
2424 | *caller_jiffies = jiffies; | |
2425 | return true; | |
2426 | } | |
2427 | return false; | |
2428 | } | |
2429 | EXPORT_SYMBOL(printk_timed_ratelimit); | |
2430 | ||
2431 | static DEFINE_SPINLOCK(dump_list_lock); | |
2432 | static LIST_HEAD(dump_list); | |
2433 | ||
2434 | /** | |
2435 | * kmsg_dump_register - register a kernel log dumper. | |
2436 | * @dumper: pointer to the kmsg_dumper structure | |
2437 | * | |
2438 | * Adds a kernel log dumper to the system. The dump callback in the | |
2439 | * structure will be called when the kernel oopses or panics and must be | |
2440 | * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise. | |
2441 | */ | |
2442 | int kmsg_dump_register(struct kmsg_dumper *dumper) | |
2443 | { | |
2444 | unsigned long flags; | |
2445 | int err = -EBUSY; | |
2446 | ||
2447 | /* The dump callback needs to be set */ | |
2448 | if (!dumper->dump) | |
2449 | return -EINVAL; | |
2450 | ||
2451 | spin_lock_irqsave(&dump_list_lock, flags); | |
2452 | /* Don't allow registering multiple times */ | |
2453 | if (!dumper->registered) { | |
2454 | dumper->registered = 1; | |
2455 | list_add_tail_rcu(&dumper->list, &dump_list); | |
2456 | err = 0; | |
2457 | } | |
2458 | spin_unlock_irqrestore(&dump_list_lock, flags); | |
2459 | ||
2460 | return err; | |
2461 | } | |
2462 | EXPORT_SYMBOL_GPL(kmsg_dump_register); | |
2463 | ||
2464 | /** | |
2465 | * kmsg_dump_unregister - unregister a kmsg dumper. | |
2466 | * @dumper: pointer to the kmsg_dumper structure | |
2467 | * | |
2468 | * Removes a dump device from the system. Returns zero on success and | |
2469 | * %-EINVAL otherwise. | |
2470 | */ | |
2471 | int kmsg_dump_unregister(struct kmsg_dumper *dumper) | |
2472 | { | |
2473 | unsigned long flags; | |
2474 | int err = -EINVAL; | |
2475 | ||
2476 | spin_lock_irqsave(&dump_list_lock, flags); | |
2477 | if (dumper->registered) { | |
2478 | dumper->registered = 0; | |
2479 | list_del_rcu(&dumper->list); | |
2480 | err = 0; | |
2481 | } | |
2482 | spin_unlock_irqrestore(&dump_list_lock, flags); | |
2483 | synchronize_rcu(); | |
2484 | ||
2485 | return err; | |
2486 | } | |
2487 | EXPORT_SYMBOL_GPL(kmsg_dump_unregister); | |
2488 | ||
2489 | static bool always_kmsg_dump; | |
2490 | module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR); | |
2491 | ||
2492 | /** | |
2493 | * kmsg_dump - dump kernel log to kernel message dumpers. | |
2494 | * @reason: the reason (oops, panic etc) for dumping | |
2495 | * | |
2496 | * Call each of the registered dumper's dump() callback, which can | |
2497 | * retrieve the kmsg records with kmsg_dump_get_line() or | |
2498 | * kmsg_dump_get_buffer(). | |
2499 | */ | |
2500 | void kmsg_dump(enum kmsg_dump_reason reason) | |
2501 | { | |
2502 | struct kmsg_dumper *dumper; | |
2503 | unsigned long flags; | |
2504 | ||
2505 | if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump) | |
2506 | return; | |
2507 | ||
2508 | rcu_read_lock(); | |
2509 | list_for_each_entry_rcu(dumper, &dump_list, list) { | |
2510 | if (dumper->max_reason && reason > dumper->max_reason) | |
2511 | continue; | |
2512 | ||
2513 | /* initialize iterator with data about the stored records */ | |
2514 | dumper->active = true; | |
2515 | ||
2516 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
2517 | dumper->cur_seq = clear_seq; | |
2518 | dumper->cur_idx = clear_idx; | |
2519 | dumper->next_seq = log_next_seq; | |
2520 | dumper->next_idx = log_next_idx; | |
2521 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2522 | ||
2523 | /* invoke dumper which will iterate over records */ | |
2524 | dumper->dump(dumper, reason); | |
2525 | ||
2526 | /* reset iterator */ | |
2527 | dumper->active = false; | |
2528 | } | |
2529 | rcu_read_unlock(); | |
2530 | } | |
2531 | ||
2532 | /** | |
2533 | * kmsg_dump_get_line - retrieve one kmsg log line | |
2534 | * @dumper: registered kmsg dumper | |
2535 | * @syslog: include the "<4>" prefixes | |
2536 | * @line: buffer to copy the line to | |
2537 | * @size: maximum size of the buffer | |
2538 | * @len: length of line placed into buffer | |
2539 | * | |
2540 | * Start at the beginning of the kmsg buffer, with the oldest kmsg | |
2541 | * record, and copy one record into the provided buffer. | |
2542 | * | |
2543 | * Consecutive calls will return the next available record moving | |
2544 | * towards the end of the buffer with the youngest messages. | |
2545 | * | |
2546 | * A return value of FALSE indicates that there are no more records to | |
2547 | * read. | |
2548 | */ | |
2549 | bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog, | |
2550 | char *line, size_t size, size_t *len) | |
2551 | { | |
2552 | unsigned long flags; | |
2553 | struct log *msg; | |
2554 | size_t l = 0; | |
2555 | bool ret = false; | |
2556 | ||
2557 | if (!dumper->active) | |
2558 | goto out; | |
2559 | ||
2560 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
2561 | if (dumper->cur_seq < log_first_seq) { | |
2562 | /* messages are gone, move to first available one */ | |
2563 | dumper->cur_seq = log_first_seq; | |
2564 | dumper->cur_idx = log_first_idx; | |
2565 | } | |
2566 | ||
2567 | /* last entry */ | |
2568 | if (dumper->cur_seq >= log_next_seq) { | |
2569 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2570 | goto out; | |
2571 | } | |
2572 | ||
2573 | msg = log_from_idx(dumper->cur_idx); | |
2574 | l = msg_print_text(msg, 0, syslog, line, size); | |
2575 | ||
2576 | dumper->cur_idx = log_next(dumper->cur_idx); | |
2577 | dumper->cur_seq++; | |
2578 | ret = true; | |
2579 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2580 | out: | |
2581 | if (len) | |
2582 | *len = l; | |
2583 | return ret; | |
2584 | } | |
2585 | EXPORT_SYMBOL_GPL(kmsg_dump_get_line); | |
2586 | ||
2587 | /** | |
2588 | * kmsg_dump_get_buffer - copy kmsg log lines | |
2589 | * @dumper: registered kmsg dumper | |
2590 | * @syslog: include the "<4>" prefixes | |
2591 | * @buf: buffer to copy the line to | |
2592 | * @size: maximum size of the buffer | |
2593 | * @len: length of line placed into buffer | |
2594 | * | |
2595 | * Start at the end of the kmsg buffer and fill the provided buffer | |
2596 | * with as many of the the *youngest* kmsg records that fit into it. | |
2597 | * If the buffer is large enough, all available kmsg records will be | |
2598 | * copied with a single call. | |
2599 | * | |
2600 | * Consecutive calls will fill the buffer with the next block of | |
2601 | * available older records, not including the earlier retrieved ones. | |
2602 | * | |
2603 | * A return value of FALSE indicates that there are no more records to | |
2604 | * read. | |
2605 | */ | |
2606 | bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog, | |
2607 | char *buf, size_t size, size_t *len) | |
2608 | { | |
2609 | unsigned long flags; | |
2610 | u64 seq; | |
2611 | u32 idx; | |
2612 | u64 next_seq; | |
2613 | u32 next_idx; | |
2614 | enum log_flags prev; | |
2615 | size_t l = 0; | |
2616 | bool ret = false; | |
2617 | ||
2618 | if (!dumper->active) | |
2619 | goto out; | |
2620 | ||
2621 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
2622 | if (dumper->cur_seq < log_first_seq) { | |
2623 | /* messages are gone, move to first available one */ | |
2624 | dumper->cur_seq = log_first_seq; | |
2625 | dumper->cur_idx = log_first_idx; | |
2626 | } | |
2627 | ||
2628 | /* last entry */ | |
2629 | if (dumper->cur_seq >= dumper->next_seq) { | |
2630 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2631 | goto out; | |
2632 | } | |
2633 | ||
2634 | /* calculate length of entire buffer */ | |
2635 | seq = dumper->cur_seq; | |
2636 | idx = dumper->cur_idx; | |
2637 | prev = 0; | |
2638 | while (seq < dumper->next_seq) { | |
2639 | struct log *msg = log_from_idx(idx); | |
2640 | ||
2641 | l += msg_print_text(msg, prev, true, NULL, 0); | |
2642 | idx = log_next(idx); | |
2643 | seq++; | |
2644 | prev = msg->flags; | |
2645 | } | |
2646 | ||
2647 | /* move first record forward until length fits into the buffer */ | |
2648 | seq = dumper->cur_seq; | |
2649 | idx = dumper->cur_idx; | |
2650 | prev = 0; | |
2651 | while (l > size && seq < dumper->next_seq) { | |
2652 | struct log *msg = log_from_idx(idx); | |
2653 | ||
2654 | l -= msg_print_text(msg, prev, true, NULL, 0); | |
2655 | idx = log_next(idx); | |
2656 | seq++; | |
2657 | prev = msg->flags; | |
2658 | } | |
2659 | ||
2660 | /* last message in next interation */ | |
2661 | next_seq = seq; | |
2662 | next_idx = idx; | |
2663 | ||
2664 | l = 0; | |
2665 | prev = 0; | |
2666 | while (seq < dumper->next_seq) { | |
2667 | struct log *msg = log_from_idx(idx); | |
2668 | ||
2669 | l += msg_print_text(msg, prev, syslog, buf + l, size - l); | |
2670 | idx = log_next(idx); | |
2671 | seq++; | |
2672 | prev = msg->flags; | |
2673 | } | |
2674 | ||
2675 | dumper->next_seq = next_seq; | |
2676 | dumper->next_idx = next_idx; | |
2677 | ret = true; | |
2678 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2679 | out: | |
2680 | if (len) | |
2681 | *len = l; | |
2682 | return ret; | |
2683 | } | |
2684 | EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer); | |
2685 | ||
2686 | /** | |
2687 | * kmsg_dump_rewind - reset the interator | |
2688 | * @dumper: registered kmsg dumper | |
2689 | * | |
2690 | * Reset the dumper's iterator so that kmsg_dump_get_line() and | |
2691 | * kmsg_dump_get_buffer() can be called again and used multiple | |
2692 | * times within the same dumper.dump() callback. | |
2693 | */ | |
2694 | void kmsg_dump_rewind(struct kmsg_dumper *dumper) | |
2695 | { | |
2696 | unsigned long flags; | |
2697 | ||
2698 | raw_spin_lock_irqsave(&logbuf_lock, flags); | |
2699 | dumper->cur_seq = clear_seq; | |
2700 | dumper->cur_idx = clear_idx; | |
2701 | dumper->next_seq = log_next_seq; | |
2702 | dumper->next_idx = log_next_idx; | |
2703 | raw_spin_unlock_irqrestore(&logbuf_lock, flags); | |
2704 | } | |
2705 | EXPORT_SYMBOL_GPL(kmsg_dump_rewind); | |
2706 | #endif |