xref: /openbmc/linux/kernel/printk/printk.c (revision 174cd4b1)
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/delay.h>
30 #include <linux/smp.h>
31 #include <linux/security.h>
32 #include <linux/bootmem.h>
33 #include <linux/memblock.h>
34 #include <linux/syscalls.h>
35 #include <linux/kexec.h>
36 #include <linux/kdb.h>
37 #include <linux/ratelimit.h>
38 #include <linux/kmsg_dump.h>
39 #include <linux/syslog.h>
40 #include <linux/cpu.h>
41 #include <linux/notifier.h>
42 #include <linux/rculist.h>
43 #include <linux/poll.h>
44 #include <linux/irq_work.h>
45 #include <linux/utsname.h>
46 #include <linux/ctype.h>
47 #include <linux/uio.h>
48 #include <linux/sched/clock.h>
49 
50 #include <linux/uaccess.h>
51 #include <asm/sections.h>
52 
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/printk.h>
55 
56 #include "console_cmdline.h"
57 #include "braille.h"
58 #include "internal.h"
59 
60 int console_printk[4] = {
61 	CONSOLE_LOGLEVEL_DEFAULT,	/* console_loglevel */
62 	MESSAGE_LOGLEVEL_DEFAULT,	/* default_message_loglevel */
63 	CONSOLE_LOGLEVEL_MIN,		/* minimum_console_loglevel */
64 	CONSOLE_LOGLEVEL_DEFAULT,	/* default_console_loglevel */
65 };
66 
67 /*
68  * Low level drivers may need that to know if they can schedule in
69  * their unblank() callback or not. So let's export it.
70  */
71 int oops_in_progress;
72 EXPORT_SYMBOL(oops_in_progress);
73 
74 /*
75  * console_sem protects the console_drivers list, and also
76  * provides serialisation for access to the entire console
77  * driver system.
78  */
79 static DEFINE_SEMAPHORE(console_sem);
80 struct console *console_drivers;
81 EXPORT_SYMBOL_GPL(console_drivers);
82 
83 #ifdef CONFIG_LOCKDEP
84 static struct lockdep_map console_lock_dep_map = {
85 	.name = "console_lock"
86 };
87 #endif
88 
89 enum devkmsg_log_bits {
90 	__DEVKMSG_LOG_BIT_ON = 0,
91 	__DEVKMSG_LOG_BIT_OFF,
92 	__DEVKMSG_LOG_BIT_LOCK,
93 };
94 
95 enum devkmsg_log_masks {
96 	DEVKMSG_LOG_MASK_ON             = BIT(__DEVKMSG_LOG_BIT_ON),
97 	DEVKMSG_LOG_MASK_OFF            = BIT(__DEVKMSG_LOG_BIT_OFF),
98 	DEVKMSG_LOG_MASK_LOCK           = BIT(__DEVKMSG_LOG_BIT_LOCK),
99 };
100 
101 /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
102 #define DEVKMSG_LOG_MASK_DEFAULT	0
103 
104 static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
105 
106 static int __control_devkmsg(char *str)
107 {
108 	if (!str)
109 		return -EINVAL;
110 
111 	if (!strncmp(str, "on", 2)) {
112 		devkmsg_log = DEVKMSG_LOG_MASK_ON;
113 		return 2;
114 	} else if (!strncmp(str, "off", 3)) {
115 		devkmsg_log = DEVKMSG_LOG_MASK_OFF;
116 		return 3;
117 	} else if (!strncmp(str, "ratelimit", 9)) {
118 		devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
119 		return 9;
120 	}
121 	return -EINVAL;
122 }
123 
124 static int __init control_devkmsg(char *str)
125 {
126 	if (__control_devkmsg(str) < 0)
127 		return 1;
128 
129 	/*
130 	 * Set sysctl string accordingly:
131 	 */
132 	if (devkmsg_log == DEVKMSG_LOG_MASK_ON) {
133 		memset(devkmsg_log_str, 0, DEVKMSG_STR_MAX_SIZE);
134 		strncpy(devkmsg_log_str, "on", 2);
135 	} else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF) {
136 		memset(devkmsg_log_str, 0, DEVKMSG_STR_MAX_SIZE);
137 		strncpy(devkmsg_log_str, "off", 3);
138 	}
139 	/* else "ratelimit" which is set by default. */
140 
141 	/*
142 	 * Sysctl cannot change it anymore. The kernel command line setting of
143 	 * this parameter is to force the setting to be permanent throughout the
144 	 * runtime of the system. This is a precation measure against userspace
145 	 * trying to be a smarta** and attempting to change it up on us.
146 	 */
147 	devkmsg_log |= DEVKMSG_LOG_MASK_LOCK;
148 
149 	return 0;
150 }
151 __setup("printk.devkmsg=", control_devkmsg);
152 
153 char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit";
154 
155 int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
156 			      void __user *buffer, size_t *lenp, loff_t *ppos)
157 {
158 	char old_str[DEVKMSG_STR_MAX_SIZE];
159 	unsigned int old;
160 	int err;
161 
162 	if (write) {
163 		if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK)
164 			return -EINVAL;
165 
166 		old = devkmsg_log;
167 		strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE);
168 	}
169 
170 	err = proc_dostring(table, write, buffer, lenp, ppos);
171 	if (err)
172 		return err;
173 
174 	if (write) {
175 		err = __control_devkmsg(devkmsg_log_str);
176 
177 		/*
178 		 * Do not accept an unknown string OR a known string with
179 		 * trailing crap...
180 		 */
181 		if (err < 0 || (err + 1 != *lenp)) {
182 
183 			/* ... and restore old setting. */
184 			devkmsg_log = old;
185 			strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE);
186 
187 			return -EINVAL;
188 		}
189 	}
190 
191 	return 0;
192 }
193 
194 /*
195  * Number of registered extended console drivers.
196  *
197  * If extended consoles are present, in-kernel cont reassembly is disabled
198  * and each fragment is stored as a separate log entry with proper
199  * continuation flag so that every emitted message has full metadata.  This
200  * doesn't change the result for regular consoles or /proc/kmsg.  For
201  * /dev/kmsg, as long as the reader concatenates messages according to
202  * consecutive continuation flags, the end result should be the same too.
203  */
204 static int nr_ext_console_drivers;
205 
206 /*
207  * Helper macros to handle lockdep when locking/unlocking console_sem. We use
208  * macros instead of functions so that _RET_IP_ contains useful information.
209  */
210 #define down_console_sem() do { \
211 	down(&console_sem);\
212 	mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
213 } while (0)
214 
215 static int __down_trylock_console_sem(unsigned long ip)
216 {
217 	int lock_failed;
218 	unsigned long flags;
219 
220 	/*
221 	 * Here and in __up_console_sem() we need to be in safe mode,
222 	 * because spindump/WARN/etc from under console ->lock will
223 	 * deadlock in printk()->down_trylock_console_sem() otherwise.
224 	 */
225 	printk_safe_enter_irqsave(flags);
226 	lock_failed = down_trylock(&console_sem);
227 	printk_safe_exit_irqrestore(flags);
228 
229 	if (lock_failed)
230 		return 1;
231 	mutex_acquire(&console_lock_dep_map, 0, 1, ip);
232 	return 0;
233 }
234 #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
235 
236 static void __up_console_sem(unsigned long ip)
237 {
238 	unsigned long flags;
239 
240 	mutex_release(&console_lock_dep_map, 1, ip);
241 
242 	printk_safe_enter_irqsave(flags);
243 	up(&console_sem);
244 	printk_safe_exit_irqrestore(flags);
245 }
246 #define up_console_sem() __up_console_sem(_RET_IP_)
247 
248 /*
249  * This is used for debugging the mess that is the VT code by
250  * keeping track if we have the console semaphore held. It's
251  * definitely not the perfect debug tool (we don't know if _WE_
252  * hold it and are racing, but it helps tracking those weird code
253  * paths in the console code where we end up in places I want
254  * locked without the console sempahore held).
255  */
256 static int console_locked, console_suspended;
257 
258 /*
259  * If exclusive_console is non-NULL then only this console is to be printed to.
260  */
261 static struct console *exclusive_console;
262 
263 /*
264  *	Array of consoles built from command line options (console=)
265  */
266 
267 #define MAX_CMDLINECONSOLES 8
268 
269 static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
270 
271 static int selected_console = -1;
272 static int preferred_console = -1;
273 int console_set_on_cmdline;
274 EXPORT_SYMBOL(console_set_on_cmdline);
275 
276 /* Flag: console code may call schedule() */
277 static int console_may_schedule;
278 
279 /*
280  * The printk log buffer consists of a chain of concatenated variable
281  * length records. Every record starts with a record header, containing
282  * the overall length of the record.
283  *
284  * The heads to the first and last entry in the buffer, as well as the
285  * sequence numbers of these entries are maintained when messages are
286  * stored.
287  *
288  * If the heads indicate available messages, the length in the header
289  * tells the start next message. A length == 0 for the next message
290  * indicates a wrap-around to the beginning of the buffer.
291  *
292  * Every record carries the monotonic timestamp in microseconds, as well as
293  * the standard userspace syslog level and syslog facility. The usual
294  * kernel messages use LOG_KERN; userspace-injected messages always carry
295  * a matching syslog facility, by default LOG_USER. The origin of every
296  * message can be reliably determined that way.
297  *
298  * The human readable log message directly follows the message header. The
299  * length of the message text is stored in the header, the stored message
300  * is not terminated.
301  *
302  * Optionally, a message can carry a dictionary of properties (key/value pairs),
303  * to provide userspace with a machine-readable message context.
304  *
305  * Examples for well-defined, commonly used property names are:
306  *   DEVICE=b12:8               device identifier
307  *                                b12:8         block dev_t
308  *                                c127:3        char dev_t
309  *                                n8            netdev ifindex
310  *                                +sound:card0  subsystem:devname
311  *   SUBSYSTEM=pci              driver-core subsystem name
312  *
313  * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
314  * follows directly after a '=' character. Every property is terminated by
315  * a '\0' character. The last property is not terminated.
316  *
317  * Example of a message structure:
318  *   0000  ff 8f 00 00 00 00 00 00      monotonic time in nsec
319  *   0008  34 00                        record is 52 bytes long
320  *   000a        0b 00                  text is 11 bytes long
321  *   000c              1f 00            dictionary is 23 bytes long
322  *   000e                    03 00      LOG_KERN (facility) LOG_ERR (level)
323  *   0010  69 74 27 73 20 61 20 6c      "it's a l"
324  *         69 6e 65                     "ine"
325  *   001b           44 45 56 49 43      "DEVIC"
326  *         45 3d 62 38 3a 32 00 44      "E=b8:2\0D"
327  *         52 49 56 45 52 3d 62 75      "RIVER=bu"
328  *         67                           "g"
329  *   0032     00 00 00                  padding to next message header
330  *
331  * The 'struct printk_log' buffer header must never be directly exported to
332  * userspace, it is a kernel-private implementation detail that might
333  * need to be changed in the future, when the requirements change.
334  *
335  * /dev/kmsg exports the structured data in the following line format:
336  *   "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
337  *
338  * Users of the export format should ignore possible additional values
339  * separated by ',', and find the message after the ';' character.
340  *
341  * The optional key/value pairs are attached as continuation lines starting
342  * with a space character and terminated by a newline. All possible
343  * non-prinatable characters are escaped in the "\xff" notation.
344  */
345 
346 enum log_flags {
347 	LOG_NOCONS	= 1,	/* already flushed, do not print to console */
348 	LOG_NEWLINE	= 2,	/* text ended with a newline */
349 	LOG_PREFIX	= 4,	/* text started with a prefix */
350 	LOG_CONT	= 8,	/* text is a fragment of a continuation line */
351 };
352 
353 struct printk_log {
354 	u64 ts_nsec;		/* timestamp in nanoseconds */
355 	u16 len;		/* length of entire record */
356 	u16 text_len;		/* length of text buffer */
357 	u16 dict_len;		/* length of dictionary buffer */
358 	u8 facility;		/* syslog facility */
359 	u8 flags:5;		/* internal record flags */
360 	u8 level:3;		/* syslog level */
361 }
362 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
363 __packed __aligned(4)
364 #endif
365 ;
366 
367 /*
368  * The logbuf_lock protects kmsg buffer, indices, counters.  This can be taken
369  * within the scheduler's rq lock. It must be released before calling
370  * console_unlock() or anything else that might wake up a process.
371  */
372 DEFINE_RAW_SPINLOCK(logbuf_lock);
373 
374 /*
375  * Helper macros to lock/unlock logbuf_lock and switch between
376  * printk-safe/unsafe modes.
377  */
378 #define logbuf_lock_irq()				\
379 	do {						\
380 		printk_safe_enter_irq();		\
381 		raw_spin_lock(&logbuf_lock);		\
382 	} while (0)
383 
384 #define logbuf_unlock_irq()				\
385 	do {						\
386 		raw_spin_unlock(&logbuf_lock);		\
387 		printk_safe_exit_irq();			\
388 	} while (0)
389 
390 #define logbuf_lock_irqsave(flags)			\
391 	do {						\
392 		printk_safe_enter_irqsave(flags);	\
393 		raw_spin_lock(&logbuf_lock);		\
394 	} while (0)
395 
396 #define logbuf_unlock_irqrestore(flags)		\
397 	do {						\
398 		raw_spin_unlock(&logbuf_lock);		\
399 		printk_safe_exit_irqrestore(flags);	\
400 	} while (0)
401 
402 #ifdef CONFIG_PRINTK
403 DECLARE_WAIT_QUEUE_HEAD(log_wait);
404 /* the next printk record to read by syslog(READ) or /proc/kmsg */
405 static u64 syslog_seq;
406 static u32 syslog_idx;
407 static size_t syslog_partial;
408 
409 /* index and sequence number of the first record stored in the buffer */
410 static u64 log_first_seq;
411 static u32 log_first_idx;
412 
413 /* index and sequence number of the next record to store in the buffer */
414 static u64 log_next_seq;
415 static u32 log_next_idx;
416 
417 /* the next printk record to write to the console */
418 static u64 console_seq;
419 static u32 console_idx;
420 
421 /* the next printk record to read after the last 'clear' command */
422 static u64 clear_seq;
423 static u32 clear_idx;
424 
425 #define PREFIX_MAX		32
426 #define LOG_LINE_MAX		(1024 - PREFIX_MAX)
427 
428 #define LOG_LEVEL(v)		((v) & 0x07)
429 #define LOG_FACILITY(v)		((v) >> 3 & 0xff)
430 
431 /* record buffer */
432 #define LOG_ALIGN __alignof__(struct printk_log)
433 #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
434 static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
435 static char *log_buf = __log_buf;
436 static u32 log_buf_len = __LOG_BUF_LEN;
437 
438 /* Return log buffer address */
439 char *log_buf_addr_get(void)
440 {
441 	return log_buf;
442 }
443 
444 /* Return log buffer size */
445 u32 log_buf_len_get(void)
446 {
447 	return log_buf_len;
448 }
449 
450 /* human readable text of the record */
451 static char *log_text(const struct printk_log *msg)
452 {
453 	return (char *)msg + sizeof(struct printk_log);
454 }
455 
456 /* optional key/value pair dictionary attached to the record */
457 static char *log_dict(const struct printk_log *msg)
458 {
459 	return (char *)msg + sizeof(struct printk_log) + msg->text_len;
460 }
461 
462 /* get record by index; idx must point to valid msg */
463 static struct printk_log *log_from_idx(u32 idx)
464 {
465 	struct printk_log *msg = (struct printk_log *)(log_buf + idx);
466 
467 	/*
468 	 * A length == 0 record is the end of buffer marker. Wrap around and
469 	 * read the message at the start of the buffer.
470 	 */
471 	if (!msg->len)
472 		return (struct printk_log *)log_buf;
473 	return msg;
474 }
475 
476 /* get next record; idx must point to valid msg */
477 static u32 log_next(u32 idx)
478 {
479 	struct printk_log *msg = (struct printk_log *)(log_buf + idx);
480 
481 	/* length == 0 indicates the end of the buffer; wrap */
482 	/*
483 	 * A length == 0 record is the end of buffer marker. Wrap around and
484 	 * read the message at the start of the buffer as *this* one, and
485 	 * return the one after that.
486 	 */
487 	if (!msg->len) {
488 		msg = (struct printk_log *)log_buf;
489 		return msg->len;
490 	}
491 	return idx + msg->len;
492 }
493 
494 /*
495  * Check whether there is enough free space for the given message.
496  *
497  * The same values of first_idx and next_idx mean that the buffer
498  * is either empty or full.
499  *
500  * If the buffer is empty, we must respect the position of the indexes.
501  * They cannot be reset to the beginning of the buffer.
502  */
503 static int logbuf_has_space(u32 msg_size, bool empty)
504 {
505 	u32 free;
506 
507 	if (log_next_idx > log_first_idx || empty)
508 		free = max(log_buf_len - log_next_idx, log_first_idx);
509 	else
510 		free = log_first_idx - log_next_idx;
511 
512 	/*
513 	 * We need space also for an empty header that signalizes wrapping
514 	 * of the buffer.
515 	 */
516 	return free >= msg_size + sizeof(struct printk_log);
517 }
518 
519 static int log_make_free_space(u32 msg_size)
520 {
521 	while (log_first_seq < log_next_seq &&
522 	       !logbuf_has_space(msg_size, false)) {
523 		/* drop old messages until we have enough contiguous space */
524 		log_first_idx = log_next(log_first_idx);
525 		log_first_seq++;
526 	}
527 
528 	if (clear_seq < log_first_seq) {
529 		clear_seq = log_first_seq;
530 		clear_idx = log_first_idx;
531 	}
532 
533 	/* sequence numbers are equal, so the log buffer is empty */
534 	if (logbuf_has_space(msg_size, log_first_seq == log_next_seq))
535 		return 0;
536 
537 	return -ENOMEM;
538 }
539 
540 /* compute the message size including the padding bytes */
541 static u32 msg_used_size(u16 text_len, u16 dict_len, u32 *pad_len)
542 {
543 	u32 size;
544 
545 	size = sizeof(struct printk_log) + text_len + dict_len;
546 	*pad_len = (-size) & (LOG_ALIGN - 1);
547 	size += *pad_len;
548 
549 	return size;
550 }
551 
552 /*
553  * Define how much of the log buffer we could take at maximum. The value
554  * must be greater than two. Note that only half of the buffer is available
555  * when the index points to the middle.
556  */
557 #define MAX_LOG_TAKE_PART 4
558 static const char trunc_msg[] = "<truncated>";
559 
560 static u32 truncate_msg(u16 *text_len, u16 *trunc_msg_len,
561 			u16 *dict_len, u32 *pad_len)
562 {
563 	/*
564 	 * The message should not take the whole buffer. Otherwise, it might
565 	 * get removed too soon.
566 	 */
567 	u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART;
568 	if (*text_len > max_text_len)
569 		*text_len = max_text_len;
570 	/* enable the warning message */
571 	*trunc_msg_len = strlen(trunc_msg);
572 	/* disable the "dict" completely */
573 	*dict_len = 0;
574 	/* compute the size again, count also the warning message */
575 	return msg_used_size(*text_len + *trunc_msg_len, 0, pad_len);
576 }
577 
578 /* insert record into the buffer, discard old ones, update heads */
579 static int log_store(int facility, int level,
580 		     enum log_flags flags, u64 ts_nsec,
581 		     const char *dict, u16 dict_len,
582 		     const char *text, u16 text_len)
583 {
584 	struct printk_log *msg;
585 	u32 size, pad_len;
586 	u16 trunc_msg_len = 0;
587 
588 	/* number of '\0' padding bytes to next message */
589 	size = msg_used_size(text_len, dict_len, &pad_len);
590 
591 	if (log_make_free_space(size)) {
592 		/* truncate the message if it is too long for empty buffer */
593 		size = truncate_msg(&text_len, &trunc_msg_len,
594 				    &dict_len, &pad_len);
595 		/* survive when the log buffer is too small for trunc_msg */
596 		if (log_make_free_space(size))
597 			return 0;
598 	}
599 
600 	if (log_next_idx + size + sizeof(struct printk_log) > log_buf_len) {
601 		/*
602 		 * This message + an additional empty header does not fit
603 		 * at the end of the buffer. Add an empty header with len == 0
604 		 * to signify a wrap around.
605 		 */
606 		memset(log_buf + log_next_idx, 0, sizeof(struct printk_log));
607 		log_next_idx = 0;
608 	}
609 
610 	/* fill message */
611 	msg = (struct printk_log *)(log_buf + log_next_idx);
612 	memcpy(log_text(msg), text, text_len);
613 	msg->text_len = text_len;
614 	if (trunc_msg_len) {
615 		memcpy(log_text(msg) + text_len, trunc_msg, trunc_msg_len);
616 		msg->text_len += trunc_msg_len;
617 	}
618 	memcpy(log_dict(msg), dict, dict_len);
619 	msg->dict_len = dict_len;
620 	msg->facility = facility;
621 	msg->level = level & 7;
622 	msg->flags = flags & 0x1f;
623 	if (ts_nsec > 0)
624 		msg->ts_nsec = ts_nsec;
625 	else
626 		msg->ts_nsec = local_clock();
627 	memset(log_dict(msg) + dict_len, 0, pad_len);
628 	msg->len = size;
629 
630 	/* insert message */
631 	log_next_idx += msg->len;
632 	log_next_seq++;
633 
634 	return msg->text_len;
635 }
636 
637 int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT);
638 
639 static int syslog_action_restricted(int type)
640 {
641 	if (dmesg_restrict)
642 		return 1;
643 	/*
644 	 * Unless restricted, we allow "read all" and "get buffer size"
645 	 * for everybody.
646 	 */
647 	return type != SYSLOG_ACTION_READ_ALL &&
648 	       type != SYSLOG_ACTION_SIZE_BUFFER;
649 }
650 
651 int check_syslog_permissions(int type, int source)
652 {
653 	/*
654 	 * If this is from /proc/kmsg and we've already opened it, then we've
655 	 * already done the capabilities checks at open time.
656 	 */
657 	if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN)
658 		goto ok;
659 
660 	if (syslog_action_restricted(type)) {
661 		if (capable(CAP_SYSLOG))
662 			goto ok;
663 		/*
664 		 * For historical reasons, accept CAP_SYS_ADMIN too, with
665 		 * a warning.
666 		 */
667 		if (capable(CAP_SYS_ADMIN)) {
668 			pr_warn_once("%s (%d): Attempt to access syslog with "
669 				     "CAP_SYS_ADMIN but no CAP_SYSLOG "
670 				     "(deprecated).\n",
671 				 current->comm, task_pid_nr(current));
672 			goto ok;
673 		}
674 		return -EPERM;
675 	}
676 ok:
677 	return security_syslog(type);
678 }
679 EXPORT_SYMBOL_GPL(check_syslog_permissions);
680 
681 static void append_char(char **pp, char *e, char c)
682 {
683 	if (*pp < e)
684 		*(*pp)++ = c;
685 }
686 
687 static ssize_t msg_print_ext_header(char *buf, size_t size,
688 				    struct printk_log *msg, u64 seq)
689 {
690 	u64 ts_usec = msg->ts_nsec;
691 
692 	do_div(ts_usec, 1000);
693 
694 	return scnprintf(buf, size, "%u,%llu,%llu,%c;",
695 		       (msg->facility << 3) | msg->level, seq, ts_usec,
696 		       msg->flags & LOG_CONT ? 'c' : '-');
697 }
698 
699 static ssize_t msg_print_ext_body(char *buf, size_t size,
700 				  char *dict, size_t dict_len,
701 				  char *text, size_t text_len)
702 {
703 	char *p = buf, *e = buf + size;
704 	size_t i;
705 
706 	/* escape non-printable characters */
707 	for (i = 0; i < text_len; i++) {
708 		unsigned char c = text[i];
709 
710 		if (c < ' ' || c >= 127 || c == '\\')
711 			p += scnprintf(p, e - p, "\\x%02x", c);
712 		else
713 			append_char(&p, e, c);
714 	}
715 	append_char(&p, e, '\n');
716 
717 	if (dict_len) {
718 		bool line = true;
719 
720 		for (i = 0; i < dict_len; i++) {
721 			unsigned char c = dict[i];
722 
723 			if (line) {
724 				append_char(&p, e, ' ');
725 				line = false;
726 			}
727 
728 			if (c == '\0') {
729 				append_char(&p, e, '\n');
730 				line = true;
731 				continue;
732 			}
733 
734 			if (c < ' ' || c >= 127 || c == '\\') {
735 				p += scnprintf(p, e - p, "\\x%02x", c);
736 				continue;
737 			}
738 
739 			append_char(&p, e, c);
740 		}
741 		append_char(&p, e, '\n');
742 	}
743 
744 	return p - buf;
745 }
746 
747 /* /dev/kmsg - userspace message inject/listen interface */
748 struct devkmsg_user {
749 	u64 seq;
750 	u32 idx;
751 	struct ratelimit_state rs;
752 	struct mutex lock;
753 	char buf[CONSOLE_EXT_LOG_MAX];
754 };
755 
756 static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from)
757 {
758 	char *buf, *line;
759 	int level = default_message_loglevel;
760 	int facility = 1;	/* LOG_USER */
761 	struct file *file = iocb->ki_filp;
762 	struct devkmsg_user *user = file->private_data;
763 	size_t len = iov_iter_count(from);
764 	ssize_t ret = len;
765 
766 	if (!user || len > LOG_LINE_MAX)
767 		return -EINVAL;
768 
769 	/* Ignore when user logging is disabled. */
770 	if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
771 		return len;
772 
773 	/* Ratelimit when not explicitly enabled. */
774 	if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) {
775 		if (!___ratelimit(&user->rs, current->comm))
776 			return ret;
777 	}
778 
779 	buf = kmalloc(len+1, GFP_KERNEL);
780 	if (buf == NULL)
781 		return -ENOMEM;
782 
783 	buf[len] = '\0';
784 	if (!copy_from_iter_full(buf, len, from)) {
785 		kfree(buf);
786 		return -EFAULT;
787 	}
788 
789 	/*
790 	 * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
791 	 * the decimal value represents 32bit, the lower 3 bit are the log
792 	 * level, the rest are the log facility.
793 	 *
794 	 * If no prefix or no userspace facility is specified, we
795 	 * enforce LOG_USER, to be able to reliably distinguish
796 	 * kernel-generated messages from userspace-injected ones.
797 	 */
798 	line = buf;
799 	if (line[0] == '<') {
800 		char *endp = NULL;
801 		unsigned int u;
802 
803 		u = simple_strtoul(line + 1, &endp, 10);
804 		if (endp && endp[0] == '>') {
805 			level = LOG_LEVEL(u);
806 			if (LOG_FACILITY(u) != 0)
807 				facility = LOG_FACILITY(u);
808 			endp++;
809 			len -= endp - line;
810 			line = endp;
811 		}
812 	}
813 
814 	printk_emit(facility, level, NULL, 0, "%s", line);
815 	kfree(buf);
816 	return ret;
817 }
818 
819 static ssize_t devkmsg_read(struct file *file, char __user *buf,
820 			    size_t count, loff_t *ppos)
821 {
822 	struct devkmsg_user *user = file->private_data;
823 	struct printk_log *msg;
824 	size_t len;
825 	ssize_t ret;
826 
827 	if (!user)
828 		return -EBADF;
829 
830 	ret = mutex_lock_interruptible(&user->lock);
831 	if (ret)
832 		return ret;
833 
834 	logbuf_lock_irq();
835 	while (user->seq == log_next_seq) {
836 		if (file->f_flags & O_NONBLOCK) {
837 			ret = -EAGAIN;
838 			logbuf_unlock_irq();
839 			goto out;
840 		}
841 
842 		logbuf_unlock_irq();
843 		ret = wait_event_interruptible(log_wait,
844 					       user->seq != log_next_seq);
845 		if (ret)
846 			goto out;
847 		logbuf_lock_irq();
848 	}
849 
850 	if (user->seq < log_first_seq) {
851 		/* our last seen message is gone, return error and reset */
852 		user->idx = log_first_idx;
853 		user->seq = log_first_seq;
854 		ret = -EPIPE;
855 		logbuf_unlock_irq();
856 		goto out;
857 	}
858 
859 	msg = log_from_idx(user->idx);
860 	len = msg_print_ext_header(user->buf, sizeof(user->buf),
861 				   msg, user->seq);
862 	len += msg_print_ext_body(user->buf + len, sizeof(user->buf) - len,
863 				  log_dict(msg), msg->dict_len,
864 				  log_text(msg), msg->text_len);
865 
866 	user->idx = log_next(user->idx);
867 	user->seq++;
868 	logbuf_unlock_irq();
869 
870 	if (len > count) {
871 		ret = -EINVAL;
872 		goto out;
873 	}
874 
875 	if (copy_to_user(buf, user->buf, len)) {
876 		ret = -EFAULT;
877 		goto out;
878 	}
879 	ret = len;
880 out:
881 	mutex_unlock(&user->lock);
882 	return ret;
883 }
884 
885 static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
886 {
887 	struct devkmsg_user *user = file->private_data;
888 	loff_t ret = 0;
889 
890 	if (!user)
891 		return -EBADF;
892 	if (offset)
893 		return -ESPIPE;
894 
895 	logbuf_lock_irq();
896 	switch (whence) {
897 	case SEEK_SET:
898 		/* the first record */
899 		user->idx = log_first_idx;
900 		user->seq = log_first_seq;
901 		break;
902 	case SEEK_DATA:
903 		/*
904 		 * The first record after the last SYSLOG_ACTION_CLEAR,
905 		 * like issued by 'dmesg -c'. Reading /dev/kmsg itself
906 		 * changes no global state, and does not clear anything.
907 		 */
908 		user->idx = clear_idx;
909 		user->seq = clear_seq;
910 		break;
911 	case SEEK_END:
912 		/* after the last record */
913 		user->idx = log_next_idx;
914 		user->seq = log_next_seq;
915 		break;
916 	default:
917 		ret = -EINVAL;
918 	}
919 	logbuf_unlock_irq();
920 	return ret;
921 }
922 
923 static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
924 {
925 	struct devkmsg_user *user = file->private_data;
926 	int ret = 0;
927 
928 	if (!user)
929 		return POLLERR|POLLNVAL;
930 
931 	poll_wait(file, &log_wait, wait);
932 
933 	logbuf_lock_irq();
934 	if (user->seq < log_next_seq) {
935 		/* return error when data has vanished underneath us */
936 		if (user->seq < log_first_seq)
937 			ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
938 		else
939 			ret = POLLIN|POLLRDNORM;
940 	}
941 	logbuf_unlock_irq();
942 
943 	return ret;
944 }
945 
946 static int devkmsg_open(struct inode *inode, struct file *file)
947 {
948 	struct devkmsg_user *user;
949 	int err;
950 
951 	if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
952 		return -EPERM;
953 
954 	/* write-only does not need any file context */
955 	if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
956 		err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
957 					       SYSLOG_FROM_READER);
958 		if (err)
959 			return err;
960 	}
961 
962 	user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
963 	if (!user)
964 		return -ENOMEM;
965 
966 	ratelimit_default_init(&user->rs);
967 	ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE);
968 
969 	mutex_init(&user->lock);
970 
971 	logbuf_lock_irq();
972 	user->idx = log_first_idx;
973 	user->seq = log_first_seq;
974 	logbuf_unlock_irq();
975 
976 	file->private_data = user;
977 	return 0;
978 }
979 
980 static int devkmsg_release(struct inode *inode, struct file *file)
981 {
982 	struct devkmsg_user *user = file->private_data;
983 
984 	if (!user)
985 		return 0;
986 
987 	ratelimit_state_exit(&user->rs);
988 
989 	mutex_destroy(&user->lock);
990 	kfree(user);
991 	return 0;
992 }
993 
994 const struct file_operations kmsg_fops = {
995 	.open = devkmsg_open,
996 	.read = devkmsg_read,
997 	.write_iter = devkmsg_write,
998 	.llseek = devkmsg_llseek,
999 	.poll = devkmsg_poll,
1000 	.release = devkmsg_release,
1001 };
1002 
1003 #ifdef CONFIG_KEXEC_CORE
1004 /*
1005  * This appends the listed symbols to /proc/vmcore
1006  *
1007  * /proc/vmcore is used by various utilities, like crash and makedumpfile to
1008  * obtain access to symbols that are otherwise very difficult to locate.  These
1009  * symbols are specifically used so that utilities can access and extract the
1010  * dmesg log from a vmcore file after a crash.
1011  */
1012 void log_buf_kexec_setup(void)
1013 {
1014 	VMCOREINFO_SYMBOL(log_buf);
1015 	VMCOREINFO_SYMBOL(log_buf_len);
1016 	VMCOREINFO_SYMBOL(log_first_idx);
1017 	VMCOREINFO_SYMBOL(clear_idx);
1018 	VMCOREINFO_SYMBOL(log_next_idx);
1019 	/*
1020 	 * Export struct printk_log size and field offsets. User space tools can
1021 	 * parse it and detect any changes to structure down the line.
1022 	 */
1023 	VMCOREINFO_STRUCT_SIZE(printk_log);
1024 	VMCOREINFO_OFFSET(printk_log, ts_nsec);
1025 	VMCOREINFO_OFFSET(printk_log, len);
1026 	VMCOREINFO_OFFSET(printk_log, text_len);
1027 	VMCOREINFO_OFFSET(printk_log, dict_len);
1028 }
1029 #endif
1030 
1031 /* requested log_buf_len from kernel cmdline */
1032 static unsigned long __initdata new_log_buf_len;
1033 
1034 /* we practice scaling the ring buffer by powers of 2 */
1035 static void __init log_buf_len_update(unsigned size)
1036 {
1037 	if (size)
1038 		size = roundup_pow_of_two(size);
1039 	if (size > log_buf_len)
1040 		new_log_buf_len = size;
1041 }
1042 
1043 /* save requested log_buf_len since it's too early to process it */
1044 static int __init log_buf_len_setup(char *str)
1045 {
1046 	unsigned size = memparse(str, &str);
1047 
1048 	log_buf_len_update(size);
1049 
1050 	return 0;
1051 }
1052 early_param("log_buf_len", log_buf_len_setup);
1053 
1054 #ifdef CONFIG_SMP
1055 #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1056 
1057 static void __init log_buf_add_cpu(void)
1058 {
1059 	unsigned int cpu_extra;
1060 
1061 	/*
1062 	 * archs should set up cpu_possible_bits properly with
1063 	 * set_cpu_possible() after setup_arch() but just in
1064 	 * case lets ensure this is valid.
1065 	 */
1066 	if (num_possible_cpus() == 1)
1067 		return;
1068 
1069 	cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
1070 
1071 	/* by default this will only continue through for large > 64 CPUs */
1072 	if (cpu_extra <= __LOG_BUF_LEN / 2)
1073 		return;
1074 
1075 	pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1076 		__LOG_CPU_MAX_BUF_LEN);
1077 	pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1078 		cpu_extra);
1079 	pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN);
1080 
1081 	log_buf_len_update(cpu_extra + __LOG_BUF_LEN);
1082 }
1083 #else /* !CONFIG_SMP */
1084 static inline void log_buf_add_cpu(void) {}
1085 #endif /* CONFIG_SMP */
1086 
1087 void __init setup_log_buf(int early)
1088 {
1089 	unsigned long flags;
1090 	char *new_log_buf;
1091 	int free;
1092 
1093 	if (log_buf != __log_buf)
1094 		return;
1095 
1096 	if (!early && !new_log_buf_len)
1097 		log_buf_add_cpu();
1098 
1099 	if (!new_log_buf_len)
1100 		return;
1101 
1102 	if (early) {
1103 		new_log_buf =
1104 			memblock_virt_alloc(new_log_buf_len, LOG_ALIGN);
1105 	} else {
1106 		new_log_buf = memblock_virt_alloc_nopanic(new_log_buf_len,
1107 							  LOG_ALIGN);
1108 	}
1109 
1110 	if (unlikely(!new_log_buf)) {
1111 		pr_err("log_buf_len: %ld bytes not available\n",
1112 			new_log_buf_len);
1113 		return;
1114 	}
1115 
1116 	logbuf_lock_irqsave(flags);
1117 	log_buf_len = new_log_buf_len;
1118 	log_buf = new_log_buf;
1119 	new_log_buf_len = 0;
1120 	free = __LOG_BUF_LEN - log_next_idx;
1121 	memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
1122 	logbuf_unlock_irqrestore(flags);
1123 
1124 	pr_info("log_buf_len: %d bytes\n", log_buf_len);
1125 	pr_info("early log buf free: %d(%d%%)\n",
1126 		free, (free * 100) / __LOG_BUF_LEN);
1127 }
1128 
1129 static bool __read_mostly ignore_loglevel;
1130 
1131 static int __init ignore_loglevel_setup(char *str)
1132 {
1133 	ignore_loglevel = true;
1134 	pr_info("debug: ignoring loglevel setting.\n");
1135 
1136 	return 0;
1137 }
1138 
1139 early_param("ignore_loglevel", ignore_loglevel_setup);
1140 module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1141 MODULE_PARM_DESC(ignore_loglevel,
1142 		 "ignore loglevel setting (prints all kernel messages to the console)");
1143 
1144 static bool suppress_message_printing(int level)
1145 {
1146 	return (level >= console_loglevel && !ignore_loglevel);
1147 }
1148 
1149 #ifdef CONFIG_BOOT_PRINTK_DELAY
1150 
1151 static int boot_delay; /* msecs delay after each printk during bootup */
1152 static unsigned long long loops_per_msec;	/* based on boot_delay */
1153 
1154 static int __init boot_delay_setup(char *str)
1155 {
1156 	unsigned long lpj;
1157 
1158 	lpj = preset_lpj ? preset_lpj : 1000000;	/* some guess */
1159 	loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
1160 
1161 	get_option(&str, &boot_delay);
1162 	if (boot_delay > 10 * 1000)
1163 		boot_delay = 0;
1164 
1165 	pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1166 		"HZ: %d, loops_per_msec: %llu\n",
1167 		boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
1168 	return 0;
1169 }
1170 early_param("boot_delay", boot_delay_setup);
1171 
1172 static void boot_delay_msec(int level)
1173 {
1174 	unsigned long long k;
1175 	unsigned long timeout;
1176 
1177 	if ((boot_delay == 0 || system_state != SYSTEM_BOOTING)
1178 		|| suppress_message_printing(level)) {
1179 		return;
1180 	}
1181 
1182 	k = (unsigned long long)loops_per_msec * boot_delay;
1183 
1184 	timeout = jiffies + msecs_to_jiffies(boot_delay);
1185 	while (k) {
1186 		k--;
1187 		cpu_relax();
1188 		/*
1189 		 * use (volatile) jiffies to prevent
1190 		 * compiler reduction; loop termination via jiffies
1191 		 * is secondary and may or may not happen.
1192 		 */
1193 		if (time_after(jiffies, timeout))
1194 			break;
1195 		touch_nmi_watchdog();
1196 	}
1197 }
1198 #else
1199 static inline void boot_delay_msec(int level)
1200 {
1201 }
1202 #endif
1203 
1204 static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME);
1205 module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
1206 
1207 static size_t print_time(u64 ts, char *buf)
1208 {
1209 	unsigned long rem_nsec;
1210 
1211 	if (!printk_time)
1212 		return 0;
1213 
1214 	rem_nsec = do_div(ts, 1000000000);
1215 
1216 	if (!buf)
1217 		return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
1218 
1219 	return sprintf(buf, "[%5lu.%06lu] ",
1220 		       (unsigned long)ts, rem_nsec / 1000);
1221 }
1222 
1223 static size_t print_prefix(const struct printk_log *msg, bool syslog, char *buf)
1224 {
1225 	size_t len = 0;
1226 	unsigned int prefix = (msg->facility << 3) | msg->level;
1227 
1228 	if (syslog) {
1229 		if (buf) {
1230 			len += sprintf(buf, "<%u>", prefix);
1231 		} else {
1232 			len += 3;
1233 			if (prefix > 999)
1234 				len += 3;
1235 			else if (prefix > 99)
1236 				len += 2;
1237 			else if (prefix > 9)
1238 				len++;
1239 		}
1240 	}
1241 
1242 	len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
1243 	return len;
1244 }
1245 
1246 static size_t msg_print_text(const struct printk_log *msg, bool syslog, char *buf, size_t size)
1247 {
1248 	const char *text = log_text(msg);
1249 	size_t text_size = msg->text_len;
1250 	size_t len = 0;
1251 
1252 	do {
1253 		const char *next = memchr(text, '\n', text_size);
1254 		size_t text_len;
1255 
1256 		if (next) {
1257 			text_len = next - text;
1258 			next++;
1259 			text_size -= next - text;
1260 		} else {
1261 			text_len = text_size;
1262 		}
1263 
1264 		if (buf) {
1265 			if (print_prefix(msg, syslog, NULL) +
1266 			    text_len + 1 >= size - len)
1267 				break;
1268 
1269 			len += print_prefix(msg, syslog, buf + len);
1270 			memcpy(buf + len, text, text_len);
1271 			len += text_len;
1272 			buf[len++] = '\n';
1273 		} else {
1274 			/* SYSLOG_ACTION_* buffer size only calculation */
1275 			len += print_prefix(msg, syslog, NULL);
1276 			len += text_len;
1277 			len++;
1278 		}
1279 
1280 		text = next;
1281 	} while (text);
1282 
1283 	return len;
1284 }
1285 
1286 static int syslog_print(char __user *buf, int size)
1287 {
1288 	char *text;
1289 	struct printk_log *msg;
1290 	int len = 0;
1291 
1292 	text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1293 	if (!text)
1294 		return -ENOMEM;
1295 
1296 	while (size > 0) {
1297 		size_t n;
1298 		size_t skip;
1299 
1300 		logbuf_lock_irq();
1301 		if (syslog_seq < log_first_seq) {
1302 			/* messages are gone, move to first one */
1303 			syslog_seq = log_first_seq;
1304 			syslog_idx = log_first_idx;
1305 			syslog_partial = 0;
1306 		}
1307 		if (syslog_seq == log_next_seq) {
1308 			logbuf_unlock_irq();
1309 			break;
1310 		}
1311 
1312 		skip = syslog_partial;
1313 		msg = log_from_idx(syslog_idx);
1314 		n = msg_print_text(msg, true, text, LOG_LINE_MAX + PREFIX_MAX);
1315 		if (n - syslog_partial <= size) {
1316 			/* message fits into buffer, move forward */
1317 			syslog_idx = log_next(syslog_idx);
1318 			syslog_seq++;
1319 			n -= syslog_partial;
1320 			syslog_partial = 0;
1321 		} else if (!len){
1322 			/* partial read(), remember position */
1323 			n = size;
1324 			syslog_partial += n;
1325 		} else
1326 			n = 0;
1327 		logbuf_unlock_irq();
1328 
1329 		if (!n)
1330 			break;
1331 
1332 		if (copy_to_user(buf, text + skip, n)) {
1333 			if (!len)
1334 				len = -EFAULT;
1335 			break;
1336 		}
1337 
1338 		len += n;
1339 		size -= n;
1340 		buf += n;
1341 	}
1342 
1343 	kfree(text);
1344 	return len;
1345 }
1346 
1347 static int syslog_print_all(char __user *buf, int size, bool clear)
1348 {
1349 	char *text;
1350 	int len = 0;
1351 
1352 	text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1353 	if (!text)
1354 		return -ENOMEM;
1355 
1356 	logbuf_lock_irq();
1357 	if (buf) {
1358 		u64 next_seq;
1359 		u64 seq;
1360 		u32 idx;
1361 
1362 		/*
1363 		 * Find first record that fits, including all following records,
1364 		 * into the user-provided buffer for this dump.
1365 		 */
1366 		seq = clear_seq;
1367 		idx = clear_idx;
1368 		while (seq < log_next_seq) {
1369 			struct printk_log *msg = log_from_idx(idx);
1370 
1371 			len += msg_print_text(msg, true, NULL, 0);
1372 			idx = log_next(idx);
1373 			seq++;
1374 		}
1375 
1376 		/* move first record forward until length fits into the buffer */
1377 		seq = clear_seq;
1378 		idx = clear_idx;
1379 		while (len > size && seq < log_next_seq) {
1380 			struct printk_log *msg = log_from_idx(idx);
1381 
1382 			len -= msg_print_text(msg, true, NULL, 0);
1383 			idx = log_next(idx);
1384 			seq++;
1385 		}
1386 
1387 		/* last message fitting into this dump */
1388 		next_seq = log_next_seq;
1389 
1390 		len = 0;
1391 		while (len >= 0 && seq < next_seq) {
1392 			struct printk_log *msg = log_from_idx(idx);
1393 			int textlen;
1394 
1395 			textlen = msg_print_text(msg, true, text,
1396 						 LOG_LINE_MAX + PREFIX_MAX);
1397 			if (textlen < 0) {
1398 				len = textlen;
1399 				break;
1400 			}
1401 			idx = log_next(idx);
1402 			seq++;
1403 
1404 			logbuf_unlock_irq();
1405 			if (copy_to_user(buf + len, text, textlen))
1406 				len = -EFAULT;
1407 			else
1408 				len += textlen;
1409 			logbuf_lock_irq();
1410 
1411 			if (seq < log_first_seq) {
1412 				/* messages are gone, move to next one */
1413 				seq = log_first_seq;
1414 				idx = log_first_idx;
1415 			}
1416 		}
1417 	}
1418 
1419 	if (clear) {
1420 		clear_seq = log_next_seq;
1421 		clear_idx = log_next_idx;
1422 	}
1423 	logbuf_unlock_irq();
1424 
1425 	kfree(text);
1426 	return len;
1427 }
1428 
1429 int do_syslog(int type, char __user *buf, int len, int source)
1430 {
1431 	bool clear = false;
1432 	static int saved_console_loglevel = LOGLEVEL_DEFAULT;
1433 	int error;
1434 
1435 	error = check_syslog_permissions(type, source);
1436 	if (error)
1437 		goto out;
1438 
1439 	switch (type) {
1440 	case SYSLOG_ACTION_CLOSE:	/* Close log */
1441 		break;
1442 	case SYSLOG_ACTION_OPEN:	/* Open log */
1443 		break;
1444 	case SYSLOG_ACTION_READ:	/* Read from log */
1445 		error = -EINVAL;
1446 		if (!buf || len < 0)
1447 			goto out;
1448 		error = 0;
1449 		if (!len)
1450 			goto out;
1451 		if (!access_ok(VERIFY_WRITE, buf, len)) {
1452 			error = -EFAULT;
1453 			goto out;
1454 		}
1455 		error = wait_event_interruptible(log_wait,
1456 						 syslog_seq != log_next_seq);
1457 		if (error)
1458 			goto out;
1459 		error = syslog_print(buf, len);
1460 		break;
1461 	/* Read/clear last kernel messages */
1462 	case SYSLOG_ACTION_READ_CLEAR:
1463 		clear = true;
1464 		/* FALL THRU */
1465 	/* Read last kernel messages */
1466 	case SYSLOG_ACTION_READ_ALL:
1467 		error = -EINVAL;
1468 		if (!buf || len < 0)
1469 			goto out;
1470 		error = 0;
1471 		if (!len)
1472 			goto out;
1473 		if (!access_ok(VERIFY_WRITE, buf, len)) {
1474 			error = -EFAULT;
1475 			goto out;
1476 		}
1477 		error = syslog_print_all(buf, len, clear);
1478 		break;
1479 	/* Clear ring buffer */
1480 	case SYSLOG_ACTION_CLEAR:
1481 		syslog_print_all(NULL, 0, true);
1482 		break;
1483 	/* Disable logging to console */
1484 	case SYSLOG_ACTION_CONSOLE_OFF:
1485 		if (saved_console_loglevel == LOGLEVEL_DEFAULT)
1486 			saved_console_loglevel = console_loglevel;
1487 		console_loglevel = minimum_console_loglevel;
1488 		break;
1489 	/* Enable logging to console */
1490 	case SYSLOG_ACTION_CONSOLE_ON:
1491 		if (saved_console_loglevel != LOGLEVEL_DEFAULT) {
1492 			console_loglevel = saved_console_loglevel;
1493 			saved_console_loglevel = LOGLEVEL_DEFAULT;
1494 		}
1495 		break;
1496 	/* Set level of messages printed to console */
1497 	case SYSLOG_ACTION_CONSOLE_LEVEL:
1498 		error = -EINVAL;
1499 		if (len < 1 || len > 8)
1500 			goto out;
1501 		if (len < minimum_console_loglevel)
1502 			len = minimum_console_loglevel;
1503 		console_loglevel = len;
1504 		/* Implicitly re-enable logging to console */
1505 		saved_console_loglevel = LOGLEVEL_DEFAULT;
1506 		error = 0;
1507 		break;
1508 	/* Number of chars in the log buffer */
1509 	case SYSLOG_ACTION_SIZE_UNREAD:
1510 		logbuf_lock_irq();
1511 		if (syslog_seq < log_first_seq) {
1512 			/* messages are gone, move to first one */
1513 			syslog_seq = log_first_seq;
1514 			syslog_idx = log_first_idx;
1515 			syslog_partial = 0;
1516 		}
1517 		if (source == SYSLOG_FROM_PROC) {
1518 			/*
1519 			 * Short-cut for poll(/"proc/kmsg") which simply checks
1520 			 * for pending data, not the size; return the count of
1521 			 * records, not the length.
1522 			 */
1523 			error = log_next_seq - syslog_seq;
1524 		} else {
1525 			u64 seq = syslog_seq;
1526 			u32 idx = syslog_idx;
1527 
1528 			error = 0;
1529 			while (seq < log_next_seq) {
1530 				struct printk_log *msg = log_from_idx(idx);
1531 
1532 				error += msg_print_text(msg, true, NULL, 0);
1533 				idx = log_next(idx);
1534 				seq++;
1535 			}
1536 			error -= syslog_partial;
1537 		}
1538 		logbuf_unlock_irq();
1539 		break;
1540 	/* Size of the log buffer */
1541 	case SYSLOG_ACTION_SIZE_BUFFER:
1542 		error = log_buf_len;
1543 		break;
1544 	default:
1545 		error = -EINVAL;
1546 		break;
1547 	}
1548 out:
1549 	return error;
1550 }
1551 
1552 SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1553 {
1554 	return do_syslog(type, buf, len, SYSLOG_FROM_READER);
1555 }
1556 
1557 /*
1558  * Call the console drivers, asking them to write out
1559  * log_buf[start] to log_buf[end - 1].
1560  * The console_lock must be held.
1561  */
1562 static void call_console_drivers(const char *ext_text, size_t ext_len,
1563 				 const char *text, size_t len)
1564 {
1565 	struct console *con;
1566 
1567 	trace_console_rcuidle(text, len);
1568 
1569 	if (!console_drivers)
1570 		return;
1571 
1572 	for_each_console(con) {
1573 		if (exclusive_console && con != exclusive_console)
1574 			continue;
1575 		if (!(con->flags & CON_ENABLED))
1576 			continue;
1577 		if (!con->write)
1578 			continue;
1579 		if (!cpu_online(smp_processor_id()) &&
1580 		    !(con->flags & CON_ANYTIME))
1581 			continue;
1582 		if (con->flags & CON_EXTENDED)
1583 			con->write(con, ext_text, ext_len);
1584 		else
1585 			con->write(con, text, len);
1586 	}
1587 }
1588 
1589 int printk_delay_msec __read_mostly;
1590 
1591 static inline void printk_delay(void)
1592 {
1593 	if (unlikely(printk_delay_msec)) {
1594 		int m = printk_delay_msec;
1595 
1596 		while (m--) {
1597 			mdelay(1);
1598 			touch_nmi_watchdog();
1599 		}
1600 	}
1601 }
1602 
1603 /*
1604  * Continuation lines are buffered, and not committed to the record buffer
1605  * until the line is complete, or a race forces it. The line fragments
1606  * though, are printed immediately to the consoles to ensure everything has
1607  * reached the console in case of a kernel crash.
1608  */
1609 static struct cont {
1610 	char buf[LOG_LINE_MAX];
1611 	size_t len;			/* length == 0 means unused buffer */
1612 	struct task_struct *owner;	/* task of first print*/
1613 	u64 ts_nsec;			/* time of first print */
1614 	u8 level;			/* log level of first message */
1615 	u8 facility;			/* log facility of first message */
1616 	enum log_flags flags;		/* prefix, newline flags */
1617 } cont;
1618 
1619 static void cont_flush(void)
1620 {
1621 	if (cont.len == 0)
1622 		return;
1623 
1624 	log_store(cont.facility, cont.level, cont.flags, cont.ts_nsec,
1625 		  NULL, 0, cont.buf, cont.len);
1626 	cont.len = 0;
1627 }
1628 
1629 static bool cont_add(int facility, int level, enum log_flags flags, const char *text, size_t len)
1630 {
1631 	/*
1632 	 * If ext consoles are present, flush and skip in-kernel
1633 	 * continuation.  See nr_ext_console_drivers definition.  Also, if
1634 	 * the line gets too long, split it up in separate records.
1635 	 */
1636 	if (nr_ext_console_drivers || cont.len + len > sizeof(cont.buf)) {
1637 		cont_flush();
1638 		return false;
1639 	}
1640 
1641 	if (!cont.len) {
1642 		cont.facility = facility;
1643 		cont.level = level;
1644 		cont.owner = current;
1645 		cont.ts_nsec = local_clock();
1646 		cont.flags = flags;
1647 	}
1648 
1649 	memcpy(cont.buf + cont.len, text, len);
1650 	cont.len += len;
1651 
1652 	// The original flags come from the first line,
1653 	// but later continuations can add a newline.
1654 	if (flags & LOG_NEWLINE) {
1655 		cont.flags |= LOG_NEWLINE;
1656 		cont_flush();
1657 	}
1658 
1659 	if (cont.len > (sizeof(cont.buf) * 80) / 100)
1660 		cont_flush();
1661 
1662 	return true;
1663 }
1664 
1665 static size_t log_output(int facility, int level, enum log_flags lflags, const char *dict, size_t dictlen, char *text, size_t text_len)
1666 {
1667 	/*
1668 	 * If an earlier line was buffered, and we're a continuation
1669 	 * write from the same process, try to add it to the buffer.
1670 	 */
1671 	if (cont.len) {
1672 		if (cont.owner == current && (lflags & LOG_CONT)) {
1673 			if (cont_add(facility, level, lflags, text, text_len))
1674 				return text_len;
1675 		}
1676 		/* Otherwise, make sure it's flushed */
1677 		cont_flush();
1678 	}
1679 
1680 	/* Skip empty continuation lines that couldn't be added - they just flush */
1681 	if (!text_len && (lflags & LOG_CONT))
1682 		return 0;
1683 
1684 	/* If it doesn't end in a newline, try to buffer the current line */
1685 	if (!(lflags & LOG_NEWLINE)) {
1686 		if (cont_add(facility, level, lflags, text, text_len))
1687 			return text_len;
1688 	}
1689 
1690 	/* Store it in the record log */
1691 	return log_store(facility, level, lflags, 0, dict, dictlen, text, text_len);
1692 }
1693 
1694 asmlinkage int vprintk_emit(int facility, int level,
1695 			    const char *dict, size_t dictlen,
1696 			    const char *fmt, va_list args)
1697 {
1698 	static char textbuf[LOG_LINE_MAX];
1699 	char *text = textbuf;
1700 	size_t text_len = 0;
1701 	enum log_flags lflags = 0;
1702 	unsigned long flags;
1703 	int printed_len = 0;
1704 	bool in_sched = false;
1705 
1706 	if (level == LOGLEVEL_SCHED) {
1707 		level = LOGLEVEL_DEFAULT;
1708 		in_sched = true;
1709 	}
1710 
1711 	boot_delay_msec(level);
1712 	printk_delay();
1713 
1714 	/* This stops the holder of console_sem just where we want him */
1715 	logbuf_lock_irqsave(flags);
1716 	/*
1717 	 * The printf needs to come first; we need the syslog
1718 	 * prefix which might be passed-in as a parameter.
1719 	 */
1720 	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1721 
1722 	/* mark and strip a trailing newline */
1723 	if (text_len && text[text_len-1] == '\n') {
1724 		text_len--;
1725 		lflags |= LOG_NEWLINE;
1726 	}
1727 
1728 	/* strip kernel syslog prefix and extract log level or control flags */
1729 	if (facility == 0) {
1730 		int kern_level;
1731 
1732 		while ((kern_level = printk_get_level(text)) != 0) {
1733 			switch (kern_level) {
1734 			case '0' ... '7':
1735 				if (level == LOGLEVEL_DEFAULT)
1736 					level = kern_level - '0';
1737 				/* fallthrough */
1738 			case 'd':	/* KERN_DEFAULT */
1739 				lflags |= LOG_PREFIX;
1740 				break;
1741 			case 'c':	/* KERN_CONT */
1742 				lflags |= LOG_CONT;
1743 			}
1744 
1745 			text_len -= 2;
1746 			text += 2;
1747 		}
1748 	}
1749 
1750 	if (level == LOGLEVEL_DEFAULT)
1751 		level = default_message_loglevel;
1752 
1753 	if (dict)
1754 		lflags |= LOG_PREFIX|LOG_NEWLINE;
1755 
1756 	printed_len += log_output(facility, level, lflags, dict, dictlen, text, text_len);
1757 
1758 	logbuf_unlock_irqrestore(flags);
1759 
1760 	/* If called from the scheduler, we can not call up(). */
1761 	if (!in_sched) {
1762 		/*
1763 		 * Try to acquire and then immediately release the console
1764 		 * semaphore.  The release will print out buffers and wake up
1765 		 * /dev/kmsg and syslog() users.
1766 		 */
1767 		if (console_trylock())
1768 			console_unlock();
1769 	}
1770 
1771 	return printed_len;
1772 }
1773 EXPORT_SYMBOL(vprintk_emit);
1774 
1775 asmlinkage int vprintk(const char *fmt, va_list args)
1776 {
1777 	return vprintk_func(fmt, args);
1778 }
1779 EXPORT_SYMBOL(vprintk);
1780 
1781 asmlinkage int printk_emit(int facility, int level,
1782 			   const char *dict, size_t dictlen,
1783 			   const char *fmt, ...)
1784 {
1785 	va_list args;
1786 	int r;
1787 
1788 	va_start(args, fmt);
1789 	r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1790 	va_end(args);
1791 
1792 	return r;
1793 }
1794 EXPORT_SYMBOL(printk_emit);
1795 
1796 int vprintk_default(const char *fmt, va_list args)
1797 {
1798 	int r;
1799 
1800 #ifdef CONFIG_KGDB_KDB
1801 	/* Allow to pass printk() to kdb but avoid a recursion. */
1802 	if (unlikely(kdb_trap_printk && kdb_printf_cpu < 0)) {
1803 		r = vkdb_printf(KDB_MSGSRC_PRINTK, fmt, args);
1804 		return r;
1805 	}
1806 #endif
1807 	r = vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
1808 
1809 	return r;
1810 }
1811 EXPORT_SYMBOL_GPL(vprintk_default);
1812 
1813 /**
1814  * printk - print a kernel message
1815  * @fmt: format string
1816  *
1817  * This is printk(). It can be called from any context. We want it to work.
1818  *
1819  * We try to grab the console_lock. If we succeed, it's easy - we log the
1820  * output and call the console drivers.  If we fail to get the semaphore, we
1821  * place the output into the log buffer and return. The current holder of
1822  * the console_sem will notice the new output in console_unlock(); and will
1823  * send it to the consoles before releasing the lock.
1824  *
1825  * One effect of this deferred printing is that code which calls printk() and
1826  * then changes console_loglevel may break. This is because console_loglevel
1827  * is inspected when the actual printing occurs.
1828  *
1829  * See also:
1830  * printf(3)
1831  *
1832  * See the vsnprintf() documentation for format string extensions over C99.
1833  */
1834 asmlinkage __visible int printk(const char *fmt, ...)
1835 {
1836 	va_list args;
1837 	int r;
1838 
1839 	va_start(args, fmt);
1840 	r = vprintk_func(fmt, args);
1841 	va_end(args);
1842 
1843 	return r;
1844 }
1845 EXPORT_SYMBOL(printk);
1846 
1847 #else /* CONFIG_PRINTK */
1848 
1849 #define LOG_LINE_MAX		0
1850 #define PREFIX_MAX		0
1851 
1852 static u64 syslog_seq;
1853 static u32 syslog_idx;
1854 static u64 console_seq;
1855 static u32 console_idx;
1856 static u64 log_first_seq;
1857 static u32 log_first_idx;
1858 static u64 log_next_seq;
1859 static char *log_text(const struct printk_log *msg) { return NULL; }
1860 static char *log_dict(const struct printk_log *msg) { return NULL; }
1861 static struct printk_log *log_from_idx(u32 idx) { return NULL; }
1862 static u32 log_next(u32 idx) { return 0; }
1863 static ssize_t msg_print_ext_header(char *buf, size_t size,
1864 				    struct printk_log *msg,
1865 				    u64 seq) { return 0; }
1866 static ssize_t msg_print_ext_body(char *buf, size_t size,
1867 				  char *dict, size_t dict_len,
1868 				  char *text, size_t text_len) { return 0; }
1869 static void call_console_drivers(const char *ext_text, size_t ext_len,
1870 				 const char *text, size_t len) {}
1871 static size_t msg_print_text(const struct printk_log *msg,
1872 			     bool syslog, char *buf, size_t size) { return 0; }
1873 static bool suppress_message_printing(int level) { return false; }
1874 
1875 #endif /* CONFIG_PRINTK */
1876 
1877 #ifdef CONFIG_EARLY_PRINTK
1878 struct console *early_console;
1879 
1880 asmlinkage __visible void early_printk(const char *fmt, ...)
1881 {
1882 	va_list ap;
1883 	char buf[512];
1884 	int n;
1885 
1886 	if (!early_console)
1887 		return;
1888 
1889 	va_start(ap, fmt);
1890 	n = vscnprintf(buf, sizeof(buf), fmt, ap);
1891 	va_end(ap);
1892 
1893 	early_console->write(early_console, buf, n);
1894 }
1895 #endif
1896 
1897 static int __add_preferred_console(char *name, int idx, char *options,
1898 				   char *brl_options)
1899 {
1900 	struct console_cmdline *c;
1901 	int i;
1902 
1903 	/*
1904 	 *	See if this tty is not yet registered, and
1905 	 *	if we have a slot free.
1906 	 */
1907 	for (i = 0, c = console_cmdline;
1908 	     i < MAX_CMDLINECONSOLES && c->name[0];
1909 	     i++, c++) {
1910 		if (strcmp(c->name, name) == 0 && c->index == idx) {
1911 			if (!brl_options)
1912 				selected_console = i;
1913 			return 0;
1914 		}
1915 	}
1916 	if (i == MAX_CMDLINECONSOLES)
1917 		return -E2BIG;
1918 	if (!brl_options)
1919 		selected_console = i;
1920 	strlcpy(c->name, name, sizeof(c->name));
1921 	c->options = options;
1922 	braille_set_options(c, brl_options);
1923 
1924 	c->index = idx;
1925 	return 0;
1926 }
1927 /*
1928  * Set up a console.  Called via do_early_param() in init/main.c
1929  * for each "console=" parameter in the boot command line.
1930  */
1931 static int __init console_setup(char *str)
1932 {
1933 	char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */
1934 	char *s, *options, *brl_options = NULL;
1935 	int idx;
1936 
1937 	if (_braille_console_setup(&str, &brl_options))
1938 		return 1;
1939 
1940 	/*
1941 	 * Decode str into name, index, options.
1942 	 */
1943 	if (str[0] >= '0' && str[0] <= '9') {
1944 		strcpy(buf, "ttyS");
1945 		strncpy(buf + 4, str, sizeof(buf) - 5);
1946 	} else {
1947 		strncpy(buf, str, sizeof(buf) - 1);
1948 	}
1949 	buf[sizeof(buf) - 1] = 0;
1950 	options = strchr(str, ',');
1951 	if (options)
1952 		*(options++) = 0;
1953 #ifdef __sparc__
1954 	if (!strcmp(str, "ttya"))
1955 		strcpy(buf, "ttyS0");
1956 	if (!strcmp(str, "ttyb"))
1957 		strcpy(buf, "ttyS1");
1958 #endif
1959 	for (s = buf; *s; s++)
1960 		if (isdigit(*s) || *s == ',')
1961 			break;
1962 	idx = simple_strtoul(s, NULL, 10);
1963 	*s = 0;
1964 
1965 	__add_preferred_console(buf, idx, options, brl_options);
1966 	console_set_on_cmdline = 1;
1967 	return 1;
1968 }
1969 __setup("console=", console_setup);
1970 
1971 /**
1972  * add_preferred_console - add a device to the list of preferred consoles.
1973  * @name: device name
1974  * @idx: device index
1975  * @options: options for this console
1976  *
1977  * The last preferred console added will be used for kernel messages
1978  * and stdin/out/err for init.  Normally this is used by console_setup
1979  * above to handle user-supplied console arguments; however it can also
1980  * be used by arch-specific code either to override the user or more
1981  * commonly to provide a default console (ie from PROM variables) when
1982  * the user has not supplied one.
1983  */
1984 int add_preferred_console(char *name, int idx, char *options)
1985 {
1986 	return __add_preferred_console(name, idx, options, NULL);
1987 }
1988 
1989 bool console_suspend_enabled = true;
1990 EXPORT_SYMBOL(console_suspend_enabled);
1991 
1992 static int __init console_suspend_disable(char *str)
1993 {
1994 	console_suspend_enabled = false;
1995 	return 1;
1996 }
1997 __setup("no_console_suspend", console_suspend_disable);
1998 module_param_named(console_suspend, console_suspend_enabled,
1999 		bool, S_IRUGO | S_IWUSR);
2000 MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
2001 	" and hibernate operations");
2002 
2003 /**
2004  * suspend_console - suspend the console subsystem
2005  *
2006  * This disables printk() while we go into suspend states
2007  */
2008 void suspend_console(void)
2009 {
2010 	if (!console_suspend_enabled)
2011 		return;
2012 	printk("Suspending console(s) (use no_console_suspend to debug)\n");
2013 	console_lock();
2014 	console_suspended = 1;
2015 	up_console_sem();
2016 }
2017 
2018 void resume_console(void)
2019 {
2020 	if (!console_suspend_enabled)
2021 		return;
2022 	down_console_sem();
2023 	console_suspended = 0;
2024 	console_unlock();
2025 }
2026 
2027 /**
2028  * console_cpu_notify - print deferred console messages after CPU hotplug
2029  * @cpu: unused
2030  *
2031  * If printk() is called from a CPU that is not online yet, the messages
2032  * will be spooled but will not show up on the console.  This function is
2033  * called when a new CPU comes online (or fails to come up), and ensures
2034  * that any such output gets printed.
2035  */
2036 static int console_cpu_notify(unsigned int cpu)
2037 {
2038 	if (!cpuhp_tasks_frozen) {
2039 		console_lock();
2040 		console_unlock();
2041 	}
2042 	return 0;
2043 }
2044 
2045 /**
2046  * console_lock - lock the console system for exclusive use.
2047  *
2048  * Acquires a lock which guarantees that the caller has
2049  * exclusive access to the console system and the console_drivers list.
2050  *
2051  * Can sleep, returns nothing.
2052  */
2053 void console_lock(void)
2054 {
2055 	might_sleep();
2056 
2057 	down_console_sem();
2058 	if (console_suspended)
2059 		return;
2060 	console_locked = 1;
2061 	console_may_schedule = 1;
2062 }
2063 EXPORT_SYMBOL(console_lock);
2064 
2065 /**
2066  * console_trylock - try to lock the console system for exclusive use.
2067  *
2068  * Try to acquire a lock which guarantees that the caller has exclusive
2069  * access to the console system and the console_drivers list.
2070  *
2071  * returns 1 on success, and 0 on failure to acquire the lock.
2072  */
2073 int console_trylock(void)
2074 {
2075 	if (down_trylock_console_sem())
2076 		return 0;
2077 	if (console_suspended) {
2078 		up_console_sem();
2079 		return 0;
2080 	}
2081 	console_locked = 1;
2082 	/*
2083 	 * When PREEMPT_COUNT disabled we can't reliably detect if it's
2084 	 * safe to schedule (e.g. calling printk while holding a spin_lock),
2085 	 * because preempt_disable()/preempt_enable() are just barriers there
2086 	 * and preempt_count() is always 0.
2087 	 *
2088 	 * RCU read sections have a separate preemption counter when
2089 	 * PREEMPT_RCU enabled thus we must take extra care and check
2090 	 * rcu_preempt_depth(), otherwise RCU read sections modify
2091 	 * preempt_count().
2092 	 */
2093 	console_may_schedule = !oops_in_progress &&
2094 			preemptible() &&
2095 			!rcu_preempt_depth();
2096 	return 1;
2097 }
2098 EXPORT_SYMBOL(console_trylock);
2099 
2100 int is_console_locked(void)
2101 {
2102 	return console_locked;
2103 }
2104 
2105 /*
2106  * Check if we have any console that is capable of printing while cpu is
2107  * booting or shutting down. Requires console_sem.
2108  */
2109 static int have_callable_console(void)
2110 {
2111 	struct console *con;
2112 
2113 	for_each_console(con)
2114 		if ((con->flags & CON_ENABLED) &&
2115 				(con->flags & CON_ANYTIME))
2116 			return 1;
2117 
2118 	return 0;
2119 }
2120 
2121 /*
2122  * Can we actually use the console at this time on this cpu?
2123  *
2124  * Console drivers may assume that per-cpu resources have been allocated. So
2125  * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2126  * call them until this CPU is officially up.
2127  */
2128 static inline int can_use_console(void)
2129 {
2130 	return cpu_online(raw_smp_processor_id()) || have_callable_console();
2131 }
2132 
2133 /**
2134  * console_unlock - unlock the console system
2135  *
2136  * Releases the console_lock which the caller holds on the console system
2137  * and the console driver list.
2138  *
2139  * While the console_lock was held, console output may have been buffered
2140  * by printk().  If this is the case, console_unlock(); emits
2141  * the output prior to releasing the lock.
2142  *
2143  * If there is output waiting, we wake /dev/kmsg and syslog() users.
2144  *
2145  * console_unlock(); may be called from any context.
2146  */
2147 void console_unlock(void)
2148 {
2149 	static char ext_text[CONSOLE_EXT_LOG_MAX];
2150 	static char text[LOG_LINE_MAX + PREFIX_MAX];
2151 	static u64 seen_seq;
2152 	unsigned long flags;
2153 	bool wake_klogd = false;
2154 	bool do_cond_resched, retry;
2155 
2156 	if (console_suspended) {
2157 		up_console_sem();
2158 		return;
2159 	}
2160 
2161 	/*
2162 	 * Console drivers are called under logbuf_lock, so
2163 	 * @console_may_schedule should be cleared before; however, we may
2164 	 * end up dumping a lot of lines, for example, if called from
2165 	 * console registration path, and should invoke cond_resched()
2166 	 * between lines if allowable.  Not doing so can cause a very long
2167 	 * scheduling stall on a slow console leading to RCU stall and
2168 	 * softlockup warnings which exacerbate the issue with more
2169 	 * messages practically incapacitating the system.
2170 	 */
2171 	do_cond_resched = console_may_schedule;
2172 	console_may_schedule = 0;
2173 
2174 again:
2175 	/*
2176 	 * We released the console_sem lock, so we need to recheck if
2177 	 * cpu is online and (if not) is there at least one CON_ANYTIME
2178 	 * console.
2179 	 */
2180 	if (!can_use_console()) {
2181 		console_locked = 0;
2182 		up_console_sem();
2183 		return;
2184 	}
2185 
2186 	for (;;) {
2187 		struct printk_log *msg;
2188 		size_t ext_len = 0;
2189 		size_t len;
2190 
2191 		printk_safe_enter_irqsave(flags);
2192 		raw_spin_lock(&logbuf_lock);
2193 		if (seen_seq != log_next_seq) {
2194 			wake_klogd = true;
2195 			seen_seq = log_next_seq;
2196 		}
2197 
2198 		if (console_seq < log_first_seq) {
2199 			len = sprintf(text, "** %u printk messages dropped ** ",
2200 				      (unsigned)(log_first_seq - console_seq));
2201 
2202 			/* messages are gone, move to first one */
2203 			console_seq = log_first_seq;
2204 			console_idx = log_first_idx;
2205 		} else {
2206 			len = 0;
2207 		}
2208 skip:
2209 		if (console_seq == log_next_seq)
2210 			break;
2211 
2212 		msg = log_from_idx(console_idx);
2213 		if (suppress_message_printing(msg->level)) {
2214 			/*
2215 			 * Skip record we have buffered and already printed
2216 			 * directly to the console when we received it, and
2217 			 * record that has level above the console loglevel.
2218 			 */
2219 			console_idx = log_next(console_idx);
2220 			console_seq++;
2221 			goto skip;
2222 		}
2223 
2224 		len += msg_print_text(msg, false, text + len, sizeof(text) - len);
2225 		if (nr_ext_console_drivers) {
2226 			ext_len = msg_print_ext_header(ext_text,
2227 						sizeof(ext_text),
2228 						msg, console_seq);
2229 			ext_len += msg_print_ext_body(ext_text + ext_len,
2230 						sizeof(ext_text) - ext_len,
2231 						log_dict(msg), msg->dict_len,
2232 						log_text(msg), msg->text_len);
2233 		}
2234 		console_idx = log_next(console_idx);
2235 		console_seq++;
2236 		raw_spin_unlock(&logbuf_lock);
2237 
2238 		stop_critical_timings();	/* don't trace print latency */
2239 		call_console_drivers(ext_text, ext_len, text, len);
2240 		start_critical_timings();
2241 		printk_safe_exit_irqrestore(flags);
2242 
2243 		if (do_cond_resched)
2244 			cond_resched();
2245 	}
2246 	console_locked = 0;
2247 
2248 	/* Release the exclusive_console once it is used */
2249 	if (unlikely(exclusive_console))
2250 		exclusive_console = NULL;
2251 
2252 	raw_spin_unlock(&logbuf_lock);
2253 
2254 	up_console_sem();
2255 
2256 	/*
2257 	 * Someone could have filled up the buffer again, so re-check if there's
2258 	 * something to flush. In case we cannot trylock the console_sem again,
2259 	 * there's a new owner and the console_unlock() from them will do the
2260 	 * flush, no worries.
2261 	 */
2262 	raw_spin_lock(&logbuf_lock);
2263 	retry = console_seq != log_next_seq;
2264 	raw_spin_unlock(&logbuf_lock);
2265 	printk_safe_exit_irqrestore(flags);
2266 
2267 	if (retry && console_trylock())
2268 		goto again;
2269 
2270 	if (wake_klogd)
2271 		wake_up_klogd();
2272 }
2273 EXPORT_SYMBOL(console_unlock);
2274 
2275 /**
2276  * console_conditional_schedule - yield the CPU if required
2277  *
2278  * If the console code is currently allowed to sleep, and
2279  * if this CPU should yield the CPU to another task, do
2280  * so here.
2281  *
2282  * Must be called within console_lock();.
2283  */
2284 void __sched console_conditional_schedule(void)
2285 {
2286 	if (console_may_schedule)
2287 		cond_resched();
2288 }
2289 EXPORT_SYMBOL(console_conditional_schedule);
2290 
2291 void console_unblank(void)
2292 {
2293 	struct console *c;
2294 
2295 	/*
2296 	 * console_unblank can no longer be called in interrupt context unless
2297 	 * oops_in_progress is set to 1..
2298 	 */
2299 	if (oops_in_progress) {
2300 		if (down_trylock_console_sem() != 0)
2301 			return;
2302 	} else
2303 		console_lock();
2304 
2305 	console_locked = 1;
2306 	console_may_schedule = 0;
2307 	for_each_console(c)
2308 		if ((c->flags & CON_ENABLED) && c->unblank)
2309 			c->unblank();
2310 	console_unlock();
2311 }
2312 
2313 /**
2314  * console_flush_on_panic - flush console content on panic
2315  *
2316  * Immediately output all pending messages no matter what.
2317  */
2318 void console_flush_on_panic(void)
2319 {
2320 	/*
2321 	 * If someone else is holding the console lock, trylock will fail
2322 	 * and may_schedule may be set.  Ignore and proceed to unlock so
2323 	 * that messages are flushed out.  As this can be called from any
2324 	 * context and we don't want to get preempted while flushing,
2325 	 * ensure may_schedule is cleared.
2326 	 */
2327 	console_trylock();
2328 	console_may_schedule = 0;
2329 	console_unlock();
2330 }
2331 
2332 /*
2333  * Return the console tty driver structure and its associated index
2334  */
2335 struct tty_driver *console_device(int *index)
2336 {
2337 	struct console *c;
2338 	struct tty_driver *driver = NULL;
2339 
2340 	console_lock();
2341 	for_each_console(c) {
2342 		if (!c->device)
2343 			continue;
2344 		driver = c->device(c, index);
2345 		if (driver)
2346 			break;
2347 	}
2348 	console_unlock();
2349 	return driver;
2350 }
2351 
2352 /*
2353  * Prevent further output on the passed console device so that (for example)
2354  * serial drivers can disable console output before suspending a port, and can
2355  * re-enable output afterwards.
2356  */
2357 void console_stop(struct console *console)
2358 {
2359 	console_lock();
2360 	console->flags &= ~CON_ENABLED;
2361 	console_unlock();
2362 }
2363 EXPORT_SYMBOL(console_stop);
2364 
2365 void console_start(struct console *console)
2366 {
2367 	console_lock();
2368 	console->flags |= CON_ENABLED;
2369 	console_unlock();
2370 }
2371 EXPORT_SYMBOL(console_start);
2372 
2373 static int __read_mostly keep_bootcon;
2374 
2375 static int __init keep_bootcon_setup(char *str)
2376 {
2377 	keep_bootcon = 1;
2378 	pr_info("debug: skip boot console de-registration.\n");
2379 
2380 	return 0;
2381 }
2382 
2383 early_param("keep_bootcon", keep_bootcon_setup);
2384 
2385 /*
2386  * The console driver calls this routine during kernel initialization
2387  * to register the console printing procedure with printk() and to
2388  * print any messages that were printed by the kernel before the
2389  * console driver was initialized.
2390  *
2391  * This can happen pretty early during the boot process (because of
2392  * early_printk) - sometimes before setup_arch() completes - be careful
2393  * of what kernel features are used - they may not be initialised yet.
2394  *
2395  * There are two types of consoles - bootconsoles (early_printk) and
2396  * "real" consoles (everything which is not a bootconsole) which are
2397  * handled differently.
2398  *  - Any number of bootconsoles can be registered at any time.
2399  *  - As soon as a "real" console is registered, all bootconsoles
2400  *    will be unregistered automatically.
2401  *  - Once a "real" console is registered, any attempt to register a
2402  *    bootconsoles will be rejected
2403  */
2404 void register_console(struct console *newcon)
2405 {
2406 	int i;
2407 	unsigned long flags;
2408 	struct console *bcon = NULL;
2409 	struct console_cmdline *c;
2410 
2411 	if (console_drivers)
2412 		for_each_console(bcon)
2413 			if (WARN(bcon == newcon,
2414 					"console '%s%d' already registered\n",
2415 					bcon->name, bcon->index))
2416 				return;
2417 
2418 	/*
2419 	 * before we register a new CON_BOOT console, make sure we don't
2420 	 * already have a valid console
2421 	 */
2422 	if (console_drivers && newcon->flags & CON_BOOT) {
2423 		/* find the last or real console */
2424 		for_each_console(bcon) {
2425 			if (!(bcon->flags & CON_BOOT)) {
2426 				pr_info("Too late to register bootconsole %s%d\n",
2427 					newcon->name, newcon->index);
2428 				return;
2429 			}
2430 		}
2431 	}
2432 
2433 	if (console_drivers && console_drivers->flags & CON_BOOT)
2434 		bcon = console_drivers;
2435 
2436 	if (preferred_console < 0 || bcon || !console_drivers)
2437 		preferred_console = selected_console;
2438 
2439 	/*
2440 	 *	See if we want to use this console driver. If we
2441 	 *	didn't select a console we take the first one
2442 	 *	that registers here.
2443 	 */
2444 	if (preferred_console < 0) {
2445 		if (newcon->index < 0)
2446 			newcon->index = 0;
2447 		if (newcon->setup == NULL ||
2448 		    newcon->setup(newcon, NULL) == 0) {
2449 			newcon->flags |= CON_ENABLED;
2450 			if (newcon->device) {
2451 				newcon->flags |= CON_CONSDEV;
2452 				preferred_console = 0;
2453 			}
2454 		}
2455 	}
2456 
2457 	/*
2458 	 *	See if this console matches one we selected on
2459 	 *	the command line.
2460 	 */
2461 	for (i = 0, c = console_cmdline;
2462 	     i < MAX_CMDLINECONSOLES && c->name[0];
2463 	     i++, c++) {
2464 		if (!newcon->match ||
2465 		    newcon->match(newcon, c->name, c->index, c->options) != 0) {
2466 			/* default matching */
2467 			BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name));
2468 			if (strcmp(c->name, newcon->name) != 0)
2469 				continue;
2470 			if (newcon->index >= 0 &&
2471 			    newcon->index != c->index)
2472 				continue;
2473 			if (newcon->index < 0)
2474 				newcon->index = c->index;
2475 
2476 			if (_braille_register_console(newcon, c))
2477 				return;
2478 
2479 			if (newcon->setup &&
2480 			    newcon->setup(newcon, c->options) != 0)
2481 				break;
2482 		}
2483 
2484 		newcon->flags |= CON_ENABLED;
2485 		if (i == selected_console) {
2486 			newcon->flags |= CON_CONSDEV;
2487 			preferred_console = selected_console;
2488 		}
2489 		break;
2490 	}
2491 
2492 	if (!(newcon->flags & CON_ENABLED))
2493 		return;
2494 
2495 	/*
2496 	 * If we have a bootconsole, and are switching to a real console,
2497 	 * don't print everything out again, since when the boot console, and
2498 	 * the real console are the same physical device, it's annoying to
2499 	 * see the beginning boot messages twice
2500 	 */
2501 	if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2502 		newcon->flags &= ~CON_PRINTBUFFER;
2503 
2504 	/*
2505 	 *	Put this console in the list - keep the
2506 	 *	preferred driver at the head of the list.
2507 	 */
2508 	console_lock();
2509 	if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2510 		newcon->next = console_drivers;
2511 		console_drivers = newcon;
2512 		if (newcon->next)
2513 			newcon->next->flags &= ~CON_CONSDEV;
2514 	} else {
2515 		newcon->next = console_drivers->next;
2516 		console_drivers->next = newcon;
2517 	}
2518 
2519 	if (newcon->flags & CON_EXTENDED)
2520 		if (!nr_ext_console_drivers++)
2521 			pr_info("printk: continuation disabled due to ext consoles, expect more fragments in /dev/kmsg\n");
2522 
2523 	if (newcon->flags & CON_PRINTBUFFER) {
2524 		/*
2525 		 * console_unlock(); will print out the buffered messages
2526 		 * for us.
2527 		 */
2528 		logbuf_lock_irqsave(flags);
2529 		console_seq = syslog_seq;
2530 		console_idx = syslog_idx;
2531 		logbuf_unlock_irqrestore(flags);
2532 		/*
2533 		 * We're about to replay the log buffer.  Only do this to the
2534 		 * just-registered console to avoid excessive message spam to
2535 		 * the already-registered consoles.
2536 		 */
2537 		exclusive_console = newcon;
2538 	}
2539 	console_unlock();
2540 	console_sysfs_notify();
2541 
2542 	/*
2543 	 * By unregistering the bootconsoles after we enable the real console
2544 	 * we get the "console xxx enabled" message on all the consoles -
2545 	 * boot consoles, real consoles, etc - this is to ensure that end
2546 	 * users know there might be something in the kernel's log buffer that
2547 	 * went to the bootconsole (that they do not see on the real console)
2548 	 */
2549 	pr_info("%sconsole [%s%d] enabled\n",
2550 		(newcon->flags & CON_BOOT) ? "boot" : "" ,
2551 		newcon->name, newcon->index);
2552 	if (bcon &&
2553 	    ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2554 	    !keep_bootcon) {
2555 		/* We need to iterate through all boot consoles, to make
2556 		 * sure we print everything out, before we unregister them.
2557 		 */
2558 		for_each_console(bcon)
2559 			if (bcon->flags & CON_BOOT)
2560 				unregister_console(bcon);
2561 	}
2562 }
2563 EXPORT_SYMBOL(register_console);
2564 
2565 int unregister_console(struct console *console)
2566 {
2567         struct console *a, *b;
2568 	int res;
2569 
2570 	pr_info("%sconsole [%s%d] disabled\n",
2571 		(console->flags & CON_BOOT) ? "boot" : "" ,
2572 		console->name, console->index);
2573 
2574 	res = _braille_unregister_console(console);
2575 	if (res)
2576 		return res;
2577 
2578 	res = 1;
2579 	console_lock();
2580 	if (console_drivers == console) {
2581 		console_drivers=console->next;
2582 		res = 0;
2583 	} else if (console_drivers) {
2584 		for (a=console_drivers->next, b=console_drivers ;
2585 		     a; b=a, a=b->next) {
2586 			if (a == console) {
2587 				b->next = a->next;
2588 				res = 0;
2589 				break;
2590 			}
2591 		}
2592 	}
2593 
2594 	if (!res && (console->flags & CON_EXTENDED))
2595 		nr_ext_console_drivers--;
2596 
2597 	/*
2598 	 * If this isn't the last console and it has CON_CONSDEV set, we
2599 	 * need to set it on the next preferred console.
2600 	 */
2601 	if (console_drivers != NULL && console->flags & CON_CONSDEV)
2602 		console_drivers->flags |= CON_CONSDEV;
2603 
2604 	console->flags &= ~CON_ENABLED;
2605 	console_unlock();
2606 	console_sysfs_notify();
2607 	return res;
2608 }
2609 EXPORT_SYMBOL(unregister_console);
2610 
2611 /*
2612  * Some boot consoles access data that is in the init section and which will
2613  * be discarded after the initcalls have been run. To make sure that no code
2614  * will access this data, unregister the boot consoles in a late initcall.
2615  *
2616  * If for some reason, such as deferred probe or the driver being a loadable
2617  * module, the real console hasn't registered yet at this point, there will
2618  * be a brief interval in which no messages are logged to the console, which
2619  * makes it difficult to diagnose problems that occur during this time.
2620  *
2621  * To mitigate this problem somewhat, only unregister consoles whose memory
2622  * intersects with the init section. Note that code exists elsewhere to get
2623  * rid of the boot console as soon as the proper console shows up, so there
2624  * won't be side-effects from postponing the removal.
2625  */
2626 static int __init printk_late_init(void)
2627 {
2628 	struct console *con;
2629 	int ret;
2630 
2631 	for_each_console(con) {
2632 		if (!keep_bootcon && con->flags & CON_BOOT) {
2633 			/*
2634 			 * Make sure to unregister boot consoles whose data
2635 			 * resides in the init section before the init section
2636 			 * is discarded. Boot consoles whose data will stick
2637 			 * around will automatically be unregistered when the
2638 			 * proper console replaces them.
2639 			 */
2640 			if (init_section_intersects(con, sizeof(*con)))
2641 				unregister_console(con);
2642 		}
2643 	}
2644 	ret = cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD, "printk:dead", NULL,
2645 					console_cpu_notify);
2646 	WARN_ON(ret < 0);
2647 	ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "printk:online",
2648 					console_cpu_notify, NULL);
2649 	WARN_ON(ret < 0);
2650 	return 0;
2651 }
2652 late_initcall(printk_late_init);
2653 
2654 #if defined CONFIG_PRINTK
2655 /*
2656  * Delayed printk version, for scheduler-internal messages:
2657  */
2658 #define PRINTK_PENDING_WAKEUP	0x01
2659 #define PRINTK_PENDING_OUTPUT	0x02
2660 
2661 static DEFINE_PER_CPU(int, printk_pending);
2662 
2663 static void wake_up_klogd_work_func(struct irq_work *irq_work)
2664 {
2665 	int pending = __this_cpu_xchg(printk_pending, 0);
2666 
2667 	if (pending & PRINTK_PENDING_OUTPUT) {
2668 		/* If trylock fails, someone else is doing the printing */
2669 		if (console_trylock())
2670 			console_unlock();
2671 	}
2672 
2673 	if (pending & PRINTK_PENDING_WAKEUP)
2674 		wake_up_interruptible(&log_wait);
2675 }
2676 
2677 static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
2678 	.func = wake_up_klogd_work_func,
2679 	.flags = IRQ_WORK_LAZY,
2680 };
2681 
2682 void wake_up_klogd(void)
2683 {
2684 	preempt_disable();
2685 	if (waitqueue_active(&log_wait)) {
2686 		this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
2687 		irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2688 	}
2689 	preempt_enable();
2690 }
2691 
2692 int printk_deferred(const char *fmt, ...)
2693 {
2694 	va_list args;
2695 	int r;
2696 
2697 	preempt_disable();
2698 	va_start(args, fmt);
2699 	r = vprintk_emit(0, LOGLEVEL_SCHED, NULL, 0, fmt, args);
2700 	va_end(args);
2701 
2702 	__this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
2703 	irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2704 	preempt_enable();
2705 
2706 	return r;
2707 }
2708 
2709 /*
2710  * printk rate limiting, lifted from the networking subsystem.
2711  *
2712  * This enforces a rate limit: not more than 10 kernel messages
2713  * every 5s to make a denial-of-service attack impossible.
2714  */
2715 DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2716 
2717 int __printk_ratelimit(const char *func)
2718 {
2719 	return ___ratelimit(&printk_ratelimit_state, func);
2720 }
2721 EXPORT_SYMBOL(__printk_ratelimit);
2722 
2723 /**
2724  * printk_timed_ratelimit - caller-controlled printk ratelimiting
2725  * @caller_jiffies: pointer to caller's state
2726  * @interval_msecs: minimum interval between prints
2727  *
2728  * printk_timed_ratelimit() returns true if more than @interval_msecs
2729  * milliseconds have elapsed since the last time printk_timed_ratelimit()
2730  * returned true.
2731  */
2732 bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2733 			unsigned int interval_msecs)
2734 {
2735 	unsigned long elapsed = jiffies - *caller_jiffies;
2736 
2737 	if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs))
2738 		return false;
2739 
2740 	*caller_jiffies = jiffies;
2741 	return true;
2742 }
2743 EXPORT_SYMBOL(printk_timed_ratelimit);
2744 
2745 static DEFINE_SPINLOCK(dump_list_lock);
2746 static LIST_HEAD(dump_list);
2747 
2748 /**
2749  * kmsg_dump_register - register a kernel log dumper.
2750  * @dumper: pointer to the kmsg_dumper structure
2751  *
2752  * Adds a kernel log dumper to the system. The dump callback in the
2753  * structure will be called when the kernel oopses or panics and must be
2754  * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2755  */
2756 int kmsg_dump_register(struct kmsg_dumper *dumper)
2757 {
2758 	unsigned long flags;
2759 	int err = -EBUSY;
2760 
2761 	/* The dump callback needs to be set */
2762 	if (!dumper->dump)
2763 		return -EINVAL;
2764 
2765 	spin_lock_irqsave(&dump_list_lock, flags);
2766 	/* Don't allow registering multiple times */
2767 	if (!dumper->registered) {
2768 		dumper->registered = 1;
2769 		list_add_tail_rcu(&dumper->list, &dump_list);
2770 		err = 0;
2771 	}
2772 	spin_unlock_irqrestore(&dump_list_lock, flags);
2773 
2774 	return err;
2775 }
2776 EXPORT_SYMBOL_GPL(kmsg_dump_register);
2777 
2778 /**
2779  * kmsg_dump_unregister - unregister a kmsg dumper.
2780  * @dumper: pointer to the kmsg_dumper structure
2781  *
2782  * Removes a dump device from the system. Returns zero on success and
2783  * %-EINVAL otherwise.
2784  */
2785 int kmsg_dump_unregister(struct kmsg_dumper *dumper)
2786 {
2787 	unsigned long flags;
2788 	int err = -EINVAL;
2789 
2790 	spin_lock_irqsave(&dump_list_lock, flags);
2791 	if (dumper->registered) {
2792 		dumper->registered = 0;
2793 		list_del_rcu(&dumper->list);
2794 		err = 0;
2795 	}
2796 	spin_unlock_irqrestore(&dump_list_lock, flags);
2797 	synchronize_rcu();
2798 
2799 	return err;
2800 }
2801 EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
2802 
2803 static bool always_kmsg_dump;
2804 module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
2805 
2806 /**
2807  * kmsg_dump - dump kernel log to kernel message dumpers.
2808  * @reason: the reason (oops, panic etc) for dumping
2809  *
2810  * Call each of the registered dumper's dump() callback, which can
2811  * retrieve the kmsg records with kmsg_dump_get_line() or
2812  * kmsg_dump_get_buffer().
2813  */
2814 void kmsg_dump(enum kmsg_dump_reason reason)
2815 {
2816 	struct kmsg_dumper *dumper;
2817 	unsigned long flags;
2818 
2819 	if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
2820 		return;
2821 
2822 	rcu_read_lock();
2823 	list_for_each_entry_rcu(dumper, &dump_list, list) {
2824 		if (dumper->max_reason && reason > dumper->max_reason)
2825 			continue;
2826 
2827 		/* initialize iterator with data about the stored records */
2828 		dumper->active = true;
2829 
2830 		logbuf_lock_irqsave(flags);
2831 		dumper->cur_seq = clear_seq;
2832 		dumper->cur_idx = clear_idx;
2833 		dumper->next_seq = log_next_seq;
2834 		dumper->next_idx = log_next_idx;
2835 		logbuf_unlock_irqrestore(flags);
2836 
2837 		/* invoke dumper which will iterate over records */
2838 		dumper->dump(dumper, reason);
2839 
2840 		/* reset iterator */
2841 		dumper->active = false;
2842 	}
2843 	rcu_read_unlock();
2844 }
2845 
2846 /**
2847  * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
2848  * @dumper: registered kmsg dumper
2849  * @syslog: include the "<4>" prefixes
2850  * @line: buffer to copy the line to
2851  * @size: maximum size of the buffer
2852  * @len: length of line placed into buffer
2853  *
2854  * Start at the beginning of the kmsg buffer, with the oldest kmsg
2855  * record, and copy one record into the provided buffer.
2856  *
2857  * Consecutive calls will return the next available record moving
2858  * towards the end of the buffer with the youngest messages.
2859  *
2860  * A return value of FALSE indicates that there are no more records to
2861  * read.
2862  *
2863  * The function is similar to kmsg_dump_get_line(), but grabs no locks.
2864  */
2865 bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
2866 			       char *line, size_t size, size_t *len)
2867 {
2868 	struct printk_log *msg;
2869 	size_t l = 0;
2870 	bool ret = false;
2871 
2872 	if (!dumper->active)
2873 		goto out;
2874 
2875 	if (dumper->cur_seq < log_first_seq) {
2876 		/* messages are gone, move to first available one */
2877 		dumper->cur_seq = log_first_seq;
2878 		dumper->cur_idx = log_first_idx;
2879 	}
2880 
2881 	/* last entry */
2882 	if (dumper->cur_seq >= log_next_seq)
2883 		goto out;
2884 
2885 	msg = log_from_idx(dumper->cur_idx);
2886 	l = msg_print_text(msg, syslog, line, size);
2887 
2888 	dumper->cur_idx = log_next(dumper->cur_idx);
2889 	dumper->cur_seq++;
2890 	ret = true;
2891 out:
2892 	if (len)
2893 		*len = l;
2894 	return ret;
2895 }
2896 
2897 /**
2898  * kmsg_dump_get_line - retrieve one kmsg log line
2899  * @dumper: registered kmsg dumper
2900  * @syslog: include the "<4>" prefixes
2901  * @line: buffer to copy the line to
2902  * @size: maximum size of the buffer
2903  * @len: length of line placed into buffer
2904  *
2905  * Start at the beginning of the kmsg buffer, with the oldest kmsg
2906  * record, and copy one record into the provided buffer.
2907  *
2908  * Consecutive calls will return the next available record moving
2909  * towards the end of the buffer with the youngest messages.
2910  *
2911  * A return value of FALSE indicates that there are no more records to
2912  * read.
2913  */
2914 bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
2915 			char *line, size_t size, size_t *len)
2916 {
2917 	unsigned long flags;
2918 	bool ret;
2919 
2920 	logbuf_lock_irqsave(flags);
2921 	ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
2922 	logbuf_unlock_irqrestore(flags);
2923 
2924 	return ret;
2925 }
2926 EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
2927 
2928 /**
2929  * kmsg_dump_get_buffer - copy kmsg log lines
2930  * @dumper: registered kmsg dumper
2931  * @syslog: include the "<4>" prefixes
2932  * @buf: buffer to copy the line to
2933  * @size: maximum size of the buffer
2934  * @len: length of line placed into buffer
2935  *
2936  * Start at the end of the kmsg buffer and fill the provided buffer
2937  * with as many of the the *youngest* kmsg records that fit into it.
2938  * If the buffer is large enough, all available kmsg records will be
2939  * copied with a single call.
2940  *
2941  * Consecutive calls will fill the buffer with the next block of
2942  * available older records, not including the earlier retrieved ones.
2943  *
2944  * A return value of FALSE indicates that there are no more records to
2945  * read.
2946  */
2947 bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
2948 			  char *buf, size_t size, size_t *len)
2949 {
2950 	unsigned long flags;
2951 	u64 seq;
2952 	u32 idx;
2953 	u64 next_seq;
2954 	u32 next_idx;
2955 	size_t l = 0;
2956 	bool ret = false;
2957 
2958 	if (!dumper->active)
2959 		goto out;
2960 
2961 	logbuf_lock_irqsave(flags);
2962 	if (dumper->cur_seq < log_first_seq) {
2963 		/* messages are gone, move to first available one */
2964 		dumper->cur_seq = log_first_seq;
2965 		dumper->cur_idx = log_first_idx;
2966 	}
2967 
2968 	/* last entry */
2969 	if (dumper->cur_seq >= dumper->next_seq) {
2970 		logbuf_unlock_irqrestore(flags);
2971 		goto out;
2972 	}
2973 
2974 	/* calculate length of entire buffer */
2975 	seq = dumper->cur_seq;
2976 	idx = dumper->cur_idx;
2977 	while (seq < dumper->next_seq) {
2978 		struct printk_log *msg = log_from_idx(idx);
2979 
2980 		l += msg_print_text(msg, true, NULL, 0);
2981 		idx = log_next(idx);
2982 		seq++;
2983 	}
2984 
2985 	/* move first record forward until length fits into the buffer */
2986 	seq = dumper->cur_seq;
2987 	idx = dumper->cur_idx;
2988 	while (l > size && seq < dumper->next_seq) {
2989 		struct printk_log *msg = log_from_idx(idx);
2990 
2991 		l -= msg_print_text(msg, true, NULL, 0);
2992 		idx = log_next(idx);
2993 		seq++;
2994 	}
2995 
2996 	/* last message in next interation */
2997 	next_seq = seq;
2998 	next_idx = idx;
2999 
3000 	l = 0;
3001 	while (seq < dumper->next_seq) {
3002 		struct printk_log *msg = log_from_idx(idx);
3003 
3004 		l += msg_print_text(msg, syslog, buf + l, size - l);
3005 		idx = log_next(idx);
3006 		seq++;
3007 	}
3008 
3009 	dumper->next_seq = next_seq;
3010 	dumper->next_idx = next_idx;
3011 	ret = true;
3012 	logbuf_unlock_irqrestore(flags);
3013 out:
3014 	if (len)
3015 		*len = l;
3016 	return ret;
3017 }
3018 EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
3019 
3020 /**
3021  * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3022  * @dumper: registered kmsg dumper
3023  *
3024  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3025  * kmsg_dump_get_buffer() can be called again and used multiple
3026  * times within the same dumper.dump() callback.
3027  *
3028  * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3029  */
3030 void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
3031 {
3032 	dumper->cur_seq = clear_seq;
3033 	dumper->cur_idx = clear_idx;
3034 	dumper->next_seq = log_next_seq;
3035 	dumper->next_idx = log_next_idx;
3036 }
3037 
3038 /**
3039  * kmsg_dump_rewind - reset the interator
3040  * @dumper: registered kmsg dumper
3041  *
3042  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3043  * kmsg_dump_get_buffer() can be called again and used multiple
3044  * times within the same dumper.dump() callback.
3045  */
3046 void kmsg_dump_rewind(struct kmsg_dumper *dumper)
3047 {
3048 	unsigned long flags;
3049 
3050 	logbuf_lock_irqsave(flags);
3051 	kmsg_dump_rewind_nolock(dumper);
3052 	logbuf_unlock_irqrestore(flags);
3053 }
3054 EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
3055 
3056 static char dump_stack_arch_desc_str[128];
3057 
3058 /**
3059  * dump_stack_set_arch_desc - set arch-specific str to show with task dumps
3060  * @fmt: printf-style format string
3061  * @...: arguments for the format string
3062  *
3063  * The configured string will be printed right after utsname during task
3064  * dumps.  Usually used to add arch-specific system identifiers.  If an
3065  * arch wants to make use of such an ID string, it should initialize this
3066  * as soon as possible during boot.
3067  */
3068 void __init dump_stack_set_arch_desc(const char *fmt, ...)
3069 {
3070 	va_list args;
3071 
3072 	va_start(args, fmt);
3073 	vsnprintf(dump_stack_arch_desc_str, sizeof(dump_stack_arch_desc_str),
3074 		  fmt, args);
3075 	va_end(args);
3076 }
3077 
3078 /**
3079  * dump_stack_print_info - print generic debug info for dump_stack()
3080  * @log_lvl: log level
3081  *
3082  * Arch-specific dump_stack() implementations can use this function to
3083  * print out the same debug information as the generic dump_stack().
3084  */
3085 void dump_stack_print_info(const char *log_lvl)
3086 {
3087 	printk("%sCPU: %d PID: %d Comm: %.20s %s %s %.*s\n",
3088 	       log_lvl, raw_smp_processor_id(), current->pid, current->comm,
3089 	       print_tainted(), init_utsname()->release,
3090 	       (int)strcspn(init_utsname()->version, " "),
3091 	       init_utsname()->version);
3092 
3093 	if (dump_stack_arch_desc_str[0] != '\0')
3094 		printk("%sHardware name: %s\n",
3095 		       log_lvl, dump_stack_arch_desc_str);
3096 
3097 	print_worker_info(log_lvl, current);
3098 }
3099 
3100 /**
3101  * show_regs_print_info - print generic debug info for show_regs()
3102  * @log_lvl: log level
3103  *
3104  * show_regs() implementations can use this function to print out generic
3105  * debug information.
3106  */
3107 void show_regs_print_info(const char *log_lvl)
3108 {
3109 	dump_stack_print_info(log_lvl);
3110 
3111 	printk("%stask: %p task.stack: %p\n",
3112 	       log_lvl, current, task_stack_page(current));
3113 }
3114 
3115 #endif
3116