path: root/kernel/printk/printk.c (plain)
blob: 8e236155157cba54c4c3b38c6742fe192aca9421

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
49	#include <asm/uaccess.h>
50	#include <asm/sections.h>
51
52	#define CREATE_TRACE_POINTS
53	#include <trace/events/printk.h>
54
55	#include "console_cmdline.h"
56	#include "braille.h"
57	#include "internal.h"
58
59	#ifdef CONFIG_EARLY_PRINTK_DIRECT
60	extern void printascii(char *);
61	#endif
62
63	int console_printk[4] = {
64	CONSOLE_LOGLEVEL_DEFAULT, /* console_loglevel */
65	MESSAGE_LOGLEVEL_DEFAULT, /* default_message_loglevel */
66	CONSOLE_LOGLEVEL_MIN, /* minimum_console_loglevel */
67	CONSOLE_LOGLEVEL_DEFAULT, /* default_console_loglevel */
68	};
69
70	/*
71	* Low level drivers may need that to know if they can schedule in
72	* their unblank() callback or not. So let's export it.
73	*/
74	int oops_in_progress;
75	EXPORT_SYMBOL(oops_in_progress);
76
77	/*
78	* console_sem protects the console_drivers list, and also
79	* provides serialisation for access to the entire console
80	* driver system.
81	*/
82	static DEFINE_SEMAPHORE(console_sem);
83	struct console *console_drivers;
84	EXPORT_SYMBOL_GPL(console_drivers);
85
86	#ifdef CONFIG_LOCKDEP
87	static struct lockdep_map console_lock_dep_map = {
88	.name = "console_lock"
89	};
90	#endif
91
92	enum devkmsg_log_bits {
93	__DEVKMSG_LOG_BIT_ON = 0,
94	__DEVKMSG_LOG_BIT_OFF,
95	__DEVKMSG_LOG_BIT_LOCK,
96	};
97
98	enum devkmsg_log_masks {
99	DEVKMSG_LOG_MASK_ON = BIT(__DEVKMSG_LOG_BIT_ON),
100	DEVKMSG_LOG_MASK_OFF = BIT(__DEVKMSG_LOG_BIT_OFF),
101	DEVKMSG_LOG_MASK_LOCK = BIT(__DEVKMSG_LOG_BIT_LOCK),
102	};
103
104	/* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
105	#define DEVKMSG_LOG_MASK_DEFAULT 0
106
107	static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
108
109	static int __control_devkmsg(char *str)
110	{
111	if (!str)
112	return -EINVAL;
113
114	if (!strncmp(str, "on", 2)) {
115	devkmsg_log = DEVKMSG_LOG_MASK_ON;
116	return 2;
117	} else if (!strncmp(str, "off", 3)) {
118	devkmsg_log = DEVKMSG_LOG_MASK_OFF;
119	return 3;
120	} else if (!strncmp(str, "ratelimit", 9)) {
121	devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
122	return 9;
123	}
124	return -EINVAL;
125	}
126
127	static int __init control_devkmsg(char *str)
128	{
129	if (__control_devkmsg(str) < 0)
130	return 1;
131
132	/*
133	* Set sysctl string accordingly:
134	*/
135	if (devkmsg_log == DEVKMSG_LOG_MASK_ON) {
136	memset(devkmsg_log_str, 0, DEVKMSG_STR_MAX_SIZE);
137	strncpy(devkmsg_log_str, "on", 2);
138	} else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF) {
139	memset(devkmsg_log_str, 0, DEVKMSG_STR_MAX_SIZE);
140	strncpy(devkmsg_log_str, "off", 3);
141	}
142	/* else "ratelimit" which is set by default. */
143
144	/*
145	* Sysctl cannot change it anymore. The kernel command line setting of
146	* this parameter is to force the setting to be permanent throughout the
147	* runtime of the system. This is a precation measure against userspace
148	* trying to be a smarta** and attempting to change it up on us.
149	*/
150	devkmsg_log |= DEVKMSG_LOG_MASK_LOCK;
151
152	return 0;
153	}
154	__setup("printk.devkmsg=", control_devkmsg);
155
156	char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit";
157
158	int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
159	void __user *buffer, size_t *lenp, loff_t *ppos)
160	{
161	char old_str[DEVKMSG_STR_MAX_SIZE];
162	unsigned int old;
163	int err;
164
165	if (write) {
166	if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK)
167	return -EINVAL;
168
169	old = devkmsg_log;
170	strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE);
171	}
172
173	err = proc_dostring(table, write, buffer, lenp, ppos);
174	if (err)
175	return err;
176
177	if (write) {
178	err = __control_devkmsg(devkmsg_log_str);
179
180	/*
181	* Do not accept an unknown string OR a known string with
182	* trailing crap...
183	*/
184	if (err < 0 || (err + 1 != *lenp)) {
185
186	/* ... and restore old setting. */
187	devkmsg_log = old;
188	strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE);
189
190	return -EINVAL;
191	}
192	}
193
194	return 0;
195	}
196
197	/*
198	* Number of registered extended console drivers.
199	*
200	* If extended consoles are present, in-kernel cont reassembly is disabled
201	* and each fragment is stored as a separate log entry with proper
202	* continuation flag so that every emitted message has full metadata. This
203	* doesn't change the result for regular consoles or /proc/kmsg. For
204	* /dev/kmsg, as long as the reader concatenates messages according to
205	* consecutive continuation flags, the end result should be the same too.
206	*/
207	static int nr_ext_console_drivers;
208
209	/*
210	* Helper macros to handle lockdep when locking/unlocking console_sem. We use
211	* macros instead of functions so that _RET_IP_ contains useful information.
212	*/
213	#define down_console_sem() do { \
214	down(&console_sem);\
215	mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
216	} while (0)
217
218	static int __down_trylock_console_sem(unsigned long ip)
219	{
220	if (down_trylock(&console_sem))
221	return 1;
222	mutex_acquire(&console_lock_dep_map, 0, 1, ip);
223	return 0;
224	}
225	#define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
226
227	#define up_console_sem() do { \
228	mutex_release(&console_lock_dep_map, 1, _RET_IP_);\
229	up(&console_sem);\
230	} while (0)
231
232	/*
233	* This is used for debugging the mess that is the VT code by
234	* keeping track if we have the console semaphore held. It's
235	* definitely not the perfect debug tool (we don't know if _WE_
236	* hold it and are racing, but it helps tracking those weird code
237	* paths in the console code where we end up in places I want
238	* locked without the console sempahore held).
239	*/
240	static int console_locked, console_suspended;
241
242	/*
243	* If exclusive_console is non-NULL then only this console is to be printed to.
244	*/
245	static struct console *exclusive_console;
246
247	/*
248	* Array of consoles built from command line options (console=)
249	*/
250
251	#define MAX_CMDLINECONSOLES 8
252
253	static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
254
255	static int selected_console = -1;
256	static int preferred_console = -1;
257	int console_set_on_cmdline;
258	EXPORT_SYMBOL(console_set_on_cmdline);
259
260	/* Flag: console code may call schedule() */
261	static int console_may_schedule;
262
263	/*
264	* The printk log buffer consists of a chain of concatenated variable
265	* length records. Every record starts with a record header, containing
266	* the overall length of the record.
267	*
268	* The heads to the first and last entry in the buffer, as well as the
269	* sequence numbers of these entries are maintained when messages are
270	* stored.
271	*
272	* If the heads indicate available messages, the length in the header
273	* tells the start next message. A length == 0 for the next message
274	* indicates a wrap-around to the beginning of the buffer.
275	*
276	* Every record carries the monotonic timestamp in microseconds, as well as
277	* the standard userspace syslog level and syslog facility. The usual
278	* kernel messages use LOG_KERN; userspace-injected messages always carry
279	* a matching syslog facility, by default LOG_USER. The origin of every
280	* message can be reliably determined that way.
281	*
282	* The human readable log message directly follows the message header. The
283	* length of the message text is stored in the header, the stored message
284	* is not terminated.
285	*
286	* Optionally, a message can carry a dictionary of properties (key/value pairs),
287	* to provide userspace with a machine-readable message context.
288	*
289	* Examples for well-defined, commonly used property names are:
290	* DEVICE=b12:8 device identifier
291	* b12:8 block dev_t
292	* c127:3 char dev_t
293	* n8 netdev ifindex
294	* +sound:card0 subsystem:devname
295	* SUBSYSTEM=pci driver-core subsystem name
296	*
297	* Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
298	* follows directly after a '=' character. Every property is terminated by
299	* a '\0' character. The last property is not terminated.
300	*
301	* Example of a message structure:
302	* 0000 ff 8f 00 00 00 00 00 00 monotonic time in nsec
303	* 0008 34 00 record is 52 bytes long
304	* 000a 0b 00 text is 11 bytes long
305	* 000c 1f 00 dictionary is 23 bytes long
306	* 000e 03 00 LOG_KERN (facility) LOG_ERR (level)
307	* 0010 69 74 27 73 20 61 20 6c "it's a l"
308	* 69 6e 65 "ine"
309	* 001b 44 45 56 49 43 "DEVIC"
310	* 45 3d 62 38 3a 32 00 44 "E=b8:2\0D"
311	* 52 49 56 45 52 3d 62 75 "RIVER=bu"
312	* 67 "g"
313	* 0032 00 00 00 padding to next message header
314	*
315	* The 'struct printk_log' buffer header must never be directly exported to
316	* userspace, it is a kernel-private implementation detail that might
317	* need to be changed in the future, when the requirements change.
318	*
319	* /dev/kmsg exports the structured data in the following line format:
320	* "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
321	*
322	* Users of the export format should ignore possible additional values
323	* separated by ',', and find the message after the ';' character.
324	*
325	* The optional key/value pairs are attached as continuation lines starting
326	* with a space character and terminated by a newline. All possible
327	* non-prinatable characters are escaped in the "\xff" notation.
328	*/
329
330	enum log_flags {
331	LOG_NOCONS = 1, /* already flushed, do not print to console */
332	LOG_NEWLINE = 2, /* text ended with a newline */
333	LOG_PREFIX = 4, /* text started with a prefix */
334	LOG_CONT = 8, /* text is a fragment of a continuation line */
335	};
336
337	struct printk_log {
338	u64 ts_nsec; /* timestamp in nanoseconds */
339	u16 len; /* length of entire record */
340	u16 text_len; /* length of text buffer */
341	u16 dict_len; /* length of dictionary buffer */
342	u8 facility; /* syslog facility */
343	u8 flags:5; /* internal record flags */
344	u8 level:3; /* syslog level */
345	#ifdef CONFIG_AMLOGIC_DRIVER
346	int cpu;
347	#endif
348	}
349	#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
350	__packed __aligned(4)
351	#endif
352	;
353
354	/*
355	* The logbuf_lock protects kmsg buffer, indices, counters. This can be taken
356	* within the scheduler's rq lock. It must be released before calling
357	* console_unlock() or anything else that might wake up a process.
358	*/
359	DEFINE_RAW_SPINLOCK(logbuf_lock);
360
361	#ifdef CONFIG_PRINTK
362	DECLARE_WAIT_QUEUE_HEAD(log_wait);
363	#ifdef CONFIG_AMLOGIC_DRIVER
364	static int current_cpu;
365	#endif
366	/* the next printk record to read by syslog(READ) or /proc/kmsg */
367	static u64 syslog_seq;
368	static u32 syslog_idx;
369	static enum log_flags syslog_prev;
370	static size_t syslog_partial;
371
372	/* index and sequence number of the first record stored in the buffer */
373	static u64 log_first_seq;
374	static u32 log_first_idx;
375
376	/* index and sequence number of the next record to store in the buffer */
377	static u64 log_next_seq;
378	static u32 log_next_idx;
379
380	/* the next printk record to write to the console */
381	static u64 console_seq;
382	static u32 console_idx;
383	static enum log_flags console_prev;
384
385	/* the next printk record to read after the last 'clear' command */
386	static u64 clear_seq;
387	static u32 clear_idx;
388
389	#define PREFIX_MAX 32
390	#define LOG_LINE_MAX (1024 - PREFIX_MAX)
391
392	#define LOG_LEVEL(v) ((v) & 0x07)
393	#define LOG_FACILITY(v) ((v) >> 3 & 0xff)
394
395	/* record buffer */
396	#define LOG_ALIGN __alignof__(struct printk_log)
397	#define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
398	static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
399	static char *log_buf = __log_buf;
400	static u32 log_buf_len = __LOG_BUF_LEN;
401
402	/* Return log buffer address */
403	char *log_buf_addr_get(void)
404	{
405	return log_buf;
406	}
407
408	/* Return log buffer size */
409	u32 log_buf_len_get(void)
410	{
411	return log_buf_len;
412	}
413
414	/* human readable text of the record */
415	static char *log_text(const struct printk_log *msg)
416	{
417	return (char *)msg + sizeof(struct printk_log);
418	}
419
420	/* optional key/value pair dictionary attached to the record */
421	static char *log_dict(const struct printk_log *msg)
422	{
423	return (char *)msg + sizeof(struct printk_log) + msg->text_len;
424	}
425
426	/* get record by index; idx must point to valid msg */
427	static struct printk_log *log_from_idx(u32 idx)
428	{
429	struct printk_log *msg = (struct printk_log *)(log_buf + idx);
430
431	/*
432	* A length == 0 record is the end of buffer marker. Wrap around and
433	* read the message at the start of the buffer.
434	*/
435	if (!msg->len)
436	return (struct printk_log *)log_buf;
437	return msg;
438	}
439
440	/* get next record; idx must point to valid msg */
441	static u32 log_next(u32 idx)
442	{
443	struct printk_log *msg = (struct printk_log *)(log_buf + idx);
444
445	/* length == 0 indicates the end of the buffer; wrap */
446	/*
447	* A length == 0 record is the end of buffer marker. Wrap around and
448	* read the message at the start of the buffer as *this* one, and
449	* return the one after that.
450	*/
451	if (!msg->len) {
452	msg = (struct printk_log *)log_buf;
453	return msg->len;
454	}
455	return idx + msg->len;
456	}
457
458	/*
459	* Check whether there is enough free space for the given message.
460	*
461	* The same values of first_idx and next_idx mean that the buffer
462	* is either empty or full.
463	*
464	* If the buffer is empty, we must respect the position of the indexes.
465	* They cannot be reset to the beginning of the buffer.
466	*/
467	static int logbuf_has_space(u32 msg_size, bool empty)
468	{
469	u32 free;
470
471	if (log_next_idx > log_first_idx || empty)
472	free = max(log_buf_len - log_next_idx, log_first_idx);
473	else
474	free = log_first_idx - log_next_idx;
475
476	/*
477	* We need space also for an empty header that signalizes wrapping
478	* of the buffer.
479	*/
480	return free >= msg_size + sizeof(struct printk_log);
481	}
482
483	static int log_make_free_space(u32 msg_size)
484	{
485	while (log_first_seq < log_next_seq &&
486	!logbuf_has_space(msg_size, false)) {
487	/* drop old messages until we have enough contiguous space */
488	log_first_idx = log_next(log_first_idx);
489	log_first_seq++;
490	}
491
492	if (clear_seq < log_first_seq) {
493	clear_seq = log_first_seq;
494	clear_idx = log_first_idx;
495	}
496
497	/* sequence numbers are equal, so the log buffer is empty */
498	if (logbuf_has_space(msg_size, log_first_seq == log_next_seq))
499	return 0;
500
501	return -ENOMEM;
502	}
503
504	/* compute the message size including the padding bytes */
505	static u32 msg_used_size(u16 text_len, u16 dict_len, u32 *pad_len)
506	{
507	u32 size;
508
509	size = sizeof(struct printk_log) + text_len + dict_len;
510	*pad_len = (-size) & (LOG_ALIGN - 1);
511	size += *pad_len;
512
513	return size;
514	}
515
516	/*
517	* Define how much of the log buffer we could take at maximum. The value
518	* must be greater than two. Note that only half of the buffer is available
519	* when the index points to the middle.
520	*/
521	#define MAX_LOG_TAKE_PART 4
522	static const char trunc_msg[] = "<truncated>";
523
524	static u32 truncate_msg(u16 *text_len, u16 *trunc_msg_len,
525	u16 *dict_len, u32 *pad_len)
526	{
527	/*
528	* The message should not take the whole buffer. Otherwise, it might
529	* get removed too soon.
530	*/
531	u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART;
532	if (*text_len > max_text_len)
533	*text_len = max_text_len;
534	/* enable the warning message */
535	*trunc_msg_len = strlen(trunc_msg);
536	/* disable the "dict" completely */
537	*dict_len = 0;
538	/* compute the size again, count also the warning message */
539	return msg_used_size(*text_len + *trunc_msg_len, 0, pad_len);
540	}
541
542	/* insert record into the buffer, discard old ones, update heads */
543	static int log_store(int facility, int level,
544	enum log_flags flags, u64 ts_nsec,
545	const char *dict, u16 dict_len,
546	const char *text, u16 text_len)
547	{
548	struct printk_log *msg;
549	u32 size, pad_len;
550	u16 trunc_msg_len = 0;
551
552	/* number of '\0' padding bytes to next message */
553	size = msg_used_size(text_len, dict_len, &pad_len);
554
555	if (log_make_free_space(size)) {
556	/* truncate the message if it is too long for empty buffer */
557	size = truncate_msg(&text_len, &trunc_msg_len,
558	&dict_len, &pad_len);
559	/* survive when the log buffer is too small for trunc_msg */
560	if (log_make_free_space(size))
561	return 0;
562	}
563
564	if (log_next_idx + size + sizeof(struct printk_log) > log_buf_len) {
565	/*
566	* This message + an additional empty header does not fit
567	* at the end of the buffer. Add an empty header with len == 0
568	* to signify a wrap around.
569	*/
570	memset(log_buf + log_next_idx, 0, sizeof(struct printk_log));
571	log_next_idx = 0;
572	}
573
574	/* fill message */
575	msg = (struct printk_log *)(log_buf + log_next_idx);
576	memcpy(log_text(msg), text, text_len);
577	msg->text_len = text_len;
578	if (trunc_msg_len) {
579	memcpy(log_text(msg) + text_len, trunc_msg, trunc_msg_len);
580	msg->text_len += trunc_msg_len;
581	}
582	memcpy(log_dict(msg), dict, dict_len);
583	msg->dict_len = dict_len;
584	msg->facility = facility;
585	msg->level = level & 7;
586	msg->flags = flags & 0x1f;
587	#ifdef CONFIG_AMLOGIC_DRIVER
588	msg->cpu = smp_processor_id();
589	#endif
590	if (ts_nsec > 0)
591	msg->ts_nsec = ts_nsec;
592	else
593	msg->ts_nsec = local_clock();
594	memset(log_dict(msg) + dict_len, 0, pad_len);
595	msg->len = size;
596
597	/* insert message */
598	log_next_idx += msg->len;
599	log_next_seq++;
600
601	return msg->text_len;
602	}
603
604	int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT);
605
606	static int syslog_action_restricted(int type)
607	{
608	if (dmesg_restrict)
609	return 1;
610	/*
611	* Unless restricted, we allow "read all" and "get buffer size"
612	* for everybody.
613	*/
614	return type != SYSLOG_ACTION_READ_ALL &&
615	type != SYSLOG_ACTION_SIZE_BUFFER;
616	}
617
618	int check_syslog_permissions(int type, int source)
619	{
620	/*
621	* If this is from /proc/kmsg and we've already opened it, then we've
622	* already done the capabilities checks at open time.
623	*/
624	if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN)
625	goto ok;
626
627	if (syslog_action_restricted(type)) {
628	if (capable(CAP_SYSLOG))
629	goto ok;
630	/*
631	* For historical reasons, accept CAP_SYS_ADMIN too, with
632	* a warning.
633	*/
634	if (capable(CAP_SYS_ADMIN)) {
635	pr_warn_once("%s (%d): Attempt to access syslog with "
636	"CAP_SYS_ADMIN but no CAP_SYSLOG "
637	"(deprecated).\n",
638	current->comm, task_pid_nr(current));
639	goto ok;
640	}
641	return -EPERM;
642	}
643	ok:
644	return security_syslog(type);
645	}
646	EXPORT_SYMBOL_GPL(check_syslog_permissions);
647
648	static void append_char(char **pp, char *e, char c)
649	{
650	if (*pp < e)
651	*(*pp)++ = c;
652	}
653
654	static ssize_t msg_print_ext_header(char *buf, size_t size,
655	struct printk_log *msg, u64 seq,
656	enum log_flags prev_flags)
657	{
658	u64 ts_usec = msg->ts_nsec;
659	char cont = '-';
660
661	do_div(ts_usec, 1000);
662
663	/*
664	* If we couldn't merge continuation line fragments during the print,
665	* export the stored flags to allow an optional external merge of the
666	* records. Merging the records isn't always neccessarily correct, like
667	* when we hit a race during printing. In most cases though, it produces
668	* better readable output. 'c' in the record flags mark the first
669	* fragment of a line, '+' the following.
670	*/
671	if (msg->flags & LOG_CONT)
672	cont = (prev_flags & LOG_CONT) ? '+' : 'c';
673
674	return scnprintf(buf, size, "%u,%llu,%llu,%c;",
675	(msg->facility << 3) | msg->level, seq, ts_usec, cont);
676	}
677
678	static ssize_t msg_print_ext_body(char *buf, size_t size,
679	char *dict, size_t dict_len,
680	char *text, size_t text_len)
681	{
682	char *p = buf, *e = buf + size;
683	size_t i;
684
685	/* escape non-printable characters */
686	for (i = 0; i < text_len; i++) {
687	unsigned char c = text[i];
688
689	if (c < ' ' || c >= 127 || c == '\\')
690	p += scnprintf(p, e - p, "\\x%02x", c);
691	else
692	append_char(&p, e, c);
693	}
694	append_char(&p, e, '\n');
695
696	if (dict_len) {
697	bool line = true;
698
699	for (i = 0; i < dict_len; i++) {
700	unsigned char c = dict[i];
701
702	if (line) {
703	append_char(&p, e, ' ');
704	line = false;
705	}
706
707	if (c == '\0') {
708	append_char(&p, e, '\n');
709	line = true;
710	continue;
711	}
712
713	if (c < ' ' || c >= 127 || c == '\\') {
714	p += scnprintf(p, e - p, "\\x%02x", c);
715	continue;
716	}
717
718	append_char(&p, e, c);
719	}
720	append_char(&p, e, '\n');
721	}
722
723	return p - buf;
724	}
725
726	/* /dev/kmsg - userspace message inject/listen interface */
727	struct devkmsg_user {
728	u64 seq;
729	u32 idx;
730	enum log_flags prev;
731	struct ratelimit_state rs;
732	struct mutex lock;
733	char buf[CONSOLE_EXT_LOG_MAX];
734	};
735
736	static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from)
737	{
738	char *buf, *line;
739	int level = default_message_loglevel;
740	int facility = 1; /* LOG_USER */
741	struct file *file = iocb->ki_filp;
742	struct devkmsg_user *user = file->private_data;
743	size_t len = iov_iter_count(from);
744	ssize_t ret = len;
745
746	if (!user || len > LOG_LINE_MAX)
747	return -EINVAL;
748
749	/* Ignore when user logging is disabled. */
750	if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
751	return len;
752
753	/* Ratelimit when not explicitly enabled. */
754	if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) {
755	if (!___ratelimit(&user->rs, current->comm))
756	return ret;
757	}
758
759	buf = kmalloc(len+1, GFP_KERNEL);
760	if (buf == NULL)
761	return -ENOMEM;
762
763	buf[len] = '\0';
764	if (copy_from_iter(buf, len, from) != len) {
765	kfree(buf);
766	return -EFAULT;
767	}
768
769	/*
770	* Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
771	* the decimal value represents 32bit, the lower 3 bit are the log
772	* level, the rest are the log facility.
773	*
774	* If no prefix or no userspace facility is specified, we
775	* enforce LOG_USER, to be able to reliably distinguish
776	* kernel-generated messages from userspace-injected ones.
777	*/
778	line = buf;
779	if (line[0] == '<') {
780	char *endp = NULL;
781	unsigned int u;
782
783	u = simple_strtoul(line + 1, &endp, 10);
784	if (endp && endp[0] == '>') {
785	level = LOG_LEVEL(u);
786	if (LOG_FACILITY(u) != 0)
787	facility = LOG_FACILITY(u);
788	endp++;
789	len -= endp - line;
790	line = endp;
791	}
792	}
793
794	printk_emit(facility, level, NULL, 0, "%s", line);
795	kfree(buf);
796	return ret;
797	}
798
799	static ssize_t devkmsg_read(struct file *file, char __user *buf,
800	size_t count, loff_t *ppos)
801	{
802	struct devkmsg_user *user = file->private_data;
803	struct printk_log *msg;
804	size_t len;
805	ssize_t ret;
806
807	if (!user)
808	return -EBADF;
809
810	ret = mutex_lock_interruptible(&user->lock);
811	if (ret)
812	return ret;
813	raw_spin_lock_irq(&logbuf_lock);
814	while (user->seq == log_next_seq) {
815	if (file->f_flags & O_NONBLOCK) {
816	ret = -EAGAIN;
817	raw_spin_unlock_irq(&logbuf_lock);
818	goto out;
819	}
820
821	raw_spin_unlock_irq(&logbuf_lock);
822	ret = wait_event_interruptible(log_wait,
823	user->seq != log_next_seq);
824	if (ret)
825	goto out;
826	raw_spin_lock_irq(&logbuf_lock);
827	}
828
829	if (user->seq < log_first_seq) {
830	/* our last seen message is gone, return error and reset */
831	user->idx = log_first_idx;
832	user->seq = log_first_seq;
833	ret = -EPIPE;
834	raw_spin_unlock_irq(&logbuf_lock);
835	goto out;
836	}
837
838	msg = log_from_idx(user->idx);
839	len = msg_print_ext_header(user->buf, sizeof(user->buf),
840	msg, user->seq, user->prev);
841	len += msg_print_ext_body(user->buf + len, sizeof(user->buf) - len,
842	log_dict(msg), msg->dict_len,
843	log_text(msg), msg->text_len);
844
845	user->prev = msg->flags;
846	user->idx = log_next(user->idx);
847	user->seq++;
848	raw_spin_unlock_irq(&logbuf_lock);
849
850	if (len > count) {
851	ret = -EINVAL;
852	goto out;
853	}
854
855	if (copy_to_user(buf, user->buf, len)) {
856	ret = -EFAULT;
857	goto out;
858	}
859	ret = len;
860	out:
861	mutex_unlock(&user->lock);
862	return ret;
863	}
864
865	static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
866	{
867	struct devkmsg_user *user = file->private_data;
868	loff_t ret = 0;
869
870	if (!user)
871	return -EBADF;
872	if (offset)
873	return -ESPIPE;
874
875	raw_spin_lock_irq(&logbuf_lock);
876	switch (whence) {
877	case SEEK_SET:
878	/* the first record */
879	user->idx = log_first_idx;
880	user->seq = log_first_seq;
881	break;
882	case SEEK_DATA:
883	/*
884	* The first record after the last SYSLOG_ACTION_CLEAR,
885	* like issued by 'dmesg -c'. Reading /dev/kmsg itself
886	* changes no global state, and does not clear anything.
887	*/
888	user->idx = clear_idx;
889	user->seq = clear_seq;
890	break;
891	case SEEK_END:
892	/* after the last record */
893	user->idx = log_next_idx;
894	user->seq = log_next_seq;
895	break;
896	default:
897	ret = -EINVAL;
898	}
899	raw_spin_unlock_irq(&logbuf_lock);
900	return ret;
901	}
902
903	static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
904	{
905	struct devkmsg_user *user = file->private_data;
906	int ret = 0;
907
908	if (!user)
909	return POLLERR|POLLNVAL;
910
911	poll_wait(file, &log_wait, wait);
912
913	raw_spin_lock_irq(&logbuf_lock);
914	if (user->seq < log_next_seq) {
915	/* return error when data has vanished underneath us */
916	if (user->seq < log_first_seq)
917	ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
918	else
919	ret = POLLIN|POLLRDNORM;
920	}
921	raw_spin_unlock_irq(&logbuf_lock);
922
923	return ret;
924	}
925
926	static int devkmsg_open(struct inode *inode, struct file *file)
927	{
928	struct devkmsg_user *user;
929	int err;
930
931	if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
932	return -EPERM;
933
934	/* write-only does not need any file context */
935	if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
936	err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
937	SYSLOG_FROM_READER);
938	if (err)
939	return err;
940	}
941
942	user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
943	if (!user)
944	return -ENOMEM;
945
946	ratelimit_default_init(&user->rs);
947	ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE);
948
949	mutex_init(&user->lock);
950
951	raw_spin_lock_irq(&logbuf_lock);
952	user->idx = log_first_idx;
953	user->seq = log_first_seq;
954	raw_spin_unlock_irq(&logbuf_lock);
955
956	file->private_data = user;
957	return 0;
958	}
959
960	static int devkmsg_release(struct inode *inode, struct file *file)
961	{
962	struct devkmsg_user *user = file->private_data;
963
964	if (!user)
965	return 0;
966
967	ratelimit_state_exit(&user->rs);
968
969	mutex_destroy(&user->lock);
970	kfree(user);
971	return 0;
972	}
973
974	const struct file_operations kmsg_fops = {
975	.open = devkmsg_open,
976	.read = devkmsg_read,
977	.write_iter = devkmsg_write,
978	.llseek = devkmsg_llseek,
979	.poll = devkmsg_poll,
980	.release = devkmsg_release,
981	};
982
983	#ifdef CONFIG_KEXEC_CORE
984	/*
985	* This appends the listed symbols to /proc/vmcore
986	*
987	* /proc/vmcore is used by various utilities, like crash and makedumpfile to
988	* obtain access to symbols that are otherwise very difficult to locate. These
989	* symbols are specifically used so that utilities can access and extract the
990	* dmesg log from a vmcore file after a crash.
991	*/
992	void log_buf_kexec_setup(void)
993	{
994	VMCOREINFO_SYMBOL(log_buf);
995	VMCOREINFO_SYMBOL(log_buf_len);
996	VMCOREINFO_SYMBOL(log_first_idx);
997	VMCOREINFO_SYMBOL(clear_idx);
998	VMCOREINFO_SYMBOL(log_next_idx);
999	/*
1000	* Export struct printk_log size and field offsets. User space tools can
1001	* parse it and detect any changes to structure down the line.
1002	*/
1003	VMCOREINFO_STRUCT_SIZE(printk_log);
1004	VMCOREINFO_OFFSET(printk_log, ts_nsec);
1005	VMCOREINFO_OFFSET(printk_log, len);
1006	VMCOREINFO_OFFSET(printk_log, text_len);
1007	VMCOREINFO_OFFSET(printk_log, dict_len);
1008	}
1009	#endif
1010
1011	/* requested log_buf_len from kernel cmdline */
1012	static unsigned long __initdata new_log_buf_len;
1013
1014	/* we practice scaling the ring buffer by powers of 2 */
1015	static void __init log_buf_len_update(unsigned size)
1016	{
1017	if (size)
1018	size = roundup_pow_of_two(size);
1019	if (size > log_buf_len)
1020	new_log_buf_len = size;
1021	}
1022
1023	/* save requested log_buf_len since it's too early to process it */
1024	static int __init log_buf_len_setup(char *str)
1025	{
1026	unsigned size = memparse(str, &str);
1027
1028	log_buf_len_update(size);
1029
1030	return 0;
1031	}
1032	early_param("log_buf_len", log_buf_len_setup);
1033
1034	#ifdef CONFIG_SMP
1035	#define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1036
1037	static void __init log_buf_add_cpu(void)
1038	{
1039	unsigned int cpu_extra;
1040
1041	/*
1042	* archs should set up cpu_possible_bits properly with
1043	* set_cpu_possible() after setup_arch() but just in
1044	* case lets ensure this is valid.
1045	*/
1046	if (num_possible_cpus() == 1)
1047	return;
1048
1049	cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
1050
1051	/* by default this will only continue through for large > 64 CPUs */
1052	if (cpu_extra <= __LOG_BUF_LEN / 2)
1053	return;
1054
1055	pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1056	__LOG_CPU_MAX_BUF_LEN);
1057	pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1058	cpu_extra);
1059	pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN);
1060
1061	log_buf_len_update(cpu_extra + __LOG_BUF_LEN);
1062	}
1063	#else /* !CONFIG_SMP */
1064	static inline void log_buf_add_cpu(void) {}
1065	#endif /* CONFIG_SMP */
1066
1067	void __init setup_log_buf(int early)
1068	{
1069	unsigned long flags;
1070	char *new_log_buf;
1071	int free;
1072
1073	if (log_buf != __log_buf)
1074	return;
1075
1076	if (!early && !new_log_buf_len)
1077	log_buf_add_cpu();
1078
1079	if (!new_log_buf_len)
1080	return;
1081
1082	if (early) {
1083	new_log_buf =
1084	memblock_virt_alloc(new_log_buf_len, LOG_ALIGN);
1085	} else {
1086	new_log_buf = memblock_virt_alloc_nopanic(new_log_buf_len,
1087	LOG_ALIGN);
1088	}
1089
1090	if (unlikely(!new_log_buf)) {
1091	pr_err("log_buf_len: %ld bytes not available\n",
1092	new_log_buf_len);
1093	return;
1094	}
1095
1096	raw_spin_lock_irqsave(&logbuf_lock, flags);
1097	log_buf_len = new_log_buf_len;
1098	log_buf = new_log_buf;
1099	new_log_buf_len = 0;
1100	free = __LOG_BUF_LEN - log_next_idx;
1101	memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
1102	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
1103
1104	pr_info("log_buf_len: %d bytes\n", log_buf_len);
1105	pr_info("early log buf free: %d(%d%%)\n",
1106	free, (free * 100) / __LOG_BUF_LEN);
1107	}
1108
1109	static bool __read_mostly ignore_loglevel;
1110
1111	static int __init ignore_loglevel_setup(char *str)
1112	{
1113	ignore_loglevel = true;
1114	pr_info("debug: ignoring loglevel setting.\n");
1115
1116	return 0;
1117	}
1118
1119	early_param("ignore_loglevel", ignore_loglevel_setup);
1120	module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1121	MODULE_PARM_DESC(ignore_loglevel,
1122	"ignore loglevel setting (prints all kernel messages to the console)");
1123
1124	static bool suppress_message_printing(int level)
1125	{
1126	return (level >= console_loglevel && !ignore_loglevel);
1127	}
1128
1129	#ifdef CONFIG_BOOT_PRINTK_DELAY
1130
1131	static int boot_delay; /* msecs delay after each printk during bootup */
1132	static unsigned long long loops_per_msec; /* based on boot_delay */
1133
1134	static int __init boot_delay_setup(char *str)
1135	{
1136	unsigned long lpj;
1137
1138	lpj = preset_lpj ? preset_lpj : 1000000; /* some guess */
1139	loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
1140
1141	get_option(&str, &boot_delay);
1142	if (boot_delay > 10 * 1000)
1143	boot_delay = 0;
1144
1145	pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1146	"HZ: %d, loops_per_msec: %llu\n",
1147	boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
1148	return 0;
1149	}
1150	early_param("boot_delay", boot_delay_setup);
1151
1152	static void boot_delay_msec(int level)
1153	{
1154	unsigned long long k;
1155	unsigned long timeout;
1156
1157	if ((boot_delay == 0 || system_state != SYSTEM_BOOTING)
1158	|| suppress_message_printing(level)) {
1159	return;
1160	}
1161
1162	k = (unsigned long long)loops_per_msec * boot_delay;
1163
1164	timeout = jiffies + msecs_to_jiffies(boot_delay);
1165	while (k) {
1166	k--;
1167	cpu_relax();
1168	/*
1169	* use (volatile) jiffies to prevent
1170	* compiler reduction; loop termination via jiffies
1171	* is secondary and may or may not happen.
1172	*/
1173	if (time_after(jiffies, timeout))
1174	break;
1175	touch_nmi_watchdog();
1176	}
1177	}
1178	#else
1179	static inline void boot_delay_msec(int level)
1180	{
1181	}
1182	#endif
1183
1184	static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME);
1185	module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
1186
1187	static size_t print_time(u64 ts, char *buf)
1188	{
1189	unsigned long rem_nsec;
1190
1191	if (!printk_time)
1192	return 0;
1193
1194	rem_nsec = do_div(ts, 1000000000);
1195
1196	if (!buf)
1197	return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
1198
1199	#if defined(CONFIG_SMP) && defined(CONFIG_AMLOGIC_DRIVER)
1200	return sprintf(buf, "[%5lu.%06lu@%d] ",
1201	(unsigned long)ts, rem_nsec / 1000, current_cpu);
1202	#else
1203	return sprintf(buf, "[%5lu.%06lu] ",
1204	(unsigned long)ts, rem_nsec / 1000);
1205	#endif
1206	}
1207
1208	static size_t print_prefix(const struct printk_log *msg, bool syslog, char *buf)
1209	{
1210	size_t len = 0;
1211	unsigned int prefix = (msg->facility << 3) | msg->level;
1212
1213	if (syslog) {
1214	if (buf) {
1215	len += sprintf(buf, "<%u>", prefix);
1216	} else {
1217	len += 3;
1218	if (prefix > 999)
1219	len += 3;
1220	else if (prefix > 99)
1221	len += 2;
1222	else if (prefix > 9)
1223	len++;
1224	}
1225	}
1226	#ifdef CONFIG_AMLOGIC_DRIVER
1227	current_cpu = msg->cpu;
1228	#endif
1229	len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
1230	return len;
1231	}
1232
1233	static size_t msg_print_text(const struct printk_log *msg, enum log_flags prev,
1234	bool syslog, char *buf, size_t size)
1235	{
1236	const char *text = log_text(msg);
1237	size_t text_size = msg->text_len;
1238	bool prefix = true;
1239	bool newline = true;
1240	size_t len = 0;
1241
1242	if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
1243	prefix = false;
1244
1245	if (msg->flags & LOG_CONT) {
1246	if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
1247	prefix = false;
1248
1249	if (!(msg->flags & LOG_NEWLINE))
1250	newline = false;
1251	}
1252
1253	do {
1254	const char *next = memchr(text, '\n', text_size);
1255	size_t text_len;
1256
1257	if (next) {
1258	text_len = next - text;
1259	next++;
1260	text_size -= next - text;
1261	} else {
1262	text_len = text_size;
1263	}
1264
1265	if (buf) {
1266	if (print_prefix(msg, syslog, NULL) +
1267	text_len + 1 >= size - len)
1268	break;
1269
1270	if (prefix)
1271	len += print_prefix(msg, syslog, buf + len);
1272	memcpy(buf + len, text, text_len);
1273	len += text_len;
1274	if (next || newline)
1275	buf[len++] = '\n';
1276	} else {
1277	/* SYSLOG_ACTION_* buffer size only calculation */
1278	if (prefix)
1279	len += print_prefix(msg, syslog, NULL);
1280	len += text_len;
1281	if (next || newline)
1282	len++;
1283	}
1284
1285	prefix = true;
1286	text = next;
1287	} while (text);
1288
1289	return len;
1290	}
1291
1292	static int syslog_print(char __user *buf, int size)
1293	{
1294	char *text;
1295	struct printk_log *msg;
1296	int len = 0;
1297
1298	text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1299	if (!text)
1300	return -ENOMEM;
1301
1302	while (size > 0) {
1303	size_t n;
1304	size_t skip;
1305
1306	raw_spin_lock_irq(&logbuf_lock);
1307	if (syslog_seq < log_first_seq) {
1308	/* messages are gone, move to first one */
1309	syslog_seq = log_first_seq;
1310	syslog_idx = log_first_idx;
1311	syslog_prev = 0;
1312	syslog_partial = 0;
1313	}
1314	if (syslog_seq == log_next_seq) {
1315	raw_spin_unlock_irq(&logbuf_lock);
1316	break;
1317	}
1318
1319	skip = syslog_partial;
1320	msg = log_from_idx(syslog_idx);
1321	n = msg_print_text(msg, syslog_prev, true, text,
1322	LOG_LINE_MAX + PREFIX_MAX);
1323	if (n - syslog_partial <= size) {
1324	/* message fits into buffer, move forward */
1325	syslog_idx = log_next(syslog_idx);
1326	syslog_seq++;
1327	syslog_prev = msg->flags;
1328	n -= syslog_partial;
1329	syslog_partial = 0;
1330	} else if (!len){
1331	/* partial read(), remember position */
1332	n = size;
1333	syslog_partial += n;
1334	} else
1335	n = 0;
1336	raw_spin_unlock_irq(&logbuf_lock);
1337
1338	if (!n)
1339	break;
1340
1341	if (copy_to_user(buf, text + skip, n)) {
1342	if (!len)
1343	len = -EFAULT;
1344	break;
1345	}
1346
1347	len += n;
1348	size -= n;
1349	buf += n;
1350	}
1351
1352	kfree(text);
1353	return len;
1354	}
1355
1356	static int syslog_print_all(char __user *buf, int size, bool clear)
1357	{
1358	char *text;
1359	int len = 0;
1360
1361	text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1362	if (!text)
1363	return -ENOMEM;
1364
1365	raw_spin_lock_irq(&logbuf_lock);
1366	if (buf) {
1367	u64 next_seq;
1368	u64 seq;
1369	u32 idx;
1370	enum log_flags prev;
1371
1372	/*
1373	* Find first record that fits, including all following records,
1374	* into the user-provided buffer for this dump.
1375	*/
1376	seq = clear_seq;
1377	idx = clear_idx;
1378	prev = 0;
1379	while (seq < log_next_seq) {
1380	struct printk_log *msg = log_from_idx(idx);
1381
1382	len += msg_print_text(msg, prev, true, NULL, 0);
1383	prev = msg->flags;
1384	idx = log_next(idx);
1385	seq++;
1386	}
1387
1388	/* move first record forward until length fits into the buffer */
1389	seq = clear_seq;
1390	idx = clear_idx;
1391	prev = 0;
1392	while (len > size && seq < log_next_seq) {
1393	struct printk_log *msg = log_from_idx(idx);
1394
1395	len -= msg_print_text(msg, prev, true, NULL, 0);
1396	prev = msg->flags;
1397	idx = log_next(idx);
1398	seq++;
1399	}
1400
1401	/* last message fitting into this dump */
1402	next_seq = log_next_seq;
1403
1404	len = 0;
1405	while (len >= 0 && seq < next_seq) {
1406	struct printk_log *msg = log_from_idx(idx);
1407	int textlen;
1408
1409	textlen = msg_print_text(msg, prev, true, text,
1410	LOG_LINE_MAX + PREFIX_MAX);
1411	if (textlen < 0) {
1412	len = textlen;
1413	break;
1414	}
1415	idx = log_next(idx);
1416	seq++;
1417	prev = msg->flags;
1418
1419	raw_spin_unlock_irq(&logbuf_lock);
1420	if (copy_to_user(buf + len, text, textlen))
1421	len = -EFAULT;
1422	else
1423	len += textlen;
1424	raw_spin_lock_irq(&logbuf_lock);
1425
1426	if (seq < log_first_seq) {
1427	/* messages are gone, move to next one */
1428	seq = log_first_seq;
1429	idx = log_first_idx;
1430	prev = 0;
1431	}
1432	}
1433	}
1434
1435	if (clear) {
1436	clear_seq = log_next_seq;
1437	clear_idx = log_next_idx;
1438	}
1439	raw_spin_unlock_irq(&logbuf_lock);
1440
1441	kfree(text);
1442	return len;
1443	}
1444
1445	int do_syslog(int type, char __user *buf, int len, int source)
1446	{
1447	bool clear = false;
1448	static int saved_console_loglevel = LOGLEVEL_DEFAULT;
1449	int error;
1450
1451	error = check_syslog_permissions(type, source);
1452	if (error)
1453	goto out;
1454
1455	switch (type) {
1456	case SYSLOG_ACTION_CLOSE: /* Close log */
1457	break;
1458	case SYSLOG_ACTION_OPEN: /* Open log */
1459	break;
1460	case SYSLOG_ACTION_READ: /* Read from log */
1461	error = -EINVAL;
1462	if (!buf || len < 0)
1463	goto out;
1464	error = 0;
1465	if (!len)
1466	goto out;
1467	if (!access_ok(VERIFY_WRITE, buf, len)) {
1468	error = -EFAULT;
1469	goto out;
1470	}
1471	error = wait_event_interruptible(log_wait,
1472	syslog_seq != log_next_seq);
1473	if (error)
1474	goto out;
1475	error = syslog_print(buf, len);
1476	break;
1477	/* Read/clear last kernel messages */
1478	case SYSLOG_ACTION_READ_CLEAR:
1479	clear = true;
1480	/* FALL THRU */
1481	/* Read last kernel messages */
1482	case SYSLOG_ACTION_READ_ALL:
1483	error = -EINVAL;
1484	if (!buf || len < 0)
1485	goto out;
1486	error = 0;
1487	if (!len)
1488	goto out;
1489	if (!access_ok(VERIFY_WRITE, buf, len)) {
1490	error = -EFAULT;
1491	goto out;
1492	}
1493	error = syslog_print_all(buf, len, clear);
1494	break;
1495	/* Clear ring buffer */
1496	case SYSLOG_ACTION_CLEAR:
1497	syslog_print_all(NULL, 0, true);
1498	break;
1499	/* Disable logging to console */
1500	case SYSLOG_ACTION_CONSOLE_OFF:
1501	if (saved_console_loglevel == LOGLEVEL_DEFAULT)
1502	saved_console_loglevel = console_loglevel;
1503	console_loglevel = minimum_console_loglevel;
1504	break;
1505	/* Enable logging to console */
1506	case SYSLOG_ACTION_CONSOLE_ON:
1507	if (saved_console_loglevel != LOGLEVEL_DEFAULT) {
1508	console_loglevel = saved_console_loglevel;
1509	saved_console_loglevel = LOGLEVEL_DEFAULT;
1510	}
1511	break;
1512	/* Set level of messages printed to console */
1513	case SYSLOG_ACTION_CONSOLE_LEVEL:
1514	error = -EINVAL;
1515	if (len < 1 || len > 8)
1516	goto out;
1517	if (len < minimum_console_loglevel)
1518	len = minimum_console_loglevel;
1519	console_loglevel = len;
1520	/* Implicitly re-enable logging to console */
1521	saved_console_loglevel = LOGLEVEL_DEFAULT;
1522	error = 0;
1523	break;
1524	/* Number of chars in the log buffer */
1525	case SYSLOG_ACTION_SIZE_UNREAD:
1526	raw_spin_lock_irq(&logbuf_lock);
1527	if (syslog_seq < log_first_seq) {
1528	/* messages are gone, move to first one */
1529	syslog_seq = log_first_seq;
1530	syslog_idx = log_first_idx;
1531	syslog_prev = 0;
1532	syslog_partial = 0;
1533	}
1534	if (source == SYSLOG_FROM_PROC) {
1535	/*
1536	* Short-cut for poll(/"proc/kmsg") which simply checks
1537	* for pending data, not the size; return the count of
1538	* records, not the length.
1539	*/
1540	error = log_next_seq - syslog_seq;
1541	} else {
1542	u64 seq = syslog_seq;
1543	u32 idx = syslog_idx;
1544	enum log_flags prev = syslog_prev;
1545
1546	error = 0;
1547	while (seq < log_next_seq) {
1548	struct printk_log *msg = log_from_idx(idx);
1549
1550	error += msg_print_text(msg, prev, true, NULL, 0);
1551	idx = log_next(idx);
1552	seq++;
1553	prev = msg->flags;
1554	}
1555	error -= syslog_partial;
1556	}
1557	raw_spin_unlock_irq(&logbuf_lock);
1558	break;
1559	/* Size of the log buffer */
1560	case SYSLOG_ACTION_SIZE_BUFFER:
1561	error = log_buf_len;
1562	break;
1563	default:
1564	error = -EINVAL;
1565	break;
1566	}
1567	out:
1568	return error;
1569	}
1570
1571	SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1572	{
1573	return do_syslog(type, buf, len, SYSLOG_FROM_READER);
1574	}
1575
1576	/*
1577	* Call the console drivers, asking them to write out
1578	* log_buf[start] to log_buf[end - 1].
1579	* The console_lock must be held.
1580	*/
1581	static void call_console_drivers(int level,
1582	const char *ext_text, size_t ext_len,
1583	const char *text, size_t len)
1584	{
1585	struct console *con;
1586
1587	trace_console_rcuidle(text, len);
1588
1589	if (!console_drivers)
1590	return;
1591
1592	for_each_console(con) {
1593	if (exclusive_console && con != exclusive_console)
1594	continue;
1595	if (!(con->flags & CON_ENABLED))
1596	continue;
1597	if (!con->write)
1598	continue;
1599	if (!cpu_online(smp_processor_id()) &&
1600	!(con->flags & CON_ANYTIME))
1601	continue;
1602	if (con->flags & CON_EXTENDED)
1603	con->write(con, ext_text, ext_len);
1604	else
1605	con->write(con, text, len);
1606	}
1607	}
1608
1609	/*
1610	* Zap console related locks when oopsing.
1611	* To leave time for slow consoles to print a full oops,
1612	* only zap at most once every 30 seconds.
1613	*/
1614	static void zap_locks(void)
1615	{
1616	static unsigned long oops_timestamp;
1617
1618	if (time_after_eq(jiffies, oops_timestamp) &&
1619	!time_after(jiffies, oops_timestamp + 30 * HZ))
1620	return;
1621
1622	oops_timestamp = jiffies;
1623
1624	debug_locks_off();
1625	/* If a crash is occurring, make sure we can't deadlock */
1626	raw_spin_lock_init(&logbuf_lock);
1627	/* And make sure that we print immediately */
1628	sema_init(&console_sem, 1);
1629	}
1630
1631	int printk_delay_msec __read_mostly;
1632
1633	static inline void printk_delay(void)
1634	{
1635	if (unlikely(printk_delay_msec)) {
1636	int m = printk_delay_msec;
1637
1638	while (m--) {
1639	mdelay(1);
1640	touch_nmi_watchdog();
1641	}
1642	}
1643	}
1644
1645	/*
1646	* Continuation lines are buffered, and not committed to the record buffer
1647	* until the line is complete, or a race forces it. The line fragments
1648	* though, are printed immediately to the consoles to ensure everything has
1649	* reached the console in case of a kernel crash.
1650	*/
1651	static struct cont {
1652	char buf[LOG_LINE_MAX];
1653	size_t len; /* length == 0 means unused buffer */
1654	size_t cons; /* bytes written to console */
1655	struct task_struct *owner; /* task of first print*/
1656	u64 ts_nsec; /* time of first print */
1657	u8 level; /* log level of first message */
1658	u8 facility; /* log facility of first message */
1659	enum log_flags flags; /* prefix, newline flags */
1660	bool flushed:1; /* buffer sealed and committed */
1661	} cont;
1662
1663	static void cont_flush(void)
1664	{
1665	if (cont.flushed)
1666	return;
1667	if (cont.len == 0)
1668	return;
1669	if (cont.cons) {
1670	/*
1671	* If a fragment of this line was directly flushed to the
1672	* console; wait for the console to pick up the rest of the
1673	* line. LOG_NOCONS suppresses a duplicated output.
1674	*/
1675	log_store(cont.facility, cont.level, cont.flags | LOG_NOCONS,
1676	cont.ts_nsec, NULL, 0, cont.buf, cont.len);
1677	cont.flushed = true;
1678	} else {
1679	/*
1680	* If no fragment of this line ever reached the console,
1681	* just submit it to the store and free the buffer.
1682	*/
1683	log_store(cont.facility, cont.level, cont.flags, 0,
1684	NULL, 0, cont.buf, cont.len);
1685	cont.len = 0;
1686	}
1687	}
1688
1689	static bool cont_add(int facility, int level, enum log_flags flags, const char *text, size_t len)
1690	{
1691	if (cont.len && cont.flushed)
1692	return false;
1693
1694	/*
1695	* If ext consoles are present, flush and skip in-kernel
1696	* continuation. See nr_ext_console_drivers definition. Also, if
1697	* the line gets too long, split it up in separate records.
1698	*/
1699	if (nr_ext_console_drivers || cont.len + len > sizeof(cont.buf)) {
1700	cont_flush();
1701	return false;
1702	}
1703
1704	if (!cont.len) {
1705	cont.facility = facility;
1706	cont.level = level;
1707	cont.owner = current;
1708	cont.ts_nsec = local_clock();
1709	cont.flags = flags;
1710	cont.cons = 0;
1711	cont.flushed = false;
1712	}
1713
1714	memcpy(cont.buf + cont.len, text, len);
1715	cont.len += len;
1716
1717	// The original flags come from the first line,
1718	// but later continuations can add a newline.
1719	if (flags & LOG_NEWLINE) {
1720	cont.flags |= LOG_NEWLINE;
1721	cont_flush();
1722	}
1723
1724	if (cont.len > (sizeof(cont.buf) * 80) / 100)
1725	cont_flush();
1726
1727	return true;
1728	}
1729
1730	static size_t cont_print_text(char *text, size_t size)
1731	{
1732	size_t textlen = 0;
1733	size_t len;
1734
1735	if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
1736	textlen += print_time(cont.ts_nsec, text);
1737	size -= textlen;
1738	}
1739
1740	len = cont.len - cont.cons;
1741	if (len > 0) {
1742	if (len+1 > size)
1743	len = size-1;
1744	memcpy(text + textlen, cont.buf + cont.cons, len);
1745	textlen += len;
1746	cont.cons = cont.len;
1747	}
1748
1749	if (cont.flushed) {
1750	if (cont.flags & LOG_NEWLINE)
1751	text[textlen++] = '\n';
1752	/* got everything, release buffer */
1753	cont.len = 0;
1754	}
1755	return textlen;
1756	}
1757
1758	#define AML_LOSE_CONTLINE_DEF 1
1759
1760	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)
1761	{
1762	/*
1763	* If an earlier line was buffered, and we're a continuation
1764	* write from the same process, try to add it to the buffer.
1765	*/
1766	if (cont.len) {
1767	#if (AML_LOSE_CONTLINE_DEF == 1)
1768	if (cont.owner == current && !(lflags & LOG_PREFIX)) {
1769	#else
1770	if (cont.owner == current && (lflags & LOG_CONT)) {
1771	#endif
1772	if (cont_add(facility, level, lflags, text, text_len))
1773	return text_len;
1774	}
1775	/* Otherwise, make sure it's flushed */
1776	cont_flush();
1777	}
1778
1779	/* Skip empty continuation lines that couldn't be added - they just flush */
1780	if (!text_len && (lflags & LOG_CONT))
1781	return 0;
1782
1783	/* If it doesn't end in a newline, try to buffer the current line */
1784	if (!(lflags & LOG_NEWLINE)) {
1785	if (cont_add(facility, level, lflags, text, text_len))
1786	return text_len;
1787	}
1788
1789	/* Store it in the record log */
1790	return log_store(facility, level, lflags, 0, dict, dictlen, text, text_len);
1791	}
1792
1793	asmlinkage int vprintk_emit(int facility, int level,
1794	const char *dict, size_t dictlen,
1795	const char *fmt, va_list args)
1796	{
1797	static bool recursion_bug;
1798	static char textbuf[LOG_LINE_MAX];
1799	char *text = textbuf;
1800	size_t text_len = 0;
1801	enum log_flags lflags = 0;
1802	unsigned long flags;
1803	int this_cpu;
1804	int printed_len = 0;
1805	int nmi_message_lost;
1806	bool in_sched = false;
1807	/* cpu currently holding logbuf_lock in this function */
1808	static unsigned int logbuf_cpu = UINT_MAX;
1809
1810	if (level == LOGLEVEL_SCHED) {
1811	level = LOGLEVEL_DEFAULT;
1812	in_sched = true;
1813	}
1814
1815	boot_delay_msec(level);
1816	printk_delay();
1817
1818	local_irq_save(flags);
1819	this_cpu = smp_processor_id();
1820
1821	/*
1822	* Ouch, printk recursed into itself!
1823	*/
1824	if (unlikely(logbuf_cpu == this_cpu)) {
1825	/*
1826	* If a crash is occurring during printk() on this CPU,
1827	* then try to get the crash message out but make sure
1828	* we can't deadlock. Otherwise just return to avoid the
1829	* recursion and return - but flag the recursion so that
1830	* it can be printed at the next appropriate moment:
1831	*/
1832	if (!oops_in_progress && !lockdep_recursing(current)) {
1833	recursion_bug = true;
1834	local_irq_restore(flags);
1835	return 0;
1836	}
1837	zap_locks();
1838	}
1839
1840	lockdep_off();
1841	/* This stops the holder of console_sem just where we want him */
1842	raw_spin_lock(&logbuf_lock);
1843	logbuf_cpu = this_cpu;
1844
1845	if (unlikely(recursion_bug)) {
1846	static const char recursion_msg[] =
1847	"BUG: recent printk recursion!";
1848
1849	recursion_bug = false;
1850	/* emit KERN_CRIT message */
1851	printed_len += log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1852	NULL, 0, recursion_msg,
1853	strlen(recursion_msg));
1854	}
1855
1856	nmi_message_lost = get_nmi_message_lost();
1857	if (unlikely(nmi_message_lost)) {
1858	text_len = scnprintf(textbuf, sizeof(textbuf),
1859	"BAD LUCK: lost %d message(s) from NMI context!",
1860	nmi_message_lost);
1861	printed_len += log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1862	NULL, 0, textbuf, text_len);
1863	}
1864
1865	/*
1866	* The printf needs to come first; we need the syslog
1867	* prefix which might be passed-in as a parameter.
1868	*/
1869	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1870
1871	/* mark and strip a trailing newline */
1872	if (text_len && text[text_len-1] == '\n') {
1873	text_len--;
1874	lflags |= LOG_NEWLINE;
1875	}
1876
1877	/* strip kernel syslog prefix and extract log level or control flags */
1878	if (facility == 0) {
1879	int kern_level;
1880
1881	while ((kern_level = printk_get_level(text)) != 0) {
1882	switch (kern_level) {
1883	case '0' ... '7':
1884	if (level == LOGLEVEL_DEFAULT)
1885	level = kern_level - '0';
1886	/* fallthrough */
1887	case 'd': /* KERN_DEFAULT */
1888	lflags |= LOG_PREFIX;
1889	break;
1890	case 'c': /* KERN_CONT */
1891	lflags |= LOG_CONT;
1892	}
1893
1894	text_len -= 2;
1895	text += 2;
1896	}
1897	}
1898
1899	#ifdef CONFIG_EARLY_PRINTK_DIRECT
1900	printascii(text);
1901	#endif
1902
1903	if (level == LOGLEVEL_DEFAULT)
1904	level = default_message_loglevel;
1905
1906	if (dict)
1907	lflags |= LOG_PREFIX|LOG_NEWLINE;
1908
1909	printed_len += log_output(facility, level, lflags, dict, dictlen, text, text_len);
1910
1911	logbuf_cpu = UINT_MAX;
1912	raw_spin_unlock(&logbuf_lock);
1913	lockdep_on();
1914	local_irq_restore(flags);
1915
1916	/* If called from the scheduler, we can not call up(). */
1917	if (!in_sched) {
1918	lockdep_off();
1919	/*
1920	* Try to acquire and then immediately release the console
1921	* semaphore. The release will print out buffers and wake up
1922	* /dev/kmsg and syslog() users.
1923	*/
1924	if (console_trylock())
1925	console_unlock();
1926	lockdep_on();
1927	}
1928
1929	return printed_len;
1930	}
1931	EXPORT_SYMBOL(vprintk_emit);
1932
1933	asmlinkage int vprintk(const char *fmt, va_list args)
1934	{
1935	return vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
1936	}
1937	EXPORT_SYMBOL(vprintk);
1938
1939	asmlinkage int printk_emit(int facility, int level,
1940	const char *dict, size_t dictlen,
1941	const char *fmt, ...)
1942	{
1943	va_list args;
1944	int r;
1945
1946	va_start(args, fmt);
1947	r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1948	va_end(args);
1949
1950	return r;
1951	}
1952	EXPORT_SYMBOL(printk_emit);
1953
1954	int vprintk_default(const char *fmt, va_list args)
1955	{
1956	int r;
1957
1958	#ifdef CONFIG_KGDB_KDB
1959	if (unlikely(kdb_trap_printk)) {
1960	r = vkdb_printf(KDB_MSGSRC_PRINTK, fmt, args);
1961	return r;
1962	}
1963	#endif
1964	r = vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
1965
1966	return r;
1967	}
1968	EXPORT_SYMBOL_GPL(vprintk_default);
1969
1970	/**
1971	* printk - print a kernel message
1972	* @fmt: format string
1973	*
1974	* This is printk(). It can be called from any context. We want it to work.
1975	*
1976	* We try to grab the console_lock. If we succeed, it's easy - we log the
1977	* output and call the console drivers. If we fail to get the semaphore, we
1978	* place the output into the log buffer and return. The current holder of
1979	* the console_sem will notice the new output in console_unlock(); and will
1980	* send it to the consoles before releasing the lock.
1981	*
1982	* One effect of this deferred printing is that code which calls printk() and
1983	* then changes console_loglevel may break. This is because console_loglevel
1984	* is inspected when the actual printing occurs.
1985	*
1986	* See also:
1987	* printf(3)
1988	*
1989	* See the vsnprintf() documentation for format string extensions over C99.
1990	*/
1991	asmlinkage __visible int printk(const char *fmt, ...)
1992	{
1993	va_list args;
1994	int r;
1995
1996	va_start(args, fmt);
1997	r = vprintk_func(fmt, args);
1998	va_end(args);
1999
2000	return r;
2001	}
2002	EXPORT_SYMBOL(printk);
2003
2004	#else /* CONFIG_PRINTK */
2005
2006	#define LOG_LINE_MAX 0
2007	#define PREFIX_MAX 0
2008
2009	static u64 syslog_seq;
2010	static u32 syslog_idx;
2011	static u64 console_seq;
2012	static u32 console_idx;
2013	static enum log_flags syslog_prev;
2014	static u64 log_first_seq;
2015	static u32 log_first_idx;
2016	static u64 log_next_seq;
2017	static enum log_flags console_prev;
2018	static struct cont {
2019	size_t len;
2020	size_t cons;
2021	u8 level;
2022	bool flushed:1;
2023	} cont;
2024	static char *log_text(const struct printk_log *msg) { return NULL; }
2025	static char *log_dict(const struct printk_log *msg) { return NULL; }
2026	static struct printk_log *log_from_idx(u32 idx) { return NULL; }
2027	static u32 log_next(u32 idx) { return 0; }
2028	static ssize_t msg_print_ext_header(char *buf, size_t size,
2029	struct printk_log *msg, u64 seq,
2030	enum log_flags prev_flags) { return 0; }
2031	static ssize_t msg_print_ext_body(char *buf, size_t size,
2032	char *dict, size_t dict_len,
2033	char *text, size_t text_len) { return 0; }
2034	static void call_console_drivers(int level,
2035	const char *ext_text, size_t ext_len,
2036	const char *text, size_t len) {}
2037	static size_t msg_print_text(const struct printk_log *msg, enum log_flags prev,
2038	bool syslog, char *buf, size_t size) { return 0; }
2039	static size_t cont_print_text(char *text, size_t size) { return 0; }
2040	static bool suppress_message_printing(int level) { return false; }
2041
2042	/* Still needs to be defined for users */
2043	DEFINE_PER_CPU(printk_func_t, printk_func);
2044
2045	#endif /* CONFIG_PRINTK */
2046
2047	#ifdef CONFIG_EARLY_PRINTK
2048	struct console *early_console;
2049
2050	asmlinkage __visible void early_printk(const char *fmt, ...)
2051	{
2052	va_list ap;
2053	char buf[512];
2054	int n;
2055
2056	if (!early_console)
2057	return;
2058
2059	va_start(ap, fmt);
2060	n = vscnprintf(buf, sizeof(buf), fmt, ap);
2061	va_end(ap);
2062
2063	early_console->write(early_console, buf, n);
2064	}
2065	#endif
2066
2067	static int __add_preferred_console(char *name, int idx, char *options,
2068	char *brl_options)
2069	{
2070	struct console_cmdline *c;
2071	int i;
2072
2073	/*
2074	* See if this tty is not yet registered, and
2075	* if we have a slot free.
2076	*/
2077	for (i = 0, c = console_cmdline;
2078	i < MAX_CMDLINECONSOLES && c->name[0];
2079	i++, c++) {
2080	if (strcmp(c->name, name) == 0 && c->index == idx) {
2081	if (!brl_options)
2082	selected_console = i;
2083	return 0;
2084	}
2085	}
2086	if (i == MAX_CMDLINECONSOLES)
2087	return -E2BIG;
2088	if (!brl_options)
2089	selected_console = i;
2090	strlcpy(c->name, name, sizeof(c->name));
2091	c->options = options;
2092	braille_set_options(c, brl_options);
2093
2094	c->index = idx;
2095	return 0;
2096	}
2097	/*
2098	* Set up a console. Called via do_early_param() in init/main.c
2099	* for each "console=" parameter in the boot command line.
2100	*/
2101	static int __init console_setup(char *str)
2102	{
2103	char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */
2104	char *s, *options, *brl_options = NULL;
2105	int idx;
2106
2107	if (_braille_console_setup(&str, &brl_options))
2108	return 1;
2109
2110	/*
2111	* Decode str into name, index, options.
2112	*/
2113	if (str[0] >= '0' && str[0] <= '9') {
2114	strcpy(buf, "ttyS");
2115	strncpy(buf + 4, str, sizeof(buf) - 5);
2116	} else {
2117	strncpy(buf, str, sizeof(buf) - 1);
2118	}
2119	buf[sizeof(buf) - 1] = 0;
2120	options = strchr(str, ',');
2121	if (options)
2122	*(options++) = 0;
2123	#ifdef __sparc__
2124	if (!strcmp(str, "ttya"))
2125	strcpy(buf, "ttyS0");
2126	if (!strcmp(str, "ttyb"))
2127	strcpy(buf, "ttyS1");
2128	#endif
2129	for (s = buf; *s; s++)
2130	if (isdigit(*s) || *s == ',')
2131	break;
2132	idx = simple_strtoul(s, NULL, 10);
2133	*s = 0;
2134
2135	__add_preferred_console(buf, idx, options, brl_options);
2136	console_set_on_cmdline = 1;
2137	return 1;
2138	}
2139	__setup("console=", console_setup);
2140
2141	/**
2142	* add_preferred_console - add a device to the list of preferred consoles.
2143	* @name: device name
2144	* @idx: device index
2145	* @options: options for this console
2146	*
2147	* The last preferred console added will be used for kernel messages
2148	* and stdin/out/err for init. Normally this is used by console_setup
2149	* above to handle user-supplied console arguments; however it can also
2150	* be used by arch-specific code either to override the user or more
2151	* commonly to provide a default console (ie from PROM variables) when
2152	* the user has not supplied one.
2153	*/
2154	int add_preferred_console(char *name, int idx, char *options)
2155	{
2156	return __add_preferred_console(name, idx, options, NULL);
2157	}
2158
2159	bool console_suspend_enabled = true;
2160	EXPORT_SYMBOL(console_suspend_enabled);
2161
2162	static int __init console_suspend_disable(char *str)
2163	{
2164	console_suspend_enabled = false;
2165	return 1;
2166	}
2167	__setup("no_console_suspend", console_suspend_disable);
2168	module_param_named(console_suspend, console_suspend_enabled,
2169	bool, S_IRUGO | S_IWUSR);
2170	MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
2171	" and hibernate operations");
2172
2173	/**
2174	* suspend_console - suspend the console subsystem
2175	*
2176	* This disables printk() while we go into suspend states
2177	*/
2178	void suspend_console(void)
2179	{
2180	if (!console_suspend_enabled)
2181	return;
2182	printk("Suspending console(s) (use no_console_suspend to debug)\n");
2183	console_lock();
2184	console_suspended = 1;
2185	up_console_sem();
2186	}
2187
2188	void resume_console(void)
2189	{
2190	if (!console_suspend_enabled)
2191	return;
2192	down_console_sem();
2193	console_suspended = 0;
2194	console_unlock();
2195	}
2196
2197	/**
2198	* console_cpu_notify - print deferred console messages after CPU hotplug
2199	* @self: notifier struct
2200	* @action: CPU hotplug event
2201	* @hcpu: unused
2202	*
2203	* If printk() is called from a CPU that is not online yet, the messages
2204	* will be spooled but will not show up on the console. This function is
2205	* called when a new CPU comes online (or fails to come up), and ensures
2206	* that any such output gets printed.
2207	*/
2208	static int console_cpu_notify(struct notifier_block *self,
2209	unsigned long action, void *hcpu)
2210	{
2211	switch (action) {
2212	case CPU_ONLINE:
2213	case CPU_DEAD:
2214	case CPU_DOWN_FAILED:
2215	case CPU_UP_CANCELED:
2216	console_lock();
2217	console_unlock();
2218	}
2219	return NOTIFY_OK;
2220	}
2221
2222	/**
2223	* console_lock - lock the console system for exclusive use.
2224	*
2225	* Acquires a lock which guarantees that the caller has
2226	* exclusive access to the console system and the console_drivers list.
2227	*
2228	* Can sleep, returns nothing.
2229	*/
2230	void console_lock(void)
2231	{
2232	might_sleep();
2233
2234	down_console_sem();
2235	if (console_suspended)
2236	return;
2237	console_locked = 1;
2238	console_may_schedule = 1;
2239	}
2240	EXPORT_SYMBOL(console_lock);
2241
2242	/**
2243	* console_trylock - try to lock the console system for exclusive use.
2244	*
2245	* Try to acquire a lock which guarantees that the caller has exclusive
2246	* access to the console system and the console_drivers list.
2247	*
2248	* returns 1 on success, and 0 on failure to acquire the lock.
2249	*/
2250	int console_trylock(void)
2251	{
2252	if (down_trylock_console_sem())
2253	return 0;
2254	if (console_suspended) {
2255	up_console_sem();
2256	return 0;
2257	}
2258	console_locked = 1;
2259	/*
2260	* When PREEMPT_COUNT disabled we can't reliably detect if it's
2261	* safe to schedule (e.g. calling printk while holding a spin_lock),
2262	* because preempt_disable()/preempt_enable() are just barriers there
2263	* and preempt_count() is always 0.
2264	*
2265	* RCU read sections have a separate preemption counter when
2266	* PREEMPT_RCU enabled thus we must take extra care and check
2267	* rcu_preempt_depth(), otherwise RCU read sections modify
2268	* preempt_count().
2269	*/
2270	console_may_schedule = !oops_in_progress &&
2271	preemptible() &&
2272	!rcu_preempt_depth();
2273	return 1;
2274	}
2275	EXPORT_SYMBOL(console_trylock);
2276
2277	int is_console_locked(void)
2278	{
2279	return console_locked;
2280	}
2281
2282	/*
2283	* Check if we have any console that is capable of printing while cpu is
2284	* booting or shutting down. Requires console_sem.
2285	*/
2286	static int have_callable_console(void)
2287	{
2288	struct console *con;
2289
2290	for_each_console(con)
2291	if ((con->flags & CON_ENABLED) &&
2292	(con->flags & CON_ANYTIME))
2293	return 1;
2294
2295	return 0;
2296	}
2297
2298	/*
2299	* Can we actually use the console at this time on this cpu?
2300	*
2301	* Console drivers may assume that per-cpu resources have been allocated. So
2302	* unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2303	* call them until this CPU is officially up.
2304	*/
2305	static inline int can_use_console(void)
2306	{
2307	return cpu_online(raw_smp_processor_id()) || have_callable_console();
2308	}
2309
2310	static void console_cont_flush(char *text, size_t size)
2311	{
2312	unsigned long flags;
2313	size_t len;
2314
2315	raw_spin_lock_irqsave(&logbuf_lock, flags);
2316
2317	if (!cont.len)
2318	goto out;
2319
2320	if (suppress_message_printing(cont.level)) {
2321	cont.cons = cont.len;
2322	if (cont.flushed)
2323	cont.len = 0;
2324	goto out;
2325	}
2326
2327	/*
2328	* We still queue earlier records, likely because the console was
2329	* busy. The earlier ones need to be printed before this one, we
2330	* did not flush any fragment so far, so just let it queue up.
2331	*/
2332	if (console_seq < log_next_seq && !cont.cons)
2333	goto out;
2334
2335	len = cont_print_text(text, size);
2336	raw_spin_unlock(&logbuf_lock);
2337	stop_critical_timings();
2338	call_console_drivers(cont.level, NULL, 0, text, len);
2339	start_critical_timings();
2340	local_irq_restore(flags);
2341	return;
2342	out:
2343	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2344	}
2345
2346	/**
2347	* console_unlock - unlock the console system
2348	*
2349	* Releases the console_lock which the caller holds on the console system
2350	* and the console driver list.
2351	*
2352	* While the console_lock was held, console output may have been buffered
2353	* by printk(). If this is the case, console_unlock(); emits
2354	* the output prior to releasing the lock.
2355	*
2356	* If there is output waiting, we wake /dev/kmsg and syslog() users.
2357	*
2358	* console_unlock(); may be called from any context.
2359	*/
2360	void console_unlock(void)
2361	{
2362	static char ext_text[CONSOLE_EXT_LOG_MAX];
2363	static char text[LOG_LINE_MAX + PREFIX_MAX];
2364	static u64 seen_seq;
2365	unsigned long flags;
2366	bool wake_klogd = false;
2367	bool do_cond_resched, retry;
2368
2369	if (console_suspended) {
2370	up_console_sem();
2371	return;
2372	}
2373
2374	/*
2375	* Console drivers are called with interrupts disabled, so
2376	* @console_may_schedule should be cleared before; however, we may
2377	* end up dumping a lot of lines, for example, if called from
2378	* console registration path, and should invoke cond_resched()
2379	* between lines if allowable. Not doing so can cause a very long
2380	* scheduling stall on a slow console leading to RCU stall and
2381	* softlockup warnings which exacerbate the issue with more
2382	* messages practically incapacitating the system.
2383	*
2384	* console_trylock() is not able to detect the preemptive
2385	* context reliably. Therefore the value must be stored before
2386	* and cleared after the the "again" goto label.
2387	*/
2388	do_cond_resched = console_may_schedule;
2389	again:
2390	console_may_schedule = 0;
2391
2392	/*
2393	* We released the console_sem lock, so we need to recheck if
2394	* cpu is online and (if not) is there at least one CON_ANYTIME
2395	* console.
2396	*/
2397	if (!can_use_console()) {
2398	console_locked = 0;
2399	up_console_sem();
2400	return;
2401	}
2402
2403	/* flush buffered message fragment immediately to console */
2404	console_cont_flush(text, sizeof(text));
2405
2406	for (;;) {
2407	struct printk_log *msg;
2408	size_t ext_len = 0;
2409	size_t len;
2410	int level;
2411
2412	raw_spin_lock_irqsave(&logbuf_lock, flags);
2413	if (seen_seq != log_next_seq) {
2414	wake_klogd = true;
2415	seen_seq = log_next_seq;
2416	}
2417
2418	if (console_seq < log_first_seq) {
2419	len = sprintf(text, "** %u printk messages dropped ** ",
2420	(unsigned)(log_first_seq - console_seq));
2421
2422	/* messages are gone, move to first one */
2423	console_seq = log_first_seq;
2424	console_idx = log_first_idx;
2425	console_prev = 0;
2426	} else {
2427	len = 0;
2428	}
2429	skip:
2430	if (console_seq == log_next_seq)
2431	break;
2432
2433	msg = log_from_idx(console_idx);
2434	level = msg->level;
2435	if ((msg->flags & LOG_NOCONS) ||
2436	suppress_message_printing(level)) {
2437	/*
2438	* Skip record we have buffered and already printed
2439	* directly to the console when we received it, and
2440	* record that has level above the console loglevel.
2441	*/
2442	console_idx = log_next(console_idx);
2443	console_seq++;
2444	/*
2445	* We will get here again when we register a new
2446	* CON_PRINTBUFFER console. Clear the flag so we
2447	* will properly dump everything later.
2448	*/
2449	msg->flags &= ~LOG_NOCONS;
2450	console_prev = msg->flags;
2451	goto skip;
2452	}
2453
2454	len += msg_print_text(msg, console_prev, false,
2455	text + len, sizeof(text) - len);
2456	if (nr_ext_console_drivers) {
2457	ext_len = msg_print_ext_header(ext_text,
2458	sizeof(ext_text),
2459	msg, console_seq, console_prev);
2460	ext_len += msg_print_ext_body(ext_text + ext_len,
2461	sizeof(ext_text) - ext_len,
2462	log_dict(msg), msg->dict_len,
2463	log_text(msg), msg->text_len);
2464	}
2465	console_idx = log_next(console_idx);
2466	console_seq++;
2467	console_prev = msg->flags;
2468	raw_spin_unlock(&logbuf_lock);
2469
2470	stop_critical_timings(); /* don't trace print latency */
2471	call_console_drivers(level, ext_text, ext_len, text, len);
2472	start_critical_timings();
2473	local_irq_restore(flags);
2474
2475	if (do_cond_resched)
2476	cond_resched();
2477	}
2478	console_locked = 0;
2479
2480	/* Release the exclusive_console once it is used */
2481	if (unlikely(exclusive_console))
2482	exclusive_console = NULL;
2483
2484	raw_spin_unlock(&logbuf_lock);
2485
2486	up_console_sem();
2487
2488	/*
2489	* Someone could have filled up the buffer again, so re-check if there's
2490	* something to flush. In case we cannot trylock the console_sem again,
2491	* there's a new owner and the console_unlock() from them will do the
2492	* flush, no worries.
2493	*/
2494	raw_spin_lock(&logbuf_lock);
2495	retry = console_seq != log_next_seq;
2496	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2497
2498	if (retry && console_trylock())
2499	goto again;
2500
2501	if (wake_klogd)
2502	wake_up_klogd();
2503	}
2504	EXPORT_SYMBOL(console_unlock);
2505
2506	/**
2507	* console_conditional_schedule - yield the CPU if required
2508	*
2509	* If the console code is currently allowed to sleep, and
2510	* if this CPU should yield the CPU to another task, do
2511	* so here.
2512	*
2513	* Must be called within console_lock();.
2514	*/
2515	void __sched console_conditional_schedule(void)
2516	{
2517	if (console_may_schedule)
2518	cond_resched();
2519	}
2520	EXPORT_SYMBOL(console_conditional_schedule);
2521
2522	void console_unblank(void)
2523	{
2524	struct console *c;
2525
2526	/*
2527	* console_unblank can no longer be called in interrupt context unless
2528	* oops_in_progress is set to 1..
2529	*/
2530	if (oops_in_progress) {
2531	if (down_trylock_console_sem() != 0)
2532	return;
2533	} else
2534	console_lock();
2535
2536	console_locked = 1;
2537	console_may_schedule = 0;
2538	for_each_console(c)
2539	if ((c->flags & CON_ENABLED) && c->unblank)
2540	c->unblank();
2541	console_unlock();
2542	}
2543
2544	/**
2545	* console_flush_on_panic - flush console content on panic
2546	*
2547	* Immediately output all pending messages no matter what.
2548	*/
2549	void console_flush_on_panic(void)
2550	{
2551	/*
2552	* If someone else is holding the console lock, trylock will fail
2553	* and may_schedule may be set. Ignore and proceed to unlock so
2554	* that messages are flushed out. As this can be called from any
2555	* context and we don't want to get preempted while flushing,
2556	* ensure may_schedule is cleared.
2557	*/
2558	console_trylock();
2559	console_may_schedule = 0;
2560	console_unlock();
2561	}
2562
2563	/*
2564	* Return the console tty driver structure and its associated index
2565	*/
2566	struct tty_driver *console_device(int *index)
2567	{
2568	struct console *c;
2569	struct tty_driver *driver = NULL;
2570
2571	console_lock();
2572	for_each_console(c) {
2573	if (!c->device)
2574	continue;
2575	driver = c->device(c, index);
2576	if (driver)
2577	break;
2578	}
2579	console_unlock();
2580	return driver;
2581	}
2582
2583	/*
2584	* Prevent further output on the passed console device so that (for example)
2585	* serial drivers can disable console output before suspending a port, and can
2586	* re-enable output afterwards.
2587	*/
2588	void console_stop(struct console *console)
2589	{
2590	console_lock();
2591	console->flags &= ~CON_ENABLED;
2592	console_unlock();
2593	}
2594	EXPORT_SYMBOL(console_stop);
2595
2596	void console_start(struct console *console)
2597	{
2598	console_lock();
2599	console->flags |= CON_ENABLED;
2600	console_unlock();
2601	}
2602	EXPORT_SYMBOL(console_start);
2603
2604	static int __read_mostly keep_bootcon;
2605
2606	static int __init keep_bootcon_setup(char *str)
2607	{
2608	keep_bootcon = 1;
2609	pr_info("debug: skip boot console de-registration.\n");
2610
2611	return 0;
2612	}
2613
2614	early_param("keep_bootcon", keep_bootcon_setup);
2615
2616	/*
2617	* The console driver calls this routine during kernel initialization
2618	* to register the console printing procedure with printk() and to
2619	* print any messages that were printed by the kernel before the
2620	* console driver was initialized.
2621	*
2622	* This can happen pretty early during the boot process (because of
2623	* early_printk) - sometimes before setup_arch() completes - be careful
2624	* of what kernel features are used - they may not be initialised yet.
2625	*
2626	* There are two types of consoles - bootconsoles (early_printk) and
2627	* "real" consoles (everything which is not a bootconsole) which are
2628	* handled differently.
2629	* - Any number of bootconsoles can be registered at any time.
2630	* - As soon as a "real" console is registered, all bootconsoles
2631	* will be unregistered automatically.
2632	* - Once a "real" console is registered, any attempt to register a
2633	* bootconsoles will be rejected
2634	*/
2635	void register_console(struct console *newcon)
2636	{
2637	int i;
2638	unsigned long flags;
2639	struct console *bcon = NULL;
2640	struct console_cmdline *c;
2641
2642	if (console_drivers)
2643	for_each_console(bcon)
2644	if (WARN(bcon == newcon,
2645	"console '%s%d' already registered\n",
2646	bcon->name, bcon->index))
2647	return;
2648
2649	/*
2650	* before we register a new CON_BOOT console, make sure we don't
2651	* already have a valid console
2652	*/
2653	if (console_drivers && newcon->flags & CON_BOOT) {
2654	/* find the last or real console */
2655	for_each_console(bcon) {
2656	if (!(bcon->flags & CON_BOOT)) {
2657	pr_info("Too late to register bootconsole %s%d\n",
2658	newcon->name, newcon->index);
2659	return;
2660	}
2661	}
2662	}
2663
2664	if (console_drivers && console_drivers->flags & CON_BOOT)
2665	bcon = console_drivers;
2666
2667	if (preferred_console < 0 || bcon || !console_drivers)
2668	preferred_console = selected_console;
2669
2670	/*
2671	* See if we want to use this console driver. If we
2672	* didn't select a console we take the first one
2673	* that registers here.
2674	*/
2675	if (preferred_console < 0) {
2676	if (newcon->index < 0)
2677	newcon->index = 0;
2678	if (newcon->setup == NULL ||
2679	newcon->setup(newcon, NULL) == 0) {
2680	newcon->flags |= CON_ENABLED;
2681	if (newcon->device) {
2682	newcon->flags |= CON_CONSDEV;
2683	preferred_console = 0;
2684	}
2685	}
2686	}
2687
2688	/*
2689	* See if this console matches one we selected on
2690	* the command line.
2691	*/
2692	for (i = 0, c = console_cmdline;
2693	i < MAX_CMDLINECONSOLES && c->name[0];
2694	i++, c++) {
2695	if (!newcon->match ||
2696	newcon->match(newcon, c->name, c->index, c->options) != 0) {
2697	/* default matching */
2698	BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name));
2699	if (strcmp(c->name, newcon->name) != 0)
2700	continue;
2701	if (newcon->index >= 0 &&
2702	newcon->index != c->index)
2703	continue;
2704	if (newcon->index < 0)
2705	newcon->index = c->index;
2706
2707	if (_braille_register_console(newcon, c))
2708	return;
2709
2710	if (newcon->setup &&
2711	newcon->setup(newcon, c->options) != 0)
2712	break;
2713	}
2714
2715	newcon->flags |= CON_ENABLED;
2716	if (i == selected_console) {
2717	newcon->flags |= CON_CONSDEV;
2718	preferred_console = selected_console;
2719	}
2720	break;
2721	}
2722
2723	if (!(newcon->flags & CON_ENABLED))
2724	return;
2725
2726	/*
2727	* If we have a bootconsole, and are switching to a real console,
2728	* don't print everything out again, since when the boot console, and
2729	* the real console are the same physical device, it's annoying to
2730	* see the beginning boot messages twice
2731	*/
2732	if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2733	newcon->flags &= ~CON_PRINTBUFFER;
2734
2735	/*
2736	* Put this console in the list - keep the
2737	* preferred driver at the head of the list.
2738	*/
2739	console_lock();
2740	if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2741	newcon->next = console_drivers;
2742	console_drivers = newcon;
2743	if (newcon->next)
2744	newcon->next->flags &= ~CON_CONSDEV;
2745	} else {
2746	newcon->next = console_drivers->next;
2747	console_drivers->next = newcon;
2748	}
2749
2750	if (newcon->flags & CON_EXTENDED)
2751	if (!nr_ext_console_drivers++)
2752	pr_info("printk: continuation disabled due to ext consoles, expect more fragments in /dev/kmsg\n");
2753
2754	if (newcon->flags & CON_PRINTBUFFER) {
2755	/*
2756	* console_unlock(); will print out the buffered messages
2757	* for us.
2758	*/
2759	raw_spin_lock_irqsave(&logbuf_lock, flags);
2760	console_seq = syslog_seq;
2761	console_idx = syslog_idx;
2762	console_prev = syslog_prev;
2763	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2764	/*
2765	* We're about to replay the log buffer. Only do this to the
2766	* just-registered console to avoid excessive message spam to
2767	* the already-registered consoles.
2768	*/
2769	exclusive_console = newcon;
2770	}
2771	console_unlock();
2772	console_sysfs_notify();
2773
2774	/*
2775	* By unregistering the bootconsoles after we enable the real console
2776	* we get the "console xxx enabled" message on all the consoles -
2777	* boot consoles, real consoles, etc - this is to ensure that end
2778	* users know there might be something in the kernel's log buffer that
2779	* went to the bootconsole (that they do not see on the real console)
2780	*/
2781	pr_info("%sconsole [%s%d] enabled\n",
2782	(newcon->flags & CON_BOOT) ? "boot" : "" ,
2783	newcon->name, newcon->index);
2784	if (bcon &&
2785	((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2786	!keep_bootcon) {
2787	/* We need to iterate through all boot consoles, to make
2788	* sure we print everything out, before we unregister them.
2789	*/
2790	for_each_console(bcon)
2791	if (bcon->flags & CON_BOOT)
2792	unregister_console(bcon);
2793	}
2794	}
2795	EXPORT_SYMBOL(register_console);
2796
2797	int unregister_console(struct console *console)
2798	{
2799	struct console *a, *b;
2800	int res;
2801
2802	pr_info("%sconsole [%s%d] disabled\n",
2803	(console->flags & CON_BOOT) ? "boot" : "" ,
2804	console->name, console->index);
2805
2806	res = _braille_unregister_console(console);
2807	if (res)
2808	return res;
2809
2810	res = 1;
2811	console_lock();
2812	if (console_drivers == console) {
2813	console_drivers=console->next;
2814	res = 0;
2815	} else if (console_drivers) {
2816	for (a=console_drivers->next, b=console_drivers ;
2817	a; b=a, a=b->next) {
2818	if (a == console) {
2819	b->next = a->next;
2820	res = 0;
2821	break;
2822	}
2823	}
2824	}
2825
2826	if (!res && (console->flags & CON_EXTENDED))
2827	nr_ext_console_drivers--;
2828
2829	/*
2830	* If this isn't the last console and it has CON_CONSDEV set, we
2831	* need to set it on the next preferred console.
2832	*/
2833	if (console_drivers != NULL && console->flags & CON_CONSDEV)
2834	console_drivers->flags |= CON_CONSDEV;
2835
2836	console->flags &= ~CON_ENABLED;
2837	console_unlock();
2838	console_sysfs_notify();
2839	return res;
2840	}
2841	EXPORT_SYMBOL(unregister_console);
2842
2843	/*
2844	* Some boot consoles access data that is in the init section and which will
2845	* be discarded after the initcalls have been run. To make sure that no code
2846	* will access this data, unregister the boot consoles in a late initcall.
2847	*
2848	* If for some reason, such as deferred probe or the driver being a loadable
2849	* module, the real console hasn't registered yet at this point, there will
2850	* be a brief interval in which no messages are logged to the console, which
2851	* makes it difficult to diagnose problems that occur during this time.
2852	*
2853	* To mitigate this problem somewhat, only unregister consoles whose memory
2854	* intersects with the init section. Note that code exists elsewhere to get
2855	* rid of the boot console as soon as the proper console shows up, so there
2856	* won't be side-effects from postponing the removal.
2857	*/
2858	static int __init printk_late_init(void)
2859	{
2860	struct console *con;
2861
2862	for_each_console(con) {
2863	if (!keep_bootcon && con->flags & CON_BOOT) {
2864	/*
2865	* Make sure to unregister boot consoles whose data
2866	* resides in the init section before the init section
2867	* is discarded. Boot consoles whose data will stick
2868	* around will automatically be unregistered when the
2869	* proper console replaces them.
2870	*/
2871	if (init_section_intersects(con, sizeof(*con)))
2872	unregister_console(con);
2873	}
2874	}
2875	hotcpu_notifier(console_cpu_notify, 0);
2876	return 0;
2877	}
2878	late_initcall(printk_late_init);
2879
2880	#if defined CONFIG_PRINTK
2881	/*
2882	* Delayed printk version, for scheduler-internal messages:
2883	*/
2884	#define PRINTK_PENDING_WAKEUP 0x01
2885	#define PRINTK_PENDING_OUTPUT 0x02
2886
2887	static DEFINE_PER_CPU(int, printk_pending);
2888
2889	static void wake_up_klogd_work_func(struct irq_work *irq_work)
2890	{
2891	int pending = __this_cpu_xchg(printk_pending, 0);
2892
2893	if (pending & PRINTK_PENDING_OUTPUT) {
2894	/* If trylock fails, someone else is doing the printing */
2895	if (console_trylock())
2896	console_unlock();
2897	}
2898
2899	if (pending & PRINTK_PENDING_WAKEUP)
2900	wake_up_interruptible(&log_wait);
2901	}
2902
2903	static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
2904	.func = wake_up_klogd_work_func,
2905	.flags = IRQ_WORK_LAZY,
2906	};
2907
2908	void wake_up_klogd(void)
2909	{
2910	preempt_disable();
2911	if (waitqueue_active(&log_wait)) {
2912	this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
2913	irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2914	}
2915	preempt_enable();
2916	}
2917
2918	int printk_deferred(const char *fmt, ...)
2919	{
2920	va_list args;
2921	int r;
2922
2923	preempt_disable();
2924	va_start(args, fmt);
2925	r = vprintk_emit(0, LOGLEVEL_SCHED, NULL, 0, fmt, args);
2926	va_end(args);
2927
2928	__this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
2929	irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2930	preempt_enable();
2931
2932	return r;
2933	}
2934
2935	/*
2936	* printk rate limiting, lifted from the networking subsystem.
2937	*
2938	* This enforces a rate limit: not more than 10 kernel messages
2939	* every 5s to make a denial-of-service attack impossible.
2940	*/
2941	DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2942
2943	int __printk_ratelimit(const char *func)
2944	{
2945	return ___ratelimit(&printk_ratelimit_state, func);
2946	}
2947	EXPORT_SYMBOL(__printk_ratelimit);
2948
2949	/**
2950	* printk_timed_ratelimit - caller-controlled printk ratelimiting
2951	* @caller_jiffies: pointer to caller's state
2952	* @interval_msecs: minimum interval between prints
2953	*
2954	* printk_timed_ratelimit() returns true if more than @interval_msecs
2955	* milliseconds have elapsed since the last time printk_timed_ratelimit()
2956	* returned true.
2957	*/
2958	bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2959	unsigned int interval_msecs)
2960	{
2961	unsigned long elapsed = jiffies - *caller_jiffies;
2962
2963	if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs))
2964	return false;
2965
2966	*caller_jiffies = jiffies;
2967	return true;
2968	}
2969	EXPORT_SYMBOL(printk_timed_ratelimit);
2970
2971	static DEFINE_SPINLOCK(dump_list_lock);
2972	static LIST_HEAD(dump_list);
2973
2974	/**
2975	* kmsg_dump_register - register a kernel log dumper.
2976	* @dumper: pointer to the kmsg_dumper structure
2977	*
2978	* Adds a kernel log dumper to the system. The dump callback in the
2979	* structure will be called when the kernel oopses or panics and must be
2980	* set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2981	*/
2982	int kmsg_dump_register(struct kmsg_dumper *dumper)
2983	{
2984	unsigned long flags;
2985	int err = -EBUSY;
2986
2987	/* The dump callback needs to be set */
2988	if (!dumper->dump)
2989	return -EINVAL;
2990
2991	spin_lock_irqsave(&dump_list_lock, flags);
2992	/* Don't allow registering multiple times */
2993	if (!dumper->registered) {
2994	dumper->registered = 1;
2995	list_add_tail_rcu(&dumper->list, &dump_list);
2996	err = 0;
2997	}
2998	spin_unlock_irqrestore(&dump_list_lock, flags);
2999
3000	return err;
3001	}
3002	EXPORT_SYMBOL_GPL(kmsg_dump_register);
3003
3004	/**
3005	* kmsg_dump_unregister - unregister a kmsg dumper.
3006	* @dumper: pointer to the kmsg_dumper structure
3007	*
3008	* Removes a dump device from the system. Returns zero on success and
3009	* %-EINVAL otherwise.
3010	*/
3011	int kmsg_dump_unregister(struct kmsg_dumper *dumper)
3012	{
3013	unsigned long flags;
3014	int err = -EINVAL;
3015
3016	spin_lock_irqsave(&dump_list_lock, flags);
3017	if (dumper->registered) {
3018	dumper->registered = 0;
3019	list_del_rcu(&dumper->list);
3020	err = 0;
3021	}
3022	spin_unlock_irqrestore(&dump_list_lock, flags);
3023	synchronize_rcu();
3024
3025	return err;
3026	}
3027	EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
3028
3029	static bool always_kmsg_dump;
3030	module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
3031
3032	/**
3033	* kmsg_dump - dump kernel log to kernel message dumpers.
3034	* @reason: the reason (oops, panic etc) for dumping
3035	*
3036	* Call each of the registered dumper's dump() callback, which can
3037	* retrieve the kmsg records with kmsg_dump_get_line() or
3038	* kmsg_dump_get_buffer().
3039	*/
3040	void kmsg_dump(enum kmsg_dump_reason reason)
3041	{
3042	struct kmsg_dumper *dumper;
3043	unsigned long flags;
3044
3045	if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
3046	return;
3047
3048	rcu_read_lock();
3049	list_for_each_entry_rcu(dumper, &dump_list, list) {
3050	if (dumper->max_reason && reason > dumper->max_reason)
3051	continue;
3052
3053	/* initialize iterator with data about the stored records */
3054	dumper->active = true;
3055
3056	raw_spin_lock_irqsave(&logbuf_lock, flags);
3057	dumper->cur_seq = clear_seq;
3058	dumper->cur_idx = clear_idx;
3059	dumper->next_seq = log_next_seq;
3060	dumper->next_idx = log_next_idx;
3061	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
3062
3063	/* invoke dumper which will iterate over records */
3064	dumper->dump(dumper, reason);
3065
3066	/* reset iterator */
3067	dumper->active = false;
3068	}
3069	rcu_read_unlock();
3070	}
3071
3072	/**
3073	* kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3074	* @dumper: registered kmsg dumper
3075	* @syslog: include the "<4>" prefixes
3076	* @line: buffer to copy the line to
3077	* @size: maximum size of the buffer
3078	* @len: length of line placed into buffer
3079	*
3080	* Start at the beginning of the kmsg buffer, with the oldest kmsg
3081	* record, and copy one record into the provided buffer.
3082	*
3083	* Consecutive calls will return the next available record moving
3084	* towards the end of the buffer with the youngest messages.
3085	*
3086	* A return value of FALSE indicates that there are no more records to
3087	* read.
3088	*
3089	* The function is similar to kmsg_dump_get_line(), but grabs no locks.
3090	*/
3091	bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
3092	char *line, size_t size, size_t *len)
3093	{
3094	struct printk_log *msg;
3095	size_t l = 0;
3096	bool ret = false;
3097
3098	if (!dumper->active)
3099	goto out;
3100
3101	if (dumper->cur_seq < log_first_seq) {
3102	/* messages are gone, move to first available one */
3103	dumper->cur_seq = log_first_seq;
3104	dumper->cur_idx = log_first_idx;
3105	}
3106
3107	/* last entry */
3108	if (dumper->cur_seq >= log_next_seq)
3109	goto out;
3110
3111	msg = log_from_idx(dumper->cur_idx);
3112	l = msg_print_text(msg, 0, syslog, line, size);
3113
3114	dumper->cur_idx = log_next(dumper->cur_idx);
3115	dumper->cur_seq++;
3116	ret = true;
3117	out:
3118	if (len)
3119	*len = l;
3120	return ret;
3121	}
3122
3123	/**
3124	* kmsg_dump_get_line - retrieve one kmsg log line
3125	* @dumper: registered kmsg dumper
3126	* @syslog: include the "<4>" prefixes
3127	* @line: buffer to copy the line to
3128	* @size: maximum size of the buffer
3129	* @len: length of line placed into buffer
3130	*
3131	* Start at the beginning of the kmsg buffer, with the oldest kmsg
3132	* record, and copy one record into the provided buffer.
3133	*
3134	* Consecutive calls will return the next available record moving
3135	* towards the end of the buffer with the youngest messages.
3136	*
3137	* A return value of FALSE indicates that there are no more records to
3138	* read.
3139	*/
3140	bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
3141	char *line, size_t size, size_t *len)
3142	{
3143	unsigned long flags;
3144	bool ret;
3145
3146	raw_spin_lock_irqsave(&logbuf_lock, flags);
3147	ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
3148	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
3149
3150	return ret;
3151	}
3152	EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
3153
3154	/**
3155	* kmsg_dump_get_buffer - copy kmsg log lines
3156	* @dumper: registered kmsg dumper
3157	* @syslog: include the "<4>" prefixes
3158	* @buf: buffer to copy the line to
3159	* @size: maximum size of the buffer
3160	* @len: length of line placed into buffer
3161	*
3162	* Start at the end of the kmsg buffer and fill the provided buffer
3163	* with as many of the the *youngest* kmsg records that fit into it.
3164	* If the buffer is large enough, all available kmsg records will be
3165	* copied with a single call.
3166	*
3167	* Consecutive calls will fill the buffer with the next block of
3168	* available older records, not including the earlier retrieved ones.
3169	*
3170	* A return value of FALSE indicates that there are no more records to
3171	* read.
3172	*/
3173	bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
3174	char *buf, size_t size, size_t *len)
3175	{
3176	unsigned long flags;
3177	u64 seq;
3178	u32 idx;
3179	u64 next_seq;
3180	u32 next_idx;
3181	enum log_flags prev;
3182	size_t l = 0;
3183	bool ret = false;
3184
3185	if (!dumper->active)
3186	goto out;
3187
3188	raw_spin_lock_irqsave(&logbuf_lock, flags);
3189	if (dumper->cur_seq < log_first_seq) {
3190	/* messages are gone, move to first available one */
3191	dumper->cur_seq = log_first_seq;
3192	dumper->cur_idx = log_first_idx;
3193	}
3194
3195	/* last entry */
3196	if (dumper->cur_seq >= dumper->next_seq) {
3197	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
3198	goto out;
3199	}
3200
3201	/* calculate length of entire buffer */
3202	seq = dumper->cur_seq;
3203	idx = dumper->cur_idx;
3204	prev = 0;
3205	while (seq < dumper->next_seq) {
3206	struct printk_log *msg = log_from_idx(idx);
3207
3208	l += msg_print_text(msg, prev, true, NULL, 0);
3209	idx = log_next(idx);
3210	seq++;
3211	prev = msg->flags;
3212	}
3213
3214	/* move first record forward until length fits into the buffer */
3215	seq = dumper->cur_seq;
3216	idx = dumper->cur_idx;
3217	prev = 0;
3218	while (l > size && seq < dumper->next_seq) {
3219	struct printk_log *msg = log_from_idx(idx);
3220
3221	l -= msg_print_text(msg, prev, true, NULL, 0);
3222	idx = log_next(idx);
3223	seq++;
3224	prev = msg->flags;
3225	}
3226
3227	/* last message in next interation */
3228	next_seq = seq;
3229	next_idx = idx;
3230
3231	l = 0;
3232	while (seq < dumper->next_seq) {
3233	struct printk_log *msg = log_from_idx(idx);
3234
3235	l += msg_print_text(msg, prev, syslog, buf + l, size - l);
3236	idx = log_next(idx);
3237	seq++;
3238	prev = msg->flags;
3239	}
3240
3241	dumper->next_seq = next_seq;
3242	dumper->next_idx = next_idx;
3243	ret = true;
3244	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
3245	out:
3246	if (len)
3247	*len = l;
3248	return ret;
3249	}
3250	EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
3251
3252	/**
3253	* kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3254	* @dumper: registered kmsg dumper
3255	*
3256	* Reset the dumper's iterator so that kmsg_dump_get_line() and
3257	* kmsg_dump_get_buffer() can be called again and used multiple
3258	* times within the same dumper.dump() callback.
3259	*
3260	* The function is similar to kmsg_dump_rewind(), but grabs no locks.
3261	*/
3262	void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
3263	{
3264	dumper->cur_seq = clear_seq;
3265	dumper->cur_idx = clear_idx;
3266	dumper->next_seq = log_next_seq;
3267	dumper->next_idx = log_next_idx;
3268	}
3269
3270	/**
3271	* kmsg_dump_rewind - reset the interator
3272	* @dumper: registered kmsg dumper
3273	*
3274	* Reset the dumper's iterator so that kmsg_dump_get_line() and
3275	* kmsg_dump_get_buffer() can be called again and used multiple
3276	* times within the same dumper.dump() callback.
3277	*/
3278	void kmsg_dump_rewind(struct kmsg_dumper *dumper)
3279	{
3280	unsigned long flags;
3281
3282	raw_spin_lock_irqsave(&logbuf_lock, flags);
3283	kmsg_dump_rewind_nolock(dumper);
3284	raw_spin_unlock_irqrestore(&logbuf_lock, flags);
3285	}
3286	EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
3287
3288	static char dump_stack_arch_desc_str[128];
3289
3290	/**
3291	* dump_stack_set_arch_desc - set arch-specific str to show with task dumps
3292	* @fmt: printf-style format string
3293	* @...: arguments for the format string
3294	*
3295	* The configured string will be printed right after utsname during task
3296	* dumps. Usually used to add arch-specific system identifiers. If an
3297	* arch wants to make use of such an ID string, it should initialize this
3298	* as soon as possible during boot.
3299	*/
3300	void __init dump_stack_set_arch_desc(const char *fmt, ...)
3301	{
3302	va_list args;
3303
3304	va_start(args, fmt);
3305	vsnprintf(dump_stack_arch_desc_str, sizeof(dump_stack_arch_desc_str),
3306	fmt, args);
3307	va_end(args);
3308	}
3309
3310	/**
3311	* dump_stack_print_info - print generic debug info for dump_stack()
3312	* @log_lvl: log level
3313	*
3314	* Arch-specific dump_stack() implementations can use this function to
3315	* print out the same debug information as the generic dump_stack().
3316	*/
3317	void dump_stack_print_info(const char *log_lvl)
3318	{
3319	printk("%sCPU: %d PID: %d Comm: %.20s %s %s %.*s\n",
3320	log_lvl, raw_smp_processor_id(), current->pid, current->comm,
3321	print_tainted(), init_utsname()->release,
3322	(int)strcspn(init_utsname()->version, " "),
3323	init_utsname()->version);
3324
3325	if (dump_stack_arch_desc_str[0] != '\0')
3326	printk("%sHardware name: %s\n",
3327	log_lvl, dump_stack_arch_desc_str);
3328
3329	print_worker_info(log_lvl, current);
3330	}
3331
3332	/**
3333	* show_regs_print_info - print generic debug info for show_regs()
3334	* @log_lvl: log level
3335	*
3336	* show_regs() implementations can use this function to print out generic
3337	* debug information.
3338	*/
3339	void show_regs_print_info(const char *log_lvl)
3340	{
3341	dump_stack_print_info(log_lvl);
3342
3343	printk("%stask: %p task.stack: %p\n",
3344	log_lvl, current, task_stack_page(current));
3345	}
3346
3347	#endif
3348