1 // SPDX-License-Identifier: GPL-2.0-only
3 * linux/kernel/printk.c
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * Modified to make sys_syslog() more flexible: added commands to
8 * return the last 4k of kernel messages, regardless of whether
9 * they've been read or not. Added option to suppress kernel printk's
10 * to the console. Added hook for sending the console messages
11 * elsewhere, in preparation for a serial line console (someday).
13 * Modified for sysctl support, 1/8/97, Chris Horn.
14 * Fixed SMP synchronization, 08/08/99, Manfred Spraul
15 * manfred@colorfullife.com
16 * Rewrote bits to get rid of console_lock
17 * 01Mar01 Andrew Morton
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/kernel.h>
24 #include <linux/tty.h>
25 #include <linux/tty_driver.h>
26 #include <linux/console.h>
27 #include <linux/init.h>
28 #include <linux/jiffies.h>
29 #include <linux/nmi.h>
30 #include <linux/module.h>
31 #include <linux/moduleparam.h>
32 #include <linux/delay.h>
33 #include <linux/smp.h>
34 #include <linux/security.h>
35 #include <linux/memblock.h>
36 #include <linux/syscalls.h>
37 #include <linux/crash_core.h>
38 #include <linux/kdb.h>
39 #include <linux/ratelimit.h>
40 #include <linux/kmsg_dump.h>
41 #include <linux/syslog.h>
42 #include <linux/cpu.h>
43 #include <linux/rculist.h>
44 #include <linux/poll.h>
45 #include <linux/irq_work.h>
46 #include <linux/ctype.h>
47 #include <linux/uio.h>
48 #include <linux/sched/clock.h>
49 #include <linux/sched/debug.h>
50 #include <linux/sched/task_stack.h>
52 #include <linux/uaccess.h>
53 #include <asm/sections.h>
55 #include <trace/events/initcall.h>
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/printk.h>
59 #include "console_cmdline.h"
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 */
69 EXPORT_SYMBOL_GPL(console_printk
);
71 atomic_t ignore_console_lock_warning __read_mostly
= ATOMIC_INIT(0);
72 EXPORT_SYMBOL(ignore_console_lock_warning
);
75 * Low level drivers may need that to know if they can schedule in
76 * their unblank() callback or not. So let's export it.
79 EXPORT_SYMBOL(oops_in_progress
);
82 * console_sem protects the console_drivers list, and also
83 * provides serialisation for access to the entire console
86 static DEFINE_SEMAPHORE(console_sem
);
87 struct console
*console_drivers
;
88 EXPORT_SYMBOL_GPL(console_drivers
);
91 * System may need to suppress printk message under certain
92 * circumstances, like after kernel panic happens.
94 int __read_mostly suppress_printk
;
97 static struct lockdep_map console_lock_dep_map
= {
98 .name
= "console_lock"
102 enum devkmsg_log_bits
{
103 __DEVKMSG_LOG_BIT_ON
= 0,
104 __DEVKMSG_LOG_BIT_OFF
,
105 __DEVKMSG_LOG_BIT_LOCK
,
108 enum devkmsg_log_masks
{
109 DEVKMSG_LOG_MASK_ON
= BIT(__DEVKMSG_LOG_BIT_ON
),
110 DEVKMSG_LOG_MASK_OFF
= BIT(__DEVKMSG_LOG_BIT_OFF
),
111 DEVKMSG_LOG_MASK_LOCK
= BIT(__DEVKMSG_LOG_BIT_LOCK
),
114 /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
115 #define DEVKMSG_LOG_MASK_DEFAULT 0
117 static unsigned int __read_mostly devkmsg_log
= DEVKMSG_LOG_MASK_DEFAULT
;
119 static int __control_devkmsg(char *str
)
126 len
= str_has_prefix(str
, "on");
128 devkmsg_log
= DEVKMSG_LOG_MASK_ON
;
132 len
= str_has_prefix(str
, "off");
134 devkmsg_log
= DEVKMSG_LOG_MASK_OFF
;
138 len
= str_has_prefix(str
, "ratelimit");
140 devkmsg_log
= DEVKMSG_LOG_MASK_DEFAULT
;
147 static int __init
control_devkmsg(char *str
)
149 if (__control_devkmsg(str
) < 0)
153 * Set sysctl string accordingly:
155 if (devkmsg_log
== DEVKMSG_LOG_MASK_ON
)
156 strcpy(devkmsg_log_str
, "on");
157 else if (devkmsg_log
== DEVKMSG_LOG_MASK_OFF
)
158 strcpy(devkmsg_log_str
, "off");
159 /* else "ratelimit" which is set by default. */
162 * Sysctl cannot change it anymore. The kernel command line setting of
163 * this parameter is to force the setting to be permanent throughout the
164 * runtime of the system. This is a precation measure against userspace
165 * trying to be a smarta** and attempting to change it up on us.
167 devkmsg_log
|= DEVKMSG_LOG_MASK_LOCK
;
171 __setup("printk.devkmsg=", control_devkmsg
);
173 char devkmsg_log_str
[DEVKMSG_STR_MAX_SIZE
] = "ratelimit";
175 int devkmsg_sysctl_set_loglvl(struct ctl_table
*table
, int write
,
176 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
178 char old_str
[DEVKMSG_STR_MAX_SIZE
];
183 if (devkmsg_log
& DEVKMSG_LOG_MASK_LOCK
)
187 strncpy(old_str
, devkmsg_log_str
, DEVKMSG_STR_MAX_SIZE
);
190 err
= proc_dostring(table
, write
, buffer
, lenp
, ppos
);
195 err
= __control_devkmsg(devkmsg_log_str
);
198 * Do not accept an unknown string OR a known string with
201 if (err
< 0 || (err
+ 1 != *lenp
)) {
203 /* ... and restore old setting. */
205 strncpy(devkmsg_log_str
, old_str
, DEVKMSG_STR_MAX_SIZE
);
214 /* Number of registered extended console drivers. */
215 static int nr_ext_console_drivers
;
218 * Helper macros to handle lockdep when locking/unlocking console_sem. We use
219 * macros instead of functions so that _RET_IP_ contains useful information.
221 #define down_console_sem() do { \
223 mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
226 static int __down_trylock_console_sem(unsigned long ip
)
232 * Here and in __up_console_sem() we need to be in safe mode,
233 * because spindump/WARN/etc from under console ->lock will
234 * deadlock in printk()->down_trylock_console_sem() otherwise.
236 printk_safe_enter_irqsave(flags
);
237 lock_failed
= down_trylock(&console_sem
);
238 printk_safe_exit_irqrestore(flags
);
242 mutex_acquire(&console_lock_dep_map
, 0, 1, ip
);
245 #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
247 static void __up_console_sem(unsigned long ip
)
251 mutex_release(&console_lock_dep_map
, ip
);
253 printk_safe_enter_irqsave(flags
);
255 printk_safe_exit_irqrestore(flags
);
257 #define up_console_sem() __up_console_sem(_RET_IP_)
260 * This is used for debugging the mess that is the VT code by
261 * keeping track if we have the console semaphore held. It's
262 * definitely not the perfect debug tool (we don't know if _WE_
263 * hold it and are racing, but it helps tracking those weird code
264 * paths in the console code where we end up in places I want
265 * locked without the console sempahore held).
267 static int console_locked
, console_suspended
;
270 * If exclusive_console is non-NULL then only this console is to be printed to.
272 static struct console
*exclusive_console
;
275 * Array of consoles built from command line options (console=)
278 #define MAX_CMDLINECONSOLES 8
280 static struct console_cmdline console_cmdline
[MAX_CMDLINECONSOLES
];
282 static int preferred_console
= -1;
283 int console_set_on_cmdline
;
284 EXPORT_SYMBOL(console_set_on_cmdline
);
286 /* Flag: console code may call schedule() */
287 static int console_may_schedule
;
289 enum con_msg_format_flags
{
290 MSG_FORMAT_DEFAULT
= 0,
291 MSG_FORMAT_SYSLOG
= (1 << 0),
294 static int console_msg_format
= MSG_FORMAT_DEFAULT
;
297 * The printk log buffer consists of a chain of concatenated variable
298 * length records. Every record starts with a record header, containing
299 * the overall length of the record.
301 * The heads to the first and last entry in the buffer, as well as the
302 * sequence numbers of these entries are maintained when messages are
305 * If the heads indicate available messages, the length in the header
306 * tells the start next message. A length == 0 for the next message
307 * indicates a wrap-around to the beginning of the buffer.
309 * Every record carries the monotonic timestamp in microseconds, as well as
310 * the standard userspace syslog level and syslog facility. The usual
311 * kernel messages use LOG_KERN; userspace-injected messages always carry
312 * a matching syslog facility, by default LOG_USER. The origin of every
313 * message can be reliably determined that way.
315 * The human readable log message directly follows the message header. The
316 * length of the message text is stored in the header, the stored message
319 * Optionally, a message can carry a dictionary of properties (key/value pairs),
320 * to provide userspace with a machine-readable message context.
322 * Examples for well-defined, commonly used property names are:
323 * DEVICE=b12:8 device identifier
327 * +sound:card0 subsystem:devname
328 * SUBSYSTEM=pci driver-core subsystem name
330 * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
331 * follows directly after a '=' character. Every property is terminated by
332 * a '\0' character. The last property is not terminated.
334 * Example of a message structure:
335 * 0000 ff 8f 00 00 00 00 00 00 monotonic time in nsec
336 * 0008 34 00 record is 52 bytes long
337 * 000a 0b 00 text is 11 bytes long
338 * 000c 1f 00 dictionary is 23 bytes long
339 * 000e 03 00 LOG_KERN (facility) LOG_ERR (level)
340 * 0010 69 74 27 73 20 61 20 6c "it's a l"
342 * 001b 44 45 56 49 43 "DEVIC"
343 * 45 3d 62 38 3a 32 00 44 "E=b8:2\0D"
344 * 52 49 56 45 52 3d 62 75 "RIVER=bu"
346 * 0032 00 00 00 padding to next message header
348 * The 'struct printk_log' buffer header must never be directly exported to
349 * userspace, it is a kernel-private implementation detail that might
350 * need to be changed in the future, when the requirements change.
352 * /dev/kmsg exports the structured data in the following line format:
353 * "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
355 * Users of the export format should ignore possible additional values
356 * separated by ',', and find the message after the ';' character.
358 * The optional key/value pairs are attached as continuation lines starting
359 * with a space character and terminated by a newline. All possible
360 * non-prinatable characters are escaped in the "\xff" notation.
364 LOG_NEWLINE
= 2, /* text ended with a newline */
365 LOG_CONT
= 8, /* text is a fragment of a continuation line */
369 u64 ts_nsec
; /* timestamp in nanoseconds */
370 u16 len
; /* length of entire record */
371 u16 text_len
; /* length of text buffer */
372 u16 dict_len
; /* length of dictionary buffer */
373 u8 facility
; /* syslog facility */
374 u8 flags
:5; /* internal record flags */
375 u8 level
:3; /* syslog level */
376 #ifdef CONFIG_PRINTK_CALLER
377 u32 caller_id
; /* thread id or processor id */
380 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
381 __packed
__aligned(4)
386 * The logbuf_lock protects kmsg buffer, indices, counters. This can be taken
387 * within the scheduler's rq lock. It must be released before calling
388 * console_unlock() or anything else that might wake up a process.
390 DEFINE_RAW_SPINLOCK(logbuf_lock
);
393 * Helper macros to lock/unlock logbuf_lock and switch between
394 * printk-safe/unsafe modes.
396 #define logbuf_lock_irq() \
398 printk_safe_enter_irq(); \
399 raw_spin_lock(&logbuf_lock); \
402 #define logbuf_unlock_irq() \
404 raw_spin_unlock(&logbuf_lock); \
405 printk_safe_exit_irq(); \
408 #define logbuf_lock_irqsave(flags) \
410 printk_safe_enter_irqsave(flags); \
411 raw_spin_lock(&logbuf_lock); \
414 #define logbuf_unlock_irqrestore(flags) \
416 raw_spin_unlock(&logbuf_lock); \
417 printk_safe_exit_irqrestore(flags); \
421 DECLARE_WAIT_QUEUE_HEAD(log_wait
);
422 /* the next printk record to read by syslog(READ) or /proc/kmsg */
423 static u64 syslog_seq
;
424 static u32 syslog_idx
;
425 static size_t syslog_partial
;
426 static bool syslog_time
;
428 /* index and sequence number of the first record stored in the buffer */
429 static u64 log_first_seq
;
430 static u32 log_first_idx
;
432 /* index and sequence number of the next record to store in the buffer */
433 static u64 log_next_seq
;
434 static u32 log_next_idx
;
436 /* the next printk record to write to the console */
437 static u64 console_seq
;
438 static u32 console_idx
;
439 static u64 exclusive_console_stop_seq
;
441 /* the next printk record to read after the last 'clear' command */
442 static u64 clear_seq
;
443 static u32 clear_idx
;
445 #ifdef CONFIG_PRINTK_CALLER
446 #define PREFIX_MAX 48
448 #define PREFIX_MAX 32
450 #define LOG_LINE_MAX (1024 - PREFIX_MAX)
452 #define LOG_LEVEL(v) ((v) & 0x07)
453 #define LOG_FACILITY(v) ((v) >> 3 & 0xff)
456 #define LOG_ALIGN __alignof__(struct printk_log)
457 #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
458 #define LOG_BUF_LEN_MAX (u32)(1 << 31)
459 static char __log_buf
[__LOG_BUF_LEN
] __aligned(LOG_ALIGN
);
460 static char *log_buf
= __log_buf
;
461 static u32 log_buf_len
= __LOG_BUF_LEN
;
464 * We cannot access per-CPU data (e.g. per-CPU flush irq_work) before
465 * per_cpu_areas are initialised. This variable is set to true when
466 * it's safe to access per-CPU data.
468 static bool __printk_percpu_data_ready __read_mostly
;
470 bool printk_percpu_data_ready(void)
472 return __printk_percpu_data_ready
;
475 /* Return log buffer address */
476 char *log_buf_addr_get(void)
481 /* Return log buffer size */
482 u32
log_buf_len_get(void)
487 /* human readable text of the record */
488 static char *log_text(const struct printk_log
*msg
)
490 return (char *)msg
+ sizeof(struct printk_log
);
493 /* optional key/value pair dictionary attached to the record */
494 static char *log_dict(const struct printk_log
*msg
)
496 return (char *)msg
+ sizeof(struct printk_log
) + msg
->text_len
;
499 /* get record by index; idx must point to valid msg */
500 static struct printk_log
*log_from_idx(u32 idx
)
502 struct printk_log
*msg
= (struct printk_log
*)(log_buf
+ idx
);
505 * A length == 0 record is the end of buffer marker. Wrap around and
506 * read the message at the start of the buffer.
509 return (struct printk_log
*)log_buf
;
513 /* get next record; idx must point to valid msg */
514 static u32
log_next(u32 idx
)
516 struct printk_log
*msg
= (struct printk_log
*)(log_buf
+ idx
);
518 /* length == 0 indicates the end of the buffer; wrap */
520 * A length == 0 record is the end of buffer marker. Wrap around and
521 * read the message at the start of the buffer as *this* one, and
522 * return the one after that.
525 msg
= (struct printk_log
*)log_buf
;
528 return idx
+ msg
->len
;
532 * Check whether there is enough free space for the given message.
534 * The same values of first_idx and next_idx mean that the buffer
535 * is either empty or full.
537 * If the buffer is empty, we must respect the position of the indexes.
538 * They cannot be reset to the beginning of the buffer.
540 static int logbuf_has_space(u32 msg_size
, bool empty
)
544 if (log_next_idx
> log_first_idx
|| empty
)
545 free
= max(log_buf_len
- log_next_idx
, log_first_idx
);
547 free
= log_first_idx
- log_next_idx
;
550 * We need space also for an empty header that signalizes wrapping
553 return free
>= msg_size
+ sizeof(struct printk_log
);
556 static int log_make_free_space(u32 msg_size
)
558 while (log_first_seq
< log_next_seq
&&
559 !logbuf_has_space(msg_size
, false)) {
560 /* drop old messages until we have enough contiguous space */
561 log_first_idx
= log_next(log_first_idx
);
565 if (clear_seq
< log_first_seq
) {
566 clear_seq
= log_first_seq
;
567 clear_idx
= log_first_idx
;
570 /* sequence numbers are equal, so the log buffer is empty */
571 if (logbuf_has_space(msg_size
, log_first_seq
== log_next_seq
))
577 /* compute the message size including the padding bytes */
578 static u32
msg_used_size(u16 text_len
, u16 dict_len
, u32
*pad_len
)
582 size
= sizeof(struct printk_log
) + text_len
+ dict_len
;
583 *pad_len
= (-size
) & (LOG_ALIGN
- 1);
590 * Define how much of the log buffer we could take at maximum. The value
591 * must be greater than two. Note that only half of the buffer is available
592 * when the index points to the middle.
594 #define MAX_LOG_TAKE_PART 4
595 static const char trunc_msg
[] = "<truncated>";
597 static u32
truncate_msg(u16
*text_len
, u16
*trunc_msg_len
,
598 u16
*dict_len
, u32
*pad_len
)
601 * The message should not take the whole buffer. Otherwise, it might
602 * get removed too soon.
604 u32 max_text_len
= log_buf_len
/ MAX_LOG_TAKE_PART
;
605 if (*text_len
> max_text_len
)
606 *text_len
= max_text_len
;
607 /* enable the warning message */
608 *trunc_msg_len
= strlen(trunc_msg
);
609 /* disable the "dict" completely */
611 /* compute the size again, count also the warning message */
612 return msg_used_size(*text_len
+ *trunc_msg_len
, 0, pad_len
);
615 /* insert record into the buffer, discard old ones, update heads */
616 static int log_store(u32 caller_id
, int facility
, int level
,
617 enum log_flags flags
, u64 ts_nsec
,
618 const char *dict
, u16 dict_len
,
619 const char *text
, u16 text_len
)
621 struct printk_log
*msg
;
623 u16 trunc_msg_len
= 0;
625 /* number of '\0' padding bytes to next message */
626 size
= msg_used_size(text_len
, dict_len
, &pad_len
);
628 if (log_make_free_space(size
)) {
629 /* truncate the message if it is too long for empty buffer */
630 size
= truncate_msg(&text_len
, &trunc_msg_len
,
631 &dict_len
, &pad_len
);
632 /* survive when the log buffer is too small for trunc_msg */
633 if (log_make_free_space(size
))
637 if (log_next_idx
+ size
+ sizeof(struct printk_log
) > log_buf_len
) {
639 * This message + an additional empty header does not fit
640 * at the end of the buffer. Add an empty header with len == 0
641 * to signify a wrap around.
643 memset(log_buf
+ log_next_idx
, 0, sizeof(struct printk_log
));
648 msg
= (struct printk_log
*)(log_buf
+ log_next_idx
);
649 memcpy(log_text(msg
), text
, text_len
);
650 msg
->text_len
= text_len
;
652 memcpy(log_text(msg
) + text_len
, trunc_msg
, trunc_msg_len
);
653 msg
->text_len
+= trunc_msg_len
;
655 memcpy(log_dict(msg
), dict
, dict_len
);
656 msg
->dict_len
= dict_len
;
657 msg
->facility
= facility
;
658 msg
->level
= level
& 7;
659 msg
->flags
= flags
& 0x1f;
661 msg
->ts_nsec
= ts_nsec
;
663 msg
->ts_nsec
= local_clock();
664 #ifdef CONFIG_PRINTK_CALLER
665 msg
->caller_id
= caller_id
;
667 memset(log_dict(msg
) + dict_len
, 0, pad_len
);
671 log_next_idx
+= msg
->len
;
674 return msg
->text_len
;
677 int dmesg_restrict
= IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT
);
679 static int syslog_action_restricted(int type
)
684 * Unless restricted, we allow "read all" and "get buffer size"
687 return type
!= SYSLOG_ACTION_READ_ALL
&&
688 type
!= SYSLOG_ACTION_SIZE_BUFFER
;
691 static int check_syslog_permissions(int type
, int source
)
694 * If this is from /proc/kmsg and we've already opened it, then we've
695 * already done the capabilities checks at open time.
697 if (source
== SYSLOG_FROM_PROC
&& type
!= SYSLOG_ACTION_OPEN
)
700 if (syslog_action_restricted(type
)) {
701 if (capable(CAP_SYSLOG
))
704 * For historical reasons, accept CAP_SYS_ADMIN too, with
707 if (capable(CAP_SYS_ADMIN
)) {
708 pr_warn_once("%s (%d): Attempt to access syslog with "
709 "CAP_SYS_ADMIN but no CAP_SYSLOG "
711 current
->comm
, task_pid_nr(current
));
717 return security_syslog(type
);
720 static void append_char(char **pp
, char *e
, char c
)
726 static ssize_t
msg_print_ext_header(char *buf
, size_t size
,
727 struct printk_log
*msg
, u64 seq
)
729 u64 ts_usec
= msg
->ts_nsec
;
731 #ifdef CONFIG_PRINTK_CALLER
732 u32 id
= msg
->caller_id
;
734 snprintf(caller
, sizeof(caller
), ",caller=%c%u",
735 id
& 0x80000000 ? 'C' : 'T', id
& ~0x80000000);
740 do_div(ts_usec
, 1000);
742 return scnprintf(buf
, size
, "%u,%llu,%llu,%c%s;",
743 (msg
->facility
<< 3) | msg
->level
, seq
, ts_usec
,
744 msg
->flags
& LOG_CONT
? 'c' : '-', caller
);
747 static ssize_t
msg_print_ext_body(char *buf
, size_t size
,
748 char *dict
, size_t dict_len
,
749 char *text
, size_t text_len
)
751 char *p
= buf
, *e
= buf
+ size
;
754 /* escape non-printable characters */
755 for (i
= 0; i
< text_len
; i
++) {
756 unsigned char c
= text
[i
];
758 if (c
< ' ' || c
>= 127 || c
== '\\')
759 p
+= scnprintf(p
, e
- p
, "\\x%02x", c
);
761 append_char(&p
, e
, c
);
763 append_char(&p
, e
, '\n');
768 for (i
= 0; i
< dict_len
; i
++) {
769 unsigned char c
= dict
[i
];
772 append_char(&p
, e
, ' ');
777 append_char(&p
, e
, '\n');
782 if (c
< ' ' || c
>= 127 || c
== '\\') {
783 p
+= scnprintf(p
, e
- p
, "\\x%02x", c
);
787 append_char(&p
, e
, c
);
789 append_char(&p
, e
, '\n');
795 /* /dev/kmsg - userspace message inject/listen interface */
796 struct devkmsg_user
{
799 struct ratelimit_state rs
;
801 char buf
[CONSOLE_EXT_LOG_MAX
];
804 static __printf(3, 4) __cold
805 int devkmsg_emit(int facility
, int level
, const char *fmt
, ...)
811 r
= vprintk_emit(facility
, level
, NULL
, 0, fmt
, args
);
817 static ssize_t
devkmsg_write(struct kiocb
*iocb
, struct iov_iter
*from
)
820 int level
= default_message_loglevel
;
821 int facility
= 1; /* LOG_USER */
822 struct file
*file
= iocb
->ki_filp
;
823 struct devkmsg_user
*user
= file
->private_data
;
824 size_t len
= iov_iter_count(from
);
827 if (!user
|| len
> LOG_LINE_MAX
)
830 /* Ignore when user logging is disabled. */
831 if (devkmsg_log
& DEVKMSG_LOG_MASK_OFF
)
834 /* Ratelimit when not explicitly enabled. */
835 if (!(devkmsg_log
& DEVKMSG_LOG_MASK_ON
)) {
836 if (!___ratelimit(&user
->rs
, current
->comm
))
840 buf
= kmalloc(len
+1, GFP_KERNEL
);
845 if (!copy_from_iter_full(buf
, len
, from
)) {
851 * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
852 * the decimal value represents 32bit, the lower 3 bit are the log
853 * level, the rest are the log facility.
855 * If no prefix or no userspace facility is specified, we
856 * enforce LOG_USER, to be able to reliably distinguish
857 * kernel-generated messages from userspace-injected ones.
860 if (line
[0] == '<') {
864 u
= simple_strtoul(line
+ 1, &endp
, 10);
865 if (endp
&& endp
[0] == '>') {
866 level
= LOG_LEVEL(u
);
867 if (LOG_FACILITY(u
) != 0)
868 facility
= LOG_FACILITY(u
);
875 devkmsg_emit(facility
, level
, "%s", line
);
880 static ssize_t
devkmsg_read(struct file
*file
, char __user
*buf
,
881 size_t count
, loff_t
*ppos
)
883 struct devkmsg_user
*user
= file
->private_data
;
884 struct printk_log
*msg
;
891 ret
= mutex_lock_interruptible(&user
->lock
);
896 while (user
->seq
== log_next_seq
) {
897 if (file
->f_flags
& O_NONBLOCK
) {
904 ret
= wait_event_interruptible(log_wait
,
905 user
->seq
!= log_next_seq
);
911 if (user
->seq
< log_first_seq
) {
912 /* our last seen message is gone, return error and reset */
913 user
->idx
= log_first_idx
;
914 user
->seq
= log_first_seq
;
920 msg
= log_from_idx(user
->idx
);
921 len
= msg_print_ext_header(user
->buf
, sizeof(user
->buf
),
923 len
+= msg_print_ext_body(user
->buf
+ len
, sizeof(user
->buf
) - len
,
924 log_dict(msg
), msg
->dict_len
,
925 log_text(msg
), msg
->text_len
);
927 user
->idx
= log_next(user
->idx
);
936 if (copy_to_user(buf
, user
->buf
, len
)) {
942 mutex_unlock(&user
->lock
);
946 static loff_t
devkmsg_llseek(struct file
*file
, loff_t offset
, int whence
)
948 struct devkmsg_user
*user
= file
->private_data
;
959 /* the first record */
960 user
->idx
= log_first_idx
;
961 user
->seq
= log_first_seq
;
965 * The first record after the last SYSLOG_ACTION_CLEAR,
966 * like issued by 'dmesg -c'. Reading /dev/kmsg itself
967 * changes no global state, and does not clear anything.
969 user
->idx
= clear_idx
;
970 user
->seq
= clear_seq
;
973 /* after the last record */
974 user
->idx
= log_next_idx
;
975 user
->seq
= log_next_seq
;
984 static __poll_t
devkmsg_poll(struct file
*file
, poll_table
*wait
)
986 struct devkmsg_user
*user
= file
->private_data
;
990 return EPOLLERR
|EPOLLNVAL
;
992 poll_wait(file
, &log_wait
, wait
);
995 if (user
->seq
< log_next_seq
) {
996 /* return error when data has vanished underneath us */
997 if (user
->seq
< log_first_seq
)
998 ret
= EPOLLIN
|EPOLLRDNORM
|EPOLLERR
|EPOLLPRI
;
1000 ret
= EPOLLIN
|EPOLLRDNORM
;
1002 logbuf_unlock_irq();
1007 static int devkmsg_open(struct inode
*inode
, struct file
*file
)
1009 struct devkmsg_user
*user
;
1012 if (devkmsg_log
& DEVKMSG_LOG_MASK_OFF
)
1015 /* write-only does not need any file context */
1016 if ((file
->f_flags
& O_ACCMODE
) != O_WRONLY
) {
1017 err
= check_syslog_permissions(SYSLOG_ACTION_READ_ALL
,
1018 SYSLOG_FROM_READER
);
1023 user
= kmalloc(sizeof(struct devkmsg_user
), GFP_KERNEL
);
1027 ratelimit_default_init(&user
->rs
);
1028 ratelimit_set_flags(&user
->rs
, RATELIMIT_MSG_ON_RELEASE
);
1030 mutex_init(&user
->lock
);
1033 user
->idx
= log_first_idx
;
1034 user
->seq
= log_first_seq
;
1035 logbuf_unlock_irq();
1037 file
->private_data
= user
;
1041 static int devkmsg_release(struct inode
*inode
, struct file
*file
)
1043 struct devkmsg_user
*user
= file
->private_data
;
1048 ratelimit_state_exit(&user
->rs
);
1050 mutex_destroy(&user
->lock
);
1055 const struct file_operations kmsg_fops
= {
1056 .open
= devkmsg_open
,
1057 .read
= devkmsg_read
,
1058 .write_iter
= devkmsg_write
,
1059 .llseek
= devkmsg_llseek
,
1060 .poll
= devkmsg_poll
,
1061 .release
= devkmsg_release
,
1064 #ifdef CONFIG_CRASH_CORE
1066 * This appends the listed symbols to /proc/vmcore
1068 * /proc/vmcore is used by various utilities, like crash and makedumpfile to
1069 * obtain access to symbols that are otherwise very difficult to locate. These
1070 * symbols are specifically used so that utilities can access and extract the
1071 * dmesg log from a vmcore file after a crash.
1073 void log_buf_vmcoreinfo_setup(void)
1075 VMCOREINFO_SYMBOL(log_buf
);
1076 VMCOREINFO_SYMBOL(log_buf_len
);
1077 VMCOREINFO_SYMBOL(log_first_idx
);
1078 VMCOREINFO_SYMBOL(clear_idx
);
1079 VMCOREINFO_SYMBOL(log_next_idx
);
1081 * Export struct printk_log size and field offsets. User space tools can
1082 * parse it and detect any changes to structure down the line.
1084 VMCOREINFO_STRUCT_SIZE(printk_log
);
1085 VMCOREINFO_OFFSET(printk_log
, ts_nsec
);
1086 VMCOREINFO_OFFSET(printk_log
, len
);
1087 VMCOREINFO_OFFSET(printk_log
, text_len
);
1088 VMCOREINFO_OFFSET(printk_log
, dict_len
);
1089 #ifdef CONFIG_PRINTK_CALLER
1090 VMCOREINFO_OFFSET(printk_log
, caller_id
);
1095 /* requested log_buf_len from kernel cmdline */
1096 static unsigned long __initdata new_log_buf_len
;
1098 /* we practice scaling the ring buffer by powers of 2 */
1099 static void __init
log_buf_len_update(u64 size
)
1101 if (size
> (u64
)LOG_BUF_LEN_MAX
) {
1102 size
= (u64
)LOG_BUF_LEN_MAX
;
1103 pr_err("log_buf over 2G is not supported.\n");
1107 size
= roundup_pow_of_two(size
);
1108 if (size
> log_buf_len
)
1109 new_log_buf_len
= (unsigned long)size
;
1112 /* save requested log_buf_len since it's too early to process it */
1113 static int __init
log_buf_len_setup(char *str
)
1120 size
= memparse(str
, &str
);
1122 log_buf_len_update(size
);
1126 early_param("log_buf_len", log_buf_len_setup
);
1129 #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1131 static void __init
log_buf_add_cpu(void)
1133 unsigned int cpu_extra
;
1136 * archs should set up cpu_possible_bits properly with
1137 * set_cpu_possible() after setup_arch() but just in
1138 * case lets ensure this is valid.
1140 if (num_possible_cpus() == 1)
1143 cpu_extra
= (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN
;
1145 /* by default this will only continue through for large > 64 CPUs */
1146 if (cpu_extra
<= __LOG_BUF_LEN
/ 2)
1149 pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1150 __LOG_CPU_MAX_BUF_LEN
);
1151 pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1153 pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN
);
1155 log_buf_len_update(cpu_extra
+ __LOG_BUF_LEN
);
1157 #else /* !CONFIG_SMP */
1158 static inline void log_buf_add_cpu(void) {}
1159 #endif /* CONFIG_SMP */
1161 static void __init
set_percpu_data_ready(void)
1164 /* Make sure we set this flag only after printk_safe() init is done */
1166 __printk_percpu_data_ready
= true;
1169 void __init
setup_log_buf(int early
)
1171 unsigned long flags
;
1176 * Some archs call setup_log_buf() multiple times - first is very
1177 * early, e.g. from setup_arch(), and second - when percpu_areas
1181 set_percpu_data_ready();
1183 if (log_buf
!= __log_buf
)
1186 if (!early
&& !new_log_buf_len
)
1189 if (!new_log_buf_len
)
1192 new_log_buf
= memblock_alloc(new_log_buf_len
, LOG_ALIGN
);
1193 if (unlikely(!new_log_buf
)) {
1194 pr_err("log_buf_len: %lu bytes not available\n",
1199 logbuf_lock_irqsave(flags
);
1200 log_buf_len
= new_log_buf_len
;
1201 log_buf
= new_log_buf
;
1202 new_log_buf_len
= 0;
1203 free
= __LOG_BUF_LEN
- log_next_idx
;
1204 memcpy(log_buf
, __log_buf
, __LOG_BUF_LEN
);
1205 logbuf_unlock_irqrestore(flags
);
1207 pr_info("log_buf_len: %u bytes\n", log_buf_len
);
1208 pr_info("early log buf free: %u(%u%%)\n",
1209 free
, (free
* 100) / __LOG_BUF_LEN
);
1212 static bool __read_mostly ignore_loglevel
;
1214 static int __init
ignore_loglevel_setup(char *str
)
1216 ignore_loglevel
= true;
1217 pr_info("debug: ignoring loglevel setting.\n");
1222 early_param("ignore_loglevel", ignore_loglevel_setup
);
1223 module_param(ignore_loglevel
, bool, S_IRUGO
| S_IWUSR
);
1224 MODULE_PARM_DESC(ignore_loglevel
,
1225 "ignore loglevel setting (prints all kernel messages to the console)");
1227 static bool suppress_message_printing(int level
)
1229 return (level
>= console_loglevel
&& !ignore_loglevel
);
1232 #ifdef CONFIG_BOOT_PRINTK_DELAY
1234 static int boot_delay
; /* msecs delay after each printk during bootup */
1235 static unsigned long long loops_per_msec
; /* based on boot_delay */
1237 static int __init
boot_delay_setup(char *str
)
1241 lpj
= preset_lpj
? preset_lpj
: 1000000; /* some guess */
1242 loops_per_msec
= (unsigned long long)lpj
/ 1000 * HZ
;
1244 get_option(&str
, &boot_delay
);
1245 if (boot_delay
> 10 * 1000)
1248 pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1249 "HZ: %d, loops_per_msec: %llu\n",
1250 boot_delay
, preset_lpj
, lpj
, HZ
, loops_per_msec
);
1253 early_param("boot_delay", boot_delay_setup
);
1255 static void boot_delay_msec(int level
)
1257 unsigned long long k
;
1258 unsigned long timeout
;
1260 if ((boot_delay
== 0 || system_state
>= SYSTEM_RUNNING
)
1261 || suppress_message_printing(level
)) {
1265 k
= (unsigned long long)loops_per_msec
* boot_delay
;
1267 timeout
= jiffies
+ msecs_to_jiffies(boot_delay
);
1272 * use (volatile) jiffies to prevent
1273 * compiler reduction; loop termination via jiffies
1274 * is secondary and may or may not happen.
1276 if (time_after(jiffies
, timeout
))
1278 touch_nmi_watchdog();
1282 static inline void boot_delay_msec(int level
)
1287 static bool printk_time
= IS_ENABLED(CONFIG_PRINTK_TIME
);
1288 module_param_named(time
, printk_time
, bool, S_IRUGO
| S_IWUSR
);
1290 static size_t print_syslog(unsigned int level
, char *buf
)
1292 return sprintf(buf
, "<%u>", level
);
1295 static size_t print_time(u64 ts
, char *buf
)
1297 unsigned long rem_nsec
= do_div(ts
, 1000000000);
1299 return sprintf(buf
, "[%5lu.%06lu]",
1300 (unsigned long)ts
, rem_nsec
/ 1000);
1303 #ifdef CONFIG_PRINTK_CALLER
1304 static size_t print_caller(u32 id
, char *buf
)
1308 snprintf(caller
, sizeof(caller
), "%c%u",
1309 id
& 0x80000000 ? 'C' : 'T', id
& ~0x80000000);
1310 return sprintf(buf
, "[%6s]", caller
);
1313 #define print_caller(id, buf) 0
1316 static size_t print_prefix(const struct printk_log
*msg
, bool syslog
,
1317 bool time
, char *buf
)
1322 len
= print_syslog((msg
->facility
<< 3) | msg
->level
, buf
);
1325 len
+= print_time(msg
->ts_nsec
, buf
+ len
);
1327 len
+= print_caller(msg
->caller_id
, buf
+ len
);
1329 if (IS_ENABLED(CONFIG_PRINTK_CALLER
) || time
) {
1337 static size_t msg_print_text(const struct printk_log
*msg
, bool syslog
,
1338 bool time
, char *buf
, size_t size
)
1340 const char *text
= log_text(msg
);
1341 size_t text_size
= msg
->text_len
;
1343 char prefix
[PREFIX_MAX
];
1344 const size_t prefix_len
= print_prefix(msg
, syslog
, time
, prefix
);
1347 const char *next
= memchr(text
, '\n', text_size
);
1351 text_len
= next
- text
;
1353 text_size
-= next
- text
;
1355 text_len
= text_size
;
1359 if (prefix_len
+ text_len
+ 1 >= size
- len
)
1362 memcpy(buf
+ len
, prefix
, prefix_len
);
1364 memcpy(buf
+ len
, text
, text_len
);
1368 /* SYSLOG_ACTION_* buffer size only calculation */
1369 len
+= prefix_len
+ text_len
+ 1;
1378 static int syslog_print(char __user
*buf
, int size
)
1381 struct printk_log
*msg
;
1384 text
= kmalloc(LOG_LINE_MAX
+ PREFIX_MAX
, GFP_KERNEL
);
1393 if (syslog_seq
< log_first_seq
) {
1394 /* messages are gone, move to first one */
1395 syslog_seq
= log_first_seq
;
1396 syslog_idx
= log_first_idx
;
1399 if (syslog_seq
== log_next_seq
) {
1400 logbuf_unlock_irq();
1405 * To keep reading/counting partial line consistent,
1406 * use printk_time value as of the beginning of a line.
1408 if (!syslog_partial
)
1409 syslog_time
= printk_time
;
1411 skip
= syslog_partial
;
1412 msg
= log_from_idx(syslog_idx
);
1413 n
= msg_print_text(msg
, true, syslog_time
, text
,
1414 LOG_LINE_MAX
+ PREFIX_MAX
);
1415 if (n
- syslog_partial
<= size
) {
1416 /* message fits into buffer, move forward */
1417 syslog_idx
= log_next(syslog_idx
);
1419 n
-= syslog_partial
;
1422 /* partial read(), remember position */
1424 syslog_partial
+= n
;
1427 logbuf_unlock_irq();
1432 if (copy_to_user(buf
, text
+ skip
, n
)) {
1447 static int syslog_print_all(char __user
*buf
, int size
, bool clear
)
1456 text
= kmalloc(LOG_LINE_MAX
+ PREFIX_MAX
, GFP_KERNEL
);
1463 * Find first record that fits, including all following records,
1464 * into the user-provided buffer for this dump.
1468 while (seq
< log_next_seq
) {
1469 struct printk_log
*msg
= log_from_idx(idx
);
1471 len
+= msg_print_text(msg
, true, time
, NULL
, 0);
1472 idx
= log_next(idx
);
1476 /* move first record forward until length fits into the buffer */
1479 while (len
> size
&& seq
< log_next_seq
) {
1480 struct printk_log
*msg
= log_from_idx(idx
);
1482 len
-= msg_print_text(msg
, true, time
, NULL
, 0);
1483 idx
= log_next(idx
);
1487 /* last message fitting into this dump */
1488 next_seq
= log_next_seq
;
1491 while (len
>= 0 && seq
< next_seq
) {
1492 struct printk_log
*msg
= log_from_idx(idx
);
1493 int textlen
= msg_print_text(msg
, true, time
, text
,
1494 LOG_LINE_MAX
+ PREFIX_MAX
);
1496 idx
= log_next(idx
);
1499 logbuf_unlock_irq();
1500 if (copy_to_user(buf
+ len
, text
, textlen
))
1506 if (seq
< log_first_seq
) {
1507 /* messages are gone, move to next one */
1508 seq
= log_first_seq
;
1509 idx
= log_first_idx
;
1514 clear_seq
= log_next_seq
;
1515 clear_idx
= log_next_idx
;
1517 logbuf_unlock_irq();
1523 static void syslog_clear(void)
1526 clear_seq
= log_next_seq
;
1527 clear_idx
= log_next_idx
;
1528 logbuf_unlock_irq();
1531 int do_syslog(int type
, char __user
*buf
, int len
, int source
)
1534 static int saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1537 error
= check_syslog_permissions(type
, source
);
1542 case SYSLOG_ACTION_CLOSE
: /* Close log */
1544 case SYSLOG_ACTION_OPEN
: /* Open log */
1546 case SYSLOG_ACTION_READ
: /* Read from log */
1547 if (!buf
|| len
< 0)
1551 if (!access_ok(buf
, len
))
1553 error
= wait_event_interruptible(log_wait
,
1554 syslog_seq
!= log_next_seq
);
1557 error
= syslog_print(buf
, len
);
1559 /* Read/clear last kernel messages */
1560 case SYSLOG_ACTION_READ_CLEAR
:
1563 /* Read last kernel messages */
1564 case SYSLOG_ACTION_READ_ALL
:
1565 if (!buf
|| len
< 0)
1569 if (!access_ok(buf
, len
))
1571 error
= syslog_print_all(buf
, len
, clear
);
1573 /* Clear ring buffer */
1574 case SYSLOG_ACTION_CLEAR
:
1577 /* Disable logging to console */
1578 case SYSLOG_ACTION_CONSOLE_OFF
:
1579 if (saved_console_loglevel
== LOGLEVEL_DEFAULT
)
1580 saved_console_loglevel
= console_loglevel
;
1581 console_loglevel
= minimum_console_loglevel
;
1583 /* Enable logging to console */
1584 case SYSLOG_ACTION_CONSOLE_ON
:
1585 if (saved_console_loglevel
!= LOGLEVEL_DEFAULT
) {
1586 console_loglevel
= saved_console_loglevel
;
1587 saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1590 /* Set level of messages printed to console */
1591 case SYSLOG_ACTION_CONSOLE_LEVEL
:
1592 if (len
< 1 || len
> 8)
1594 if (len
< minimum_console_loglevel
)
1595 len
= minimum_console_loglevel
;
1596 console_loglevel
= len
;
1597 /* Implicitly re-enable logging to console */
1598 saved_console_loglevel
= LOGLEVEL_DEFAULT
;
1600 /* Number of chars in the log buffer */
1601 case SYSLOG_ACTION_SIZE_UNREAD
:
1603 if (syslog_seq
< log_first_seq
) {
1604 /* messages are gone, move to first one */
1605 syslog_seq
= log_first_seq
;
1606 syslog_idx
= log_first_idx
;
1609 if (source
== SYSLOG_FROM_PROC
) {
1611 * Short-cut for poll(/"proc/kmsg") which simply checks
1612 * for pending data, not the size; return the count of
1613 * records, not the length.
1615 error
= log_next_seq
- syslog_seq
;
1617 u64 seq
= syslog_seq
;
1618 u32 idx
= syslog_idx
;
1619 bool time
= syslog_partial
? syslog_time
: printk_time
;
1621 while (seq
< log_next_seq
) {
1622 struct printk_log
*msg
= log_from_idx(idx
);
1624 error
+= msg_print_text(msg
, true, time
, NULL
,
1627 idx
= log_next(idx
);
1630 error
-= syslog_partial
;
1632 logbuf_unlock_irq();
1634 /* Size of the log buffer */
1635 case SYSLOG_ACTION_SIZE_BUFFER
:
1636 error
= log_buf_len
;
1646 SYSCALL_DEFINE3(syslog
, int, type
, char __user
*, buf
, int, len
)
1648 return do_syslog(type
, buf
, len
, SYSLOG_FROM_READER
);
1652 * Special console_lock variants that help to reduce the risk of soft-lockups.
1653 * They allow to pass console_lock to another printk() call using a busy wait.
1656 #ifdef CONFIG_LOCKDEP
1657 static struct lockdep_map console_owner_dep_map
= {
1658 .name
= "console_owner"
1662 static DEFINE_RAW_SPINLOCK(console_owner_lock
);
1663 static struct task_struct
*console_owner
;
1664 static bool console_waiter
;
1667 * console_lock_spinning_enable - mark beginning of code where another
1668 * thread might safely busy wait
1670 * This basically converts console_lock into a spinlock. This marks
1671 * the section where the console_lock owner can not sleep, because
1672 * there may be a waiter spinning (like a spinlock). Also it must be
1673 * ready to hand over the lock at the end of the section.
1675 static void console_lock_spinning_enable(void)
1677 raw_spin_lock(&console_owner_lock
);
1678 console_owner
= current
;
1679 raw_spin_unlock(&console_owner_lock
);
1681 /* The waiter may spin on us after setting console_owner */
1682 spin_acquire(&console_owner_dep_map
, 0, 0, _THIS_IP_
);
1686 * console_lock_spinning_disable_and_check - mark end of code where another
1687 * thread was able to busy wait and check if there is a waiter
1689 * This is called at the end of the section where spinning is allowed.
1690 * It has two functions. First, it is a signal that it is no longer
1691 * safe to start busy waiting for the lock. Second, it checks if
1692 * there is a busy waiter and passes the lock rights to her.
1694 * Important: Callers lose the lock if there was a busy waiter.
1695 * They must not touch items synchronized by console_lock
1698 * Return: 1 if the lock rights were passed, 0 otherwise.
1700 static int console_lock_spinning_disable_and_check(void)
1704 raw_spin_lock(&console_owner_lock
);
1705 waiter
= READ_ONCE(console_waiter
);
1706 console_owner
= NULL
;
1707 raw_spin_unlock(&console_owner_lock
);
1710 spin_release(&console_owner_dep_map
, _THIS_IP_
);
1714 /* The waiter is now free to continue */
1715 WRITE_ONCE(console_waiter
, false);
1717 spin_release(&console_owner_dep_map
, _THIS_IP_
);
1720 * Hand off console_lock to waiter. The waiter will perform
1721 * the up(). After this, the waiter is the console_lock owner.
1723 mutex_release(&console_lock_dep_map
, _THIS_IP_
);
1728 * console_trylock_spinning - try to get console_lock by busy waiting
1730 * This allows to busy wait for the console_lock when the current
1731 * owner is running in specially marked sections. It means that
1732 * the current owner is running and cannot reschedule until it
1733 * is ready to lose the lock.
1735 * Return: 1 if we got the lock, 0 othrewise
1737 static int console_trylock_spinning(void)
1739 struct task_struct
*owner
= NULL
;
1742 unsigned long flags
;
1744 if (console_trylock())
1747 printk_safe_enter_irqsave(flags
);
1749 raw_spin_lock(&console_owner_lock
);
1750 owner
= READ_ONCE(console_owner
);
1751 waiter
= READ_ONCE(console_waiter
);
1752 if (!waiter
&& owner
&& owner
!= current
) {
1753 WRITE_ONCE(console_waiter
, true);
1756 raw_spin_unlock(&console_owner_lock
);
1759 * If there is an active printk() writing to the
1760 * consoles, instead of having it write our data too,
1761 * see if we can offload that load from the active
1762 * printer, and do some printing ourselves.
1763 * Go into a spin only if there isn't already a waiter
1764 * spinning, and there is an active printer, and
1765 * that active printer isn't us (recursive printk?).
1768 printk_safe_exit_irqrestore(flags
);
1772 /* We spin waiting for the owner to release us */
1773 spin_acquire(&console_owner_dep_map
, 0, 0, _THIS_IP_
);
1774 /* Owner will clear console_waiter on hand off */
1775 while (READ_ONCE(console_waiter
))
1777 spin_release(&console_owner_dep_map
, _THIS_IP_
);
1779 printk_safe_exit_irqrestore(flags
);
1781 * The owner passed the console lock to us.
1782 * Since we did not spin on console lock, annotate
1783 * this as a trylock. Otherwise lockdep will
1786 mutex_acquire(&console_lock_dep_map
, 0, 1, _THIS_IP_
);
1792 * Call the console drivers, asking them to write out
1793 * log_buf[start] to log_buf[end - 1].
1794 * The console_lock must be held.
1796 static void call_console_drivers(const char *ext_text
, size_t ext_len
,
1797 const char *text
, size_t len
)
1799 struct console
*con
;
1801 trace_console_rcuidle(text
, len
);
1803 for_each_console(con
) {
1804 if (exclusive_console
&& con
!= exclusive_console
)
1806 if (!(con
->flags
& CON_ENABLED
))
1810 if (!cpu_online(smp_processor_id()) &&
1811 !(con
->flags
& CON_ANYTIME
))
1813 if (con
->flags
& CON_EXTENDED
)
1814 con
->write(con
, ext_text
, ext_len
);
1816 con
->write(con
, text
, len
);
1820 int printk_delay_msec __read_mostly
;
1822 static inline void printk_delay(void)
1824 if (unlikely(printk_delay_msec
)) {
1825 int m
= printk_delay_msec
;
1829 touch_nmi_watchdog();
1834 static inline u32
printk_caller_id(void)
1836 return in_task() ? task_pid_nr(current
) :
1837 0x80000000 + raw_smp_processor_id();
1841 * Continuation lines are buffered, and not committed to the record buffer
1842 * until the line is complete, or a race forces it. The line fragments
1843 * though, are printed immediately to the consoles to ensure everything has
1844 * reached the console in case of a kernel crash.
1846 static struct cont
{
1847 char buf
[LOG_LINE_MAX
];
1848 size_t len
; /* length == 0 means unused buffer */
1849 u32 caller_id
; /* printk_caller_id() of first print */
1850 u64 ts_nsec
; /* time of first print */
1851 u8 level
; /* log level of first message */
1852 u8 facility
; /* log facility of first message */
1853 enum log_flags flags
; /* prefix, newline flags */
1856 static void cont_flush(void)
1861 log_store(cont
.caller_id
, cont
.facility
, cont
.level
, cont
.flags
,
1862 cont
.ts_nsec
, NULL
, 0, cont
.buf
, cont
.len
);
1866 static bool cont_add(u32 caller_id
, int facility
, int level
,
1867 enum log_flags flags
, const char *text
, size_t len
)
1869 /* If the line gets too long, split it up in separate records. */
1870 if (cont
.len
+ len
> sizeof(cont
.buf
)) {
1876 cont
.facility
= facility
;
1878 cont
.caller_id
= caller_id
;
1879 cont
.ts_nsec
= local_clock();
1883 memcpy(cont
.buf
+ cont
.len
, text
, len
);
1886 // The original flags come from the first line,
1887 // but later continuations can add a newline.
1888 if (flags
& LOG_NEWLINE
) {
1889 cont
.flags
|= LOG_NEWLINE
;
1896 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
)
1898 const u32 caller_id
= printk_caller_id();
1901 * If an earlier line was buffered, and we're a continuation
1902 * write from the same context, try to add it to the buffer.
1905 if (cont
.caller_id
== caller_id
&& (lflags
& LOG_CONT
)) {
1906 if (cont_add(caller_id
, facility
, level
, lflags
, text
, text_len
))
1909 /* Otherwise, make sure it's flushed */
1913 /* Skip empty continuation lines that couldn't be added - they just flush */
1914 if (!text_len
&& (lflags
& LOG_CONT
))
1917 /* If it doesn't end in a newline, try to buffer the current line */
1918 if (!(lflags
& LOG_NEWLINE
)) {
1919 if (cont_add(caller_id
, facility
, level
, lflags
, text
, text_len
))
1923 /* Store it in the record log */
1924 return log_store(caller_id
, facility
, level
, lflags
, 0,
1925 dict
, dictlen
, text
, text_len
);
1928 /* Must be called under logbuf_lock. */
1929 int vprintk_store(int facility
, int level
,
1930 const char *dict
, size_t dictlen
,
1931 const char *fmt
, va_list args
)
1933 static char textbuf
[LOG_LINE_MAX
];
1934 char *text
= textbuf
;
1936 enum log_flags lflags
= 0;
1939 * The printf needs to come first; we need the syslog
1940 * prefix which might be passed-in as a parameter.
1942 text_len
= vscnprintf(text
, sizeof(textbuf
), fmt
, args
);
1944 /* mark and strip a trailing newline */
1945 if (text_len
&& text
[text_len
-1] == '\n') {
1947 lflags
|= LOG_NEWLINE
;
1950 /* strip kernel syslog prefix and extract log level or control flags */
1951 if (facility
== 0) {
1954 while ((kern_level
= printk_get_level(text
)) != 0) {
1955 switch (kern_level
) {
1957 if (level
== LOGLEVEL_DEFAULT
)
1958 level
= kern_level
- '0';
1960 case 'c': /* KERN_CONT */
1969 if (level
== LOGLEVEL_DEFAULT
)
1970 level
= default_message_loglevel
;
1973 lflags
|= LOG_NEWLINE
;
1975 return log_output(facility
, level
, lflags
,
1976 dict
, dictlen
, text
, text_len
);
1979 asmlinkage
int vprintk_emit(int facility
, int level
,
1980 const char *dict
, size_t dictlen
,
1981 const char *fmt
, va_list args
)
1984 bool in_sched
= false, pending_output
;
1985 unsigned long flags
;
1988 /* Suppress unimportant messages after panic happens */
1989 if (unlikely(suppress_printk
))
1992 if (level
== LOGLEVEL_SCHED
) {
1993 level
= LOGLEVEL_DEFAULT
;
1997 boot_delay_msec(level
);
2000 /* This stops the holder of console_sem just where we want him */
2001 logbuf_lock_irqsave(flags
);
2002 curr_log_seq
= log_next_seq
;
2003 printed_len
= vprintk_store(facility
, level
, dict
, dictlen
, fmt
, args
);
2004 pending_output
= (curr_log_seq
!= log_next_seq
);
2005 logbuf_unlock_irqrestore(flags
);
2007 /* If called from the scheduler, we can not call up(). */
2008 if (!in_sched
&& pending_output
) {
2010 * Disable preemption to avoid being preempted while holding
2011 * console_sem which would prevent anyone from printing to
2016 * Try to acquire and then immediately release the console
2017 * semaphore. The release will print out buffers and wake up
2018 * /dev/kmsg and syslog() users.
2020 if (console_trylock_spinning())
2029 EXPORT_SYMBOL(vprintk_emit
);
2031 asmlinkage
int vprintk(const char *fmt
, va_list args
)
2033 return vprintk_func(fmt
, args
);
2035 EXPORT_SYMBOL(vprintk
);
2037 int vprintk_default(const char *fmt
, va_list args
)
2041 #ifdef CONFIG_KGDB_KDB
2042 /* Allow to pass printk() to kdb but avoid a recursion. */
2043 if (unlikely(kdb_trap_printk
&& kdb_printf_cpu
< 0)) {
2044 r
= vkdb_printf(KDB_MSGSRC_PRINTK
, fmt
, args
);
2048 r
= vprintk_emit(0, LOGLEVEL_DEFAULT
, NULL
, 0, fmt
, args
);
2052 EXPORT_SYMBOL_GPL(vprintk_default
);
2055 * printk - print a kernel message
2056 * @fmt: format string
2058 * This is printk(). It can be called from any context. We want it to work.
2060 * We try to grab the console_lock. If we succeed, it's easy - we log the
2061 * output and call the console drivers. If we fail to get the semaphore, we
2062 * place the output into the log buffer and return. The current holder of
2063 * the console_sem will notice the new output in console_unlock(); and will
2064 * send it to the consoles before releasing the lock.
2066 * One effect of this deferred printing is that code which calls printk() and
2067 * then changes console_loglevel may break. This is because console_loglevel
2068 * is inspected when the actual printing occurs.
2073 * See the vsnprintf() documentation for format string extensions over C99.
2075 asmlinkage __visible
int printk(const char *fmt
, ...)
2080 va_start(args
, fmt
);
2081 r
= vprintk_func(fmt
, args
);
2086 EXPORT_SYMBOL(printk
);
2088 #else /* CONFIG_PRINTK */
2090 #define LOG_LINE_MAX 0
2091 #define PREFIX_MAX 0
2092 #define printk_time false
2094 static u64 syslog_seq
;
2095 static u32 syslog_idx
;
2096 static u64 console_seq
;
2097 static u32 console_idx
;
2098 static u64 exclusive_console_stop_seq
;
2099 static u64 log_first_seq
;
2100 static u32 log_first_idx
;
2101 static u64 log_next_seq
;
2102 static char *log_text(const struct printk_log
*msg
) { return NULL
; }
2103 static char *log_dict(const struct printk_log
*msg
) { return NULL
; }
2104 static struct printk_log
*log_from_idx(u32 idx
) { return NULL
; }
2105 static u32
log_next(u32 idx
) { return 0; }
2106 static ssize_t
msg_print_ext_header(char *buf
, size_t size
,
2107 struct printk_log
*msg
,
2108 u64 seq
) { return 0; }
2109 static ssize_t
msg_print_ext_body(char *buf
, size_t size
,
2110 char *dict
, size_t dict_len
,
2111 char *text
, size_t text_len
) { return 0; }
2112 static void console_lock_spinning_enable(void) { }
2113 static int console_lock_spinning_disable_and_check(void) { return 0; }
2114 static void call_console_drivers(const char *ext_text
, size_t ext_len
,
2115 const char *text
, size_t len
) {}
2116 static size_t msg_print_text(const struct printk_log
*msg
, bool syslog
,
2117 bool time
, char *buf
, size_t size
) { return 0; }
2118 static bool suppress_message_printing(int level
) { return false; }
2120 #endif /* CONFIG_PRINTK */
2122 #ifdef CONFIG_EARLY_PRINTK
2123 struct console
*early_console
;
2125 asmlinkage __visible
void early_printk(const char *fmt
, ...)
2135 n
= vscnprintf(buf
, sizeof(buf
), fmt
, ap
);
2138 early_console
->write(early_console
, buf
, n
);
2142 static int __add_preferred_console(char *name
, int idx
, char *options
,
2145 struct console_cmdline
*c
;
2149 * See if this tty is not yet registered, and
2150 * if we have a slot free.
2152 for (i
= 0, c
= console_cmdline
;
2153 i
< MAX_CMDLINECONSOLES
&& c
->name
[0];
2155 if (strcmp(c
->name
, name
) == 0 && c
->index
== idx
) {
2157 preferred_console
= i
;
2161 if (i
== MAX_CMDLINECONSOLES
)
2164 preferred_console
= i
;
2165 strlcpy(c
->name
, name
, sizeof(c
->name
));
2166 c
->options
= options
;
2167 braille_set_options(c
, brl_options
);
2173 static int __init
console_msg_format_setup(char *str
)
2175 if (!strcmp(str
, "syslog"))
2176 console_msg_format
= MSG_FORMAT_SYSLOG
;
2177 if (!strcmp(str
, "default"))
2178 console_msg_format
= MSG_FORMAT_DEFAULT
;
2181 __setup("console_msg_format=", console_msg_format_setup
);
2184 * Set up a console. Called via do_early_param() in init/main.c
2185 * for each "console=" parameter in the boot command line.
2187 static int __init
console_setup(char *str
)
2189 char buf
[sizeof(console_cmdline
[0].name
) + 4]; /* 4 for "ttyS" */
2190 char *s
, *options
, *brl_options
= NULL
;
2193 if (_braille_console_setup(&str
, &brl_options
))
2197 * Decode str into name, index, options.
2199 if (str
[0] >= '0' && str
[0] <= '9') {
2200 strcpy(buf
, "ttyS");
2201 strncpy(buf
+ 4, str
, sizeof(buf
) - 5);
2203 strncpy(buf
, str
, sizeof(buf
) - 1);
2205 buf
[sizeof(buf
) - 1] = 0;
2206 options
= strchr(str
, ',');
2210 if (!strcmp(str
, "ttya"))
2211 strcpy(buf
, "ttyS0");
2212 if (!strcmp(str
, "ttyb"))
2213 strcpy(buf
, "ttyS1");
2215 for (s
= buf
; *s
; s
++)
2216 if (isdigit(*s
) || *s
== ',')
2218 idx
= simple_strtoul(s
, NULL
, 10);
2221 __add_preferred_console(buf
, idx
, options
, brl_options
);
2222 console_set_on_cmdline
= 1;
2225 __setup("console=", console_setup
);
2228 * add_preferred_console - add a device to the list of preferred consoles.
2229 * @name: device name
2230 * @idx: device index
2231 * @options: options for this console
2233 * The last preferred console added will be used for kernel messages
2234 * and stdin/out/err for init. Normally this is used by console_setup
2235 * above to handle user-supplied console arguments; however it can also
2236 * be used by arch-specific code either to override the user or more
2237 * commonly to provide a default console (ie from PROM variables) when
2238 * the user has not supplied one.
2240 int add_preferred_console(char *name
, int idx
, char *options
)
2242 return __add_preferred_console(name
, idx
, options
, NULL
);
2245 bool console_suspend_enabled
= true;
2246 EXPORT_SYMBOL(console_suspend_enabled
);
2248 static int __init
console_suspend_disable(char *str
)
2250 console_suspend_enabled
= false;
2253 __setup("no_console_suspend", console_suspend_disable
);
2254 module_param_named(console_suspend
, console_suspend_enabled
,
2255 bool, S_IRUGO
| S_IWUSR
);
2256 MODULE_PARM_DESC(console_suspend
, "suspend console during suspend"
2257 " and hibernate operations");
2260 * suspend_console - suspend the console subsystem
2262 * This disables printk() while we go into suspend states
2264 void suspend_console(void)
2266 if (!console_suspend_enabled
)
2268 pr_info("Suspending console(s) (use no_console_suspend to debug)\n");
2270 console_suspended
= 1;
2274 void resume_console(void)
2276 if (!console_suspend_enabled
)
2279 console_suspended
= 0;
2284 * console_cpu_notify - print deferred console messages after CPU hotplug
2287 * If printk() is called from a CPU that is not online yet, the messages
2288 * will be printed on the console only if there are CON_ANYTIME consoles.
2289 * This function is called when a new CPU comes online (or fails to come
2290 * up) or goes offline.
2292 static int console_cpu_notify(unsigned int cpu
)
2294 if (!cpuhp_tasks_frozen
) {
2295 /* If trylock fails, someone else is doing the printing */
2296 if (console_trylock())
2303 * console_lock - lock the console system for exclusive use.
2305 * Acquires a lock which guarantees that the caller has
2306 * exclusive access to the console system and the console_drivers list.
2308 * Can sleep, returns nothing.
2310 void console_lock(void)
2315 if (console_suspended
)
2318 console_may_schedule
= 1;
2320 EXPORT_SYMBOL(console_lock
);
2323 * console_trylock - try to lock the console system for exclusive use.
2325 * Try to acquire a lock which guarantees that the caller has exclusive
2326 * access to the console system and the console_drivers list.
2328 * returns 1 on success, and 0 on failure to acquire the lock.
2330 int console_trylock(void)
2332 if (down_trylock_console_sem())
2334 if (console_suspended
) {
2339 console_may_schedule
= 0;
2342 EXPORT_SYMBOL(console_trylock
);
2344 int is_console_locked(void)
2346 return console_locked
;
2348 EXPORT_SYMBOL(is_console_locked
);
2351 * Check if we have any console that is capable of printing while cpu is
2352 * booting or shutting down. Requires console_sem.
2354 static int have_callable_console(void)
2356 struct console
*con
;
2358 for_each_console(con
)
2359 if ((con
->flags
& CON_ENABLED
) &&
2360 (con
->flags
& CON_ANYTIME
))
2367 * Can we actually use the console at this time on this cpu?
2369 * Console drivers may assume that per-cpu resources have been allocated. So
2370 * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2371 * call them until this CPU is officially up.
2373 static inline int can_use_console(void)
2375 return cpu_online(raw_smp_processor_id()) || have_callable_console();
2379 * console_unlock - unlock the console system
2381 * Releases the console_lock which the caller holds on the console system
2382 * and the console driver list.
2384 * While the console_lock was held, console output may have been buffered
2385 * by printk(). If this is the case, console_unlock(); emits
2386 * the output prior to releasing the lock.
2388 * If there is output waiting, we wake /dev/kmsg and syslog() users.
2390 * console_unlock(); may be called from any context.
2392 void console_unlock(void)
2394 static char ext_text
[CONSOLE_EXT_LOG_MAX
];
2395 static char text
[LOG_LINE_MAX
+ PREFIX_MAX
];
2396 unsigned long flags
;
2397 bool do_cond_resched
, retry
;
2399 if (console_suspended
) {
2405 * Console drivers are called with interrupts disabled, so
2406 * @console_may_schedule should be cleared before; however, we may
2407 * end up dumping a lot of lines, for example, if called from
2408 * console registration path, and should invoke cond_resched()
2409 * between lines if allowable. Not doing so can cause a very long
2410 * scheduling stall on a slow console leading to RCU stall and
2411 * softlockup warnings which exacerbate the issue with more
2412 * messages practically incapacitating the system.
2414 * console_trylock() is not able to detect the preemptive
2415 * context reliably. Therefore the value must be stored before
2416 * and cleared after the the "again" goto label.
2418 do_cond_resched
= console_may_schedule
;
2420 console_may_schedule
= 0;
2423 * We released the console_sem lock, so we need to recheck if
2424 * cpu is online and (if not) is there at least one CON_ANYTIME
2427 if (!can_use_console()) {
2434 struct printk_log
*msg
;
2438 printk_safe_enter_irqsave(flags
);
2439 raw_spin_lock(&logbuf_lock
);
2440 if (console_seq
< log_first_seq
) {
2442 "** %llu printk messages dropped **\n",
2443 log_first_seq
- console_seq
);
2445 /* messages are gone, move to first one */
2446 console_seq
= log_first_seq
;
2447 console_idx
= log_first_idx
;
2452 if (console_seq
== log_next_seq
)
2455 msg
= log_from_idx(console_idx
);
2456 if (suppress_message_printing(msg
->level
)) {
2458 * Skip record we have buffered and already printed
2459 * directly to the console when we received it, and
2460 * record that has level above the console loglevel.
2462 console_idx
= log_next(console_idx
);
2467 /* Output to all consoles once old messages replayed. */
2468 if (unlikely(exclusive_console
&&
2469 console_seq
>= exclusive_console_stop_seq
)) {
2470 exclusive_console
= NULL
;
2473 len
+= msg_print_text(msg
,
2474 console_msg_format
& MSG_FORMAT_SYSLOG
,
2475 printk_time
, text
+ len
, sizeof(text
) - len
);
2476 if (nr_ext_console_drivers
) {
2477 ext_len
= msg_print_ext_header(ext_text
,
2480 ext_len
+= msg_print_ext_body(ext_text
+ ext_len
,
2481 sizeof(ext_text
) - ext_len
,
2482 log_dict(msg
), msg
->dict_len
,
2483 log_text(msg
), msg
->text_len
);
2485 console_idx
= log_next(console_idx
);
2487 raw_spin_unlock(&logbuf_lock
);
2490 * While actively printing out messages, if another printk()
2491 * were to occur on another CPU, it may wait for this one to
2492 * finish. This task can not be preempted if there is a
2493 * waiter waiting to take over.
2495 console_lock_spinning_enable();
2497 stop_critical_timings(); /* don't trace print latency */
2498 call_console_drivers(ext_text
, ext_len
, text
, len
);
2499 start_critical_timings();
2501 if (console_lock_spinning_disable_and_check()) {
2502 printk_safe_exit_irqrestore(flags
);
2506 printk_safe_exit_irqrestore(flags
);
2508 if (do_cond_resched
)
2514 raw_spin_unlock(&logbuf_lock
);
2519 * Someone could have filled up the buffer again, so re-check if there's
2520 * something to flush. In case we cannot trylock the console_sem again,
2521 * there's a new owner and the console_unlock() from them will do the
2522 * flush, no worries.
2524 raw_spin_lock(&logbuf_lock
);
2525 retry
= console_seq
!= log_next_seq
;
2526 raw_spin_unlock(&logbuf_lock
);
2527 printk_safe_exit_irqrestore(flags
);
2529 if (retry
&& console_trylock())
2532 EXPORT_SYMBOL(console_unlock
);
2535 * console_conditional_schedule - yield the CPU if required
2537 * If the console code is currently allowed to sleep, and
2538 * if this CPU should yield the CPU to another task, do
2541 * Must be called within console_lock();.
2543 void __sched
console_conditional_schedule(void)
2545 if (console_may_schedule
)
2548 EXPORT_SYMBOL(console_conditional_schedule
);
2550 void console_unblank(void)
2555 * console_unblank can no longer be called in interrupt context unless
2556 * oops_in_progress is set to 1..
2558 if (oops_in_progress
) {
2559 if (down_trylock_console_sem() != 0)
2565 console_may_schedule
= 0;
2567 if ((c
->flags
& CON_ENABLED
) && c
->unblank
)
2573 * console_flush_on_panic - flush console content on panic
2574 * @mode: flush all messages in buffer or just the pending ones
2576 * Immediately output all pending messages no matter what.
2578 void console_flush_on_panic(enum con_flush_mode mode
)
2581 * If someone else is holding the console lock, trylock will fail
2582 * and may_schedule may be set. Ignore and proceed to unlock so
2583 * that messages are flushed out. As this can be called from any
2584 * context and we don't want to get preempted while flushing,
2585 * ensure may_schedule is cleared.
2588 console_may_schedule
= 0;
2590 if (mode
== CONSOLE_REPLAY_ALL
) {
2591 unsigned long flags
;
2593 logbuf_lock_irqsave(flags
);
2594 console_seq
= log_first_seq
;
2595 console_idx
= log_first_idx
;
2596 logbuf_unlock_irqrestore(flags
);
2602 * Return the console tty driver structure and its associated index
2604 struct tty_driver
*console_device(int *index
)
2607 struct tty_driver
*driver
= NULL
;
2610 for_each_console(c
) {
2613 driver
= c
->device(c
, index
);
2622 * Prevent further output on the passed console device so that (for example)
2623 * serial drivers can disable console output before suspending a port, and can
2624 * re-enable output afterwards.
2626 void console_stop(struct console
*console
)
2629 console
->flags
&= ~CON_ENABLED
;
2632 EXPORT_SYMBOL(console_stop
);
2634 void console_start(struct console
*console
)
2637 console
->flags
|= CON_ENABLED
;
2640 EXPORT_SYMBOL(console_start
);
2642 static int __read_mostly keep_bootcon
;
2644 static int __init
keep_bootcon_setup(char *str
)
2647 pr_info("debug: skip boot console de-registration.\n");
2652 early_param("keep_bootcon", keep_bootcon_setup
);
2655 * The console driver calls this routine during kernel initialization
2656 * to register the console printing procedure with printk() and to
2657 * print any messages that were printed by the kernel before the
2658 * console driver was initialized.
2660 * This can happen pretty early during the boot process (because of
2661 * early_printk) - sometimes before setup_arch() completes - be careful
2662 * of what kernel features are used - they may not be initialised yet.
2664 * There are two types of consoles - bootconsoles (early_printk) and
2665 * "real" consoles (everything which is not a bootconsole) which are
2666 * handled differently.
2667 * - Any number of bootconsoles can be registered at any time.
2668 * - As soon as a "real" console is registered, all bootconsoles
2669 * will be unregistered automatically.
2670 * - Once a "real" console is registered, any attempt to register a
2671 * bootconsoles will be rejected
2673 void register_console(struct console
*newcon
)
2676 unsigned long flags
;
2677 struct console
*bcon
= NULL
;
2678 struct console_cmdline
*c
;
2679 static bool has_preferred
;
2681 for_each_console(bcon
) {
2682 if (WARN(bcon
== newcon
, "console '%s%d' already registered\n",
2683 bcon
->name
, bcon
->index
))
2688 * before we register a new CON_BOOT console, make sure we don't
2689 * already have a valid console
2691 if (newcon
->flags
& CON_BOOT
) {
2692 for_each_console(bcon
) {
2693 if (!(bcon
->flags
& CON_BOOT
)) {
2694 pr_info("Too late to register bootconsole %s%d\n",
2695 newcon
->name
, newcon
->index
);
2701 if (console_drivers
&& console_drivers
->flags
& CON_BOOT
)
2702 bcon
= console_drivers
;
2704 if (!has_preferred
|| bcon
|| !console_drivers
)
2705 has_preferred
= preferred_console
>= 0;
2708 * See if we want to use this console driver. If we
2709 * didn't select a console we take the first one
2710 * that registers here.
2712 if (!has_preferred
) {
2713 if (newcon
->index
< 0)
2715 if (newcon
->setup
== NULL
||
2716 newcon
->setup(newcon
, NULL
) == 0) {
2717 newcon
->flags
|= CON_ENABLED
;
2718 if (newcon
->device
) {
2719 newcon
->flags
|= CON_CONSDEV
;
2720 has_preferred
= true;
2726 * See if this console matches one we selected on
2729 for (i
= 0, c
= console_cmdline
;
2730 i
< MAX_CMDLINECONSOLES
&& c
->name
[0];
2732 if (!newcon
->match
||
2733 newcon
->match(newcon
, c
->name
, c
->index
, c
->options
) != 0) {
2734 /* default matching */
2735 BUILD_BUG_ON(sizeof(c
->name
) != sizeof(newcon
->name
));
2736 if (strcmp(c
->name
, newcon
->name
) != 0)
2738 if (newcon
->index
>= 0 &&
2739 newcon
->index
!= c
->index
)
2741 if (newcon
->index
< 0)
2742 newcon
->index
= c
->index
;
2744 if (_braille_register_console(newcon
, c
))
2747 if (newcon
->setup
&&
2748 newcon
->setup(newcon
, c
->options
) != 0)
2752 newcon
->flags
|= CON_ENABLED
;
2753 if (i
== preferred_console
) {
2754 newcon
->flags
|= CON_CONSDEV
;
2755 has_preferred
= true;
2760 if (!(newcon
->flags
& CON_ENABLED
))
2764 * If we have a bootconsole, and are switching to a real console,
2765 * don't print everything out again, since when the boot console, and
2766 * the real console are the same physical device, it's annoying to
2767 * see the beginning boot messages twice
2769 if (bcon
&& ((newcon
->flags
& (CON_CONSDEV
| CON_BOOT
)) == CON_CONSDEV
))
2770 newcon
->flags
&= ~CON_PRINTBUFFER
;
2773 * Put this console in the list - keep the
2774 * preferred driver at the head of the list.
2777 if ((newcon
->flags
& CON_CONSDEV
) || console_drivers
== NULL
) {
2778 newcon
->next
= console_drivers
;
2779 console_drivers
= newcon
;
2781 newcon
->next
->flags
&= ~CON_CONSDEV
;
2783 newcon
->next
= console_drivers
->next
;
2784 console_drivers
->next
= newcon
;
2787 if (newcon
->flags
& CON_EXTENDED
)
2788 nr_ext_console_drivers
++;
2790 if (newcon
->flags
& CON_PRINTBUFFER
) {
2792 * console_unlock(); will print out the buffered messages
2795 logbuf_lock_irqsave(flags
);
2797 * We're about to replay the log buffer. Only do this to the
2798 * just-registered console to avoid excessive message spam to
2799 * the already-registered consoles.
2801 * Set exclusive_console with disabled interrupts to reduce
2802 * race window with eventual console_flush_on_panic() that
2803 * ignores console_lock.
2805 exclusive_console
= newcon
;
2806 exclusive_console_stop_seq
= console_seq
;
2807 console_seq
= syslog_seq
;
2808 console_idx
= syslog_idx
;
2809 logbuf_unlock_irqrestore(flags
);
2812 console_sysfs_notify();
2815 * By unregistering the bootconsoles after we enable the real console
2816 * we get the "console xxx enabled" message on all the consoles -
2817 * boot consoles, real consoles, etc - this is to ensure that end
2818 * users know there might be something in the kernel's log buffer that
2819 * went to the bootconsole (that they do not see on the real console)
2821 pr_info("%sconsole [%s%d] enabled\n",
2822 (newcon
->flags
& CON_BOOT
) ? "boot" : "" ,
2823 newcon
->name
, newcon
->index
);
2825 ((newcon
->flags
& (CON_CONSDEV
| CON_BOOT
)) == CON_CONSDEV
) &&
2827 /* We need to iterate through all boot consoles, to make
2828 * sure we print everything out, before we unregister them.
2830 for_each_console(bcon
)
2831 if (bcon
->flags
& CON_BOOT
)
2832 unregister_console(bcon
);
2835 EXPORT_SYMBOL(register_console
);
2837 int unregister_console(struct console
*console
)
2839 struct console
*con
;
2842 pr_info("%sconsole [%s%d] disabled\n",
2843 (console
->flags
& CON_BOOT
) ? "boot" : "" ,
2844 console
->name
, console
->index
);
2846 res
= _braille_unregister_console(console
);
2854 if (console_drivers
== console
) {
2855 console_drivers
=console
->next
;
2858 for_each_console(con
) {
2859 if (con
->next
== console
) {
2860 con
->next
= console
->next
;
2868 goto out_disable_unlock
;
2870 if (console
->flags
& CON_EXTENDED
)
2871 nr_ext_console_drivers
--;
2874 * If this isn't the last console and it has CON_CONSDEV set, we
2875 * need to set it on the next preferred console.
2877 if (console_drivers
!= NULL
&& console
->flags
& CON_CONSDEV
)
2878 console_drivers
->flags
|= CON_CONSDEV
;
2880 console
->flags
&= ~CON_ENABLED
;
2882 console_sysfs_notify();
2885 res
= console
->exit(console
);
2890 console
->flags
&= ~CON_ENABLED
;
2895 EXPORT_SYMBOL(unregister_console
);
2898 * Initialize the console device. This is called *early*, so
2899 * we can't necessarily depend on lots of kernel help here.
2900 * Just do some early initializations, and do the complex setup
2903 void __init
console_init(void)
2907 initcall_entry_t
*ce
;
2909 /* Setup the default TTY line discipline. */
2913 * set up the console device so that later boot sequences can
2914 * inform about problems etc..
2916 ce
= __con_initcall_start
;
2917 trace_initcall_level("console");
2918 while (ce
< __con_initcall_end
) {
2919 call
= initcall_from_entry(ce
);
2920 trace_initcall_start(call
);
2922 trace_initcall_finish(call
, ret
);
2928 * Some boot consoles access data that is in the init section and which will
2929 * be discarded after the initcalls have been run. To make sure that no code
2930 * will access this data, unregister the boot consoles in a late initcall.
2932 * If for some reason, such as deferred probe or the driver being a loadable
2933 * module, the real console hasn't registered yet at this point, there will
2934 * be a brief interval in which no messages are logged to the console, which
2935 * makes it difficult to diagnose problems that occur during this time.
2937 * To mitigate this problem somewhat, only unregister consoles whose memory
2938 * intersects with the init section. Note that all other boot consoles will
2939 * get unregistred when the real preferred console is registered.
2941 static int __init
printk_late_init(void)
2943 struct console
*con
;
2946 for_each_console(con
) {
2947 if (!(con
->flags
& CON_BOOT
))
2950 /* Check addresses that might be used for enabled consoles. */
2951 if (init_section_intersects(con
, sizeof(*con
)) ||
2952 init_section_contains(con
->write
, 0) ||
2953 init_section_contains(con
->read
, 0) ||
2954 init_section_contains(con
->device
, 0) ||
2955 init_section_contains(con
->unblank
, 0) ||
2956 init_section_contains(con
->data
, 0)) {
2958 * Please, consider moving the reported consoles out
2959 * of the init section.
2961 pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n",
2962 con
->name
, con
->index
);
2963 unregister_console(con
);
2966 ret
= cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD
, "printk:dead", NULL
,
2967 console_cpu_notify
);
2969 ret
= cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN
, "printk:online",
2970 console_cpu_notify
, NULL
);
2974 late_initcall(printk_late_init
);
2976 #if defined CONFIG_PRINTK
2978 * Delayed printk version, for scheduler-internal messages:
2980 #define PRINTK_PENDING_WAKEUP 0x01
2981 #define PRINTK_PENDING_OUTPUT 0x02
2983 static DEFINE_PER_CPU(int, printk_pending
);
2985 static void wake_up_klogd_work_func(struct irq_work
*irq_work
)
2987 int pending
= __this_cpu_xchg(printk_pending
, 0);
2989 if (pending
& PRINTK_PENDING_OUTPUT
) {
2990 /* If trylock fails, someone else is doing the printing */
2991 if (console_trylock())
2995 if (pending
& PRINTK_PENDING_WAKEUP
)
2996 wake_up_interruptible(&log_wait
);
2999 static DEFINE_PER_CPU(struct irq_work
, wake_up_klogd_work
) = {
3000 .func
= wake_up_klogd_work_func
,
3001 .flags
= ATOMIC_INIT(IRQ_WORK_LAZY
),
3004 void wake_up_klogd(void)
3006 if (!printk_percpu_data_ready())
3010 if (waitqueue_active(&log_wait
)) {
3011 this_cpu_or(printk_pending
, PRINTK_PENDING_WAKEUP
);
3012 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work
));
3017 void defer_console_output(void)
3019 if (!printk_percpu_data_ready())
3023 __this_cpu_or(printk_pending
, PRINTK_PENDING_OUTPUT
);
3024 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work
));
3028 int vprintk_deferred(const char *fmt
, va_list args
)
3032 r
= vprintk_emit(0, LOGLEVEL_SCHED
, NULL
, 0, fmt
, args
);
3033 defer_console_output();
3038 int printk_deferred(const char *fmt
, ...)
3043 va_start(args
, fmt
);
3044 r
= vprintk_deferred(fmt
, args
);
3051 * printk rate limiting, lifted from the networking subsystem.
3053 * This enforces a rate limit: not more than 10 kernel messages
3054 * every 5s to make a denial-of-service attack impossible.
3056 DEFINE_RATELIMIT_STATE(printk_ratelimit_state
, 5 * HZ
, 10);
3058 int __printk_ratelimit(const char *func
)
3060 return ___ratelimit(&printk_ratelimit_state
, func
);
3062 EXPORT_SYMBOL(__printk_ratelimit
);
3065 * printk_timed_ratelimit - caller-controlled printk ratelimiting
3066 * @caller_jiffies: pointer to caller's state
3067 * @interval_msecs: minimum interval between prints
3069 * printk_timed_ratelimit() returns true if more than @interval_msecs
3070 * milliseconds have elapsed since the last time printk_timed_ratelimit()
3073 bool printk_timed_ratelimit(unsigned long *caller_jiffies
,
3074 unsigned int interval_msecs
)
3076 unsigned long elapsed
= jiffies
- *caller_jiffies
;
3078 if (*caller_jiffies
&& elapsed
<= msecs_to_jiffies(interval_msecs
))
3081 *caller_jiffies
= jiffies
;
3084 EXPORT_SYMBOL(printk_timed_ratelimit
);
3086 static DEFINE_SPINLOCK(dump_list_lock
);
3087 static LIST_HEAD(dump_list
);
3090 * kmsg_dump_register - register a kernel log dumper.
3091 * @dumper: pointer to the kmsg_dumper structure
3093 * Adds a kernel log dumper to the system. The dump callback in the
3094 * structure will be called when the kernel oopses or panics and must be
3095 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
3097 int kmsg_dump_register(struct kmsg_dumper
*dumper
)
3099 unsigned long flags
;
3102 /* The dump callback needs to be set */
3106 spin_lock_irqsave(&dump_list_lock
, flags
);
3107 /* Don't allow registering multiple times */
3108 if (!dumper
->registered
) {
3109 dumper
->registered
= 1;
3110 list_add_tail_rcu(&dumper
->list
, &dump_list
);
3113 spin_unlock_irqrestore(&dump_list_lock
, flags
);
3117 EXPORT_SYMBOL_GPL(kmsg_dump_register
);
3120 * kmsg_dump_unregister - unregister a kmsg dumper.
3121 * @dumper: pointer to the kmsg_dumper structure
3123 * Removes a dump device from the system. Returns zero on success and
3124 * %-EINVAL otherwise.
3126 int kmsg_dump_unregister(struct kmsg_dumper
*dumper
)
3128 unsigned long flags
;
3131 spin_lock_irqsave(&dump_list_lock
, flags
);
3132 if (dumper
->registered
) {
3133 dumper
->registered
= 0;
3134 list_del_rcu(&dumper
->list
);
3137 spin_unlock_irqrestore(&dump_list_lock
, flags
);
3142 EXPORT_SYMBOL_GPL(kmsg_dump_unregister
);
3144 static bool always_kmsg_dump
;
3145 module_param_named(always_kmsg_dump
, always_kmsg_dump
, bool, S_IRUGO
| S_IWUSR
);
3148 * kmsg_dump - dump kernel log to kernel message dumpers.
3149 * @reason: the reason (oops, panic etc) for dumping
3151 * Call each of the registered dumper's dump() callback, which can
3152 * retrieve the kmsg records with kmsg_dump_get_line() or
3153 * kmsg_dump_get_buffer().
3155 void kmsg_dump(enum kmsg_dump_reason reason
)
3157 struct kmsg_dumper
*dumper
;
3158 unsigned long flags
;
3160 if ((reason
> KMSG_DUMP_OOPS
) && !always_kmsg_dump
)
3164 list_for_each_entry_rcu(dumper
, &dump_list
, list
) {
3165 if (dumper
->max_reason
&& reason
> dumper
->max_reason
)
3168 /* initialize iterator with data about the stored records */
3169 dumper
->active
= true;
3171 logbuf_lock_irqsave(flags
);
3172 dumper
->cur_seq
= clear_seq
;
3173 dumper
->cur_idx
= clear_idx
;
3174 dumper
->next_seq
= log_next_seq
;
3175 dumper
->next_idx
= log_next_idx
;
3176 logbuf_unlock_irqrestore(flags
);
3178 /* invoke dumper which will iterate over records */
3179 dumper
->dump(dumper
, reason
);
3181 /* reset iterator */
3182 dumper
->active
= false;
3188 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3189 * @dumper: registered kmsg dumper
3190 * @syslog: include the "<4>" prefixes
3191 * @line: buffer to copy the line to
3192 * @size: maximum size of the buffer
3193 * @len: length of line placed into buffer
3195 * Start at the beginning of the kmsg buffer, with the oldest kmsg
3196 * record, and copy one record into the provided buffer.
3198 * Consecutive calls will return the next available record moving
3199 * towards the end of the buffer with the youngest messages.
3201 * A return value of FALSE indicates that there are no more records to
3204 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
3206 bool kmsg_dump_get_line_nolock(struct kmsg_dumper
*dumper
, bool syslog
,
3207 char *line
, size_t size
, size_t *len
)
3209 struct printk_log
*msg
;
3213 if (!dumper
->active
)
3216 if (dumper
->cur_seq
< log_first_seq
) {
3217 /* messages are gone, move to first available one */
3218 dumper
->cur_seq
= log_first_seq
;
3219 dumper
->cur_idx
= log_first_idx
;
3223 if (dumper
->cur_seq
>= log_next_seq
)
3226 msg
= log_from_idx(dumper
->cur_idx
);
3227 l
= msg_print_text(msg
, syslog
, printk_time
, line
, size
);
3229 dumper
->cur_idx
= log_next(dumper
->cur_idx
);
3239 * kmsg_dump_get_line - retrieve one kmsg log line
3240 * @dumper: registered kmsg dumper
3241 * @syslog: include the "<4>" prefixes
3242 * @line: buffer to copy the line to
3243 * @size: maximum size of the buffer
3244 * @len: length of line placed into buffer
3246 * Start at the beginning of the kmsg buffer, with the oldest kmsg
3247 * record, and copy one record into the provided buffer.
3249 * Consecutive calls will return the next available record moving
3250 * towards the end of the buffer with the youngest messages.
3252 * A return value of FALSE indicates that there are no more records to
3255 bool kmsg_dump_get_line(struct kmsg_dumper
*dumper
, bool syslog
,
3256 char *line
, size_t size
, size_t *len
)
3258 unsigned long flags
;
3261 logbuf_lock_irqsave(flags
);
3262 ret
= kmsg_dump_get_line_nolock(dumper
, syslog
, line
, size
, len
);
3263 logbuf_unlock_irqrestore(flags
);
3267 EXPORT_SYMBOL_GPL(kmsg_dump_get_line
);
3270 * kmsg_dump_get_buffer - copy kmsg log lines
3271 * @dumper: registered kmsg dumper
3272 * @syslog: include the "<4>" prefixes
3273 * @buf: buffer to copy the line to
3274 * @size: maximum size of the buffer
3275 * @len: length of line placed into buffer
3277 * Start at the end of the kmsg buffer and fill the provided buffer
3278 * with as many of the the *youngest* kmsg records that fit into it.
3279 * If the buffer is large enough, all available kmsg records will be
3280 * copied with a single call.
3282 * Consecutive calls will fill the buffer with the next block of
3283 * available older records, not including the earlier retrieved ones.
3285 * A return value of FALSE indicates that there are no more records to
3288 bool kmsg_dump_get_buffer(struct kmsg_dumper
*dumper
, bool syslog
,
3289 char *buf
, size_t size
, size_t *len
)
3291 unsigned long flags
;
3298 bool time
= printk_time
;
3300 if (!dumper
->active
)
3303 logbuf_lock_irqsave(flags
);
3304 if (dumper
->cur_seq
< log_first_seq
) {
3305 /* messages are gone, move to first available one */
3306 dumper
->cur_seq
= log_first_seq
;
3307 dumper
->cur_idx
= log_first_idx
;
3311 if (dumper
->cur_seq
>= dumper
->next_seq
) {
3312 logbuf_unlock_irqrestore(flags
);
3316 /* calculate length of entire buffer */
3317 seq
= dumper
->cur_seq
;
3318 idx
= dumper
->cur_idx
;
3319 while (seq
< dumper
->next_seq
) {
3320 struct printk_log
*msg
= log_from_idx(idx
);
3322 l
+= msg_print_text(msg
, true, time
, NULL
, 0);
3323 idx
= log_next(idx
);
3327 /* move first record forward until length fits into the buffer */
3328 seq
= dumper
->cur_seq
;
3329 idx
= dumper
->cur_idx
;
3330 while (l
>= size
&& seq
< dumper
->next_seq
) {
3331 struct printk_log
*msg
= log_from_idx(idx
);
3333 l
-= msg_print_text(msg
, true, time
, NULL
, 0);
3334 idx
= log_next(idx
);
3338 /* last message in next interation */
3343 while (seq
< dumper
->next_seq
) {
3344 struct printk_log
*msg
= log_from_idx(idx
);
3346 l
+= msg_print_text(msg
, syslog
, time
, buf
+ l
, size
- l
);
3347 idx
= log_next(idx
);
3351 dumper
->next_seq
= next_seq
;
3352 dumper
->next_idx
= next_idx
;
3354 logbuf_unlock_irqrestore(flags
);
3360 EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer
);
3363 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3364 * @dumper: registered kmsg dumper
3366 * Reset the dumper's iterator so that kmsg_dump_get_line() and
3367 * kmsg_dump_get_buffer() can be called again and used multiple
3368 * times within the same dumper.dump() callback.
3370 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3372 void kmsg_dump_rewind_nolock(struct kmsg_dumper
*dumper
)
3374 dumper
->cur_seq
= clear_seq
;
3375 dumper
->cur_idx
= clear_idx
;
3376 dumper
->next_seq
= log_next_seq
;
3377 dumper
->next_idx
= log_next_idx
;
3381 * kmsg_dump_rewind - reset the interator
3382 * @dumper: registered kmsg dumper
3384 * Reset the dumper's iterator so that kmsg_dump_get_line() and
3385 * kmsg_dump_get_buffer() can be called again and used multiple
3386 * times within the same dumper.dump() callback.
3388 void kmsg_dump_rewind(struct kmsg_dumper
*dumper
)
3390 unsigned long flags
;
3392 logbuf_lock_irqsave(flags
);
3393 kmsg_dump_rewind_nolock(dumper
);
3394 logbuf_unlock_irqrestore(flags
);
3396 EXPORT_SYMBOL_GPL(kmsg_dump_rewind
);