path: root/kernel/printk/printk.c (plain)
blob: b67e84cfd085c8cee1e7fd352ef9df404f3180e4

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