2 * linux/kernel/printk.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
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).
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
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/kernel.h>
23 #include <linux/tty.h>
24 #include <linux/tty_driver.h>
25 #include <linux/console.h>
26 #include <linux/init.h>
27 #include <linux/jiffies.h>
28 #include <linux/nmi.h>
29 #include <linux/module.h>
30 #include <linux/moduleparam.h>
31 #include <linux/delay.h>
32 #include <linux/smp.h>
33 #include <linux/security.h>
34 #include <linux/memblock.h>
35 #include <linux/syscalls.h>
36 #include <linux/crash_core.h>
37 #include <linux/kdb.h>
38 #include <linux/ratelimit.h>
39 #include <linux/kmsg_dump.h>
40 #include <linux/syslog.h>
41 #include <linux/cpu.h>
42 #include <linux/rculist.h>
43 #include <linux/poll.h>
44 #include <linux/irq_work.h>
45 #include <linux/ctype.h>
46 #include <linux/uio.h>
47 #include <linux/sched/clock.h>
48 #include <linux/sched/debug.h>
49 #include <linux/sched/task_stack.h>
51 #include <linux/uaccess.h>
52 #include <asm/sections.h>
54 #include <trace/events/initcall.h>
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/printk.h>
58 #include "console_cmdline.h"
62 int console_printk
[4] = {
63 CONSOLE_LOGLEVEL_DEFAULT
, /* console_loglevel */
64 MESSAGE_LOGLEVEL_DEFAULT
, /* default_message_loglevel */
65 CONSOLE_LOGLEVEL_MIN
, /* minimum_console_loglevel */
66 CONSOLE_LOGLEVEL_DEFAULT
, /* default_console_loglevel */
68 EXPORT_SYMBOL_GPL(console_printk
);
70 atomic_t ignore_console_lock_warning __read_mostly
= ATOMIC_INIT(0);
71 EXPORT_SYMBOL(ignore_console_lock_warning
);
74 * Low level drivers may need that to know if they can schedule in
75 * their unblank() callback or not. So let's export it.
78 EXPORT_SYMBOL(oops_in_progress
);
81 * console_sem protects the console_drivers list, and also
82 * provides serialisation for access to the entire console
85 static DEFINE_SEMAPHORE(console_sem
);
86 struct console
*console_drivers
;
87 EXPORT_SYMBOL_GPL(console_drivers
);
90 static struct lockdep_map console_lock_dep_map
= {
91 .name
= "console_lock"
95 enum devkmsg_log_bits
{
96 __DEVKMSG_LOG_BIT_ON
= 0,
97 __DEVKMSG_LOG_BIT_OFF
,
98 __DEVKMSG_LOG_BIT_LOCK
,
101 enum devkmsg_log_masks
{
102 DEVKMSG_LOG_MASK_ON
= BIT(__DEVKMSG_LOG_BIT_ON
),
103 DEVKMSG_LOG_MASK_OFF
= BIT(__DEVKMSG_LOG_BIT_OFF
),
104 DEVKMSG_LOG_MASK_LOCK
= BIT(__DEVKMSG_LOG_BIT_LOCK
),
107 /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
108 #define DEVKMSG_LOG_MASK_DEFAULT 0
110 static unsigned int __read_mostly devkmsg_log
= DEVKMSG_LOG_MASK_DEFAULT
;
112 static int __control_devkmsg(char *str
)
117 if (!strncmp(str
, "on", 2)) {
118 devkmsg_log
= DEVKMSG_LOG_MASK_ON
;
120 } else if (!strncmp(str
, "off", 3)) {
121 devkmsg_log
= DEVKMSG_LOG_MASK_OFF
;
123 } else if (!strncmp(str
, "ratelimit", 9)) {
124 devkmsg_log
= DEVKMSG_LOG_MASK_DEFAULT
;
130 static int __init
control_devkmsg(char *str
)
132 if (__control_devkmsg(str
) < 0)
136 * Set sysctl string accordingly:
138 if (devkmsg_log
== DEVKMSG_LOG_MASK_ON
)
139 strcpy(devkmsg_log_str
, "on");
140 else if (devkmsg_log
== DEVKMSG_LOG_MASK_OFF
)
141 strcpy(devkmsg_log_str
, "off");
142 /* else "ratelimit" which is set by default. */
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.
150 devkmsg_log
|= DEVKMSG_LOG_MASK_LOCK
;
154 __setup("printk.devkmsg=", control_devkmsg
);
156 char devkmsg_log_str
[DEVKMSG_STR_MAX_SIZE
] = "ratelimit";
158 int devkmsg_sysctl_set_loglvl(struct ctl_table
*table
, int write
,
159 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
161 char old_str
[DEVKMSG_STR_MAX_SIZE
];
166 if (devkmsg_log
& DEVKMSG_LOG_MASK_LOCK
)
170 strncpy(old_str
, devkmsg_log_str
, DEVKMSG_STR_MAX_SIZE
);
173 err
= proc_dostring(table
, write
, buffer
, lenp
, ppos
);
178 err
= __control_devkmsg(devkmsg_log_str
);
181 * Do not accept an unknown string OR a known string with
184 if (err
< 0 || (err
+ 1 != *lenp
)) {
186 /* ... and restore old setting. */
188 strncpy(devkmsg_log_str
, old_str
, DEVKMSG_STR_MAX_SIZE
);
197 /* Number of registered extended console drivers. */
198 static int nr_ext_console_drivers
;
201 * Helper macros to handle lockdep when locking/unlocking console_sem. We use
202 * macros instead of functions so that _RET_IP_ contains useful information.
204 #define down_console_sem() do { \
206 mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
209 static int __down_trylock_console_sem(unsigned long ip
)
215 * Here and in __up_console_sem() we need to be in safe mode,
216 * because spindump/WARN/etc from under console ->lock will
217 * deadlock in printk()->down_trylock_console_sem() otherwise.
219 printk_safe_enter_irqsave(flags
);
220 lock_failed
= down_trylock(&console_sem
);
221 printk_safe_exit_irqrestore(flags
);
225 mutex_acquire(&console_lock_dep_map
, 0, 1, ip
);
228 #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
230 static void __up_console_sem(unsigned long ip
)
234 mutex_release(&console_lock_dep_map
, 1, ip
);
236 printk_safe_enter_irqsave(flags
);
238 printk_safe_exit_irqrestore(flags
);
240 #define up_console_sem() __up_console_sem(_RET_IP_)
243 * This is used for debugging the mess that is the VT code by
244 * keeping track if we have the console semaphore held. It's
245 * definitely not the perfect debug tool (we don't know if _WE_
246 * hold it and are racing, but it helps tracking those weird code
247 * paths in the console code where we end up in places I want
248 * locked without the console sempahore held).
250 static int console_locked
, console_suspended
;
253 * If exclusive_console is non-NULL then only this console is to be printed to.
255 static struct console
*exclusive_console
;
258 * Array of consoles built from command line options (console=)
261 #define MAX_CMDLINECONSOLES 8
263 static struct console_cmdline console_cmdline
[MAX_CMDLINECONSOLES
];
265 static int preferred_console
= -1;
266 int console_set_on_cmdline
;
267 EXPORT_SYMBOL(console_set_on_cmdline
);
269 /* Flag: console code may call schedule() */
270 static int console_may_schedule
;
272 enum con_msg_format_flags
{
273 MSG_FORMAT_DEFAULT
= 0,
274 MSG_FORMAT_SYSLOG
= (1 << 0),
277 static int console_msg_format
= MSG_FORMAT_DEFAULT
;
280 * The printk log buffer consists of a chain of concatenated variable
281 * length records. Every record starts with a record header, containing
282 * the overall length of the record.
284 * The heads to the first and last entry in the buffer, as well as the
285 * sequence numbers of these entries are maintained when messages are
288 * If the heads indicate available messages, the length in the header
289 * tells the start next message. A length == 0 for the next message
290 * indicates a wrap-around to the beginning of the buffer.
292 * Every record carries the monotonic timestamp in microseconds, as well as
293 * the standard userspace syslog level and syslog facility. The usual
294 * kernel messages use LOG_KERN; userspace-injected messages always carry
295 * a matching syslog facility, by default LOG_USER. The origin of every
296 * message can be reliably determined that way.
298 * The human readable log message directly follows the message header. The
299 * length of the message text is stored in the header, the stored message
302 * Optionally, a message can carry a dictionary of properties (key/value pairs),
303 * to provide userspace with a machine-readable message context.
305 * Examples for well-defined, commonly used property names are:
306 * DEVICE=b12:8 device identifier
310 * +sound:card0 subsystem:devname
311 * SUBSYSTEM=pci driver-core subsystem name
313 * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
314 * follows directly after a '=' character. Every property is terminated by
315 * a '\0' character. The last property is not terminated.
317 * Example of a message structure:
318 * 0000 ff 8f 00 00 00 00 00 00 monotonic time in nsec
319 * 0008 34 00 record is 52 bytes long
320 * 000a 0b 00 text is 11 bytes long
321 * 000c 1f 00 dictionary is 23 bytes long
322 * 000e 03 00 LOG_KERN (facility) LOG_ERR (level)
323 * 0010 69 74 27 73 20 61 20 6c "it's a l"
325 * 001b 44 45 56 49 43 "DEVIC"
326 * 45 3d 62 38 3a 32 00 44 "E=b8:2\0D"
327 * 52 49 56 45 52 3d 62 75 "RIVER=bu"
329 * 0032 00 00 00 padding to next message header
331 * The 'struct printk_log' buffer header must never be directly exported to
332 * userspace, it is a kernel-private implementation detail that might
333 * need to be changed in the future, when the requirements change.
335 * /dev/kmsg exports the structured data in the following line format:
336 * "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
338 * Users of the export format should ignore possible additional values
339 * separated by ',', and find the message after the ';' character.
341 * The optional key/value pairs are attached as continuation lines starting
342 * with a space character and terminated by a newline. All possible
343 * non-prinatable characters are escaped in the "\xff" notation.
347 LOG_NEWLINE
= 2, /* text ended with a newline */
348 LOG_CONT
= 8, /* text is a fragment of a continuation line */
352 u64 ts_nsec
; /* timestamp in nanoseconds */
353 u16 len
; /* length of entire record */
354 u16 text_len
; /* length of text buffer */
355 u16 dict_len
; /* length of dictionary buffer */
356 u8 facility
; /* syslog facility */
357 u8 flags
:5; /* internal record flags */
358 u8 level
:3; /* syslog level */
359 #ifdef CONFIG_PRINTK_CALLER
360 u32 caller_id
; /* thread id or processor id */
363 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
364 __packed
__aligned(4)
369 * The logbuf_lock protects kmsg buffer, indices, counters. This can be taken
370 * within the scheduler's rq lock. It must be released before calling
371 * console_unlock() or anything else that might wake up a process.
373 DEFINE_RAW_SPINLOCK(logbuf_lock
);
376 * Helper macros to lock/unlock logbuf_lock and switch between
377 * printk-safe/unsafe modes.
379 #define logbuf_lock_irq() \
381 printk_safe_enter_irq(); \
382 raw_spin_lock(&logbuf_lock); \
385 #define logbuf_unlock_irq() \
387 raw_spin_unlock(&logbuf_lock); \
388 printk_safe_exit_irq(); \
391 #define logbuf_lock_irqsave(flags) \
393 printk_safe_enter_irqsave(flags); \
394 raw_spin_lock(&logbuf_lock); \
397 #define logbuf_unlock_irqrestore(flags) \
399 raw_spin_unlock(&logbuf_lock); \
400 printk_safe_exit_irqrestore(flags); \
404 DECLARE_WAIT_QUEUE_HEAD(log_wait
);
405 /* the next printk record to read by syslog(READ) or /proc/kmsg */
406 static u64 syslog_seq
;
407 static u32 syslog_idx
;
408 static size_t syslog_partial
;
409 static bool syslog_time
;
411 /* index and sequence number of the first record stored in the buffer */
412 static u64 log_first_seq
;
413 static u32 log_first_idx
;
415 /* index and sequence number of the next record to store in the buffer */
416 static u64 log_next_seq
;
417 static u32 log_next_idx
;
419 /* the next printk record to write to the console */
420 static u64 console_seq
;
421 static u32 console_idx
;
422 static u64 exclusive_console_stop_seq
;
424 /* the next printk record to read after the last 'clear' command */
425 static u64 clear_seq
;
426 static u32 clear_idx
;
428 #ifdef CONFIG_PRINTK_CALLER
429 #define PREFIX_MAX 48
431 #define PREFIX_MAX 32
433 #define LOG_LINE_MAX (1024 - PREFIX_MAX)
435 #define LOG_LEVEL(v) ((v) & 0x07)
436 #define LOG_FACILITY(v) ((v) >> 3 & 0xff)
439 #define LOG_ALIGN __alignof__(struct printk_log)
440 #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
441 #define LOG_BUF_LEN_MAX (u32)(1 << 31)
442 static char __log_buf
[__LOG_BUF_LEN
] __aligned(LOG_ALIGN
);
443 static char *log_buf
= __log_buf
;
444 static u32 log_buf_len
= __LOG_BUF_LEN
;
446 /* Return log buffer address */
447 char *log_buf_addr_get(void)
452 /* Return log buffer size */
453 u32
log_buf_len_get(void)
458 /* human readable text of the record */
459 static char *log_text(const struct printk_log
*msg
)
461 return (char *)msg
+ sizeof(struct printk_log
);
464 /* optional key/value pair dictionary attached to the record */
465 static char *log_dict(const struct printk_log
*msg
)
467 return (char *)msg
+ sizeof(struct printk_log
) + msg
->text_len
;
470 /* get record by index; idx must point to valid msg */
471 static struct printk_log
*log_from_idx(u32 idx
)
473 struct printk_log
*msg
= (struct printk_log
*)(log_buf
+ idx
);
476 * A length == 0 record is the end of buffer marker. Wrap around and
477 * read the message at the start of the buffer.
480 return (struct printk_log
*)log_buf
;
484 /* get next record; idx must point to valid msg */
485 static u32
log_next(u32 idx
)
487 struct printk_log
*msg
= (struct printk_log
*)(log_buf
+ idx
);
489 /* length == 0 indicates the end of the buffer; wrap */
491 * A length == 0 record is the end of buffer marker. Wrap around and
492 * read the message at the start of the buffer as *this* one, and
493 * return the one after that.
496 msg
= (struct printk_log
*)log_buf
;
499 return idx
+ msg
->len
;
503 * Check whether there is enough free space for the given message.
505 * The same values of first_idx and next_idx mean that the buffer
506 * is either empty or full.
508 * If the buffer is empty, we must respect the position of the indexes.
509 * They cannot be reset to the beginning of the buffer.
511 static int logbuf_has_space(u32 msg_size
, bool empty
)
515 if (log_next_idx
> log_first_idx
|| empty
)
516 free
= max(log_buf_len
- log_next_idx
, log_first_idx
);
518 free
= log_first_idx
- log_next_idx
;
521 * We need space also for an empty header that signalizes wrapping
524 return free
>= msg_size
+ sizeof(struct printk_log
);
527 static int log_make_free_space(u32 msg_size
)
529 while (log_first_seq
< log_next_seq
&&
530 !logbuf_has_space(msg_size
, false)) {
531 /* drop old messages until we have enough contiguous space */
532 log_first_idx
= log_next(log_first_idx
);
536 if (clear_seq
< log_first_seq
) {
537 clear_seq
= log_first_seq
;
538 clear_idx
= log_first_idx
;
541 /* sequence numbers are equal, so the log buffer is empty */
542 if (logbuf_has_space(msg_size
, log_first_seq
== log_next_seq
))
548 /* compute the message size including the padding bytes */
549 static u32
msg_used_size(u16 text_len
, u16 dict_len
, u32
*pad_len
)
553 size
= sizeof(struct printk_log
) + text_len
+ dict_len
;
554 *pad_len
= (-size
) & (LOG_ALIGN
- 1);
561 * Define how much of the log buffer we could take at maximum. The value
562 * must be greater than two. Note that only half of the buffer is available
563 * when the index points to the middle.
565 #define MAX_LOG_TAKE_PART 4
566 static const char trunc_msg
[] = "<truncated>";
568 static u32
truncate_msg(u16
*text_len
, u16
*trunc_msg_len
,
569 u16
*dict_len
, u32
*pad_len
)
572 * The message should not take the whole buffer. Otherwise, it might
573 * get removed too soon.
575 u32 max_text_len
= log_buf_len
/ MAX_LOG_TAKE_PART
;
576 if (*text_len
> max_text_len
)
577 *text_len
= max_text_len
;
578 /* enable the warning message */
579 *trunc_msg_len
= strlen(trunc_msg
);
580 /* disable the "dict" completely */
582 /* compute the size again, count also the warning message */
583 return msg_used_size(*text_len
+ *trunc_msg_len
, 0, pad_len
);
586 /* insert record into the buffer, discard old ones, update heads */
587 static int log_store(u32 caller_id
, int facility
, int level
,
588 enum log_flags flags
, u64 ts_nsec
,
589 const char *dict
, u16 dict_len
,
590 const char *text
, u16 text_len
)
592 struct printk_log
*msg
;
594 u16 trunc_msg_len
= 0;
596 /* number of '\0' padding bytes to next message */
597 size
= msg_used_size(text_len
, dict_len
, &pad_len
);
599 if (log_make_free_space(size
)) {
600 /* truncate the message if it is too long for empty buffer */
601 size
= truncate_msg(&text_len
, &trunc_msg_len
,
602 &dict_len
, &pad_len
);
603 /* survive when the log buffer is too small for trunc_msg */
604 if (log_make_free_space(size
))
608 if (log_next_idx
+ size
+ sizeof(struct printk_log
) > log_buf_len
) {
610 * This message + an additional empty header does not fit
611 * at the end of the buffer. Add an empty header with len == 0
612 * to signify a wrap around.
614 memset(log_buf
+ log_next_idx
, 0, sizeof(struct printk_log
));
619 msg
= (struct printk_log
*)(log_buf
+ log_next_idx
);
620 memcpy(log_text(msg
), text
, text_len
);
621 msg
->text_len
= text_len
;
623 memcpy(log_text(msg
) + text_len
, trunc_msg
, trunc_msg_len
);
624 msg
->text_len
+= trunc_msg_len
;
626 memcpy(log_dict(msg
), dict
, dict_len
);
627 msg
->dict_len
= dict_len
;
628 msg
->facility
= facility
;
629 msg
->level
= level
& 7;
630 msg
->flags
= flags
& 0x1f;
632 msg
->ts_nsec
= ts_nsec
;
634 msg
->ts_nsec
= local_clock();
635 #ifdef CONFIG_PRINTK_CALLER
636 msg
->caller_id
= caller_id
;
638 memset(log_dict(msg
) + dict_len
, 0, pad_len
);
642 log_next_idx
+= msg
->len
;
645 return msg
->text_len
;
648 int dmesg_restrict
= IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT
);
650 static int syslog_action_restricted(int type
)
655 * Unless restricted, we allow "read all" and "get buffer size"
658 return type
!= SYSLOG_ACTION_READ_ALL
&&
659 type
!= SYSLOG_ACTION_SIZE_BUFFER
;
662 static int check_syslog_permissions(int type
, int source
)
665 * If this is from /proc/kmsg and we've already opened it, then we've
666 * already done the capabilities checks at open time.
668 if (source
== SYSLOG_FROM_PROC
&& type
!= SYSLOG_ACTION_OPEN
)
671 if (syslog_action_restricted(type
)) {
672 if (capable(CAP_SYSLOG
))
675 * For historical reasons, accept CAP_SYS_ADMIN too, with
678 if (capable(CAP_SYS_ADMIN
)) {
679 pr_warn_once("%s (%d): Attempt to access syslog with "
680 "CAP_SYS_ADMIN but no CAP_SYSLOG "
682 current
->comm
, task_pid_nr(current
));
688 return security_syslog(type
);
691 static void append_char(char **pp
, char *e
, char c
)
697 static ssize_t
msg_print_ext_header(char *buf
, size_t size
,
698 struct printk_log
*msg
, u64 seq
)
700 u64 ts_usec
= msg
->ts_nsec
;
702 #ifdef CONFIG_PRINTK_CALLER
703 u32 id
= msg
->caller_id
;
705 snprintf(caller
, sizeof(caller
), ",caller=%c%u",
706 id
& 0x80000000 ? 'C' : 'T', id
& ~0x80000000);
711 do_div(ts_usec
, 1000);
713 return scnprintf(buf
, size
, "%u,%llu,%llu,%c%s;",
714 (msg
->facility
<< 3) | msg
->level
, seq
, ts_usec
,
715 msg
->flags
& LOG_CONT
? 'c' : '-', caller
);
718 static ssize_t
msg_print_ext_body(char *buf
, size_t size
,
719 char *dict
, size_t dict_len
,
720 char *text
, size_t text_len
)
722 char *p
= buf
, *e
= buf
+ size
;
725 /* escape non-printable characters */
726 for (i
= 0; i
< text_len
; i
++) {
727 unsigned char c
= text
[i
];
729 if (c
< ' ' || c
>= 127 || c
== '\\')
730 p
+= scnprintf(p
, e
- p
, "\\x%02x", c
);
732 append_char(&p
, e
, c
);
734 append_char(&p
, e
, '\n');
739 for (i
= 0; i
< dict_len
; i
++) {
740 unsigned char c
= dict
[i
];
743 append_char(&p
, e
, ' ');
748 append_char(&p
, e
, '\n');
753 if (c
< ' ' || c
>= 127 || c
== '\\') {
754 p
+= scnprintf(p
, e
- p
, "\\x%02x", c
);
758 append_char(&p
, e
, c
);
760 append_char(&p
, e
, '\n');
766 /* /dev/kmsg - userspace message inject/listen interface */
767 struct devkmsg_user
{
770 struct ratelimit_state rs
;
772 char buf
[CONSOLE_EXT_LOG_MAX
];
775 static __printf(3, 4) __cold
776 int devkmsg_emit(int facility
, int level
, const char *fmt
, ...)
782 r
= vprintk_emit(facility
, level
, NULL
, 0, fmt
, args
);
788 static ssize_t
devkmsg_write(struct kiocb
*iocb
, struct iov_iter
*from
)
791 int level
= default_message_loglevel
;
792 int facility
= 1; /* LOG_USER */
793 struct file
*file
= iocb
->ki_filp
;
794 struct devkmsg_user
*user
= file
->private_data
;
795 size_t len
= iov_iter_count(from
);
798 if (!user
|| len
> LOG_LINE_MAX
)
801 /* Ignore when user logging is disabled. */
802 if (devkmsg_log
& DEVKMSG_LOG_MASK_OFF
)
805 /* Ratelimit when not explicitly enabled. */
806 if (!(devkmsg_log
& DEVKMSG_LOG_MASK_ON
)) {
807 if (!___ratelimit(&user
->rs
, current
->comm
))
811 buf
= kmalloc(len
+1, GFP_KERNEL
);
816 if (!copy_from_iter_full(buf
, len
, from
)) {
822 * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
823 * the decimal value represents 32bit, the lower 3 bit are the log
824 * level, the rest are the log facility.
826 * If no prefix or no userspace facility is specified, we
827 * enforce LOG_USER, to be able to reliably distinguish
828 * kernel-generated messages from userspace-injected ones.
831 if (line
[0] == '<') {
835 u
= simple_strtoul(line
+ 1, &endp
, 10);
836 if (endp
&& endp
[0] == '>') {
837 level
= LOG_LEVEL(u
);
838 if (LOG_FACILITY(u
) != 0)
839 facility
= LOG_FACILITY(u
);
846 devkmsg_emit(facility
, level
, "%s", line
);
851 static ssize_t
devkmsg_read(struct file
*file
, char __user
*buf
,
852 size_t count
, loff_t
*ppos
)
854 struct devkmsg_user
*user
= file
->private_data
;
855 struct printk_log
*msg
;
862 ret
= mutex_lock_interruptible(&user
->lock
);
867 while (user
->seq
== log_next_seq
) {
868 if (file
->f_flags
& O_NONBLOCK
) {
875 ret
= wait_event_interruptible(log_wait
,
876 user
->seq
!= log_next_seq
);
882 if (user
->seq
< log_first_seq
) {
883 /* our last seen message is gone, return error and reset */
884 user
->idx
= log_first_idx
;
885 user
->seq
= log_first_seq
;
891 msg
= log_from_idx(user
->idx
);
892 len
= msg_print_ext_header(user
->buf
, sizeof(user
->buf
),
894 len
+= msg_print_ext_body(user
->buf
+ len
, sizeof(user
->buf
) - len
,
895 log_dict(msg
), msg
->dict_len
,
896 log_text(msg
), msg
->text_len
);
898 user
->idx
= log_next(user
->idx
);
907 if (copy_to_user(buf
, user
->buf
, len
)) {
913 mutex_unlock(&user
->lock
);
917 static loff_t
devkmsg_llseek(struct file
*file
, loff_t offset
, int whence
)
919 struct devkmsg_user
*user
= file
->private_data
;
930 /* the first record */
931 user
->idx
= log_first_idx
;
932 user
->seq
= log_first_seq
;
936 * The first record after the last SYSLOG_ACTION_CLEAR,
937 * like issued by 'dmesg -c'. Reading /dev/kmsg itself
938 * changes no global state, and does not clear anything.
940 user
->idx
= clear_idx
;
941 user
->seq
= clear_seq
;
944 /* after the last record */
945 user
->idx
= log_next_idx
;
946 user
->seq
= log_next_seq
;
955 static __poll_t
devkmsg_poll(struct file
*file
, poll_table
*wait
)
957 struct devkmsg_user
*user
= file
->private_data
;
961 return EPOLLERR
|EPOLLNVAL
;
963 poll_wait(file
, &log_wait
, wait
);
966 if (user
->seq
< log_next_seq
) {
967 /* return error when data has vanished underneath us */
968 if (user
->seq
< log_first_seq
)
969 ret
= EPOLLIN
|EPOLLRDNORM
|EPOLLERR
|EPOLLPRI
;
971 ret
= EPOLLIN
|EPOLLRDNORM
;
978 static int devkmsg_open(struct inode
*inode
, struct file
*file
)
980 struct devkmsg_user
*user
;
983 if (devkmsg_log
& DEVKMSG_LOG_MASK_OFF
)
986 /* write-only does not need any file context */
987 if ((file
->f_flags
& O_ACCMODE
) != O_WRONLY
) {
988 err
= check_syslog_permissions(SYSLOG_ACTION_READ_ALL
,
994 user
= kmalloc(sizeof(struct devkmsg_user
), GFP_KERNEL
);
998 ratelimit_default_init(&user
->rs
);
999 ratelimit_set_flags(&user
->rs
, RATELIMIT_MSG_ON_RELEASE
);
1001 mutex_init(&user
->lock
);
1004 user
->idx
= log_first_idx
;
1005 user
->seq
= log_first_seq
;
1006 logbuf_unlock_irq();
1008 file
->private_data
= user
;
1012 static int devkmsg_release(struct inode
*inode
, struct file
*file
)
1014 struct devkmsg_user
*user
= file
->private_data
;
1019 ratelimit_state_exit(&user
->rs
);
1021 mutex_destroy(&user
->lock
);
1026 const struct file_operations kmsg_fops
= {
1027 .open
= devkmsg_open
,
1028 .read
= devkmsg_read
,
1029 .write_iter
= devkmsg_write
,
1030 .llseek
= devkmsg_llseek
,
1031 .poll
= devkmsg_poll
,
1032 .release
= devkmsg_release
,
1035 #ifdef CONFIG_CRASH_CORE
1037 * This appends the listed symbols to /proc/vmcore
1039 * /proc/vmcore is used by various utilities, like crash and makedumpfile to
1040 * obtain access to symbols that are otherwise very difficult to locate. These
1041 * symbols are specifically used so that utilities can access and extract the
1042 * dmesg log from a vmcore file after a crash.
1044 void log_buf_vmcoreinfo_setup(void)
1046 VMCOREINFO_SYMBOL(log_buf
);
1047 VMCOREINFO_SYMBOL(log_buf_len
);
1048 VMCOREINFO_SYMBOL(log_first_idx
);
1049 VMCOREINFO_SYMBOL(clear_idx
);
1050 VMCOREINFO_SYMBOL(log_next_idx
);
1052 * Export struct printk_log size and field offsets. User space tools can
1053 * parse it and detect any changes to structure down the line.
1055 VMCOREINFO_STRUCT_SIZE(printk_log
);
1056 VMCOREINFO_OFFSET(printk_log
, ts_nsec
);
1057 VMCOREINFO_OFFSET(printk_log
, len
);
1058 VMCOREINFO_OFFSET(printk_log
, text_len
);
1059 VMCOREINFO_OFFSET(printk_log
, dict_len
);
1060 #ifdef CONFIG_PRINTK_CALLER
1061 VMCOREINFO_OFFSET(printk_log
, caller_id
);
1066 /* requested log_buf_len from kernel cmdline */
1067 static unsigned long __initdata new_log_buf_len
;
1069 /* we practice scaling the ring buffer by powers of 2 */
1070 static void __init
log_buf_len_update(u64 size
)
1072 if (size
> (u64
)LOG_BUF_LEN_MAX
) {
1073 size
= (u64
)LOG_BUF_LEN_MAX
;
1074 pr_err("log_buf over 2G is not supported.\n");
1078 size
= roundup_pow_of_two(size
);
1079 if (size
> log_buf_len
)
1080 new_log_buf_len
= (unsigned long)size
;
1083 /* save requested log_buf_len since it's too early to process it */
1084 static int __init
log_buf_len_setup(char *str
)
1091 size
= memparse(str
, &str
);
1093 log_buf_len_update(size
);
1097 early_param("log_buf_len", log_buf_len_setup
);
1100 #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1102 static void __init
log_buf_add_cpu(void)
1104 unsigned int cpu_extra
;
1107 * archs should set up cpu_possible_bits properly with
1108 * set_cpu_possible() after setup_arch() but just in
1109 * case lets ensure this is valid.
1111 if (num_possible_cpus() == 1)
1114 cpu_extra
= (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN
;
1116 /* by default this will only continue through for large > 64 CPUs */
1117 if (cpu_extra
<= __LOG_BUF_LEN
/ 2)
1120 pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1121 __LOG_CPU_MAX_BUF_LEN
);
1122 pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1124 pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN
);
1126 log_buf_len_update(cpu_extra
+ __LOG_BUF_LEN
);
1128 #else /* !CONFIG_SMP */
1129 static inline void log_buf_add_cpu(void) {}
1130 #endif /* CONFIG_SMP */
1132 void __init
setup_log_buf(int early
)
1134 unsigned long flags
;
1138 if (log_buf
!= __log_buf
)
1141 if (!early
&& !new_log_buf_len
)
1144 if (!new_log_buf_len
)
1147 new_log_buf
= memblock_alloc(new_log_buf_len
, LOG_ALIGN
);
1148 if (unlikely(!new_log_buf
)) {
1149 pr_err("log_buf_len: %lu bytes not available\n",
1154 logbuf_lock_irqsave(flags
);
1155 log_buf_len
= new_log_buf_len
;
1156 log_buf
= new_log_buf
;
1157 new_log_buf_len
= 0;
1158 free
= __LOG_BUF_LEN
- log_next_idx
;
1159 memcpy(log_buf
, __log_buf
, __LOG_BUF_LEN
);
1160 logbuf_unlock_irqrestore(flags
);
1162 pr_info("log_buf_len: %u bytes\n", log_buf_len
);
1163 pr_info("early log buf free: %u(%u%%)\n",
1164 free
, (free
* 100) / __LOG_BUF_LEN
);
1167 static bool __read_mostly ignore_loglevel
;
1169 static int __init
ignore_loglevel_setup(char *str
)
1171 ignore_loglevel
= true;
1172 pr_info("debug: ignoring loglevel setting.\n");
1177 early_param("ignore_loglevel", ignore_loglevel_setup
);
1178 module_param(ignore_loglevel
, bool, S_IRUGO
| S_IWUSR
);
1179 MODULE_PARM_DESC(ignore_loglevel
,
1180 "ignore loglevel setting (prints all kernel messages to the console)");
1182 static bool suppress_message_printing(int level
)
1184 return (level
>= console_loglevel
&& !ignore_loglevel
);
1187 #ifdef CONFIG_BOOT_PRINTK_DELAY
1189 static int boot_delay
; /* msecs delay after each printk during bootup */
1190 static unsigned long long loops_per_msec
; /* based on boot_delay */
1192 static int __init
boot_delay_setup(char *str
)
1196 lpj
= preset_lpj
? preset_lpj
: 1000000; /* some guess */
1197 loops_per_msec
= (unsigned long long)lpj
/ 1000 * HZ
;
1199 get_option(&str
, &boot_delay
);
1200 if (boot_delay
> 10 * 1000)
1203 pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1204 "HZ: %d, loops_per_msec: %llu\n",
1205 boot_delay
, preset_lpj
, lpj
, HZ
, loops_per_msec
);
1208 early_param("boot_delay", boot_delay_setup
);
1210 static void boot_delay_msec(int level
)
1212 unsigned long long k
;
1213 unsigned long timeout
;
1215 if ((boot_delay
== 0 || system_state
>= SYSTEM_RUNNING
)
1216 || suppress_message_printing(level
)) {
1220 k
= (unsigned long long)loops_per_msec
* boot_delay
;
1222 timeout
= jiffies
+ msecs_to_jiffies(boot_delay
);
1227 * use (volatile) jiffies to prevent
1228 * compiler reduction; loop termination via jiffies
1229 * is secondary and may or may not happen.
1231 if (time_after(jiffies
, timeout
))
1233 touch_nmi_watchdog();
1237 static inline void boot_delay_msec(int level
)
1242 static bool printk_time
= IS_ENABLED(CONFIG_PRINTK_TIME
);
1243 module_param_named(time
, printk_time
, bool, S_IRUGO
| S_IWUSR
);
1245 static size_t print_syslog(unsigned int level
, char *buf
)
1247 return sprintf(buf
, "<%u>", level
);
1250 static size_t print_time(u64 ts
, char *buf
)
1252 unsigned long rem_nsec
= do_div(ts
, 1000000000);
1254 return sprintf(buf
, "[%5lu.%06lu]",
1255 (unsigned long)ts
, rem_nsec
/ 1000);
1258 #ifdef CONFIG_PRINTK_CALLER
1259 static size_t print_caller(u32 id
, char *buf
)
1263 snprintf(caller
, sizeof(caller
), "%c%u",
1264 id
& 0x80000000 ? 'C' : 'T', id
& ~0x80000000);
1265 return sprintf(buf
, "[%6s]", caller
);
1268 #define print_caller(id, buf) 0
1271 static size_t print_prefix(const struct printk_log
*msg
, bool syslog
,
1272 bool time
, char *buf
)
1277 len
= print_syslog((msg
->facility
<< 3) | msg
->level
, buf
);
1280 len
+= print_time(msg
->ts_nsec
, buf
+ len
);
1282 len
+= print_caller(msg
->caller_id
, buf
+ len
);
1284 if (IS_ENABLED(CONFIG_PRINTK_CALLER
) || time
) {
1292 static size_t msg_print_text(const struct printk_log
*msg
, bool syslog
,
1293 bool time
, char *buf
, size_t size
)
1295 const char *text
= log_text(msg
);
1296 size_t text_size
= msg
->text_len
;
1298 char prefix
[PREFIX_MAX
];
1299 const size_t prefix_len
= print_prefix(msg
, syslog
, time
, prefix
);
1302 const char *next
= memchr(text
, '\n', text_size
);
1306 text_len
= next
- text
;
1308 text_size
-= next
- text
;
1310 text_len
= text_size
;
1314 if (prefix_len
+ text_len
+ 1 >= size
- len
)
1317 memcpy(buf
+ len
, prefix
, prefix_len
);
1319 memcpy(buf
+ len
, text
, text_len
);
1323 /* SYSLOG_ACTION_* buffer size only calculation */
1324 len
+= prefix_len
+ text_len
+ 1;
1333 static int syslog_print(char __user
*buf
, int size
)
1336 struct printk_log
*msg
;
1339 text
= kmalloc(LOG_LINE_MAX
+ PREFIX_MAX
, GFP_KERNEL
);
1348 if (syslog_seq
< log_first_seq
) {
1349 /* messages are gone, move to first one */
1350 syslog_seq
= log_first_seq
;
1351 syslog_idx
= log_first_idx
;
1354 if (syslog_seq
== log_next_seq
) {
1355 logbuf_unlock_irq();
1360 * To keep reading/counting partial line consistent,
1361 * use printk_time value as of the beginning of a line.
1363 if (!syslog_partial
)
1364 syslog_time
= printk_time
;
1366 skip
= syslog_partial
;
1367 msg
= log_from_idx(syslog_idx
);
1368 n
= msg_print_text(msg
, true, syslog_time
, text
,
1369 LOG_LINE_MAX
+ PREFIX_MAX
);
1370 if (n
- syslog_partial
<= size
) {
1371 /* message fits into buffer, move forward */
1372 syslog_idx
= log_next(syslog_idx
);
1374 n
-= syslog_partial
;
1377 /* partial read(), remember position */
1379 syslog_partial
+= n
;
1382 logbuf_unlock_irq();
1387 if (copy_to_user(buf
, text
+ skip
, n
)) {
1402 static int syslog_print_all(char __user
*buf
, int size
, bool clear
)
1411 text
= kmalloc(LOG_LINE_MAX
+ PREFIX_MAX
, GFP_KERNEL
);
1418 * Find first record that fits, including all following records,
1419 * into the user-provided buffer for this dump.
1423 while (seq
< log_next_seq
) {
1424 struct printk_log
*msg
= log_from_idx(idx
);
1426 len
+= msg_print_text(msg
, true, time
, NULL
, 0);
1427 idx
= log_next(idx
);
1431 /* move first record forward until length fits into the buffer */
1434 while (len
> size
&& seq
< log_next_seq
) {
1435 struct printk_log
*msg
= log_from_idx(idx
);
1437 len
-= msg_print_text(msg
, true, time
, NULL
, 0);
1438 idx
= log_next(idx
);
1442 /* last message fitting into this dump */
1443 next_seq
= log_next_seq
;
1446 while (len
>= 0 && seq
< next_seq
) {
1447 struct printk_log
*msg
= log_from_idx(idx
);
1448 int textlen
= msg_print_text(msg
, true, time
, text
,
1449 LOG_LINE_MAX
+ PREFIX_MAX
);
1451 idx
= log_next(idx
);
1454 logbuf_unlock_irq();
1455 if (copy_to_user(buf
+ len
, text
, textlen
))
1461 if (seq
< log_first_seq
) {
1462 /* messages are gone, move to next one */
1463 seq
= log_first_seq
;
1464 idx
= log_first_idx
;
1469 clear_seq
= log_next_seq
;
1470 clear_idx
= log_next_idx
;
1472 logbuf_unlock_irq();
1478 static void syslog_clear(void)
1481 clear_seq
= log_next_seq
;
1482 clear_idx
= log_next_idx
;
1483 logbuf_unlock_irq();
1486 int do_syslog(int type
, char __user
*buf
, int len
, int source
)
1489 static int saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1492 error
= check_syslog_permissions(type
, source
);
1497 case SYSLOG_ACTION_CLOSE
: /* Close log */
1499 case SYSLOG_ACTION_OPEN
: /* Open log */
1501 case SYSLOG_ACTION_READ
: /* Read from log */
1502 if (!buf
|| len
< 0)
1506 if (!access_ok(buf
, len
))
1508 error
= wait_event_interruptible(log_wait
,
1509 syslog_seq
!= log_next_seq
);
1512 error
= syslog_print(buf
, len
);
1514 /* Read/clear last kernel messages */
1515 case SYSLOG_ACTION_READ_CLEAR
:
1518 /* Read last kernel messages */
1519 case SYSLOG_ACTION_READ_ALL
:
1520 if (!buf
|| len
< 0)
1524 if (!access_ok(buf
, len
))
1526 error
= syslog_print_all(buf
, len
, clear
);
1528 /* Clear ring buffer */
1529 case SYSLOG_ACTION_CLEAR
:
1532 /* Disable logging to console */
1533 case SYSLOG_ACTION_CONSOLE_OFF
:
1534 if (saved_console_loglevel
== LOGLEVEL_DEFAULT
)
1535 saved_console_loglevel
= console_loglevel
;
1536 console_loglevel
= minimum_console_loglevel
;
1538 /* Enable logging to console */
1539 case SYSLOG_ACTION_CONSOLE_ON
:
1540 if (saved_console_loglevel
!= LOGLEVEL_DEFAULT
) {
1541 console_loglevel
= saved_console_loglevel
;
1542 saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1545 /* Set level of messages printed to console */
1546 case SYSLOG_ACTION_CONSOLE_LEVEL
:
1547 if (len
< 1 || len
> 8)
1549 if (len
< minimum_console_loglevel
)
1550 len
= minimum_console_loglevel
;
1551 console_loglevel
= len
;
1552 /* Implicitly re-enable logging to console */
1553 saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1555 /* Number of chars in the log buffer */
1556 case SYSLOG_ACTION_SIZE_UNREAD
:
1558 if (syslog_seq
< log_first_seq
) {
1559 /* messages are gone, move to first one */
1560 syslog_seq
= log_first_seq
;
1561 syslog_idx
= log_first_idx
;
1564 if (source
== SYSLOG_FROM_PROC
) {
1566 * Short-cut for poll(/"proc/kmsg") which simply checks
1567 * for pending data, not the size; return the count of
1568 * records, not the length.
1570 error
= log_next_seq
- syslog_seq
;
1572 u64 seq
= syslog_seq
;
1573 u32 idx
= syslog_idx
;
1574 bool time
= syslog_partial
? syslog_time
: printk_time
;
1576 while (seq
< log_next_seq
) {
1577 struct printk_log
*msg
= log_from_idx(idx
);
1579 error
+= msg_print_text(msg
, true, time
, NULL
,
1582 idx
= log_next(idx
);
1585 error
-= syslog_partial
;
1587 logbuf_unlock_irq();
1589 /* Size of the log buffer */
1590 case SYSLOG_ACTION_SIZE_BUFFER
:
1591 error
= log_buf_len
;
1601 SYSCALL_DEFINE3(syslog
, int, type
, char __user
*, buf
, int, len
)
1603 return do_syslog(type
, buf
, len
, SYSLOG_FROM_READER
);
1607 * Special console_lock variants that help to reduce the risk of soft-lockups.
1608 * They allow to pass console_lock to another printk() call using a busy wait.
1611 #ifdef CONFIG_LOCKDEP
1612 static struct lockdep_map console_owner_dep_map
= {
1613 .name
= "console_owner"
1617 static DEFINE_RAW_SPINLOCK(console_owner_lock
);
1618 static struct task_struct
*console_owner
;
1619 static bool console_waiter
;
1622 * console_lock_spinning_enable - mark beginning of code where another
1623 * thread might safely busy wait
1625 * This basically converts console_lock into a spinlock. This marks
1626 * the section where the console_lock owner can not sleep, because
1627 * there may be a waiter spinning (like a spinlock). Also it must be
1628 * ready to hand over the lock at the end of the section.
1630 static void console_lock_spinning_enable(void)
1632 raw_spin_lock(&console_owner_lock
);
1633 console_owner
= current
;
1634 raw_spin_unlock(&console_owner_lock
);
1636 /* The waiter may spin on us after setting console_owner */
1637 spin_acquire(&console_owner_dep_map
, 0, 0, _THIS_IP_
);
1641 * console_lock_spinning_disable_and_check - mark end of code where another
1642 * thread was able to busy wait and check if there is a waiter
1644 * This is called at the end of the section where spinning is allowed.
1645 * It has two functions. First, it is a signal that it is no longer
1646 * safe to start busy waiting for the lock. Second, it checks if
1647 * there is a busy waiter and passes the lock rights to her.
1649 * Important: Callers lose the lock if there was a busy waiter.
1650 * They must not touch items synchronized by console_lock
1653 * Return: 1 if the lock rights were passed, 0 otherwise.
1655 static int console_lock_spinning_disable_and_check(void)
1659 raw_spin_lock(&console_owner_lock
);
1660 waiter
= READ_ONCE(console_waiter
);
1661 console_owner
= NULL
;
1662 raw_spin_unlock(&console_owner_lock
);
1665 spin_release(&console_owner_dep_map
, 1, _THIS_IP_
);
1669 /* The waiter is now free to continue */
1670 WRITE_ONCE(console_waiter
, false);
1672 spin_release(&console_owner_dep_map
, 1, _THIS_IP_
);
1675 * Hand off console_lock to waiter. The waiter will perform
1676 * the up(). After this, the waiter is the console_lock owner.
1678 mutex_release(&console_lock_dep_map
, 1, _THIS_IP_
);
1683 * console_trylock_spinning - try to get console_lock by busy waiting
1685 * This allows to busy wait for the console_lock when the current
1686 * owner is running in specially marked sections. It means that
1687 * the current owner is running and cannot reschedule until it
1688 * is ready to lose the lock.
1690 * Return: 1 if we got the lock, 0 othrewise
1692 static int console_trylock_spinning(void)
1694 struct task_struct
*owner
= NULL
;
1697 unsigned long flags
;
1699 if (console_trylock())
1702 printk_safe_enter_irqsave(flags
);
1704 raw_spin_lock(&console_owner_lock
);
1705 owner
= READ_ONCE(console_owner
);
1706 waiter
= READ_ONCE(console_waiter
);
1707 if (!waiter
&& owner
&& owner
!= current
) {
1708 WRITE_ONCE(console_waiter
, true);
1711 raw_spin_unlock(&console_owner_lock
);
1714 * If there is an active printk() writing to the
1715 * consoles, instead of having it write our data too,
1716 * see if we can offload that load from the active
1717 * printer, and do some printing ourselves.
1718 * Go into a spin only if there isn't already a waiter
1719 * spinning, and there is an active printer, and
1720 * that active printer isn't us (recursive printk?).
1723 printk_safe_exit_irqrestore(flags
);
1727 /* We spin waiting for the owner to release us */
1728 spin_acquire(&console_owner_dep_map
, 0, 0, _THIS_IP_
);
1729 /* Owner will clear console_waiter on hand off */
1730 while (READ_ONCE(console_waiter
))
1732 spin_release(&console_owner_dep_map
, 1, _THIS_IP_
);
1734 printk_safe_exit_irqrestore(flags
);
1736 * The owner passed the console lock to us.
1737 * Since we did not spin on console lock, annotate
1738 * this as a trylock. Otherwise lockdep will
1741 mutex_acquire(&console_lock_dep_map
, 0, 1, _THIS_IP_
);
1747 * Call the console drivers, asking them to write out
1748 * log_buf[start] to log_buf[end - 1].
1749 * The console_lock must be held.
1751 static void call_console_drivers(const char *ext_text
, size_t ext_len
,
1752 const char *text
, size_t len
)
1754 struct console
*con
;
1756 trace_console_rcuidle(text
, len
);
1758 if (!console_drivers
)
1761 for_each_console(con
) {
1762 if (exclusive_console
&& con
!= exclusive_console
)
1764 if (!(con
->flags
& CON_ENABLED
))
1768 if (!cpu_online(smp_processor_id()) &&
1769 !(con
->flags
& CON_ANYTIME
))
1771 if (con
->flags
& CON_EXTENDED
)
1772 con
->write(con
, ext_text
, ext_len
);
1774 con
->write(con
, text
, len
);
1778 int printk_delay_msec __read_mostly
;
1780 static inline void printk_delay(void)
1782 if (unlikely(printk_delay_msec
)) {
1783 int m
= printk_delay_msec
;
1787 touch_nmi_watchdog();
1792 static inline u32
printk_caller_id(void)
1794 return in_task() ? task_pid_nr(current
) :
1795 0x80000000 + raw_smp_processor_id();
1799 * Continuation lines are buffered, and not committed to the record buffer
1800 * until the line is complete, or a race forces it. The line fragments
1801 * though, are printed immediately to the consoles to ensure everything has
1802 * reached the console in case of a kernel crash.
1804 static struct cont
{
1805 char buf
[LOG_LINE_MAX
];
1806 size_t len
; /* length == 0 means unused buffer */
1807 u32 caller_id
; /* printk_caller_id() of first print */
1808 u64 ts_nsec
; /* time of first print */
1809 u8 level
; /* log level of first message */
1810 u8 facility
; /* log facility of first message */
1811 enum log_flags flags
; /* prefix, newline flags */
1814 static void cont_flush(void)
1819 log_store(cont
.caller_id
, cont
.facility
, cont
.level
, cont
.flags
,
1820 cont
.ts_nsec
, NULL
, 0, cont
.buf
, cont
.len
);
1824 static bool cont_add(u32 caller_id
, int facility
, int level
,
1825 enum log_flags flags
, const char *text
, size_t len
)
1827 /* If the line gets too long, split it up in separate records. */
1828 if (cont
.len
+ len
> sizeof(cont
.buf
)) {
1834 cont
.facility
= facility
;
1836 cont
.caller_id
= caller_id
;
1837 cont
.ts_nsec
= local_clock();
1841 memcpy(cont
.buf
+ cont
.len
, text
, len
);
1844 // The original flags come from the first line,
1845 // but later continuations can add a newline.
1846 if (flags
& LOG_NEWLINE
) {
1847 cont
.flags
|= LOG_NEWLINE
;
1854 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
)
1856 const u32 caller_id
= printk_caller_id();
1859 * If an earlier line was buffered, and we're a continuation
1860 * write from the same context, try to add it to the buffer.
1863 if (cont
.caller_id
== caller_id
&& (lflags
& LOG_CONT
)) {
1864 if (cont_add(caller_id
, facility
, level
, lflags
, text
, text_len
))
1867 /* Otherwise, make sure it's flushed */
1871 /* Skip empty continuation lines that couldn't be added - they just flush */
1872 if (!text_len
&& (lflags
& LOG_CONT
))
1875 /* If it doesn't end in a newline, try to buffer the current line */
1876 if (!(lflags
& LOG_NEWLINE
)) {
1877 if (cont_add(caller_id
, facility
, level
, lflags
, text
, text_len
))
1881 /* Store it in the record log */
1882 return log_store(caller_id
, facility
, level
, lflags
, 0,
1883 dict
, dictlen
, text
, text_len
);
1886 /* Must be called under logbuf_lock. */
1887 int vprintk_store(int facility
, int level
,
1888 const char *dict
, size_t dictlen
,
1889 const char *fmt
, va_list args
)
1891 static char textbuf
[LOG_LINE_MAX
];
1892 char *text
= textbuf
;
1894 enum log_flags lflags
= 0;
1897 * The printf needs to come first; we need the syslog
1898 * prefix which might be passed-in as a parameter.
1900 text_len
= vscnprintf(text
, sizeof(textbuf
), fmt
, args
);
1902 /* mark and strip a trailing newline */
1903 if (text_len
&& text
[text_len
-1] == '\n') {
1905 lflags
|= LOG_NEWLINE
;
1908 /* strip kernel syslog prefix and extract log level or control flags */
1909 if (facility
== 0) {
1912 while ((kern_level
= printk_get_level(text
)) != 0) {
1913 switch (kern_level
) {
1915 if (level
== LOGLEVEL_DEFAULT
)
1916 level
= kern_level
- '0';
1918 case 'c': /* KERN_CONT */
1927 if (level
== LOGLEVEL_DEFAULT
)
1928 level
= default_message_loglevel
;
1931 lflags
|= LOG_NEWLINE
;
1933 return log_output(facility
, level
, lflags
,
1934 dict
, dictlen
, text
, text_len
);
1937 asmlinkage
int vprintk_emit(int facility
, int level
,
1938 const char *dict
, size_t dictlen
,
1939 const char *fmt
, va_list args
)
1942 bool in_sched
= false, pending_output
;
1943 unsigned long flags
;
1946 if (level
== LOGLEVEL_SCHED
) {
1947 level
= LOGLEVEL_DEFAULT
;
1951 boot_delay_msec(level
);
1954 /* This stops the holder of console_sem just where we want him */
1955 logbuf_lock_irqsave(flags
);
1956 curr_log_seq
= log_next_seq
;
1957 printed_len
= vprintk_store(facility
, level
, dict
, dictlen
, fmt
, args
);
1958 pending_output
= (curr_log_seq
!= log_next_seq
);
1959 logbuf_unlock_irqrestore(flags
);
1961 /* If called from the scheduler, we can not call up(). */
1962 if (!in_sched
&& pending_output
) {
1964 * Disable preemption to avoid being preempted while holding
1965 * console_sem which would prevent anyone from printing to
1970 * Try to acquire and then immediately release the console
1971 * semaphore. The release will print out buffers and wake up
1972 * /dev/kmsg and syslog() users.
1974 if (console_trylock_spinning())
1983 EXPORT_SYMBOL(vprintk_emit
);
1985 asmlinkage
int vprintk(const char *fmt
, va_list args
)
1987 return vprintk_func(fmt
, args
);
1989 EXPORT_SYMBOL(vprintk
);
1991 int vprintk_default(const char *fmt
, va_list args
)
1995 #ifdef CONFIG_KGDB_KDB
1996 /* Allow to pass printk() to kdb but avoid a recursion. */
1997 if (unlikely(kdb_trap_printk
&& kdb_printf_cpu
< 0)) {
1998 r
= vkdb_printf(KDB_MSGSRC_PRINTK
, fmt
, args
);
2002 r
= vprintk_emit(0, LOGLEVEL_DEFAULT
, NULL
, 0, fmt
, args
);
2006 EXPORT_SYMBOL_GPL(vprintk_default
);
2009 * printk - print a kernel message
2010 * @fmt: format string
2012 * This is printk(). It can be called from any context. We want it to work.
2014 * We try to grab the console_lock. If we succeed, it's easy - we log the
2015 * output and call the console drivers. If we fail to get the semaphore, we
2016 * place the output into the log buffer and return. The current holder of
2017 * the console_sem will notice the new output in console_unlock(); and will
2018 * send it to the consoles before releasing the lock.
2020 * One effect of this deferred printing is that code which calls printk() and
2021 * then changes console_loglevel may break. This is because console_loglevel
2022 * is inspected when the actual printing occurs.
2027 * See the vsnprintf() documentation for format string extensions over C99.
2029 asmlinkage __visible
int printk(const char *fmt
, ...)
2034 va_start(args
, fmt
);
2035 r
= vprintk_func(fmt
, args
);
2040 EXPORT_SYMBOL(printk
);
2042 #else /* CONFIG_PRINTK */
2044 #define LOG_LINE_MAX 0
2045 #define PREFIX_MAX 0
2046 #define printk_time false
2048 static u64 syslog_seq
;
2049 static u32 syslog_idx
;
2050 static u64 console_seq
;
2051 static u32 console_idx
;
2052 static u64 exclusive_console_stop_seq
;
2053 static u64 log_first_seq
;
2054 static u32 log_first_idx
;
2055 static u64 log_next_seq
;
2056 static char *log_text(const struct printk_log
*msg
) { return NULL
; }
2057 static char *log_dict(const struct printk_log
*msg
) { return NULL
; }
2058 static struct printk_log
*log_from_idx(u32 idx
) { return NULL
; }
2059 static u32
log_next(u32 idx
) { return 0; }
2060 static ssize_t
msg_print_ext_header(char *buf
, size_t size
,
2061 struct printk_log
*msg
,
2062 u64 seq
) { return 0; }
2063 static ssize_t
msg_print_ext_body(char *buf
, size_t size
,
2064 char *dict
, size_t dict_len
,
2065 char *text
, size_t text_len
) { return 0; }
2066 static void console_lock_spinning_enable(void) { }
2067 static int console_lock_spinning_disable_and_check(void) { return 0; }
2068 static void call_console_drivers(const char *ext_text
, size_t ext_len
,
2069 const char *text
, size_t len
) {}
2070 static size_t msg_print_text(const struct printk_log
*msg
, bool syslog
,
2071 bool time
, char *buf
, size_t size
) { return 0; }
2072 static bool suppress_message_printing(int level
) { return false; }
2074 #endif /* CONFIG_PRINTK */
2076 #ifdef CONFIG_EARLY_PRINTK
2077 struct console
*early_console
;
2079 asmlinkage __visible
void early_printk(const char *fmt
, ...)
2089 n
= vscnprintf(buf
, sizeof(buf
), fmt
, ap
);
2092 early_console
->write(early_console
, buf
, n
);
2096 static int __add_preferred_console(char *name
, int idx
, char *options
,
2099 struct console_cmdline
*c
;
2103 * See if this tty is not yet registered, and
2104 * if we have a slot free.
2106 for (i
= 0, c
= console_cmdline
;
2107 i
< MAX_CMDLINECONSOLES
&& c
->name
[0];
2109 if (strcmp(c
->name
, name
) == 0 && c
->index
== idx
) {
2111 preferred_console
= i
;
2115 if (i
== MAX_CMDLINECONSOLES
)
2118 preferred_console
= i
;
2119 strlcpy(c
->name
, name
, sizeof(c
->name
));
2120 c
->options
= options
;
2121 braille_set_options(c
, brl_options
);
2127 static int __init
console_msg_format_setup(char *str
)
2129 if (!strcmp(str
, "syslog"))
2130 console_msg_format
= MSG_FORMAT_SYSLOG
;
2131 if (!strcmp(str
, "default"))
2132 console_msg_format
= MSG_FORMAT_DEFAULT
;
2135 __setup("console_msg_format=", console_msg_format_setup
);
2138 * Set up a console. Called via do_early_param() in init/main.c
2139 * for each "console=" parameter in the boot command line.
2141 static int __init
console_setup(char *str
)
2143 char buf
[sizeof(console_cmdline
[0].name
) + 4]; /* 4 for "ttyS" */
2144 char *s
, *options
, *brl_options
= NULL
;
2147 if (_braille_console_setup(&str
, &brl_options
))
2151 * Decode str into name, index, options.
2153 if (str
[0] >= '0' && str
[0] <= '9') {
2154 strcpy(buf
, "ttyS");
2155 strncpy(buf
+ 4, str
, sizeof(buf
) - 5);
2157 strncpy(buf
, str
, sizeof(buf
) - 1);
2159 buf
[sizeof(buf
) - 1] = 0;
2160 options
= strchr(str
, ',');
2164 if (!strcmp(str
, "ttya"))
2165 strcpy(buf
, "ttyS0");
2166 if (!strcmp(str
, "ttyb"))
2167 strcpy(buf
, "ttyS1");
2169 for (s
= buf
; *s
; s
++)
2170 if (isdigit(*s
) || *s
== ',')
2172 idx
= simple_strtoul(s
, NULL
, 10);
2175 __add_preferred_console(buf
, idx
, options
, brl_options
);
2176 console_set_on_cmdline
= 1;
2179 __setup("console=", console_setup
);
2182 * add_preferred_console - add a device to the list of preferred consoles.
2183 * @name: device name
2184 * @idx: device index
2185 * @options: options for this console
2187 * The last preferred console added will be used for kernel messages
2188 * and stdin/out/err for init. Normally this is used by console_setup
2189 * above to handle user-supplied console arguments; however it can also
2190 * be used by arch-specific code either to override the user or more
2191 * commonly to provide a default console (ie from PROM variables) when
2192 * the user has not supplied one.
2194 int add_preferred_console(char *name
, int idx
, char *options
)
2196 return __add_preferred_console(name
, idx
, options
, NULL
);
2199 bool console_suspend_enabled
= true;
2200 EXPORT_SYMBOL(console_suspend_enabled
);
2202 static int __init
console_suspend_disable(char *str
)
2204 console_suspend_enabled
= false;
2207 __setup("no_console_suspend", console_suspend_disable
);
2208 module_param_named(console_suspend
, console_suspend_enabled
,
2209 bool, S_IRUGO
| S_IWUSR
);
2210 MODULE_PARM_DESC(console_suspend
, "suspend console during suspend"
2211 " and hibernate operations");
2214 * suspend_console - suspend the console subsystem
2216 * This disables printk() while we go into suspend states
2218 void suspend_console(void)
2220 if (!console_suspend_enabled
)
2222 pr_info("Suspending console(s) (use no_console_suspend to debug)\n");
2224 console_suspended
= 1;
2228 void resume_console(void)
2230 if (!console_suspend_enabled
)
2233 console_suspended
= 0;
2238 * console_cpu_notify - print deferred console messages after CPU hotplug
2241 * If printk() is called from a CPU that is not online yet, the messages
2242 * will be printed on the console only if there are CON_ANYTIME consoles.
2243 * This function is called when a new CPU comes online (or fails to come
2244 * up) or goes offline.
2246 static int console_cpu_notify(unsigned int cpu
)
2248 if (!cpuhp_tasks_frozen
) {
2249 /* If trylock fails, someone else is doing the printing */
2250 if (console_trylock())
2257 * console_lock - lock the console system for exclusive use.
2259 * Acquires a lock which guarantees that the caller has
2260 * exclusive access to the console system and the console_drivers list.
2262 * Can sleep, returns nothing.
2264 void console_lock(void)
2269 if (console_suspended
)
2272 console_may_schedule
= 1;
2274 EXPORT_SYMBOL(console_lock
);
2277 * console_trylock - try to lock the console system for exclusive use.
2279 * Try to acquire a lock which guarantees that the caller has exclusive
2280 * access to the console system and the console_drivers list.
2282 * returns 1 on success, and 0 on failure to acquire the lock.
2284 int console_trylock(void)
2286 if (down_trylock_console_sem())
2288 if (console_suspended
) {
2293 console_may_schedule
= 0;
2296 EXPORT_SYMBOL(console_trylock
);
2298 int is_console_locked(void)
2300 return console_locked
;
2302 EXPORT_SYMBOL(is_console_locked
);
2305 * Check if we have any console that is capable of printing while cpu is
2306 * booting or shutting down. Requires console_sem.
2308 static int have_callable_console(void)
2310 struct console
*con
;
2312 for_each_console(con
)
2313 if ((con
->flags
& CON_ENABLED
) &&
2314 (con
->flags
& CON_ANYTIME
))
2321 * Can we actually use the console at this time on this cpu?
2323 * Console drivers may assume that per-cpu resources have been allocated. So
2324 * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2325 * call them until this CPU is officially up.
2327 static inline int can_use_console(void)
2329 return cpu_online(raw_smp_processor_id()) || have_callable_console();
2333 * console_unlock - unlock the console system
2335 * Releases the console_lock which the caller holds on the console system
2336 * and the console driver list.
2338 * While the console_lock was held, console output may have been buffered
2339 * by printk(). If this is the case, console_unlock(); emits
2340 * the output prior to releasing the lock.
2342 * If there is output waiting, we wake /dev/kmsg and syslog() users.
2344 * console_unlock(); may be called from any context.
2346 void console_unlock(void)
2348 static char ext_text
[CONSOLE_EXT_LOG_MAX
];
2349 static char text
[LOG_LINE_MAX
+ PREFIX_MAX
];
2350 unsigned long flags
;
2351 bool do_cond_resched
, retry
;
2353 if (console_suspended
) {
2359 * Console drivers are called with interrupts disabled, so
2360 * @console_may_schedule should be cleared before; however, we may
2361 * end up dumping a lot of lines, for example, if called from
2362 * console registration path, and should invoke cond_resched()
2363 * between lines if allowable. Not doing so can cause a very long
2364 * scheduling stall on a slow console leading to RCU stall and
2365 * softlockup warnings which exacerbate the issue with more
2366 * messages practically incapacitating the system.
2368 * console_trylock() is not able to detect the preemptive
2369 * context reliably. Therefore the value must be stored before
2370 * and cleared after the the "again" goto label.
2372 do_cond_resched
= console_may_schedule
;
2374 console_may_schedule
= 0;
2377 * We released the console_sem lock, so we need to recheck if
2378 * cpu is online and (if not) is there at least one CON_ANYTIME
2381 if (!can_use_console()) {
2388 struct printk_log
*msg
;
2392 printk_safe_enter_irqsave(flags
);
2393 raw_spin_lock(&logbuf_lock
);
2394 if (console_seq
< log_first_seq
) {
2396 "** %llu printk messages dropped **\n",
2397 log_first_seq
- console_seq
);
2399 /* messages are gone, move to first one */
2400 console_seq
= log_first_seq
;
2401 console_idx
= log_first_idx
;
2406 if (console_seq
== log_next_seq
)
2409 msg
= log_from_idx(console_idx
);
2410 if (suppress_message_printing(msg
->level
)) {
2412 * Skip record we have buffered and already printed
2413 * directly to the console when we received it, and
2414 * record that has level above the console loglevel.
2416 console_idx
= log_next(console_idx
);
2421 /* Output to all consoles once old messages replayed. */
2422 if (unlikely(exclusive_console
&&
2423 console_seq
>= exclusive_console_stop_seq
)) {
2424 exclusive_console
= NULL
;
2427 len
+= msg_print_text(msg
,
2428 console_msg_format
& MSG_FORMAT_SYSLOG
,
2429 printk_time
, text
+ len
, sizeof(text
) - len
);
2430 if (nr_ext_console_drivers
) {
2431 ext_len
= msg_print_ext_header(ext_text
,
2434 ext_len
+= msg_print_ext_body(ext_text
+ ext_len
,
2435 sizeof(ext_text
) - ext_len
,
2436 log_dict(msg
), msg
->dict_len
,
2437 log_text(msg
), msg
->text_len
);
2439 console_idx
= log_next(console_idx
);
2441 raw_spin_unlock(&logbuf_lock
);
2444 * While actively printing out messages, if another printk()
2445 * were to occur on another CPU, it may wait for this one to
2446 * finish. This task can not be preempted if there is a
2447 * waiter waiting to take over.
2449 console_lock_spinning_enable();
2451 stop_critical_timings(); /* don't trace print latency */
2452 call_console_drivers(ext_text
, ext_len
, text
, len
);
2453 start_critical_timings();
2455 if (console_lock_spinning_disable_and_check()) {
2456 printk_safe_exit_irqrestore(flags
);
2460 printk_safe_exit_irqrestore(flags
);
2462 if (do_cond_resched
)
2468 raw_spin_unlock(&logbuf_lock
);
2473 * Someone could have filled up the buffer again, so re-check if there's
2474 * something to flush. In case we cannot trylock the console_sem again,
2475 * there's a new owner and the console_unlock() from them will do the
2476 * flush, no worries.
2478 raw_spin_lock(&logbuf_lock
);
2479 retry
= console_seq
!= log_next_seq
;
2480 raw_spin_unlock(&logbuf_lock
);
2481 printk_safe_exit_irqrestore(flags
);
2483 if (retry
&& console_trylock())
2486 EXPORT_SYMBOL(console_unlock
);
2489 * console_conditional_schedule - yield the CPU if required
2491 * If the console code is currently allowed to sleep, and
2492 * if this CPU should yield the CPU to another task, do
2495 * Must be called within console_lock();.
2497 void __sched
console_conditional_schedule(void)
2499 if (console_may_schedule
)
2502 EXPORT_SYMBOL(console_conditional_schedule
);
2504 void console_unblank(void)
2509 * console_unblank can no longer be called in interrupt context unless
2510 * oops_in_progress is set to 1..
2512 if (oops_in_progress
) {
2513 if (down_trylock_console_sem() != 0)
2519 console_may_schedule
= 0;
2521 if ((c
->flags
& CON_ENABLED
) && c
->unblank
)
2527 * console_flush_on_panic - flush console content on panic
2529 * Immediately output all pending messages no matter what.
2531 void console_flush_on_panic(void)
2534 * If someone else is holding the console lock, trylock will fail
2535 * and may_schedule may be set. Ignore and proceed to unlock so
2536 * that messages are flushed out. As this can be called from any
2537 * context and we don't want to get preempted while flushing,
2538 * ensure may_schedule is cleared.
2541 console_may_schedule
= 0;
2546 * Return the console tty driver structure and its associated index
2548 struct tty_driver
*console_device(int *index
)
2551 struct tty_driver
*driver
= NULL
;
2554 for_each_console(c
) {
2557 driver
= c
->device(c
, index
);
2566 * Prevent further output on the passed console device so that (for example)
2567 * serial drivers can disable console output before suspending a port, and can
2568 * re-enable output afterwards.
2570 void console_stop(struct console
*console
)
2573 console
->flags
&= ~CON_ENABLED
;
2576 EXPORT_SYMBOL(console_stop
);
2578 void console_start(struct console
*console
)
2581 console
->flags
|= CON_ENABLED
;
2584 EXPORT_SYMBOL(console_start
);
2586 static int __read_mostly keep_bootcon
;
2588 static int __init
keep_bootcon_setup(char *str
)
2591 pr_info("debug: skip boot console de-registration.\n");
2596 early_param("keep_bootcon", keep_bootcon_setup
);
2599 * The console driver calls this routine during kernel initialization
2600 * to register the console printing procedure with printk() and to
2601 * print any messages that were printed by the kernel before the
2602 * console driver was initialized.
2604 * This can happen pretty early during the boot process (because of
2605 * early_printk) - sometimes before setup_arch() completes - be careful
2606 * of what kernel features are used - they may not be initialised yet.
2608 * There are two types of consoles - bootconsoles (early_printk) and
2609 * "real" consoles (everything which is not a bootconsole) which are
2610 * handled differently.
2611 * - Any number of bootconsoles can be registered at any time.
2612 * - As soon as a "real" console is registered, all bootconsoles
2613 * will be unregistered automatically.
2614 * - Once a "real" console is registered, any attempt to register a
2615 * bootconsoles will be rejected
2617 void register_console(struct console
*newcon
)
2620 unsigned long flags
;
2621 struct console
*bcon
= NULL
;
2622 struct console_cmdline
*c
;
2623 static bool has_preferred
;
2625 if (console_drivers
)
2626 for_each_console(bcon
)
2627 if (WARN(bcon
== newcon
,
2628 "console '%s%d' already registered\n",
2629 bcon
->name
, bcon
->index
))
2633 * before we register a new CON_BOOT console, make sure we don't
2634 * already have a valid console
2636 if (console_drivers
&& newcon
->flags
& CON_BOOT
) {
2637 /* find the last or real console */
2638 for_each_console(bcon
) {
2639 if (!(bcon
->flags
& CON_BOOT
)) {
2640 pr_info("Too late to register bootconsole %s%d\n",
2641 newcon
->name
, newcon
->index
);
2647 if (console_drivers
&& console_drivers
->flags
& CON_BOOT
)
2648 bcon
= console_drivers
;
2650 if (!has_preferred
|| bcon
|| !console_drivers
)
2651 has_preferred
= preferred_console
>= 0;
2654 * See if we want to use this console driver. If we
2655 * didn't select a console we take the first one
2656 * that registers here.
2658 if (!has_preferred
) {
2659 if (newcon
->index
< 0)
2661 if (newcon
->setup
== NULL
||
2662 newcon
->setup(newcon
, NULL
) == 0) {
2663 newcon
->flags
|= CON_ENABLED
;
2664 if (newcon
->device
) {
2665 newcon
->flags
|= CON_CONSDEV
;
2666 has_preferred
= true;
2672 * See if this console matches one we selected on
2675 for (i
= 0, c
= console_cmdline
;
2676 i
< MAX_CMDLINECONSOLES
&& c
->name
[0];
2678 if (!newcon
->match
||
2679 newcon
->match(newcon
, c
->name
, c
->index
, c
->options
) != 0) {
2680 /* default matching */
2681 BUILD_BUG_ON(sizeof(c
->name
) != sizeof(newcon
->name
));
2682 if (strcmp(c
->name
, newcon
->name
) != 0)
2684 if (newcon
->index
>= 0 &&
2685 newcon
->index
!= c
->index
)
2687 if (newcon
->index
< 0)
2688 newcon
->index
= c
->index
;
2690 if (_braille_register_console(newcon
, c
))
2693 if (newcon
->setup
&&
2694 newcon
->setup(newcon
, c
->options
) != 0)
2698 newcon
->flags
|= CON_ENABLED
;
2699 if (i
== preferred_console
) {
2700 newcon
->flags
|= CON_CONSDEV
;
2701 has_preferred
= true;
2706 if (!(newcon
->flags
& CON_ENABLED
))
2710 * If we have a bootconsole, and are switching to a real console,
2711 * don't print everything out again, since when the boot console, and
2712 * the real console are the same physical device, it's annoying to
2713 * see the beginning boot messages twice
2715 if (bcon
&& ((newcon
->flags
& (CON_CONSDEV
| CON_BOOT
)) == CON_CONSDEV
))
2716 newcon
->flags
&= ~CON_PRINTBUFFER
;
2719 * Put this console in the list - keep the
2720 * preferred driver at the head of the list.
2723 if ((newcon
->flags
& CON_CONSDEV
) || console_drivers
== NULL
) {
2724 newcon
->next
= console_drivers
;
2725 console_drivers
= newcon
;
2727 newcon
->next
->flags
&= ~CON_CONSDEV
;
2729 newcon
->next
= console_drivers
->next
;
2730 console_drivers
->next
= newcon
;
2733 if (newcon
->flags
& CON_EXTENDED
)
2734 nr_ext_console_drivers
++;
2736 if (newcon
->flags
& CON_PRINTBUFFER
) {
2738 * console_unlock(); will print out the buffered messages
2741 logbuf_lock_irqsave(flags
);
2742 console_seq
= syslog_seq
;
2743 console_idx
= syslog_idx
;
2745 * We're about to replay the log buffer. Only do this to the
2746 * just-registered console to avoid excessive message spam to
2747 * the already-registered consoles.
2749 * Set exclusive_console with disabled interrupts to reduce
2750 * race window with eventual console_flush_on_panic() that
2751 * ignores console_lock.
2753 exclusive_console
= newcon
;
2754 exclusive_console_stop_seq
= console_seq
;
2755 logbuf_unlock_irqrestore(flags
);
2758 console_sysfs_notify();
2761 * By unregistering the bootconsoles after we enable the real console
2762 * we get the "console xxx enabled" message on all the consoles -
2763 * boot consoles, real consoles, etc - this is to ensure that end
2764 * users know there might be something in the kernel's log buffer that
2765 * went to the bootconsole (that they do not see on the real console)
2767 pr_info("%sconsole [%s%d] enabled\n",
2768 (newcon
->flags
& CON_BOOT
) ? "boot" : "" ,
2769 newcon
->name
, newcon
->index
);
2771 ((newcon
->flags
& (CON_CONSDEV
| CON_BOOT
)) == CON_CONSDEV
) &&
2773 /* We need to iterate through all boot consoles, to make
2774 * sure we print everything out, before we unregister them.
2776 for_each_console(bcon
)
2777 if (bcon
->flags
& CON_BOOT
)
2778 unregister_console(bcon
);
2781 EXPORT_SYMBOL(register_console
);
2783 int unregister_console(struct console
*console
)
2785 struct console
*a
, *b
;
2788 pr_info("%sconsole [%s%d] disabled\n",
2789 (console
->flags
& CON_BOOT
) ? "boot" : "" ,
2790 console
->name
, console
->index
);
2792 res
= _braille_unregister_console(console
);
2798 if (console_drivers
== console
) {
2799 console_drivers
=console
->next
;
2801 } else if (console_drivers
) {
2802 for (a
=console_drivers
->next
, b
=console_drivers
;
2803 a
; b
=a
, a
=b
->next
) {
2812 if (!res
&& (console
->flags
& CON_EXTENDED
))
2813 nr_ext_console_drivers
--;
2816 * If this isn't the last console and it has CON_CONSDEV set, we
2817 * need to set it on the next preferred console.
2819 if (console_drivers
!= NULL
&& console
->flags
& CON_CONSDEV
)
2820 console_drivers
->flags
|= CON_CONSDEV
;
2822 console
->flags
&= ~CON_ENABLED
;
2824 console_sysfs_notify();
2827 EXPORT_SYMBOL(unregister_console
);
2830 * Initialize the console device. This is called *early*, so
2831 * we can't necessarily depend on lots of kernel help here.
2832 * Just do some early initializations, and do the complex setup
2835 void __init
console_init(void)
2839 initcall_entry_t
*ce
;
2841 /* Setup the default TTY line discipline. */
2845 * set up the console device so that later boot sequences can
2846 * inform about problems etc..
2848 ce
= __con_initcall_start
;
2849 trace_initcall_level("console");
2850 while (ce
< __con_initcall_end
) {
2851 call
= initcall_from_entry(ce
);
2852 trace_initcall_start(call
);
2854 trace_initcall_finish(call
, ret
);
2860 * Some boot consoles access data that is in the init section and which will
2861 * be discarded after the initcalls have been run. To make sure that no code
2862 * will access this data, unregister the boot consoles in a late initcall.
2864 * If for some reason, such as deferred probe or the driver being a loadable
2865 * module, the real console hasn't registered yet at this point, there will
2866 * be a brief interval in which no messages are logged to the console, which
2867 * makes it difficult to diagnose problems that occur during this time.
2869 * To mitigate this problem somewhat, only unregister consoles whose memory
2870 * intersects with the init section. Note that all other boot consoles will
2871 * get unregistred when the real preferred console is registered.
2873 static int __init
printk_late_init(void)
2875 struct console
*con
;
2878 for_each_console(con
) {
2879 if (!(con
->flags
& CON_BOOT
))
2882 /* Check addresses that might be used for enabled consoles. */
2883 if (init_section_intersects(con
, sizeof(*con
)) ||
2884 init_section_contains(con
->write
, 0) ||
2885 init_section_contains(con
->read
, 0) ||
2886 init_section_contains(con
->device
, 0) ||
2887 init_section_contains(con
->unblank
, 0) ||
2888 init_section_contains(con
->data
, 0)) {
2890 * Please, consider moving the reported consoles out
2891 * of the init section.
2893 pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n",
2894 con
->name
, con
->index
);
2895 unregister_console(con
);
2898 ret
= cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD
, "printk:dead", NULL
,
2899 console_cpu_notify
);
2901 ret
= cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN
, "printk:online",
2902 console_cpu_notify
, NULL
);
2906 late_initcall(printk_late_init
);
2908 #if defined CONFIG_PRINTK
2910 * Delayed printk version, for scheduler-internal messages:
2912 #define PRINTK_PENDING_WAKEUP 0x01
2913 #define PRINTK_PENDING_OUTPUT 0x02
2915 static DEFINE_PER_CPU(int, printk_pending
);
2917 static void wake_up_klogd_work_func(struct irq_work
*irq_work
)
2919 int pending
= __this_cpu_xchg(printk_pending
, 0);
2921 if (pending
& PRINTK_PENDING_OUTPUT
) {
2922 /* If trylock fails, someone else is doing the printing */
2923 if (console_trylock())
2927 if (pending
& PRINTK_PENDING_WAKEUP
)
2928 wake_up_interruptible(&log_wait
);
2931 static DEFINE_PER_CPU(struct irq_work
, wake_up_klogd_work
) = {
2932 .func
= wake_up_klogd_work_func
,
2933 .flags
= IRQ_WORK_LAZY
,
2936 void wake_up_klogd(void)
2939 if (waitqueue_active(&log_wait
)) {
2940 this_cpu_or(printk_pending
, PRINTK_PENDING_WAKEUP
);
2941 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work
));
2946 void defer_console_output(void)
2949 __this_cpu_or(printk_pending
, PRINTK_PENDING_OUTPUT
);
2950 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work
));
2954 int vprintk_deferred(const char *fmt
, va_list args
)
2958 r
= vprintk_emit(0, LOGLEVEL_SCHED
, NULL
, 0, fmt
, args
);
2959 defer_console_output();
2964 int printk_deferred(const char *fmt
, ...)
2969 va_start(args
, fmt
);
2970 r
= vprintk_deferred(fmt
, args
);
2977 * printk rate limiting, lifted from the networking subsystem.
2979 * This enforces a rate limit: not more than 10 kernel messages
2980 * every 5s to make a denial-of-service attack impossible.
2982 DEFINE_RATELIMIT_STATE(printk_ratelimit_state
, 5 * HZ
, 10);
2984 int __printk_ratelimit(const char *func
)
2986 return ___ratelimit(&printk_ratelimit_state
, func
);
2988 EXPORT_SYMBOL(__printk_ratelimit
);
2991 * printk_timed_ratelimit - caller-controlled printk ratelimiting
2992 * @caller_jiffies: pointer to caller's state
2993 * @interval_msecs: minimum interval between prints
2995 * printk_timed_ratelimit() returns true if more than @interval_msecs
2996 * milliseconds have elapsed since the last time printk_timed_ratelimit()
2999 bool printk_timed_ratelimit(unsigned long *caller_jiffies
,
3000 unsigned int interval_msecs
)
3002 unsigned long elapsed
= jiffies
- *caller_jiffies
;
3004 if (*caller_jiffies
&& elapsed
<= msecs_to_jiffies(interval_msecs
))
3007 *caller_jiffies
= jiffies
;
3010 EXPORT_SYMBOL(printk_timed_ratelimit
);
3012 static DEFINE_SPINLOCK(dump_list_lock
);
3013 static LIST_HEAD(dump_list
);
3016 * kmsg_dump_register - register a kernel log dumper.
3017 * @dumper: pointer to the kmsg_dumper structure
3019 * Adds a kernel log dumper to the system. The dump callback in the
3020 * structure will be called when the kernel oopses or panics and must be
3021 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
3023 int kmsg_dump_register(struct kmsg_dumper
*dumper
)
3025 unsigned long flags
;
3028 /* The dump callback needs to be set */
3032 spin_lock_irqsave(&dump_list_lock
, flags
);
3033 /* Don't allow registering multiple times */
3034 if (!dumper
->registered
) {
3035 dumper
->registered
= 1;
3036 list_add_tail_rcu(&dumper
->list
, &dump_list
);
3039 spin_unlock_irqrestore(&dump_list_lock
, flags
);
3043 EXPORT_SYMBOL_GPL(kmsg_dump_register
);
3046 * kmsg_dump_unregister - unregister a kmsg dumper.
3047 * @dumper: pointer to the kmsg_dumper structure
3049 * Removes a dump device from the system. Returns zero on success and
3050 * %-EINVAL otherwise.
3052 int kmsg_dump_unregister(struct kmsg_dumper
*dumper
)
3054 unsigned long flags
;
3057 spin_lock_irqsave(&dump_list_lock
, flags
);
3058 if (dumper
->registered
) {
3059 dumper
->registered
= 0;
3060 list_del_rcu(&dumper
->list
);
3063 spin_unlock_irqrestore(&dump_list_lock
, flags
);
3068 EXPORT_SYMBOL_GPL(kmsg_dump_unregister
);
3070 static bool always_kmsg_dump
;
3071 module_param_named(always_kmsg_dump
, always_kmsg_dump
, bool, S_IRUGO
| S_IWUSR
);
3074 * kmsg_dump - dump kernel log to kernel message dumpers.
3075 * @reason: the reason (oops, panic etc) for dumping
3077 * Call each of the registered dumper's dump() callback, which can
3078 * retrieve the kmsg records with kmsg_dump_get_line() or
3079 * kmsg_dump_get_buffer().
3081 void kmsg_dump(enum kmsg_dump_reason reason
)
3083 struct kmsg_dumper
*dumper
;
3084 unsigned long flags
;
3086 if ((reason
> KMSG_DUMP_OOPS
) && !always_kmsg_dump
)
3090 list_for_each_entry_rcu(dumper
, &dump_list
, list
) {
3091 if (dumper
->max_reason
&& reason
> dumper
->max_reason
)
3094 /* initialize iterator with data about the stored records */
3095 dumper
->active
= true;
3097 logbuf_lock_irqsave(flags
);
3098 dumper
->cur_seq
= clear_seq
;
3099 dumper
->cur_idx
= clear_idx
;
3100 dumper
->next_seq
= log_next_seq
;
3101 dumper
->next_idx
= log_next_idx
;
3102 logbuf_unlock_irqrestore(flags
);
3104 /* invoke dumper which will iterate over records */
3105 dumper
->dump(dumper
, reason
);
3107 /* reset iterator */
3108 dumper
->active
= false;
3114 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3115 * @dumper: registered kmsg dumper
3116 * @syslog: include the "<4>" prefixes
3117 * @line: buffer to copy the line to
3118 * @size: maximum size of the buffer
3119 * @len: length of line placed into buffer
3121 * Start at the beginning of the kmsg buffer, with the oldest kmsg
3122 * record, and copy one record into the provided buffer.
3124 * Consecutive calls will return the next available record moving
3125 * towards the end of the buffer with the youngest messages.
3127 * A return value of FALSE indicates that there are no more records to
3130 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
3132 bool kmsg_dump_get_line_nolock(struct kmsg_dumper
*dumper
, bool syslog
,
3133 char *line
, size_t size
, size_t *len
)
3135 struct printk_log
*msg
;
3139 if (!dumper
->active
)
3142 if (dumper
->cur_seq
< log_first_seq
) {
3143 /* messages are gone, move to first available one */
3144 dumper
->cur_seq
= log_first_seq
;
3145 dumper
->cur_idx
= log_first_idx
;
3149 if (dumper
->cur_seq
>= log_next_seq
)
3152 msg
= log_from_idx(dumper
->cur_idx
);
3153 l
= msg_print_text(msg
, syslog
, printk_time
, line
, size
);
3155 dumper
->cur_idx
= log_next(dumper
->cur_idx
);
3165 * kmsg_dump_get_line - retrieve one kmsg log line
3166 * @dumper: registered kmsg dumper
3167 * @syslog: include the "<4>" prefixes
3168 * @line: buffer to copy the line to
3169 * @size: maximum size of the buffer
3170 * @len: length of line placed into buffer
3172 * Start at the beginning of the kmsg buffer, with the oldest kmsg
3173 * record, and copy one record into the provided buffer.
3175 * Consecutive calls will return the next available record moving
3176 * towards the end of the buffer with the youngest messages.
3178 * A return value of FALSE indicates that there are no more records to
3181 bool kmsg_dump_get_line(struct kmsg_dumper
*dumper
, bool syslog
,
3182 char *line
, size_t size
, size_t *len
)
3184 unsigned long flags
;
3187 logbuf_lock_irqsave(flags
);
3188 ret
= kmsg_dump_get_line_nolock(dumper
, syslog
, line
, size
, len
);
3189 logbuf_unlock_irqrestore(flags
);
3193 EXPORT_SYMBOL_GPL(kmsg_dump_get_line
);
3196 * kmsg_dump_get_buffer - copy kmsg log lines
3197 * @dumper: registered kmsg dumper
3198 * @syslog: include the "<4>" prefixes
3199 * @buf: buffer to copy the line to
3200 * @size: maximum size of the buffer
3201 * @len: length of line placed into buffer
3203 * Start at the end of the kmsg buffer and fill the provided buffer
3204 * with as many of the the *youngest* kmsg records that fit into it.
3205 * If the buffer is large enough, all available kmsg records will be
3206 * copied with a single call.
3208 * Consecutive calls will fill the buffer with the next block of
3209 * available older records, not including the earlier retrieved ones.
3211 * A return value of FALSE indicates that there are no more records to
3214 bool kmsg_dump_get_buffer(struct kmsg_dumper
*dumper
, bool syslog
,
3215 char *buf
, size_t size
, size_t *len
)
3217 unsigned long flags
;
3224 bool time
= printk_time
;
3226 if (!dumper
->active
)
3229 logbuf_lock_irqsave(flags
);
3230 if (dumper
->cur_seq
< log_first_seq
) {
3231 /* messages are gone, move to first available one */
3232 dumper
->cur_seq
= log_first_seq
;
3233 dumper
->cur_idx
= log_first_idx
;
3237 if (dumper
->cur_seq
>= dumper
->next_seq
) {
3238 logbuf_unlock_irqrestore(flags
);
3242 /* calculate length of entire buffer */
3243 seq
= dumper
->cur_seq
;
3244 idx
= dumper
->cur_idx
;
3245 while (seq
< dumper
->next_seq
) {
3246 struct printk_log
*msg
= log_from_idx(idx
);
3248 l
+= msg_print_text(msg
, true, time
, NULL
, 0);
3249 idx
= log_next(idx
);
3253 /* move first record forward until length fits into the buffer */
3254 seq
= dumper
->cur_seq
;
3255 idx
= dumper
->cur_idx
;
3256 while (l
> size
&& seq
< dumper
->next_seq
) {
3257 struct printk_log
*msg
= log_from_idx(idx
);
3259 l
-= msg_print_text(msg
, true, time
, NULL
, 0);
3260 idx
= log_next(idx
);
3264 /* last message in next interation */
3269 while (seq
< dumper
->next_seq
) {
3270 struct printk_log
*msg
= log_from_idx(idx
);
3272 l
+= msg_print_text(msg
, syslog
, time
, buf
+ l
, size
- l
);
3273 idx
= log_next(idx
);
3277 dumper
->next_seq
= next_seq
;
3278 dumper
->next_idx
= next_idx
;
3280 logbuf_unlock_irqrestore(flags
);
3286 EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer
);
3289 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3290 * @dumper: registered kmsg dumper
3292 * Reset the dumper's iterator so that kmsg_dump_get_line() and
3293 * kmsg_dump_get_buffer() can be called again and used multiple
3294 * times within the same dumper.dump() callback.
3296 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3298 void kmsg_dump_rewind_nolock(struct kmsg_dumper
*dumper
)
3300 dumper
->cur_seq
= clear_seq
;
3301 dumper
->cur_idx
= clear_idx
;
3302 dumper
->next_seq
= log_next_seq
;
3303 dumper
->next_idx
= log_next_idx
;
3307 * kmsg_dump_rewind - reset the interator
3308 * @dumper: registered kmsg dumper
3310 * Reset the dumper's iterator so that kmsg_dump_get_line() and
3311 * kmsg_dump_get_buffer() can be called again and used multiple
3312 * times within the same dumper.dump() callback.
3314 void kmsg_dump_rewind(struct kmsg_dumper
*dumper
)
3316 unsigned long flags
;
3318 logbuf_lock_irqsave(flags
);
3319 kmsg_dump_rewind_nolock(dumper
);
3320 logbuf_unlock_irqrestore(flags
);
3322 EXPORT_SYMBOL_GPL(kmsg_dump_rewind
);