Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0-only
2 : /*
3 : * linux/kernel/printk.c
4 : *
5 : * Copyright (C) 1991, 1992 Linus Torvalds
6 : *
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).
12 : * Ted Ts'o, 2/11/93.
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
18 : */
19 :
20 : #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 :
22 : #include <linux/kernel.h>
23 : #include <linux/mm.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/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>
50 :
51 : #include <linux/uaccess.h>
52 : #include <asm/sections.h>
53 :
54 : #include <trace/events/initcall.h>
55 : #define CREATE_TRACE_POINTS
56 : #include <trace/events/printk.h>
57 :
58 : #include "printk_ringbuffer.h"
59 : #include "console_cmdline.h"
60 : #include "braille.h"
61 : #include "internal.h"
62 :
63 : int console_printk[4] = {
64 : CONSOLE_LOGLEVEL_DEFAULT, /* console_loglevel */
65 : MESSAGE_LOGLEVEL_DEFAULT, /* default_message_loglevel */
66 : CONSOLE_LOGLEVEL_MIN, /* minimum_console_loglevel */
67 : CONSOLE_LOGLEVEL_DEFAULT, /* default_console_loglevel */
68 : };
69 : EXPORT_SYMBOL_GPL(console_printk);
70 :
71 : atomic_t ignore_console_lock_warning __read_mostly = ATOMIC_INIT(0);
72 : EXPORT_SYMBOL(ignore_console_lock_warning);
73 :
74 : /*
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.
77 : */
78 : int oops_in_progress;
79 : EXPORT_SYMBOL(oops_in_progress);
80 :
81 : /*
82 : * console_sem protects the console_drivers list, and also
83 : * provides serialisation for access to the entire console
84 : * driver system.
85 : */
86 : static DEFINE_SEMAPHORE(console_sem);
87 : struct console *console_drivers;
88 : EXPORT_SYMBOL_GPL(console_drivers);
89 :
90 : /*
91 : * System may need to suppress printk message under certain
92 : * circumstances, like after kernel panic happens.
93 : */
94 : int __read_mostly suppress_printk;
95 :
96 : #ifdef CONFIG_LOCKDEP
97 : static struct lockdep_map console_lock_dep_map = {
98 : .name = "console_lock"
99 : };
100 : #endif
101 :
102 : enum devkmsg_log_bits {
103 : __DEVKMSG_LOG_BIT_ON = 0,
104 : __DEVKMSG_LOG_BIT_OFF,
105 : __DEVKMSG_LOG_BIT_LOCK,
106 : };
107 :
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),
112 : };
113 :
114 : /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
115 : #define DEVKMSG_LOG_MASK_DEFAULT 0
116 :
117 : static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
118 :
119 0 : static int __control_devkmsg(char *str)
120 : {
121 0 : size_t len;
122 :
123 0 : if (!str)
124 : return -EINVAL;
125 :
126 0 : len = str_has_prefix(str, "on");
127 0 : if (len) {
128 0 : devkmsg_log = DEVKMSG_LOG_MASK_ON;
129 0 : return len;
130 : }
131 :
132 0 : len = str_has_prefix(str, "off");
133 0 : if (len) {
134 0 : devkmsg_log = DEVKMSG_LOG_MASK_OFF;
135 0 : return len;
136 : }
137 :
138 0 : len = str_has_prefix(str, "ratelimit");
139 0 : if (len) {
140 0 : devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
141 0 : return len;
142 : }
143 :
144 : return -EINVAL;
145 : }
146 :
147 0 : static int __init control_devkmsg(char *str)
148 : {
149 0 : if (__control_devkmsg(str) < 0)
150 : return 1;
151 :
152 : /*
153 : * Set sysctl string accordingly:
154 : */
155 0 : if (devkmsg_log == DEVKMSG_LOG_MASK_ON)
156 0 : strcpy(devkmsg_log_str, "on");
157 0 : else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF)
158 0 : strcpy(devkmsg_log_str, "off");
159 : /* else "ratelimit" which is set by default. */
160 :
161 : /*
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.
166 : */
167 0 : devkmsg_log |= DEVKMSG_LOG_MASK_LOCK;
168 :
169 0 : return 0;
170 : }
171 : __setup("printk.devkmsg=", control_devkmsg);
172 :
173 : char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit";
174 :
175 0 : int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
176 : void *buffer, size_t *lenp, loff_t *ppos)
177 : {
178 0 : char old_str[DEVKMSG_STR_MAX_SIZE];
179 0 : unsigned int old;
180 0 : int err;
181 :
182 0 : if (write) {
183 0 : if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK)
184 : return -EINVAL;
185 :
186 0 : old = devkmsg_log;
187 0 : strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE);
188 : }
189 :
190 0 : err = proc_dostring(table, write, buffer, lenp, ppos);
191 0 : if (err)
192 : return err;
193 :
194 0 : if (write) {
195 0 : err = __control_devkmsg(devkmsg_log_str);
196 :
197 : /*
198 : * Do not accept an unknown string OR a known string with
199 : * trailing crap...
200 : */
201 0 : if (err < 0 || (err + 1 != *lenp)) {
202 :
203 : /* ... and restore old setting. */
204 0 : devkmsg_log = old;
205 0 : strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE);
206 :
207 0 : return -EINVAL;
208 : }
209 : }
210 :
211 : return 0;
212 : }
213 :
214 : /* Number of registered extended console drivers. */
215 : static int nr_ext_console_drivers;
216 :
217 : /*
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.
220 : */
221 : #define down_console_sem() do { \
222 : down(&console_sem);\
223 : mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
224 : } while (0)
225 :
226 307 : static int __down_trylock_console_sem(unsigned long ip)
227 : {
228 307 : int lock_failed;
229 307 : unsigned long flags;
230 :
231 : /*
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.
235 : */
236 614 : printk_safe_enter_irqsave(flags);
237 307 : lock_failed = down_trylock(&console_sem);
238 307 : printk_safe_exit_irqrestore(flags);
239 :
240 307 : if (lock_failed)
241 : return 1;
242 306 : mutex_acquire(&console_lock_dep_map, 0, 1, ip);
243 306 : return 0;
244 : }
245 : #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
246 :
247 522 : static void __up_console_sem(unsigned long ip)
248 : {
249 522 : unsigned long flags;
250 :
251 522 : mutex_release(&console_lock_dep_map, ip);
252 :
253 1044 : printk_safe_enter_irqsave(flags);
254 522 : up(&console_sem);
255 522 : printk_safe_exit_irqrestore(flags);
256 522 : }
257 : #define up_console_sem() __up_console_sem(_RET_IP_)
258 :
259 : /*
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).
266 : */
267 : static int console_locked, console_suspended;
268 :
269 : /*
270 : * If exclusive_console is non-NULL then only this console is to be printed to.
271 : */
272 : static struct console *exclusive_console;
273 :
274 : /*
275 : * Array of consoles built from command line options (console=)
276 : */
277 :
278 : #define MAX_CMDLINECONSOLES 8
279 :
280 : static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
281 :
282 : static int preferred_console = -1;
283 : static bool has_preferred_console;
284 : int console_set_on_cmdline;
285 : EXPORT_SYMBOL(console_set_on_cmdline);
286 :
287 : /* Flag: console code may call schedule() */
288 : static int console_may_schedule;
289 :
290 : enum con_msg_format_flags {
291 : MSG_FORMAT_DEFAULT = 0,
292 : MSG_FORMAT_SYSLOG = (1 << 0),
293 : };
294 :
295 : static int console_msg_format = MSG_FORMAT_DEFAULT;
296 :
297 : /*
298 : * The printk log buffer consists of a sequenced collection of records, each
299 : * containing variable length message text. Every record also contains its
300 : * own meta-data (@info).
301 : *
302 : * Every record meta-data carries the timestamp in microseconds, as well as
303 : * the standard userspace syslog level and syslog facility. The usual kernel
304 : * messages use LOG_KERN; userspace-injected messages always carry a matching
305 : * syslog facility, by default LOG_USER. The origin of every message can be
306 : * reliably determined that way.
307 : *
308 : * The human readable log message of a record is available in @text, the
309 : * length of the message text in @text_len. The stored message is not
310 : * terminated.
311 : *
312 : * Optionally, a record can carry a dictionary of properties (key/value
313 : * pairs), to provide userspace with a machine-readable message context.
314 : *
315 : * Examples for well-defined, commonly used property names are:
316 : * DEVICE=b12:8 device identifier
317 : * b12:8 block dev_t
318 : * c127:3 char dev_t
319 : * n8 netdev ifindex
320 : * +sound:card0 subsystem:devname
321 : * SUBSYSTEM=pci driver-core subsystem name
322 : *
323 : * Valid characters in property names are [a-zA-Z0-9.-_]. Property names
324 : * and values are terminated by a '\0' character.
325 : *
326 : * Example of record values:
327 : * record.text_buf = "it's a line" (unterminated)
328 : * record.info.seq = 56
329 : * record.info.ts_nsec = 36863
330 : * record.info.text_len = 11
331 : * record.info.facility = 0 (LOG_KERN)
332 : * record.info.flags = 0
333 : * record.info.level = 3 (LOG_ERR)
334 : * record.info.caller_id = 299 (task 299)
335 : * record.info.dev_info.subsystem = "pci" (terminated)
336 : * record.info.dev_info.device = "+pci:0000:00:01.0" (terminated)
337 : *
338 : * The 'struct printk_info' buffer must never be directly exported to
339 : * userspace, it is a kernel-private implementation detail that might
340 : * need to be changed in the future, when the requirements change.
341 : *
342 : * /dev/kmsg exports the structured data in the following line format:
343 : * "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
344 : *
345 : * Users of the export format should ignore possible additional values
346 : * separated by ',', and find the message after the ';' character.
347 : *
348 : * The optional key/value pairs are attached as continuation lines starting
349 : * with a space character and terminated by a newline. All possible
350 : * non-prinatable characters are escaped in the "\xff" notation.
351 : */
352 :
353 : enum log_flags {
354 : LOG_NEWLINE = 2, /* text ended with a newline */
355 : LOG_CONT = 8, /* text is a fragment of a continuation line */
356 : };
357 :
358 : /*
359 : * The logbuf_lock protects kmsg buffer, indices, counters. This can be taken
360 : * within the scheduler's rq lock. It must be released before calling
361 : * console_unlock() or anything else that might wake up a process.
362 : */
363 : DEFINE_RAW_SPINLOCK(logbuf_lock);
364 :
365 : /*
366 : * Helper macros to lock/unlock logbuf_lock and switch between
367 : * printk-safe/unsafe modes.
368 : */
369 : #define logbuf_lock_irq() \
370 : do { \
371 : printk_safe_enter_irq(); \
372 : raw_spin_lock(&logbuf_lock); \
373 : } while (0)
374 :
375 : #define logbuf_unlock_irq() \
376 : do { \
377 : raw_spin_unlock(&logbuf_lock); \
378 : printk_safe_exit_irq(); \
379 : } while (0)
380 :
381 : #define logbuf_lock_irqsave(flags) \
382 : do { \
383 : printk_safe_enter_irqsave(flags); \
384 : raw_spin_lock(&logbuf_lock); \
385 : } while (0)
386 :
387 : #define logbuf_unlock_irqrestore(flags) \
388 : do { \
389 : raw_spin_unlock(&logbuf_lock); \
390 : printk_safe_exit_irqrestore(flags); \
391 : } while (0)
392 :
393 : #ifdef CONFIG_PRINTK
394 : DECLARE_WAIT_QUEUE_HEAD(log_wait);
395 : /* the next printk record to read by syslog(READ) or /proc/kmsg */
396 : static u64 syslog_seq;
397 : static size_t syslog_partial;
398 : static bool syslog_time;
399 :
400 : /* the next printk record to write to the console */
401 : static u64 console_seq;
402 : static u64 exclusive_console_stop_seq;
403 : static unsigned long console_dropped;
404 :
405 : /* the next printk record to read after the last 'clear' command */
406 : static u64 clear_seq;
407 :
408 : #ifdef CONFIG_PRINTK_CALLER
409 : #define PREFIX_MAX 48
410 : #else
411 : #define PREFIX_MAX 32
412 : #endif
413 : #define LOG_LINE_MAX (1024 - PREFIX_MAX)
414 :
415 : #define LOG_LEVEL(v) ((v) & 0x07)
416 : #define LOG_FACILITY(v) ((v) >> 3 & 0xff)
417 :
418 : /* record buffer */
419 : #define LOG_ALIGN __alignof__(unsigned long)
420 : #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
421 : #define LOG_BUF_LEN_MAX (u32)(1 << 31)
422 : static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
423 : static char *log_buf = __log_buf;
424 : static u32 log_buf_len = __LOG_BUF_LEN;
425 :
426 : /*
427 : * Define the average message size. This only affects the number of
428 : * descriptors that will be available. Underestimating is better than
429 : * overestimating (too many available descriptors is better than not enough).
430 : */
431 : #define PRB_AVGBITS 5 /* 32 character average length */
432 :
433 : #if CONFIG_LOG_BUF_SHIFT <= PRB_AVGBITS
434 : #error CONFIG_LOG_BUF_SHIFT value too small.
435 : #endif
436 : _DEFINE_PRINTKRB(printk_rb_static, CONFIG_LOG_BUF_SHIFT - PRB_AVGBITS,
437 : PRB_AVGBITS, &__log_buf[0]);
438 :
439 : static struct printk_ringbuffer printk_rb_dynamic;
440 :
441 : static struct printk_ringbuffer *prb = &printk_rb_static;
442 :
443 : /*
444 : * We cannot access per-CPU data (e.g. per-CPU flush irq_work) before
445 : * per_cpu_areas are initialised. This variable is set to true when
446 : * it's safe to access per-CPU data.
447 : */
448 : static bool __printk_percpu_data_ready __read_mostly;
449 :
450 307 : bool printk_percpu_data_ready(void)
451 : {
452 307 : return __printk_percpu_data_ready;
453 : }
454 :
455 : /* Return log buffer address */
456 0 : char *log_buf_addr_get(void)
457 : {
458 0 : return log_buf;
459 : }
460 :
461 : /* Return log buffer size */
462 0 : u32 log_buf_len_get(void)
463 : {
464 0 : return log_buf_len;
465 : }
466 :
467 : /*
468 : * Define how much of the log buffer we could take at maximum. The value
469 : * must be greater than two. Note that only half of the buffer is available
470 : * when the index points to the middle.
471 : */
472 : #define MAX_LOG_TAKE_PART 4
473 : static const char trunc_msg[] = "<truncated>";
474 :
475 0 : static void truncate_msg(u16 *text_len, u16 *trunc_msg_len)
476 : {
477 : /*
478 : * The message should not take the whole buffer. Otherwise, it might
479 : * get removed too soon.
480 : */
481 0 : u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART;
482 :
483 0 : if (*text_len > max_text_len)
484 0 : *text_len = max_text_len;
485 :
486 : /* enable the warning message (if there is room) */
487 0 : *trunc_msg_len = strlen(trunc_msg);
488 0 : if (*text_len >= *trunc_msg_len)
489 0 : *text_len -= *trunc_msg_len;
490 : else
491 0 : *trunc_msg_len = 0;
492 0 : }
493 :
494 : int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT);
495 :
496 2 : static int syslog_action_restricted(int type)
497 : {
498 2 : if (dmesg_restrict)
499 : return 1;
500 : /*
501 : * Unless restricted, we allow "read all" and "get buffer size"
502 : * for everybody.
503 : */
504 2 : return type != SYSLOG_ACTION_READ_ALL &&
505 2 : type != SYSLOG_ACTION_SIZE_BUFFER;
506 : }
507 :
508 7 : static int check_syslog_permissions(int type, int source)
509 : {
510 : /*
511 : * If this is from /proc/kmsg and we've already opened it, then we've
512 : * already done the capabilities checks at open time.
513 : */
514 7 : if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN)
515 5 : goto ok;
516 :
517 2 : if (syslog_action_restricted(type)) {
518 1 : if (capable(CAP_SYSLOG))
519 1 : goto ok;
520 : /*
521 : * For historical reasons, accept CAP_SYS_ADMIN too, with
522 : * a warning.
523 : */
524 0 : if (capable(CAP_SYS_ADMIN)) {
525 0 : pr_warn_once("%s (%d): Attempt to access syslog with "
526 : "CAP_SYS_ADMIN but no CAP_SYSLOG "
527 : "(deprecated).\n",
528 : current->comm, task_pid_nr(current));
529 0 : goto ok;
530 : }
531 : return -EPERM;
532 : }
533 1 : ok:
534 7 : return security_syslog(type);
535 : }
536 :
537 13971 : static void append_char(char **pp, char *e, char c)
538 : {
539 13971 : if (*pp < e)
540 13971 : *(*pp)++ = c;
541 : }
542 :
543 269 : static ssize_t info_print_ext_header(char *buf, size_t size,
544 : struct printk_info *info)
545 : {
546 269 : u64 ts_usec = info->ts_nsec;
547 269 : char caller[20];
548 : #ifdef CONFIG_PRINTK_CALLER
549 : u32 id = info->caller_id;
550 :
551 : snprintf(caller, sizeof(caller), ",caller=%c%u",
552 : id & 0x80000000 ? 'C' : 'T', id & ~0x80000000);
553 : #else
554 269 : caller[0] = '\0';
555 : #endif
556 :
557 269 : do_div(ts_usec, 1000);
558 :
559 807 : return scnprintf(buf, size, "%u,%llu,%llu,%c%s;",
560 269 : (info->facility << 3) | info->level, info->seq,
561 269 : ts_usec, info->flags & LOG_CONT ? 'c' : '-', caller);
562 : }
563 :
564 287 : static ssize_t msg_add_ext_text(char *buf, size_t size,
565 : const char *text, size_t text_len,
566 : unsigned char endc)
567 : {
568 287 : char *p = buf, *e = buf + size;
569 287 : size_t i;
570 :
571 : /* escape non-printable characters */
572 13974 : for (i = 0; i < text_len; i++) {
573 13687 : unsigned char c = text[i];
574 :
575 13687 : if (c < ' ' || c >= 127 || c == '\\')
576 3 : p += scnprintf(p, e - p, "\\x%02x", c);
577 : else
578 27371 : append_char(&p, e, c);
579 : }
580 287 : append_char(&p, e, endc);
581 :
582 287 : return p - buf;
583 : }
584 :
585 538 : static ssize_t msg_add_dict_text(char *buf, size_t size,
586 : const char *key, const char *val)
587 : {
588 538 : size_t val_len = strlen(val);
589 538 : ssize_t len;
590 :
591 538 : if (!val_len)
592 : return 0;
593 :
594 6 : len = msg_add_ext_text(buf, size, "", 0, ' '); /* dict prefix */
595 6 : len += msg_add_ext_text(buf + len, size - len, key, strlen(key), '=');
596 6 : len += msg_add_ext_text(buf + len, size - len, val, val_len, '\n');
597 :
598 6 : return len;
599 : }
600 :
601 269 : static ssize_t msg_print_ext_body(char *buf, size_t size,
602 : char *text, size_t text_len,
603 : struct dev_printk_info *dev_info)
604 : {
605 269 : ssize_t len;
606 :
607 269 : len = msg_add_ext_text(buf, size, text, text_len, '\n');
608 :
609 269 : if (!dev_info)
610 0 : goto out;
611 :
612 538 : len += msg_add_dict_text(buf + len, size - len, "SUBSYSTEM",
613 269 : dev_info->subsystem);
614 269 : len += msg_add_dict_text(buf + len, size - len, "DEVICE",
615 269 : dev_info->device);
616 269 : out:
617 269 : return len;
618 : }
619 :
620 : /* /dev/kmsg - userspace message inject/listen interface */
621 : struct devkmsg_user {
622 : u64 seq;
623 : struct ratelimit_state rs;
624 : struct mutex lock;
625 : char buf[CONSOLE_EXT_LOG_MAX];
626 :
627 : struct printk_info info;
628 : char text_buf[CONSOLE_EXT_LOG_MAX];
629 : struct printk_record record;
630 : };
631 :
632 : static __printf(3, 4) __cold
633 13 : int devkmsg_emit(int facility, int level, const char *fmt, ...)
634 : {
635 13 : va_list args;
636 13 : int r;
637 :
638 13 : va_start(args, fmt);
639 13 : r = vprintk_emit(facility, level, NULL, fmt, args);
640 13 : va_end(args);
641 :
642 13 : return r;
643 : }
644 :
645 50 : static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from)
646 : {
647 50 : char *buf, *line;
648 50 : int level = default_message_loglevel;
649 50 : int facility = 1; /* LOG_USER */
650 50 : struct file *file = iocb->ki_filp;
651 50 : struct devkmsg_user *user = file->private_data;
652 50 : size_t len = iov_iter_count(from);
653 50 : ssize_t ret = len;
654 :
655 50 : if (!user || len > LOG_LINE_MAX)
656 : return -EINVAL;
657 :
658 : /* Ignore when user logging is disabled. */
659 50 : if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
660 : return len;
661 :
662 : /* Ratelimit when not explicitly enabled. */
663 50 : if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) {
664 50 : if (!___ratelimit(&user->rs, current->comm))
665 : return ret;
666 : }
667 :
668 13 : buf = kmalloc(len+1, GFP_KERNEL);
669 13 : if (buf == NULL)
670 : return -ENOMEM;
671 :
672 13 : buf[len] = '\0';
673 26 : if (!copy_from_iter_full(buf, len, from)) {
674 0 : kfree(buf);
675 0 : return -EFAULT;
676 : }
677 :
678 : /*
679 : * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
680 : * the decimal value represents 32bit, the lower 3 bit are the log
681 : * level, the rest are the log facility.
682 : *
683 : * If no prefix or no userspace facility is specified, we
684 : * enforce LOG_USER, to be able to reliably distinguish
685 : * kernel-generated messages from userspace-injected ones.
686 : */
687 13 : line = buf;
688 13 : if (line[0] == '<') {
689 12 : char *endp = NULL;
690 12 : unsigned int u;
691 :
692 12 : u = simple_strtoul(line + 1, &endp, 10);
693 12 : if (endp && endp[0] == '>') {
694 12 : level = LOG_LEVEL(u);
695 12 : if (LOG_FACILITY(u) != 0)
696 12 : facility = LOG_FACILITY(u);
697 12 : endp++;
698 12 : line = endp;
699 : }
700 : }
701 :
702 13 : devkmsg_emit(facility, level, "%s", line);
703 13 : kfree(buf);
704 13 : return ret;
705 : }
706 :
707 270 : static ssize_t devkmsg_read(struct file *file, char __user *buf,
708 : size_t count, loff_t *ppos)
709 : {
710 270 : struct devkmsg_user *user = file->private_data;
711 270 : struct printk_record *r = &user->record;
712 270 : size_t len;
713 270 : ssize_t ret;
714 :
715 270 : if (!user)
716 : return -EBADF;
717 :
718 270 : ret = mutex_lock_interruptible(&user->lock);
719 270 : if (ret)
720 : return ret;
721 :
722 270 : logbuf_lock_irq();
723 270 : if (!prb_read_valid(prb, user->seq, r)) {
724 1 : if (file->f_flags & O_NONBLOCK) {
725 1 : ret = -EAGAIN;
726 1 : logbuf_unlock_irq();
727 1 : goto out;
728 : }
729 :
730 0 : logbuf_unlock_irq();
731 0 : ret = wait_event_interruptible(log_wait,
732 : prb_read_valid(prb, user->seq, r));
733 0 : if (ret)
734 0 : goto out;
735 0 : logbuf_lock_irq();
736 : }
737 :
738 269 : if (r->info->seq != user->seq) {
739 : /* our last seen message is gone, return error and reset */
740 0 : user->seq = r->info->seq;
741 0 : ret = -EPIPE;
742 0 : logbuf_unlock_irq();
743 0 : goto out;
744 : }
745 :
746 269 : len = info_print_ext_header(user->buf, sizeof(user->buf), r->info);
747 538 : len += msg_print_ext_body(user->buf + len, sizeof(user->buf) - len,
748 269 : &r->text_buf[0], r->info->text_len,
749 269 : &r->info->dev_info);
750 :
751 269 : user->seq = r->info->seq + 1;
752 269 : logbuf_unlock_irq();
753 :
754 269 : if (len > count) {
755 0 : ret = -EINVAL;
756 0 : goto out;
757 : }
758 :
759 538 : if (copy_to_user(buf, user->buf, len)) {
760 0 : ret = -EFAULT;
761 0 : goto out;
762 : }
763 269 : ret = len;
764 270 : out:
765 270 : mutex_unlock(&user->lock);
766 270 : return ret;
767 : }
768 :
769 : /*
770 : * Be careful when modifying this function!!!
771 : *
772 : * Only few operations are supported because the device works only with the
773 : * entire variable length messages (records). Non-standard values are
774 : * returned in the other cases and has been this way for quite some time.
775 : * User space applications might depend on this behavior.
776 : */
777 0 : static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
778 : {
779 0 : struct devkmsg_user *user = file->private_data;
780 0 : loff_t ret = 0;
781 :
782 0 : if (!user)
783 : return -EBADF;
784 0 : if (offset)
785 : return -ESPIPE;
786 :
787 0 : logbuf_lock_irq();
788 0 : switch (whence) {
789 0 : case SEEK_SET:
790 : /* the first record */
791 0 : user->seq = prb_first_valid_seq(prb);
792 0 : break;
793 0 : case SEEK_DATA:
794 : /*
795 : * The first record after the last SYSLOG_ACTION_CLEAR,
796 : * like issued by 'dmesg -c'. Reading /dev/kmsg itself
797 : * changes no global state, and does not clear anything.
798 : */
799 0 : user->seq = clear_seq;
800 0 : break;
801 0 : case SEEK_END:
802 : /* after the last record */
803 0 : user->seq = prb_next_seq(prb);
804 0 : break;
805 : default:
806 : ret = -EINVAL;
807 : }
808 0 : logbuf_unlock_irq();
809 0 : return ret;
810 : }
811 :
812 76 : static __poll_t devkmsg_poll(struct file *file, poll_table *wait)
813 : {
814 76 : struct devkmsg_user *user = file->private_data;
815 76 : struct printk_info info;
816 76 : __poll_t ret = 0;
817 :
818 76 : if (!user)
819 : return EPOLLERR|EPOLLNVAL;
820 :
821 76 : poll_wait(file, &log_wait, wait);
822 :
823 76 : logbuf_lock_irq();
824 76 : if (prb_read_valid_info(prb, user->seq, &info, NULL)) {
825 : /* return error when data has vanished underneath us */
826 68 : if (info.seq != user->seq)
827 : ret = EPOLLIN|EPOLLRDNORM|EPOLLERR|EPOLLPRI;
828 : else
829 68 : ret = EPOLLIN|EPOLLRDNORM;
830 : }
831 76 : logbuf_unlock_irq();
832 :
833 76 : return ret;
834 : }
835 :
836 18 : static int devkmsg_open(struct inode *inode, struct file *file)
837 : {
838 18 : struct devkmsg_user *user;
839 18 : int err;
840 :
841 18 : if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
842 : return -EPERM;
843 :
844 : /* write-only does not need any file context */
845 18 : if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
846 1 : err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
847 : SYSLOG_FROM_READER);
848 1 : if (err)
849 : return err;
850 : }
851 :
852 18 : user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
853 18 : if (!user)
854 : return -ENOMEM;
855 :
856 18 : ratelimit_default_init(&user->rs);
857 18 : ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE);
858 :
859 18 : mutex_init(&user->lock);
860 :
861 18 : prb_rec_init_rd(&user->record, &user->info,
862 : &user->text_buf[0], sizeof(user->text_buf));
863 :
864 18 : logbuf_lock_irq();
865 18 : user->seq = prb_first_valid_seq(prb);
866 18 : logbuf_unlock_irq();
867 :
868 18 : file->private_data = user;
869 18 : return 0;
870 : }
871 :
872 15 : static int devkmsg_release(struct inode *inode, struct file *file)
873 : {
874 15 : struct devkmsg_user *user = file->private_data;
875 :
876 15 : if (!user)
877 : return 0;
878 :
879 15 : ratelimit_state_exit(&user->rs);
880 :
881 15 : mutex_destroy(&user->lock);
882 15 : kfree(user);
883 15 : return 0;
884 : }
885 :
886 : const struct file_operations kmsg_fops = {
887 : .open = devkmsg_open,
888 : .read = devkmsg_read,
889 : .write_iter = devkmsg_write,
890 : .llseek = devkmsg_llseek,
891 : .poll = devkmsg_poll,
892 : .release = devkmsg_release,
893 : };
894 :
895 : #ifdef CONFIG_CRASH_CORE
896 : /*
897 : * This appends the listed symbols to /proc/vmcore
898 : *
899 : * /proc/vmcore is used by various utilities, like crash and makedumpfile to
900 : * obtain access to symbols that are otherwise very difficult to locate. These
901 : * symbols are specifically used so that utilities can access and extract the
902 : * dmesg log from a vmcore file after a crash.
903 : */
904 : void log_buf_vmcoreinfo_setup(void)
905 : {
906 : struct dev_printk_info *dev_info = NULL;
907 :
908 : VMCOREINFO_SYMBOL(prb);
909 : VMCOREINFO_SYMBOL(printk_rb_static);
910 : VMCOREINFO_SYMBOL(clear_seq);
911 :
912 : /*
913 : * Export struct size and field offsets. User space tools can
914 : * parse it and detect any changes to structure down the line.
915 : */
916 :
917 : VMCOREINFO_STRUCT_SIZE(printk_ringbuffer);
918 : VMCOREINFO_OFFSET(printk_ringbuffer, desc_ring);
919 : VMCOREINFO_OFFSET(printk_ringbuffer, text_data_ring);
920 : VMCOREINFO_OFFSET(printk_ringbuffer, fail);
921 :
922 : VMCOREINFO_STRUCT_SIZE(prb_desc_ring);
923 : VMCOREINFO_OFFSET(prb_desc_ring, count_bits);
924 : VMCOREINFO_OFFSET(prb_desc_ring, descs);
925 : VMCOREINFO_OFFSET(prb_desc_ring, infos);
926 : VMCOREINFO_OFFSET(prb_desc_ring, head_id);
927 : VMCOREINFO_OFFSET(prb_desc_ring, tail_id);
928 :
929 : VMCOREINFO_STRUCT_SIZE(prb_desc);
930 : VMCOREINFO_OFFSET(prb_desc, state_var);
931 : VMCOREINFO_OFFSET(prb_desc, text_blk_lpos);
932 :
933 : VMCOREINFO_STRUCT_SIZE(prb_data_blk_lpos);
934 : VMCOREINFO_OFFSET(prb_data_blk_lpos, begin);
935 : VMCOREINFO_OFFSET(prb_data_blk_lpos, next);
936 :
937 : VMCOREINFO_STRUCT_SIZE(printk_info);
938 : VMCOREINFO_OFFSET(printk_info, seq);
939 : VMCOREINFO_OFFSET(printk_info, ts_nsec);
940 : VMCOREINFO_OFFSET(printk_info, text_len);
941 : VMCOREINFO_OFFSET(printk_info, caller_id);
942 : VMCOREINFO_OFFSET(printk_info, dev_info);
943 :
944 : VMCOREINFO_STRUCT_SIZE(dev_printk_info);
945 : VMCOREINFO_OFFSET(dev_printk_info, subsystem);
946 : VMCOREINFO_LENGTH(printk_info_subsystem, sizeof(dev_info->subsystem));
947 : VMCOREINFO_OFFSET(dev_printk_info, device);
948 : VMCOREINFO_LENGTH(printk_info_device, sizeof(dev_info->device));
949 :
950 : VMCOREINFO_STRUCT_SIZE(prb_data_ring);
951 : VMCOREINFO_OFFSET(prb_data_ring, size_bits);
952 : VMCOREINFO_OFFSET(prb_data_ring, data);
953 : VMCOREINFO_OFFSET(prb_data_ring, head_lpos);
954 : VMCOREINFO_OFFSET(prb_data_ring, tail_lpos);
955 :
956 : VMCOREINFO_SIZE(atomic_long_t);
957 : VMCOREINFO_TYPE_OFFSET(atomic_long_t, counter);
958 : }
959 : #endif
960 :
961 : /* requested log_buf_len from kernel cmdline */
962 : static unsigned long __initdata new_log_buf_len;
963 :
964 : /* we practice scaling the ring buffer by powers of 2 */
965 0 : static void __init log_buf_len_update(u64 size)
966 : {
967 0 : if (size > (u64)LOG_BUF_LEN_MAX) {
968 0 : size = (u64)LOG_BUF_LEN_MAX;
969 0 : pr_err("log_buf over 2G is not supported.\n");
970 : }
971 :
972 0 : if (size)
973 0 : size = roundup_pow_of_two(size);
974 0 : if (size > log_buf_len)
975 0 : new_log_buf_len = (unsigned long)size;
976 0 : }
977 :
978 : /* save requested log_buf_len since it's too early to process it */
979 0 : static int __init log_buf_len_setup(char *str)
980 : {
981 0 : u64 size;
982 :
983 0 : if (!str)
984 : return -EINVAL;
985 :
986 0 : size = memparse(str, &str);
987 :
988 0 : log_buf_len_update(size);
989 :
990 0 : return 0;
991 : }
992 : early_param("log_buf_len", log_buf_len_setup);
993 :
994 : #ifdef CONFIG_SMP
995 : #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
996 :
997 1 : static void __init log_buf_add_cpu(void)
998 : {
999 1 : unsigned int cpu_extra;
1000 :
1001 : /*
1002 : * archs should set up cpu_possible_bits properly with
1003 : * set_cpu_possible() after setup_arch() but just in
1004 : * case lets ensure this is valid.
1005 : */
1006 1 : if (num_possible_cpus() == 1)
1007 : return;
1008 :
1009 1 : cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
1010 :
1011 : /* by default this will only continue through for large > 64 CPUs */
1012 1 : if (cpu_extra <= __LOG_BUF_LEN / 2)
1013 : return;
1014 :
1015 0 : pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1016 : __LOG_CPU_MAX_BUF_LEN);
1017 0 : pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1018 : cpu_extra);
1019 0 : pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN);
1020 :
1021 0 : log_buf_len_update(cpu_extra + __LOG_BUF_LEN);
1022 : }
1023 : #else /* !CONFIG_SMP */
1024 : static inline void log_buf_add_cpu(void) {}
1025 : #endif /* CONFIG_SMP */
1026 :
1027 1 : static void __init set_percpu_data_ready(void)
1028 : {
1029 1 : printk_safe_init();
1030 : /* Make sure we set this flag only after printk_safe() init is done */
1031 1 : barrier();
1032 1 : __printk_percpu_data_ready = true;
1033 1 : }
1034 :
1035 0 : static unsigned int __init add_to_rb(struct printk_ringbuffer *rb,
1036 : struct printk_record *r)
1037 : {
1038 0 : struct prb_reserved_entry e;
1039 0 : struct printk_record dest_r;
1040 :
1041 0 : prb_rec_init_wr(&dest_r, r->info->text_len);
1042 :
1043 0 : if (!prb_reserve(&e, rb, &dest_r))
1044 : return 0;
1045 :
1046 0 : memcpy(&dest_r.text_buf[0], &r->text_buf[0], r->info->text_len);
1047 0 : dest_r.info->text_len = r->info->text_len;
1048 0 : dest_r.info->facility = r->info->facility;
1049 0 : dest_r.info->level = r->info->level;
1050 0 : dest_r.info->flags = r->info->flags;
1051 0 : dest_r.info->ts_nsec = r->info->ts_nsec;
1052 0 : dest_r.info->caller_id = r->info->caller_id;
1053 0 : memcpy(&dest_r.info->dev_info, &r->info->dev_info, sizeof(dest_r.info->dev_info));
1054 :
1055 0 : prb_final_commit(&e);
1056 :
1057 0 : return prb_record_text_space(&e);
1058 : }
1059 :
1060 : static char setup_text_buf[LOG_LINE_MAX] __initdata;
1061 :
1062 2 : void __init setup_log_buf(int early)
1063 : {
1064 2 : struct printk_info *new_infos;
1065 2 : unsigned int new_descs_count;
1066 2 : struct prb_desc *new_descs;
1067 2 : struct printk_info info;
1068 2 : struct printk_record r;
1069 2 : size_t new_descs_size;
1070 2 : size_t new_infos_size;
1071 2 : unsigned long flags;
1072 2 : char *new_log_buf;
1073 2 : unsigned int free;
1074 2 : u64 seq;
1075 :
1076 : /*
1077 : * Some archs call setup_log_buf() multiple times - first is very
1078 : * early, e.g. from setup_arch(), and second - when percpu_areas
1079 : * are initialised.
1080 : */
1081 2 : if (!early)
1082 1 : set_percpu_data_ready();
1083 :
1084 2 : if (log_buf != __log_buf)
1085 2 : return;
1086 :
1087 2 : if (!early && !new_log_buf_len)
1088 1 : log_buf_add_cpu();
1089 :
1090 2 : if (!new_log_buf_len)
1091 : return;
1092 :
1093 0 : new_descs_count = new_log_buf_len >> PRB_AVGBITS;
1094 0 : if (new_descs_count == 0) {
1095 0 : pr_err("new_log_buf_len: %lu too small\n", new_log_buf_len);
1096 0 : return;
1097 : }
1098 :
1099 0 : new_log_buf = memblock_alloc(new_log_buf_len, LOG_ALIGN);
1100 0 : if (unlikely(!new_log_buf)) {
1101 0 : pr_err("log_buf_len: %lu text bytes not available\n",
1102 : new_log_buf_len);
1103 0 : return;
1104 : }
1105 :
1106 0 : new_descs_size = new_descs_count * sizeof(struct prb_desc);
1107 0 : new_descs = memblock_alloc(new_descs_size, LOG_ALIGN);
1108 0 : if (unlikely(!new_descs)) {
1109 0 : pr_err("log_buf_len: %zu desc bytes not available\n",
1110 : new_descs_size);
1111 0 : goto err_free_log_buf;
1112 : }
1113 :
1114 0 : new_infos_size = new_descs_count * sizeof(struct printk_info);
1115 0 : new_infos = memblock_alloc(new_infos_size, LOG_ALIGN);
1116 0 : if (unlikely(!new_infos)) {
1117 0 : pr_err("log_buf_len: %zu info bytes not available\n",
1118 : new_infos_size);
1119 0 : goto err_free_descs;
1120 : }
1121 :
1122 0 : prb_rec_init_rd(&r, &info, &setup_text_buf[0], sizeof(setup_text_buf));
1123 :
1124 0 : prb_init(&printk_rb_dynamic,
1125 0 : new_log_buf, ilog2(new_log_buf_len),
1126 0 : new_descs, ilog2(new_descs_count),
1127 : new_infos);
1128 :
1129 0 : printk_safe_enter_irqsave(flags);
1130 :
1131 0 : log_buf_len = new_log_buf_len;
1132 0 : log_buf = new_log_buf;
1133 0 : new_log_buf_len = 0;
1134 :
1135 0 : free = __LOG_BUF_LEN;
1136 0 : prb_for_each_record(0, &printk_rb_static, seq, &r)
1137 0 : free -= add_to_rb(&printk_rb_dynamic, &r);
1138 :
1139 : /*
1140 : * This is early enough that everything is still running on the
1141 : * boot CPU and interrupts are disabled. So no new messages will
1142 : * appear during the transition to the dynamic buffer.
1143 : */
1144 0 : prb = &printk_rb_dynamic;
1145 :
1146 0 : printk_safe_exit_irqrestore(flags);
1147 :
1148 0 : if (seq != prb_next_seq(&printk_rb_static)) {
1149 0 : pr_err("dropped %llu messages\n",
1150 : prb_next_seq(&printk_rb_static) - seq);
1151 : }
1152 :
1153 0 : pr_info("log_buf_len: %u bytes\n", log_buf_len);
1154 0 : pr_info("early log buf free: %u(%u%%)\n",
1155 : free, (free * 100) / __LOG_BUF_LEN);
1156 0 : return;
1157 :
1158 0 : err_free_descs:
1159 0 : memblock_free(__pa(new_descs), new_descs_size);
1160 0 : err_free_log_buf:
1161 0 : memblock_free(__pa(new_log_buf), new_log_buf_len);
1162 : }
1163 :
1164 : static bool __read_mostly ignore_loglevel;
1165 :
1166 0 : static int __init ignore_loglevel_setup(char *str)
1167 : {
1168 0 : ignore_loglevel = true;
1169 0 : pr_info("debug: ignoring loglevel setting.\n");
1170 :
1171 0 : return 0;
1172 : }
1173 :
1174 : early_param("ignore_loglevel", ignore_loglevel_setup);
1175 : module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1176 : MODULE_PARM_DESC(ignore_loglevel,
1177 : "ignore loglevel setting (prints all kernel messages to the console)");
1178 :
1179 357 : static bool suppress_message_printing(int level)
1180 : {
1181 21 : return (level >= console_loglevel && !ignore_loglevel);
1182 : }
1183 :
1184 : #ifdef CONFIG_BOOT_PRINTK_DELAY
1185 :
1186 : static int boot_delay; /* msecs delay after each printk during bootup */
1187 : static unsigned long long loops_per_msec; /* based on boot_delay */
1188 :
1189 : static int __init boot_delay_setup(char *str)
1190 : {
1191 : unsigned long lpj;
1192 :
1193 : lpj = preset_lpj ? preset_lpj : 1000000; /* some guess */
1194 : loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
1195 :
1196 : get_option(&str, &boot_delay);
1197 : if (boot_delay > 10 * 1000)
1198 : boot_delay = 0;
1199 :
1200 : pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1201 : "HZ: %d, loops_per_msec: %llu\n",
1202 : boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
1203 : return 0;
1204 : }
1205 : early_param("boot_delay", boot_delay_setup);
1206 :
1207 : static void boot_delay_msec(int level)
1208 : {
1209 : unsigned long long k;
1210 : unsigned long timeout;
1211 :
1212 : if ((boot_delay == 0 || system_state >= SYSTEM_RUNNING)
1213 : || suppress_message_printing(level)) {
1214 : return;
1215 : }
1216 :
1217 : k = (unsigned long long)loops_per_msec * boot_delay;
1218 :
1219 : timeout = jiffies + msecs_to_jiffies(boot_delay);
1220 : while (k) {
1221 : k--;
1222 : cpu_relax();
1223 : /*
1224 : * use (volatile) jiffies to prevent
1225 : * compiler reduction; loop termination via jiffies
1226 : * is secondary and may or may not happen.
1227 : */
1228 : if (time_after(jiffies, timeout))
1229 : break;
1230 : touch_nmi_watchdog();
1231 : }
1232 : }
1233 : #else
1234 307 : static inline void boot_delay_msec(int level)
1235 : {
1236 307 : }
1237 : #endif
1238 :
1239 : static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME);
1240 : module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
1241 :
1242 270 : static size_t print_syslog(unsigned int level, char *buf)
1243 : {
1244 270 : return sprintf(buf, "<%u>", level);
1245 : }
1246 :
1247 0 : static size_t print_time(u64 ts, char *buf)
1248 : {
1249 0 : unsigned long rem_nsec = do_div(ts, 1000000000);
1250 :
1251 0 : return sprintf(buf, "[%5lu.%06lu]",
1252 : (unsigned long)ts, rem_nsec / 1000);
1253 : }
1254 :
1255 : #ifdef CONFIG_PRINTK_CALLER
1256 : static size_t print_caller(u32 id, char *buf)
1257 : {
1258 : char caller[12];
1259 :
1260 : snprintf(caller, sizeof(caller), "%c%u",
1261 : id & 0x80000000 ? 'C' : 'T', id & ~0x80000000);
1262 : return sprintf(buf, "[%6s]", caller);
1263 : }
1264 : #else
1265 : #define print_caller(id, buf) 0
1266 : #endif
1267 :
1268 606 : static size_t info_print_prefix(const struct printk_info *info, bool syslog,
1269 : bool time, char *buf)
1270 : {
1271 606 : size_t len = 0;
1272 :
1273 606 : if (syslog)
1274 270 : len = print_syslog((info->facility << 3) | info->level, buf);
1275 :
1276 606 : if (time)
1277 0 : len += print_time(info->ts_nsec, buf + len);
1278 :
1279 606 : len += print_caller(info->caller_id, buf + len);
1280 :
1281 606 : if (IS_ENABLED(CONFIG_PRINTK_CALLER) || time) {
1282 0 : buf[len++] = ' ';
1283 0 : buf[len] = '\0';
1284 : }
1285 :
1286 606 : return len;
1287 : }
1288 :
1289 : /*
1290 : * Prepare the record for printing. The text is shifted within the given
1291 : * buffer to avoid a need for another one. The following operations are
1292 : * done:
1293 : *
1294 : * - Add prefix for each line.
1295 : * - Drop truncated lines that no longer fit into the buffer.
1296 : * - Add the trailing newline that has been removed in vprintk_store().
1297 : * - Add a string terminator.
1298 : *
1299 : * Since the produced string is always terminated, the maximum possible
1300 : * return value is @r->text_buf_size - 1;
1301 : *
1302 : * Return: The length of the updated/prepared text, including the added
1303 : * prefixes and the newline. The terminator is not counted. The dropped
1304 : * line(s) are not counted.
1305 : */
1306 606 : static size_t record_print_text(struct printk_record *r, bool syslog,
1307 : bool time)
1308 : {
1309 606 : size_t text_len = r->info->text_len;
1310 606 : size_t buf_size = r->text_buf_size;
1311 606 : char *text = r->text_buf;
1312 606 : char prefix[PREFIX_MAX];
1313 606 : bool truncated = false;
1314 606 : size_t prefix_len;
1315 606 : size_t line_len;
1316 606 : size_t len = 0;
1317 606 : char *next;
1318 :
1319 : /*
1320 : * If the message was truncated because the buffer was not large
1321 : * enough, treat the available text as if it were the full text.
1322 : */
1323 606 : if (text_len > buf_size)
1324 : text_len = buf_size;
1325 :
1326 606 : prefix_len = info_print_prefix(r->info, syslog, time, prefix);
1327 :
1328 : /*
1329 : * @text_len: bytes of unprocessed text
1330 : * @line_len: bytes of current line _without_ newline
1331 : * @text: pointer to beginning of current line
1332 : * @len: number of bytes prepared in r->text_buf
1333 : */
1334 606 : for (;;) {
1335 606 : next = memchr(text, '\n', text_len);
1336 606 : if (next) {
1337 0 : line_len = next - text;
1338 : } else {
1339 : /* Drop truncated line(s). */
1340 606 : if (truncated)
1341 : break;
1342 : line_len = text_len;
1343 : }
1344 :
1345 : /*
1346 : * Truncate the text if there is not enough space to add the
1347 : * prefix and a trailing newline and a terminator.
1348 : */
1349 606 : if (len + prefix_len + text_len + 1 + 1 > buf_size) {
1350 : /* Drop even the current line if no space. */
1351 0 : if (len + prefix_len + line_len + 1 + 1 > buf_size)
1352 : break;
1353 :
1354 0 : text_len = buf_size - len - prefix_len - 1 - 1;
1355 0 : truncated = true;
1356 : }
1357 :
1358 606 : memmove(text + prefix_len, text, text_len);
1359 606 : memcpy(text, prefix, prefix_len);
1360 :
1361 : /*
1362 : * Increment the prepared length to include the text and
1363 : * prefix that were just moved+copied. Also increment for the
1364 : * newline at the end of this line. If this is the last line,
1365 : * there is no newline, but it will be added immediately below.
1366 : */
1367 606 : len += prefix_len + line_len + 1;
1368 606 : if (text_len == line_len) {
1369 : /*
1370 : * This is the last line. Add the trailing newline
1371 : * removed in vprintk_store().
1372 : */
1373 606 : text[prefix_len + line_len] = '\n';
1374 606 : break;
1375 : }
1376 :
1377 : /*
1378 : * Advance beyond the added prefix and the related line with
1379 : * its newline.
1380 : */
1381 0 : text += prefix_len + line_len + 1;
1382 :
1383 : /*
1384 : * The remaining text has only decreased by the line with its
1385 : * newline.
1386 : *
1387 : * Note that @text_len can become zero. It happens when @text
1388 : * ended with a newline (either due to truncation or the
1389 : * original string ending with "\n\n"). The loop is correctly
1390 : * repeated and (if not truncated) an empty line with a prefix
1391 : * will be prepared.
1392 : */
1393 0 : text_len -= line_len + 1;
1394 : }
1395 :
1396 : /*
1397 : * If a buffer was provided, it will be terminated. Space for the
1398 : * string terminator is guaranteed to be available. The terminator is
1399 : * not counted in the return value.
1400 : */
1401 606 : if (buf_size > 0)
1402 606 : r->text_buf[len] = 0;
1403 :
1404 606 : return len;
1405 : }
1406 :
1407 0 : static size_t get_record_print_text_size(struct printk_info *info,
1408 : unsigned int line_count,
1409 : bool syslog, bool time)
1410 : {
1411 0 : char prefix[PREFIX_MAX];
1412 0 : size_t prefix_len;
1413 :
1414 0 : prefix_len = info_print_prefix(info, syslog, time, prefix);
1415 :
1416 : /*
1417 : * Each line will be preceded with a prefix. The intermediate
1418 : * newlines are already within the text, but a final trailing
1419 : * newline will be added.
1420 : */
1421 0 : return ((prefix_len * line_count) + info->text_len + 1);
1422 : }
1423 :
1424 3 : static int syslog_print(char __user *buf, int size)
1425 : {
1426 3 : struct printk_info info;
1427 3 : struct printk_record r;
1428 3 : char *text;
1429 3 : int len = 0;
1430 :
1431 3 : text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1432 3 : if (!text)
1433 : return -ENOMEM;
1434 :
1435 3 : prb_rec_init_rd(&r, &info, text, LOG_LINE_MAX + PREFIX_MAX);
1436 :
1437 272 : while (size > 0) {
1438 272 : size_t n;
1439 272 : size_t skip;
1440 :
1441 272 : logbuf_lock_irq();
1442 272 : if (!prb_read_valid(prb, syslog_seq, &r)) {
1443 2 : logbuf_unlock_irq();
1444 : break;
1445 : }
1446 270 : if (r.info->seq != syslog_seq) {
1447 : /* message is gone, move to next valid one */
1448 0 : syslog_seq = r.info->seq;
1449 0 : syslog_partial = 0;
1450 : }
1451 :
1452 : /*
1453 : * To keep reading/counting partial line consistent,
1454 : * use printk_time value as of the beginning of a line.
1455 : */
1456 270 : if (!syslog_partial)
1457 270 : syslog_time = printk_time;
1458 :
1459 270 : skip = syslog_partial;
1460 270 : n = record_print_text(&r, true, syslog_time);
1461 270 : if (n - syslog_partial <= size) {
1462 : /* message fits into buffer, move forward */
1463 269 : syslog_seq = r.info->seq + 1;
1464 269 : n -= syslog_partial;
1465 269 : syslog_partial = 0;
1466 1 : } else if (!len){
1467 : /* partial read(), remember position */
1468 0 : n = size;
1469 0 : syslog_partial += n;
1470 : } else
1471 : n = 0;
1472 270 : logbuf_unlock_irq();
1473 :
1474 270 : if (!n)
1475 : break;
1476 :
1477 538 : if (copy_to_user(buf, text + skip, n)) {
1478 0 : if (!len)
1479 0 : len = -EFAULT;
1480 : break;
1481 : }
1482 :
1483 269 : len += n;
1484 269 : size -= n;
1485 269 : buf += n;
1486 : }
1487 :
1488 3 : kfree(text);
1489 3 : return len;
1490 : }
1491 :
1492 0 : static int syslog_print_all(char __user *buf, int size, bool clear)
1493 : {
1494 0 : struct printk_info info;
1495 0 : unsigned int line_count;
1496 0 : struct printk_record r;
1497 0 : char *text;
1498 0 : int len = 0;
1499 0 : u64 seq;
1500 0 : bool time;
1501 :
1502 0 : text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1503 0 : if (!text)
1504 : return -ENOMEM;
1505 :
1506 0 : time = printk_time;
1507 0 : logbuf_lock_irq();
1508 : /*
1509 : * Find first record that fits, including all following records,
1510 : * into the user-provided buffer for this dump.
1511 : */
1512 0 : prb_for_each_info(clear_seq, prb, seq, &info, &line_count)
1513 0 : len += get_record_print_text_size(&info, line_count, true, time);
1514 :
1515 : /* move first record forward until length fits into the buffer */
1516 0 : prb_for_each_info(clear_seq, prb, seq, &info, &line_count) {
1517 0 : if (len <= size)
1518 : break;
1519 0 : len -= get_record_print_text_size(&info, line_count, true, time);
1520 : }
1521 :
1522 0 : prb_rec_init_rd(&r, &info, text, LOG_LINE_MAX + PREFIX_MAX);
1523 :
1524 0 : len = 0;
1525 0 : prb_for_each_record(seq, prb, seq, &r) {
1526 0 : int textlen;
1527 :
1528 0 : textlen = record_print_text(&r, true, time);
1529 :
1530 0 : if (len + textlen > size) {
1531 0 : seq--;
1532 0 : break;
1533 : }
1534 :
1535 0 : logbuf_unlock_irq();
1536 0 : if (copy_to_user(buf + len, text, textlen))
1537 : len = -EFAULT;
1538 : else
1539 0 : len += textlen;
1540 0 : logbuf_lock_irq();
1541 :
1542 0 : if (len < 0)
1543 : break;
1544 : }
1545 :
1546 0 : if (clear)
1547 0 : clear_seq = seq;
1548 0 : logbuf_unlock_irq();
1549 :
1550 0 : kfree(text);
1551 0 : return len;
1552 : }
1553 :
1554 0 : static void syslog_clear(void)
1555 : {
1556 0 : logbuf_lock_irq();
1557 0 : clear_seq = prb_next_seq(prb);
1558 0 : logbuf_unlock_irq();
1559 0 : }
1560 :
1561 6 : int do_syslog(int type, char __user *buf, int len, int source)
1562 : {
1563 6 : struct printk_info info;
1564 6 : bool clear = false;
1565 6 : static int saved_console_loglevel = LOGLEVEL_DEFAULT;
1566 6 : int error;
1567 :
1568 6 : error = check_syslog_permissions(type, source);
1569 6 : if (error)
1570 : return error;
1571 :
1572 6 : switch (type) {
1573 : case SYSLOG_ACTION_CLOSE: /* Close log */
1574 : break;
1575 : case SYSLOG_ACTION_OPEN: /* Open log */
1576 : break;
1577 5 : case SYSLOG_ACTION_READ: /* Read from log */
1578 5 : if (!buf || len < 0)
1579 : return -EINVAL;
1580 4 : if (!len)
1581 : return 0;
1582 8 : if (!access_ok(buf, len))
1583 : return -EFAULT;
1584 5 : error = wait_event_interruptible(log_wait,
1585 : prb_read_valid(prb, syslog_seq, NULL));
1586 1 : if (error)
1587 : return error;
1588 3 : error = syslog_print(buf, len);
1589 3 : break;
1590 : /* Read/clear last kernel messages */
1591 0 : case SYSLOG_ACTION_READ_CLEAR:
1592 0 : clear = true;
1593 0 : fallthrough;
1594 : /* Read last kernel messages */
1595 0 : case SYSLOG_ACTION_READ_ALL:
1596 0 : if (!buf || len < 0)
1597 : return -EINVAL;
1598 0 : if (!len)
1599 : return 0;
1600 0 : if (!access_ok(buf, len))
1601 : return -EFAULT;
1602 0 : error = syslog_print_all(buf, len, clear);
1603 0 : break;
1604 : /* Clear ring buffer */
1605 0 : case SYSLOG_ACTION_CLEAR:
1606 0 : syslog_clear();
1607 0 : break;
1608 : /* Disable logging to console */
1609 0 : case SYSLOG_ACTION_CONSOLE_OFF:
1610 0 : if (saved_console_loglevel == LOGLEVEL_DEFAULT)
1611 0 : saved_console_loglevel = console_loglevel;
1612 0 : console_loglevel = minimum_console_loglevel;
1613 0 : break;
1614 : /* Enable logging to console */
1615 0 : case SYSLOG_ACTION_CONSOLE_ON:
1616 0 : if (saved_console_loglevel != LOGLEVEL_DEFAULT) {
1617 0 : console_loglevel = saved_console_loglevel;
1618 0 : saved_console_loglevel = LOGLEVEL_DEFAULT;
1619 : }
1620 : break;
1621 : /* Set level of messages printed to console */
1622 0 : case SYSLOG_ACTION_CONSOLE_LEVEL:
1623 0 : if (len < 1 || len > 8)
1624 : return -EINVAL;
1625 0 : if (len < minimum_console_loglevel)
1626 : len = minimum_console_loglevel;
1627 0 : console_loglevel = len;
1628 : /* Implicitly re-enable logging to console */
1629 0 : saved_console_loglevel = LOGLEVEL_DEFAULT;
1630 0 : break;
1631 : /* Number of chars in the log buffer */
1632 : case SYSLOG_ACTION_SIZE_UNREAD:
1633 0 : logbuf_lock_irq();
1634 0 : if (!prb_read_valid_info(prb, syslog_seq, &info, NULL)) {
1635 : /* No unread messages. */
1636 0 : logbuf_unlock_irq();
1637 0 : return 0;
1638 : }
1639 0 : if (info.seq != syslog_seq) {
1640 : /* messages are gone, move to first one */
1641 0 : syslog_seq = info.seq;
1642 0 : syslog_partial = 0;
1643 : }
1644 0 : if (source == SYSLOG_FROM_PROC) {
1645 : /*
1646 : * Short-cut for poll(/"proc/kmsg") which simply checks
1647 : * for pending data, not the size; return the count of
1648 : * records, not the length.
1649 : */
1650 0 : error = prb_next_seq(prb) - syslog_seq;
1651 : } else {
1652 0 : bool time = syslog_partial ? syslog_time : printk_time;
1653 0 : unsigned int line_count;
1654 0 : u64 seq;
1655 :
1656 0 : prb_for_each_info(syslog_seq, prb, seq, &info,
1657 : &line_count) {
1658 0 : error += get_record_print_text_size(&info, line_count,
1659 : true, time);
1660 0 : time = printk_time;
1661 : }
1662 0 : error -= syslog_partial;
1663 : }
1664 0 : logbuf_unlock_irq();
1665 : break;
1666 : /* Size of the log buffer */
1667 0 : case SYSLOG_ACTION_SIZE_BUFFER:
1668 0 : error = log_buf_len;
1669 0 : break;
1670 0 : default:
1671 0 : error = -EINVAL;
1672 0 : break;
1673 : }
1674 :
1675 : return error;
1676 : }
1677 :
1678 0 : SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1679 : {
1680 0 : return do_syslog(type, buf, len, SYSLOG_FROM_READER);
1681 : }
1682 :
1683 : /*
1684 : * Special console_lock variants that help to reduce the risk of soft-lockups.
1685 : * They allow to pass console_lock to another printk() call using a busy wait.
1686 : */
1687 :
1688 : #ifdef CONFIG_LOCKDEP
1689 : static struct lockdep_map console_owner_dep_map = {
1690 : .name = "console_owner"
1691 : };
1692 : #endif
1693 :
1694 : static DEFINE_RAW_SPINLOCK(console_owner_lock);
1695 : static struct task_struct *console_owner;
1696 : static bool console_waiter;
1697 :
1698 : /**
1699 : * console_lock_spinning_enable - mark beginning of code where another
1700 : * thread might safely busy wait
1701 : *
1702 : * This basically converts console_lock into a spinlock. This marks
1703 : * the section where the console_lock owner can not sleep, because
1704 : * there may be a waiter spinning (like a spinlock). Also it must be
1705 : * ready to hand over the lock at the end of the section.
1706 : */
1707 336 : static void console_lock_spinning_enable(void)
1708 : {
1709 336 : raw_spin_lock(&console_owner_lock);
1710 336 : console_owner = current;
1711 336 : raw_spin_unlock(&console_owner_lock);
1712 :
1713 : /* The waiter may spin on us after setting console_owner */
1714 336 : spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1715 336 : }
1716 :
1717 : /**
1718 : * console_lock_spinning_disable_and_check - mark end of code where another
1719 : * thread was able to busy wait and check if there is a waiter
1720 : *
1721 : * This is called at the end of the section where spinning is allowed.
1722 : * It has two functions. First, it is a signal that it is no longer
1723 : * safe to start busy waiting for the lock. Second, it checks if
1724 : * there is a busy waiter and passes the lock rights to her.
1725 : *
1726 : * Important: Callers lose the lock if there was a busy waiter.
1727 : * They must not touch items synchronized by console_lock
1728 : * in this case.
1729 : *
1730 : * Return: 1 if the lock rights were passed, 0 otherwise.
1731 : */
1732 336 : static int console_lock_spinning_disable_and_check(void)
1733 : {
1734 336 : int waiter;
1735 :
1736 336 : raw_spin_lock(&console_owner_lock);
1737 336 : waiter = READ_ONCE(console_waiter);
1738 336 : console_owner = NULL;
1739 336 : raw_spin_unlock(&console_owner_lock);
1740 :
1741 336 : if (!waiter) {
1742 336 : spin_release(&console_owner_dep_map, _THIS_IP_);
1743 336 : return 0;
1744 : }
1745 :
1746 : /* The waiter is now free to continue */
1747 0 : WRITE_ONCE(console_waiter, false);
1748 :
1749 0 : spin_release(&console_owner_dep_map, _THIS_IP_);
1750 :
1751 : /*
1752 : * Hand off console_lock to waiter. The waiter will perform
1753 : * the up(). After this, the waiter is the console_lock owner.
1754 : */
1755 0 : mutex_release(&console_lock_dep_map, _THIS_IP_);
1756 0 : return 1;
1757 : }
1758 :
1759 : /**
1760 : * console_trylock_spinning - try to get console_lock by busy waiting
1761 : *
1762 : * This allows to busy wait for the console_lock when the current
1763 : * owner is running in specially marked sections. It means that
1764 : * the current owner is running and cannot reschedule until it
1765 : * is ready to lose the lock.
1766 : *
1767 : * Return: 1 if we got the lock, 0 othrewise
1768 : */
1769 307 : static int console_trylock_spinning(void)
1770 : {
1771 307 : struct task_struct *owner = NULL;
1772 307 : bool waiter;
1773 307 : bool spin = false;
1774 307 : unsigned long flags;
1775 :
1776 307 : if (console_trylock())
1777 : return 1;
1778 :
1779 2 : printk_safe_enter_irqsave(flags);
1780 :
1781 1 : raw_spin_lock(&console_owner_lock);
1782 1 : owner = READ_ONCE(console_owner);
1783 1 : waiter = READ_ONCE(console_waiter);
1784 1 : if (!waiter && owner && owner != current) {
1785 0 : WRITE_ONCE(console_waiter, true);
1786 0 : spin = true;
1787 : }
1788 1 : raw_spin_unlock(&console_owner_lock);
1789 :
1790 : /*
1791 : * If there is an active printk() writing to the
1792 : * consoles, instead of having it write our data too,
1793 : * see if we can offload that load from the active
1794 : * printer, and do some printing ourselves.
1795 : * Go into a spin only if there isn't already a waiter
1796 : * spinning, and there is an active printer, and
1797 : * that active printer isn't us (recursive printk?).
1798 : */
1799 1 : if (!spin) {
1800 1 : printk_safe_exit_irqrestore(flags);
1801 1 : return 0;
1802 : }
1803 :
1804 : /* We spin waiting for the owner to release us */
1805 0 : spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1806 : /* Owner will clear console_waiter on hand off */
1807 0 : while (READ_ONCE(console_waiter))
1808 0 : cpu_relax();
1809 0 : spin_release(&console_owner_dep_map, _THIS_IP_);
1810 :
1811 0 : printk_safe_exit_irqrestore(flags);
1812 : /*
1813 : * The owner passed the console lock to us.
1814 : * Since we did not spin on console lock, annotate
1815 : * this as a trylock. Otherwise lockdep will
1816 : * complain.
1817 : */
1818 0 : mutex_acquire(&console_lock_dep_map, 0, 1, _THIS_IP_);
1819 :
1820 0 : return 1;
1821 : }
1822 :
1823 : /*
1824 : * Call the console drivers, asking them to write out
1825 : * log_buf[start] to log_buf[end - 1].
1826 : * The console_lock must be held.
1827 : */
1828 336 : static void call_console_drivers(const char *ext_text, size_t ext_len,
1829 : const char *text, size_t len)
1830 : {
1831 336 : static char dropped_text[64];
1832 336 : size_t dropped_len = 0;
1833 336 : struct console *con;
1834 :
1835 336 : trace_console_rcuidle(text, len);
1836 :
1837 336 : if (!console_drivers)
1838 : return;
1839 :
1840 256 : if (console_dropped) {
1841 0 : dropped_len = snprintf(dropped_text, sizeof(dropped_text),
1842 : "** %lu printk messages dropped **\n",
1843 : console_dropped);
1844 0 : console_dropped = 0;
1845 : }
1846 :
1847 512 : for_each_console(con) {
1848 256 : if (exclusive_console && con != exclusive_console)
1849 0 : continue;
1850 256 : if (!(con->flags & CON_ENABLED))
1851 0 : continue;
1852 256 : if (!con->write)
1853 0 : continue;
1854 256 : if (!cpu_online(smp_processor_id()) &&
1855 9 : !(con->flags & CON_ANYTIME))
1856 0 : continue;
1857 256 : if (con->flags & CON_EXTENDED)
1858 0 : con->write(con, ext_text, ext_len);
1859 : else {
1860 256 : if (dropped_len)
1861 0 : con->write(con, dropped_text, dropped_len);
1862 256 : con->write(con, text, len);
1863 : }
1864 : }
1865 : }
1866 :
1867 : int printk_delay_msec __read_mostly;
1868 :
1869 307 : static inline void printk_delay(void)
1870 : {
1871 307 : if (unlikely(printk_delay_msec)) {
1872 : int m = printk_delay_msec;
1873 :
1874 0 : while (m--) {
1875 0 : mdelay(1);
1876 0 : touch_nmi_watchdog();
1877 : }
1878 : }
1879 307 : }
1880 :
1881 307 : static inline u32 printk_caller_id(void)
1882 : {
1883 307 : return in_task() ? task_pid_nr(current) :
1884 0 : 0x80000000 + raw_smp_processor_id();
1885 : }
1886 :
1887 : /**
1888 : * parse_prefix - Parse level and control flags.
1889 : *
1890 : * @text: The terminated text message.
1891 : * @level: A pointer to the current level value, will be updated.
1892 : * @lflags: A pointer to the current log flags, will be updated.
1893 : *
1894 : * @level may be NULL if the caller is not interested in the parsed value.
1895 : * Otherwise the variable pointed to by @level must be set to
1896 : * LOGLEVEL_DEFAULT in order to be updated with the parsed value.
1897 : *
1898 : * @lflags may be NULL if the caller is not interested in the parsed value.
1899 : * Otherwise the variable pointed to by @lflags will be OR'd with the parsed
1900 : * value.
1901 : *
1902 : * Return: The length of the parsed level and control flags.
1903 : */
1904 588 : static u16 parse_prefix(char *text, int *level, enum log_flags *lflags)
1905 : {
1906 588 : u16 prefix_len = 0;
1907 588 : int kern_level;
1908 :
1909 1026 : while (*text) {
1910 1026 : kern_level = printk_get_level(text);
1911 1026 : if (!kern_level)
1912 : break;
1913 :
1914 438 : switch (kern_level) {
1915 358 : case '0' ... '7':
1916 358 : if (level && *level == LOGLEVEL_DEFAULT)
1917 179 : *level = kern_level - '0';
1918 : break;
1919 80 : case 'c': /* KERN_CONT */
1920 80 : if (lflags)
1921 40 : *lflags |= LOG_CONT;
1922 : }
1923 :
1924 438 : prefix_len += 2;
1925 438 : text += 2;
1926 : }
1927 :
1928 588 : return prefix_len;
1929 : }
1930 :
1931 307 : static u16 printk_sprint(char *text, u16 size, int facility, enum log_flags *lflags,
1932 : const char *fmt, va_list args)
1933 : {
1934 307 : u16 text_len;
1935 :
1936 307 : text_len = vscnprintf(text, size, fmt, args);
1937 :
1938 : /* Mark and strip a trailing newline. */
1939 307 : if (text_len && text[text_len - 1] == '\n') {
1940 259 : text_len--;
1941 259 : *lflags |= LOG_NEWLINE;
1942 : }
1943 :
1944 : /* Strip log level and control flags. */
1945 307 : if (facility == 0) {
1946 294 : u16 prefix_len;
1947 :
1948 294 : prefix_len = parse_prefix(text, NULL, NULL);
1949 294 : if (prefix_len) {
1950 219 : text_len -= prefix_len;
1951 219 : memmove(text, text + prefix_len, text_len);
1952 : }
1953 : }
1954 :
1955 307 : return text_len;
1956 : }
1957 :
1958 : __printf(4, 0)
1959 307 : int vprintk_store(int facility, int level,
1960 : const struct dev_printk_info *dev_info,
1961 : const char *fmt, va_list args)
1962 : {
1963 307 : const u32 caller_id = printk_caller_id();
1964 307 : struct prb_reserved_entry e;
1965 307 : enum log_flags lflags = 0;
1966 307 : struct printk_record r;
1967 307 : u16 trunc_msg_len = 0;
1968 307 : char prefix_buf[8];
1969 307 : u16 reserve_size;
1970 307 : va_list args2;
1971 307 : u16 text_len;
1972 307 : u64 ts_nsec;
1973 :
1974 : /*
1975 : * Since the duration of printk() can vary depending on the message
1976 : * and state of the ringbuffer, grab the timestamp now so that it is
1977 : * close to the call of printk(). This provides a more deterministic
1978 : * timestamp with respect to the caller.
1979 : */
1980 307 : ts_nsec = local_clock();
1981 :
1982 : /*
1983 : * The sprintf needs to come first since the syslog prefix might be
1984 : * passed in as a parameter. An extra byte must be reserved so that
1985 : * later the vscnprintf() into the reserved buffer has room for the
1986 : * terminating '\0', which is not counted by vsnprintf().
1987 : */
1988 307 : va_copy(args2, args);
1989 307 : reserve_size = vsnprintf(&prefix_buf[0], sizeof(prefix_buf), fmt, args2) + 1;
1990 307 : va_end(args2);
1991 :
1992 307 : if (reserve_size > LOG_LINE_MAX)
1993 0 : reserve_size = LOG_LINE_MAX;
1994 :
1995 : /* Extract log level or control flags. */
1996 307 : if (facility == 0)
1997 294 : parse_prefix(&prefix_buf[0], &level, &lflags);
1998 :
1999 307 : if (level == LOGLEVEL_DEFAULT)
2000 112 : level = default_message_loglevel;
2001 :
2002 307 : if (dev_info)
2003 3 : lflags |= LOG_NEWLINE;
2004 :
2005 307 : if (lflags & LOG_CONT) {
2006 40 : prb_rec_init_wr(&r, reserve_size);
2007 40 : if (prb_reserve_in_last(&e, prb, &r, caller_id, LOG_LINE_MAX)) {
2008 38 : text_len = printk_sprint(&r.text_buf[r.info->text_len], reserve_size,
2009 : facility, &lflags, fmt, args);
2010 38 : r.info->text_len += text_len;
2011 :
2012 38 : if (lflags & LOG_NEWLINE) {
2013 17 : r.info->flags |= LOG_NEWLINE;
2014 17 : prb_final_commit(&e);
2015 : } else {
2016 21 : prb_commit(&e);
2017 : }
2018 :
2019 38 : return text_len;
2020 : }
2021 : }
2022 :
2023 : /*
2024 : * Explicitly initialize the record before every prb_reserve() call.
2025 : * prb_reserve_in_last() and prb_reserve() purposely invalidate the
2026 : * structure when they fail.
2027 : */
2028 269 : prb_rec_init_wr(&r, reserve_size);
2029 269 : if (!prb_reserve(&e, prb, &r)) {
2030 : /* truncate the message if it is too long for empty buffer */
2031 0 : truncate_msg(&reserve_size, &trunc_msg_len);
2032 :
2033 0 : prb_rec_init_wr(&r, reserve_size + trunc_msg_len);
2034 0 : if (!prb_reserve(&e, prb, &r))
2035 : return 0;
2036 : }
2037 :
2038 : /* fill message */
2039 269 : text_len = printk_sprint(&r.text_buf[0], reserve_size, facility, &lflags, fmt, args);
2040 269 : if (trunc_msg_len)
2041 0 : memcpy(&r.text_buf[text_len], trunc_msg, trunc_msg_len);
2042 269 : r.info->text_len = text_len + trunc_msg_len;
2043 269 : r.info->facility = facility;
2044 269 : r.info->level = level & 7;
2045 269 : r.info->flags = lflags & 0x1f;
2046 269 : r.info->ts_nsec = ts_nsec;
2047 269 : r.info->caller_id = caller_id;
2048 269 : if (dev_info)
2049 3 : memcpy(&r.info->dev_info, dev_info, sizeof(r.info->dev_info));
2050 :
2051 : /* A message without a trailing newline can be continued. */
2052 269 : if (!(lflags & LOG_NEWLINE))
2053 27 : prb_commit(&e);
2054 : else
2055 242 : prb_final_commit(&e);
2056 :
2057 269 : return (text_len + trunc_msg_len);
2058 : }
2059 :
2060 307 : asmlinkage int vprintk_emit(int facility, int level,
2061 : const struct dev_printk_info *dev_info,
2062 : const char *fmt, va_list args)
2063 : {
2064 307 : int printed_len;
2065 307 : bool in_sched = false;
2066 307 : unsigned long flags;
2067 :
2068 : /* Suppress unimportant messages after panic happens */
2069 307 : if (unlikely(suppress_printk))
2070 : return 0;
2071 :
2072 307 : if (level == LOGLEVEL_SCHED) {
2073 0 : level = LOGLEVEL_DEFAULT;
2074 0 : in_sched = true;
2075 : }
2076 :
2077 307 : boot_delay_msec(level);
2078 307 : printk_delay();
2079 :
2080 614 : printk_safe_enter_irqsave(flags);
2081 307 : printed_len = vprintk_store(facility, level, dev_info, fmt, args);
2082 307 : printk_safe_exit_irqrestore(flags);
2083 :
2084 : /* If called from the scheduler, we can not call up(). */
2085 307 : if (!in_sched) {
2086 : /*
2087 : * Disable preemption to avoid being preempted while holding
2088 : * console_sem which would prevent anyone from printing to
2089 : * console
2090 : */
2091 307 : preempt_disable();
2092 : /*
2093 : * Try to acquire and then immediately release the console
2094 : * semaphore. The release will print out buffers and wake up
2095 : * /dev/kmsg and syslog() users.
2096 : */
2097 307 : if (console_trylock_spinning())
2098 306 : console_unlock();
2099 307 : preempt_enable();
2100 : }
2101 :
2102 307 : wake_up_klogd();
2103 307 : return printed_len;
2104 : }
2105 : EXPORT_SYMBOL(vprintk_emit);
2106 :
2107 1 : asmlinkage int vprintk(const char *fmt, va_list args)
2108 : {
2109 1 : return vprintk_func(fmt, args);
2110 : }
2111 : EXPORT_SYMBOL(vprintk);
2112 :
2113 291 : int vprintk_default(const char *fmt, va_list args)
2114 : {
2115 291 : return vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, fmt, args);
2116 : }
2117 : EXPORT_SYMBOL_GPL(vprintk_default);
2118 :
2119 : /**
2120 : * printk - print a kernel message
2121 : * @fmt: format string
2122 : *
2123 : * This is printk(). It can be called from any context. We want it to work.
2124 : *
2125 : * We try to grab the console_lock. If we succeed, it's easy - we log the
2126 : * output and call the console drivers. If we fail to get the semaphore, we
2127 : * place the output into the log buffer and return. The current holder of
2128 : * the console_sem will notice the new output in console_unlock(); and will
2129 : * send it to the consoles before releasing the lock.
2130 : *
2131 : * One effect of this deferred printing is that code which calls printk() and
2132 : * then changes console_loglevel may break. This is because console_loglevel
2133 : * is inspected when the actual printing occurs.
2134 : *
2135 : * See also:
2136 : * printf(3)
2137 : *
2138 : * See the vsnprintf() documentation for format string extensions over C99.
2139 : */
2140 290 : asmlinkage __visible int printk(const char *fmt, ...)
2141 : {
2142 290 : va_list args;
2143 290 : int r;
2144 :
2145 290 : va_start(args, fmt);
2146 290 : r = vprintk_func(fmt, args);
2147 290 : va_end(args);
2148 :
2149 290 : return r;
2150 : }
2151 : EXPORT_SYMBOL(printk);
2152 :
2153 : #else /* CONFIG_PRINTK */
2154 :
2155 : #define LOG_LINE_MAX 0
2156 : #define PREFIX_MAX 0
2157 : #define printk_time false
2158 :
2159 : #define prb_read_valid(rb, seq, r) false
2160 : #define prb_first_valid_seq(rb) 0
2161 :
2162 : static u64 syslog_seq;
2163 : static u64 console_seq;
2164 : static u64 exclusive_console_stop_seq;
2165 : static unsigned long console_dropped;
2166 :
2167 : static size_t record_print_text(const struct printk_record *r,
2168 : bool syslog, bool time)
2169 : {
2170 : return 0;
2171 : }
2172 : static ssize_t info_print_ext_header(char *buf, size_t size,
2173 : struct printk_info *info)
2174 : {
2175 : return 0;
2176 : }
2177 : static ssize_t msg_print_ext_body(char *buf, size_t size,
2178 : char *text, size_t text_len,
2179 : struct dev_printk_info *dev_info) { return 0; }
2180 : static void console_lock_spinning_enable(void) { }
2181 : static int console_lock_spinning_disable_and_check(void) { return 0; }
2182 : static void call_console_drivers(const char *ext_text, size_t ext_len,
2183 : const char *text, size_t len) {}
2184 : static bool suppress_message_printing(int level) { return false; }
2185 :
2186 : #endif /* CONFIG_PRINTK */
2187 :
2188 : #ifdef CONFIG_EARLY_PRINTK
2189 : struct console *early_console;
2190 :
2191 0 : asmlinkage __visible void early_printk(const char *fmt, ...)
2192 : {
2193 0 : va_list ap;
2194 0 : char buf[512];
2195 0 : int n;
2196 :
2197 0 : if (!early_console)
2198 0 : return;
2199 :
2200 0 : va_start(ap, fmt);
2201 0 : n = vscnprintf(buf, sizeof(buf), fmt, ap);
2202 0 : va_end(ap);
2203 :
2204 0 : early_console->write(early_console, buf, n);
2205 : }
2206 : #endif
2207 :
2208 1 : static int __add_preferred_console(char *name, int idx, char *options,
2209 : char *brl_options, bool user_specified)
2210 : {
2211 1 : struct console_cmdline *c;
2212 1 : int i;
2213 :
2214 : /*
2215 : * See if this tty is not yet registered, and
2216 : * if we have a slot free.
2217 : */
2218 1 : for (i = 0, c = console_cmdline;
2219 1 : i < MAX_CMDLINECONSOLES && c->name[0];
2220 0 : i++, c++) {
2221 0 : if (strcmp(c->name, name) == 0 && c->index == idx) {
2222 0 : if (!brl_options)
2223 0 : preferred_console = i;
2224 0 : if (user_specified)
2225 0 : c->user_specified = true;
2226 0 : return 0;
2227 : }
2228 : }
2229 1 : if (i == MAX_CMDLINECONSOLES)
2230 : return -E2BIG;
2231 1 : if (!brl_options)
2232 1 : preferred_console = i;
2233 1 : strlcpy(c->name, name, sizeof(c->name));
2234 1 : c->options = options;
2235 1 : c->user_specified = user_specified;
2236 1 : braille_set_options(c, brl_options);
2237 :
2238 1 : c->index = idx;
2239 1 : return 0;
2240 : }
2241 :
2242 0 : static int __init console_msg_format_setup(char *str)
2243 : {
2244 0 : if (!strcmp(str, "syslog"))
2245 0 : console_msg_format = MSG_FORMAT_SYSLOG;
2246 0 : if (!strcmp(str, "default"))
2247 0 : console_msg_format = MSG_FORMAT_DEFAULT;
2248 0 : return 1;
2249 : }
2250 : __setup("console_msg_format=", console_msg_format_setup);
2251 :
2252 : /*
2253 : * Set up a console. Called via do_early_param() in init/main.c
2254 : * for each "console=" parameter in the boot command line.
2255 : */
2256 1 : static int __init console_setup(char *str)
2257 : {
2258 1 : char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */
2259 1 : char *s, *options, *brl_options = NULL;
2260 1 : int idx;
2261 :
2262 : /*
2263 : * console="" or console=null have been suggested as a way to
2264 : * disable console output. Use ttynull that has been created
2265 : * for exacly this purpose.
2266 : */
2267 1 : if (str[0] == 0 || strcmp(str, "null") == 0) {
2268 0 : __add_preferred_console("ttynull", 0, NULL, NULL, true);
2269 0 : return 1;
2270 : }
2271 :
2272 1 : if (_braille_console_setup(&str, &brl_options))
2273 : return 1;
2274 :
2275 : /*
2276 : * Decode str into name, index, options.
2277 : */
2278 1 : if (str[0] >= '0' && str[0] <= '9') {
2279 0 : strcpy(buf, "ttyS");
2280 0 : strncpy(buf + 4, str, sizeof(buf) - 5);
2281 : } else {
2282 1 : strncpy(buf, str, sizeof(buf) - 1);
2283 : }
2284 1 : buf[sizeof(buf) - 1] = 0;
2285 1 : options = strchr(str, ',');
2286 1 : if (options)
2287 0 : *(options++) = 0;
2288 : #ifdef __sparc__
2289 : if (!strcmp(str, "ttya"))
2290 : strcpy(buf, "ttyS0");
2291 : if (!strcmp(str, "ttyb"))
2292 : strcpy(buf, "ttyS1");
2293 : #endif
2294 5 : for (s = buf; *s; s++)
2295 5 : if (isdigit(*s) || *s == ',')
2296 : break;
2297 1 : idx = simple_strtoul(s, NULL, 10);
2298 1 : *s = 0;
2299 :
2300 1 : __add_preferred_console(buf, idx, options, brl_options, true);
2301 1 : console_set_on_cmdline = 1;
2302 1 : return 1;
2303 : }
2304 : __setup("console=", console_setup);
2305 :
2306 : /**
2307 : * add_preferred_console - add a device to the list of preferred consoles.
2308 : * @name: device name
2309 : * @idx: device index
2310 : * @options: options for this console
2311 : *
2312 : * The last preferred console added will be used for kernel messages
2313 : * and stdin/out/err for init. Normally this is used by console_setup
2314 : * above to handle user-supplied console arguments; however it can also
2315 : * be used by arch-specific code either to override the user or more
2316 : * commonly to provide a default console (ie from PROM variables) when
2317 : * the user has not supplied one.
2318 : */
2319 0 : int add_preferred_console(char *name, int idx, char *options)
2320 : {
2321 0 : return __add_preferred_console(name, idx, options, NULL, false);
2322 : }
2323 :
2324 : bool console_suspend_enabled = true;
2325 : EXPORT_SYMBOL(console_suspend_enabled);
2326 :
2327 0 : static int __init console_suspend_disable(char *str)
2328 : {
2329 0 : console_suspend_enabled = false;
2330 0 : return 1;
2331 : }
2332 : __setup("no_console_suspend", console_suspend_disable);
2333 : module_param_named(console_suspend, console_suspend_enabled,
2334 : bool, S_IRUGO | S_IWUSR);
2335 : MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
2336 : " and hibernate operations");
2337 :
2338 : /**
2339 : * suspend_console - suspend the console subsystem
2340 : *
2341 : * This disables printk() while we go into suspend states
2342 : */
2343 0 : void suspend_console(void)
2344 : {
2345 0 : if (!console_suspend_enabled)
2346 : return;
2347 0 : pr_info("Suspending console(s) (use no_console_suspend to debug)\n");
2348 0 : console_lock();
2349 0 : console_suspended = 1;
2350 0 : up_console_sem();
2351 : }
2352 :
2353 0 : void resume_console(void)
2354 : {
2355 0 : if (!console_suspend_enabled)
2356 : return;
2357 0 : down_console_sem();
2358 0 : console_suspended = 0;
2359 0 : console_unlock();
2360 : }
2361 :
2362 : /**
2363 : * console_cpu_notify - print deferred console messages after CPU hotplug
2364 : * @cpu: unused
2365 : *
2366 : * If printk() is called from a CPU that is not online yet, the messages
2367 : * will be printed on the console only if there are CON_ANYTIME consoles.
2368 : * This function is called when a new CPU comes online (or fails to come
2369 : * up) or goes offline.
2370 : */
2371 0 : static int console_cpu_notify(unsigned int cpu)
2372 : {
2373 0 : if (!cpuhp_tasks_frozen) {
2374 : /* If trylock fails, someone else is doing the printing */
2375 0 : if (console_trylock())
2376 0 : console_unlock();
2377 : }
2378 0 : return 0;
2379 : }
2380 :
2381 : /**
2382 : * console_lock - lock the console system for exclusive use.
2383 : *
2384 : * Acquires a lock which guarantees that the caller has
2385 : * exclusive access to the console system and the console_drivers list.
2386 : *
2387 : * Can sleep, returns nothing.
2388 : */
2389 216 : void console_lock(void)
2390 : {
2391 216 : might_sleep();
2392 :
2393 216 : down_console_sem();
2394 216 : if (console_suspended)
2395 : return;
2396 216 : console_locked = 1;
2397 216 : console_may_schedule = 1;
2398 : }
2399 : EXPORT_SYMBOL(console_lock);
2400 :
2401 : /**
2402 : * console_trylock - try to lock the console system for exclusive use.
2403 : *
2404 : * Try to acquire a lock which guarantees that the caller has exclusive
2405 : * access to the console system and the console_drivers list.
2406 : *
2407 : * returns 1 on success, and 0 on failure to acquire the lock.
2408 : */
2409 307 : int console_trylock(void)
2410 : {
2411 307 : if (down_trylock_console_sem())
2412 : return 0;
2413 306 : if (console_suspended) {
2414 0 : up_console_sem();
2415 0 : return 0;
2416 : }
2417 306 : console_locked = 1;
2418 306 : console_may_schedule = 0;
2419 306 : return 1;
2420 : }
2421 : EXPORT_SYMBOL(console_trylock);
2422 :
2423 128 : int is_console_locked(void)
2424 : {
2425 128 : return console_locked;
2426 : }
2427 : EXPORT_SYMBOL(is_console_locked);
2428 :
2429 : /*
2430 : * Check if we have any console that is capable of printing while cpu is
2431 : * booting or shutting down. Requires console_sem.
2432 : */
2433 6 : static int have_callable_console(void)
2434 : {
2435 6 : struct console *con;
2436 :
2437 6 : for_each_console(con)
2438 6 : if ((con->flags & CON_ENABLED) &&
2439 : (con->flags & CON_ANYTIME))
2440 : return 1;
2441 :
2442 : return 0;
2443 : }
2444 :
2445 : /*
2446 : * Can we actually use the console at this time on this cpu?
2447 : *
2448 : * Console drivers may assume that per-cpu resources have been allocated. So
2449 : * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2450 : * call them until this CPU is officially up.
2451 : */
2452 522 : static inline int can_use_console(void)
2453 : {
2454 528 : return cpu_online(raw_smp_processor_id()) || have_callable_console();
2455 : }
2456 :
2457 : /**
2458 : * console_unlock - unlock the console system
2459 : *
2460 : * Releases the console_lock which the caller holds on the console system
2461 : * and the console driver list.
2462 : *
2463 : * While the console_lock was held, console output may have been buffered
2464 : * by printk(). If this is the case, console_unlock(); emits
2465 : * the output prior to releasing the lock.
2466 : *
2467 : * If there is output waiting, we wake /dev/kmsg and syslog() users.
2468 : *
2469 : * console_unlock(); may be called from any context.
2470 : */
2471 522 : void console_unlock(void)
2472 : {
2473 522 : static char ext_text[CONSOLE_EXT_LOG_MAX];
2474 522 : static char text[LOG_LINE_MAX + PREFIX_MAX];
2475 522 : unsigned long flags;
2476 522 : bool do_cond_resched, retry;
2477 522 : struct printk_info info;
2478 522 : struct printk_record r;
2479 :
2480 522 : if (console_suspended) {
2481 0 : up_console_sem();
2482 0 : return;
2483 : }
2484 :
2485 522 : prb_rec_init_rd(&r, &info, text, sizeof(text));
2486 :
2487 : /*
2488 : * Console drivers are called with interrupts disabled, so
2489 : * @console_may_schedule should be cleared before; however, we may
2490 : * end up dumping a lot of lines, for example, if called from
2491 : * console registration path, and should invoke cond_resched()
2492 : * between lines if allowable. Not doing so can cause a very long
2493 : * scheduling stall on a slow console leading to RCU stall and
2494 : * softlockup warnings which exacerbate the issue with more
2495 : * messages practically incapacitating the system.
2496 : *
2497 : * console_trylock() is not able to detect the preemptive
2498 : * context reliably. Therefore the value must be stored before
2499 : * and cleared after the "again" goto label.
2500 : */
2501 522 : do_cond_resched = console_may_schedule;
2502 522 : again:
2503 522 : console_may_schedule = 0;
2504 :
2505 : /*
2506 : * We released the console_sem lock, so we need to recheck if
2507 : * cpu is online and (if not) is there at least one CON_ANYTIME
2508 : * console.
2509 : */
2510 522 : if (!can_use_console()) {
2511 0 : console_locked = 0;
2512 0 : up_console_sem();
2513 0 : return;
2514 : }
2515 :
2516 858 : for (;;) {
2517 858 : size_t ext_len = 0;
2518 858 : size_t len;
2519 :
2520 1716 : printk_safe_enter_irqsave(flags);
2521 858 : raw_spin_lock(&logbuf_lock);
2522 879 : skip:
2523 879 : if (!prb_read_valid(prb, console_seq, &r))
2524 : break;
2525 :
2526 357 : if (console_seq != r.info->seq) {
2527 0 : console_dropped += r.info->seq - console_seq;
2528 0 : console_seq = r.info->seq;
2529 : }
2530 :
2531 714 : if (suppress_message_printing(r.info->level)) {
2532 : /*
2533 : * Skip record we have buffered and already printed
2534 : * directly to the console when we received it, and
2535 : * record that has level above the console loglevel.
2536 : */
2537 21 : console_seq++;
2538 21 : goto skip;
2539 : }
2540 :
2541 : /* Output to all consoles once old messages replayed. */
2542 336 : if (unlikely(exclusive_console &&
2543 : console_seq >= exclusive_console_stop_seq)) {
2544 1 : exclusive_console = NULL;
2545 : }
2546 :
2547 : /*
2548 : * Handle extended console text first because later
2549 : * record_print_text() will modify the record buffer in-place.
2550 : */
2551 336 : if (nr_ext_console_drivers) {
2552 0 : ext_len = info_print_ext_header(ext_text,
2553 : sizeof(ext_text),
2554 : r.info);
2555 0 : ext_len += msg_print_ext_body(ext_text + ext_len,
2556 : sizeof(ext_text) - ext_len,
2557 : &r.text_buf[0],
2558 0 : r.info->text_len,
2559 0 : &r.info->dev_info);
2560 : }
2561 672 : len = record_print_text(&r,
2562 336 : console_msg_format & MSG_FORMAT_SYSLOG,
2563 : printk_time);
2564 336 : console_seq++;
2565 336 : raw_spin_unlock(&logbuf_lock);
2566 :
2567 : /*
2568 : * While actively printing out messages, if another printk()
2569 : * were to occur on another CPU, it may wait for this one to
2570 : * finish. This task can not be preempted if there is a
2571 : * waiter waiting to take over.
2572 : */
2573 336 : console_lock_spinning_enable();
2574 :
2575 336 : stop_critical_timings(); /* don't trace print latency */
2576 336 : call_console_drivers(ext_text, ext_len, text, len);
2577 336 : start_critical_timings();
2578 :
2579 336 : if (console_lock_spinning_disable_and_check()) {
2580 0 : printk_safe_exit_irqrestore(flags);
2581 0 : return;
2582 : }
2583 :
2584 336 : printk_safe_exit_irqrestore(flags);
2585 :
2586 336 : if (do_cond_resched)
2587 81 : cond_resched();
2588 : }
2589 :
2590 522 : console_locked = 0;
2591 :
2592 522 : raw_spin_unlock(&logbuf_lock);
2593 :
2594 522 : up_console_sem();
2595 :
2596 : /*
2597 : * Someone could have filled up the buffer again, so re-check if there's
2598 : * something to flush. In case we cannot trylock the console_sem again,
2599 : * there's a new owner and the console_unlock() from them will do the
2600 : * flush, no worries.
2601 : */
2602 522 : raw_spin_lock(&logbuf_lock);
2603 522 : retry = prb_read_valid(prb, console_seq, NULL);
2604 522 : raw_spin_unlock(&logbuf_lock);
2605 522 : printk_safe_exit_irqrestore(flags);
2606 :
2607 522 : if (retry && console_trylock())
2608 0 : goto again;
2609 : }
2610 : EXPORT_SYMBOL(console_unlock);
2611 :
2612 : /**
2613 : * console_conditional_schedule - yield the CPU if required
2614 : *
2615 : * If the console code is currently allowed to sleep, and
2616 : * if this CPU should yield the CPU to another task, do
2617 : * so here.
2618 : *
2619 : * Must be called within console_lock();.
2620 : */
2621 16 : void __sched console_conditional_schedule(void)
2622 : {
2623 16 : if (console_may_schedule)
2624 16 : cond_resched();
2625 16 : }
2626 : EXPORT_SYMBOL(console_conditional_schedule);
2627 :
2628 0 : void console_unblank(void)
2629 : {
2630 0 : struct console *c;
2631 :
2632 : /*
2633 : * console_unblank can no longer be called in interrupt context unless
2634 : * oops_in_progress is set to 1..
2635 : */
2636 0 : if (oops_in_progress) {
2637 0 : if (down_trylock_console_sem() != 0)
2638 : return;
2639 : } else
2640 0 : console_lock();
2641 :
2642 0 : console_locked = 1;
2643 0 : console_may_schedule = 0;
2644 0 : for_each_console(c)
2645 0 : if ((c->flags & CON_ENABLED) && c->unblank)
2646 0 : c->unblank();
2647 0 : console_unlock();
2648 : }
2649 :
2650 : /**
2651 : * console_flush_on_panic - flush console content on panic
2652 : * @mode: flush all messages in buffer or just the pending ones
2653 : *
2654 : * Immediately output all pending messages no matter what.
2655 : */
2656 0 : void console_flush_on_panic(enum con_flush_mode mode)
2657 : {
2658 : /*
2659 : * If someone else is holding the console lock, trylock will fail
2660 : * and may_schedule may be set. Ignore and proceed to unlock so
2661 : * that messages are flushed out. As this can be called from any
2662 : * context and we don't want to get preempted while flushing,
2663 : * ensure may_schedule is cleared.
2664 : */
2665 0 : console_trylock();
2666 0 : console_may_schedule = 0;
2667 :
2668 0 : if (mode == CONSOLE_REPLAY_ALL) {
2669 0 : unsigned long flags;
2670 :
2671 0 : logbuf_lock_irqsave(flags);
2672 0 : console_seq = prb_first_valid_seq(prb);
2673 0 : logbuf_unlock_irqrestore(flags);
2674 : }
2675 0 : console_unlock();
2676 0 : }
2677 :
2678 : /*
2679 : * Return the console tty driver structure and its associated index
2680 : */
2681 110 : struct tty_driver *console_device(int *index)
2682 : {
2683 110 : struct console *c;
2684 110 : struct tty_driver *driver = NULL;
2685 :
2686 110 : console_lock();
2687 110 : for_each_console(c) {
2688 110 : if (!c->device)
2689 0 : continue;
2690 110 : driver = c->device(c, index);
2691 110 : if (driver)
2692 : break;
2693 : }
2694 110 : console_unlock();
2695 110 : return driver;
2696 : }
2697 :
2698 : /*
2699 : * Prevent further output on the passed console device so that (for example)
2700 : * serial drivers can disable console output before suspending a port, and can
2701 : * re-enable output afterwards.
2702 : */
2703 0 : void console_stop(struct console *console)
2704 : {
2705 0 : console_lock();
2706 0 : console->flags &= ~CON_ENABLED;
2707 0 : console_unlock();
2708 0 : }
2709 : EXPORT_SYMBOL(console_stop);
2710 :
2711 0 : void console_start(struct console *console)
2712 : {
2713 0 : console_lock();
2714 0 : console->flags |= CON_ENABLED;
2715 0 : console_unlock();
2716 0 : }
2717 : EXPORT_SYMBOL(console_start);
2718 :
2719 : static int __read_mostly keep_bootcon;
2720 :
2721 0 : static int __init keep_bootcon_setup(char *str)
2722 : {
2723 0 : keep_bootcon = 1;
2724 0 : pr_info("debug: skip boot console de-registration.\n");
2725 :
2726 0 : return 0;
2727 : }
2728 :
2729 : early_param("keep_bootcon", keep_bootcon_setup);
2730 :
2731 : /*
2732 : * This is called by register_console() to try to match
2733 : * the newly registered console with any of the ones selected
2734 : * by either the command line or add_preferred_console() and
2735 : * setup/enable it.
2736 : *
2737 : * Care need to be taken with consoles that are statically
2738 : * enabled such as netconsole
2739 : */
2740 5 : static int try_enable_new_console(struct console *newcon, bool user_specified)
2741 : {
2742 5 : struct console_cmdline *c;
2743 5 : int i, err;
2744 :
2745 5 : for (i = 0, c = console_cmdline;
2746 9 : i < MAX_CMDLINECONSOLES && c->name[0];
2747 4 : i++, c++) {
2748 5 : if (c->user_specified != user_specified)
2749 2 : continue;
2750 4 : if (!newcon->match ||
2751 1 : newcon->match(newcon, c->name, c->index, c->options) != 0) {
2752 : /* default matching */
2753 3 : BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name));
2754 3 : if (strcmp(c->name, newcon->name) != 0)
2755 2 : continue;
2756 1 : if (newcon->index >= 0 &&
2757 0 : newcon->index != c->index)
2758 0 : continue;
2759 1 : if (newcon->index < 0)
2760 1 : newcon->index = c->index;
2761 :
2762 1 : if (_braille_register_console(newcon, c))
2763 : return 0;
2764 :
2765 1 : if (newcon->setup &&
2766 1 : (err = newcon->setup(newcon, c->options)) != 0)
2767 : return err;
2768 : }
2769 1 : newcon->flags |= CON_ENABLED;
2770 1 : if (i == preferred_console) {
2771 1 : newcon->flags |= CON_CONSDEV;
2772 1 : has_preferred_console = true;
2773 : }
2774 : return 0;
2775 : }
2776 :
2777 : /*
2778 : * Some consoles, such as pstore and netconsole, can be enabled even
2779 : * without matching. Accept the pre-enabled consoles only when match()
2780 : * and setup() had a chance to be called.
2781 : */
2782 4 : if (newcon->flags & CON_ENABLED && c->user_specified == user_specified)
2783 0 : return 0;
2784 :
2785 : return -ENOENT;
2786 : }
2787 :
2788 : /*
2789 : * The console driver calls this routine during kernel initialization
2790 : * to register the console printing procedure with printk() and to
2791 : * print any messages that were printed by the kernel before the
2792 : * console driver was initialized.
2793 : *
2794 : * This can happen pretty early during the boot process (because of
2795 : * early_printk) - sometimes before setup_arch() completes - be careful
2796 : * of what kernel features are used - they may not be initialised yet.
2797 : *
2798 : * There are two types of consoles - bootconsoles (early_printk) and
2799 : * "real" consoles (everything which is not a bootconsole) which are
2800 : * handled differently.
2801 : * - Any number of bootconsoles can be registered at any time.
2802 : * - As soon as a "real" console is registered, all bootconsoles
2803 : * will be unregistered automatically.
2804 : * - Once a "real" console is registered, any attempt to register a
2805 : * bootconsoles will be rejected
2806 : */
2807 3 : void register_console(struct console *newcon)
2808 : {
2809 3 : unsigned long flags;
2810 3 : struct console *bcon = NULL;
2811 3 : int err;
2812 :
2813 3 : for_each_console(bcon) {
2814 0 : if (WARN(bcon == newcon, "console '%s%d' already registered\n",
2815 : bcon->name, bcon->index))
2816 : return;
2817 : }
2818 :
2819 : /*
2820 : * before we register a new CON_BOOT console, make sure we don't
2821 : * already have a valid console
2822 : */
2823 3 : if (newcon->flags & CON_BOOT) {
2824 0 : for_each_console(bcon) {
2825 0 : if (!(bcon->flags & CON_BOOT)) {
2826 0 : pr_info("Too late to register bootconsole %s%d\n",
2827 : newcon->name, newcon->index);
2828 0 : return;
2829 : }
2830 : }
2831 : }
2832 :
2833 3 : if (console_drivers && console_drivers->flags & CON_BOOT)
2834 0 : bcon = console_drivers;
2835 :
2836 3 : if (!has_preferred_console || bcon || !console_drivers)
2837 3 : has_preferred_console = preferred_console >= 0;
2838 :
2839 : /*
2840 : * See if we want to use this console driver. If we
2841 : * didn't select a console we take the first one
2842 : * that registers here.
2843 : */
2844 3 : if (!has_preferred_console) {
2845 0 : if (newcon->index < 0)
2846 0 : newcon->index = 0;
2847 0 : if (newcon->setup == NULL ||
2848 0 : newcon->setup(newcon, NULL) == 0) {
2849 0 : newcon->flags |= CON_ENABLED;
2850 0 : if (newcon->device) {
2851 0 : newcon->flags |= CON_CONSDEV;
2852 0 : has_preferred_console = true;
2853 : }
2854 : }
2855 : }
2856 :
2857 : /* See if this console matches one we selected on the command line */
2858 3 : err = try_enable_new_console(newcon, true);
2859 :
2860 : /* If not, try to match against the platform default(s) */
2861 3 : if (err == -ENOENT)
2862 2 : err = try_enable_new_console(newcon, false);
2863 :
2864 : /* printk() messages are not printed to the Braille console. */
2865 3 : if (err || newcon->flags & CON_BRL)
2866 : return;
2867 :
2868 : /*
2869 : * If we have a bootconsole, and are switching to a real console,
2870 : * don't print everything out again, since when the boot console, and
2871 : * the real console are the same physical device, it's annoying to
2872 : * see the beginning boot messages twice
2873 : */
2874 1 : if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2875 0 : newcon->flags &= ~CON_PRINTBUFFER;
2876 :
2877 : /*
2878 : * Put this console in the list - keep the
2879 : * preferred driver at the head of the list.
2880 : */
2881 1 : console_lock();
2882 1 : if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2883 1 : newcon->next = console_drivers;
2884 1 : console_drivers = newcon;
2885 1 : if (newcon->next)
2886 0 : newcon->next->flags &= ~CON_CONSDEV;
2887 : /* Ensure this flag is always set for the head of the list */
2888 1 : newcon->flags |= CON_CONSDEV;
2889 : } else {
2890 0 : newcon->next = console_drivers->next;
2891 0 : console_drivers->next = newcon;
2892 : }
2893 :
2894 1 : if (newcon->flags & CON_EXTENDED)
2895 0 : nr_ext_console_drivers++;
2896 :
2897 1 : if (newcon->flags & CON_PRINTBUFFER) {
2898 : /*
2899 : * console_unlock(); will print out the buffered messages
2900 : * for us.
2901 : */
2902 2 : logbuf_lock_irqsave(flags);
2903 : /*
2904 : * We're about to replay the log buffer. Only do this to the
2905 : * just-registered console to avoid excessive message spam to
2906 : * the already-registered consoles.
2907 : *
2908 : * Set exclusive_console with disabled interrupts to reduce
2909 : * race window with eventual console_flush_on_panic() that
2910 : * ignores console_lock.
2911 : */
2912 1 : exclusive_console = newcon;
2913 1 : exclusive_console_stop_seq = console_seq;
2914 1 : console_seq = syslog_seq;
2915 1 : logbuf_unlock_irqrestore(flags);
2916 : }
2917 1 : console_unlock();
2918 1 : console_sysfs_notify();
2919 :
2920 : /*
2921 : * By unregistering the bootconsoles after we enable the real console
2922 : * we get the "console xxx enabled" message on all the consoles -
2923 : * boot consoles, real consoles, etc - this is to ensure that end
2924 : * users know there might be something in the kernel's log buffer that
2925 : * went to the bootconsole (that they do not see on the real console)
2926 : */
2927 2 : pr_info("%sconsole [%s%d] enabled\n",
2928 : (newcon->flags & CON_BOOT) ? "boot" : "" ,
2929 : newcon->name, newcon->index);
2930 1 : if (bcon &&
2931 0 : ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2932 0 : !keep_bootcon) {
2933 : /* We need to iterate through all boot consoles, to make
2934 : * sure we print everything out, before we unregister them.
2935 : */
2936 0 : for_each_console(bcon)
2937 0 : if (bcon->flags & CON_BOOT)
2938 0 : unregister_console(bcon);
2939 : }
2940 : }
2941 : EXPORT_SYMBOL(register_console);
2942 :
2943 0 : int unregister_console(struct console *console)
2944 : {
2945 0 : struct console *con;
2946 0 : int res;
2947 :
2948 0 : pr_info("%sconsole [%s%d] disabled\n",
2949 : (console->flags & CON_BOOT) ? "boot" : "" ,
2950 : console->name, console->index);
2951 :
2952 0 : res = _braille_unregister_console(console);
2953 0 : if (res < 0)
2954 : return res;
2955 0 : if (res > 0)
2956 : return 0;
2957 :
2958 0 : res = -ENODEV;
2959 0 : console_lock();
2960 0 : if (console_drivers == console) {
2961 0 : console_drivers=console->next;
2962 0 : res = 0;
2963 : } else {
2964 0 : for_each_console(con) {
2965 0 : if (con->next == console) {
2966 0 : con->next = console->next;
2967 0 : res = 0;
2968 0 : break;
2969 : }
2970 : }
2971 : }
2972 :
2973 0 : if (res)
2974 0 : goto out_disable_unlock;
2975 :
2976 0 : if (console->flags & CON_EXTENDED)
2977 0 : nr_ext_console_drivers--;
2978 :
2979 : /*
2980 : * If this isn't the last console and it has CON_CONSDEV set, we
2981 : * need to set it on the next preferred console.
2982 : */
2983 0 : if (console_drivers != NULL && console->flags & CON_CONSDEV)
2984 0 : console_drivers->flags |= CON_CONSDEV;
2985 :
2986 0 : console->flags &= ~CON_ENABLED;
2987 0 : console_unlock();
2988 0 : console_sysfs_notify();
2989 :
2990 0 : if (console->exit)
2991 0 : res = console->exit(console);
2992 :
2993 : return res;
2994 :
2995 0 : out_disable_unlock:
2996 0 : console->flags &= ~CON_ENABLED;
2997 0 : console_unlock();
2998 :
2999 0 : return res;
3000 : }
3001 : EXPORT_SYMBOL(unregister_console);
3002 :
3003 : /*
3004 : * Initialize the console device. This is called *early*, so
3005 : * we can't necessarily depend on lots of kernel help here.
3006 : * Just do some early initializations, and do the complex setup
3007 : * later.
3008 : */
3009 1 : void __init console_init(void)
3010 : {
3011 1 : int ret;
3012 1 : initcall_t call;
3013 1 : initcall_entry_t *ce;
3014 :
3015 : /* Setup the default TTY line discipline. */
3016 1 : n_tty_init();
3017 :
3018 : /*
3019 : * set up the console device so that later boot sequences can
3020 : * inform about problems etc..
3021 : */
3022 1 : ce = __con_initcall_start;
3023 1 : trace_initcall_level("console");
3024 4 : while (ce < __con_initcall_end) {
3025 3 : call = initcall_from_entry(ce);
3026 3 : trace_initcall_start(call);
3027 3 : ret = call();
3028 3 : trace_initcall_finish(call, ret);
3029 3 : ce++;
3030 : }
3031 1 : }
3032 :
3033 : /*
3034 : * Some boot consoles access data that is in the init section and which will
3035 : * be discarded after the initcalls have been run. To make sure that no code
3036 : * will access this data, unregister the boot consoles in a late initcall.
3037 : *
3038 : * If for some reason, such as deferred probe or the driver being a loadable
3039 : * module, the real console hasn't registered yet at this point, there will
3040 : * be a brief interval in which no messages are logged to the console, which
3041 : * makes it difficult to diagnose problems that occur during this time.
3042 : *
3043 : * To mitigate this problem somewhat, only unregister consoles whose memory
3044 : * intersects with the init section. Note that all other boot consoles will
3045 : * get unregistred when the real preferred console is registered.
3046 : */
3047 1 : static int __init printk_late_init(void)
3048 : {
3049 1 : struct console *con;
3050 1 : int ret;
3051 :
3052 2 : for_each_console(con) {
3053 1 : if (!(con->flags & CON_BOOT))
3054 1 : continue;
3055 :
3056 : /* Check addresses that might be used for enabled consoles. */
3057 0 : if (init_section_intersects(con, sizeof(*con)) ||
3058 0 : init_section_contains(con->write, 0) ||
3059 0 : init_section_contains(con->read, 0) ||
3060 0 : init_section_contains(con->device, 0) ||
3061 0 : init_section_contains(con->unblank, 0) ||
3062 0 : init_section_contains(con->data, 0)) {
3063 : /*
3064 : * Please, consider moving the reported consoles out
3065 : * of the init section.
3066 : */
3067 0 : pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n",
3068 : con->name, con->index);
3069 0 : unregister_console(con);
3070 : }
3071 : }
3072 1 : ret = cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD, "printk:dead", NULL,
3073 : console_cpu_notify);
3074 1 : WARN_ON(ret < 0);
3075 1 : ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "printk:online",
3076 : console_cpu_notify, NULL);
3077 1 : WARN_ON(ret < 0);
3078 1 : return 0;
3079 : }
3080 : late_initcall(printk_late_init);
3081 :
3082 : #if defined CONFIG_PRINTK
3083 : /*
3084 : * Delayed printk version, for scheduler-internal messages:
3085 : */
3086 : #define PRINTK_PENDING_WAKEUP 0x01
3087 : #define PRINTK_PENDING_OUTPUT 0x02
3088 :
3089 : static DEFINE_PER_CPU(int, printk_pending);
3090 :
3091 38 : static void wake_up_klogd_work_func(struct irq_work *irq_work)
3092 : {
3093 38 : int pending = __this_cpu_xchg(printk_pending, 0);
3094 :
3095 38 : if (pending & PRINTK_PENDING_OUTPUT) {
3096 : /* If trylock fails, someone else is doing the printing */
3097 0 : if (console_trylock())
3098 0 : console_unlock();
3099 : }
3100 :
3101 38 : if (pending & PRINTK_PENDING_WAKEUP)
3102 38 : wake_up_interruptible(&log_wait);
3103 38 : }
3104 :
3105 : static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) =
3106 : IRQ_WORK_INIT_LAZY(wake_up_klogd_work_func);
3107 :
3108 307 : void wake_up_klogd(void)
3109 : {
3110 307 : if (!printk_percpu_data_ready())
3111 : return;
3112 :
3113 207 : preempt_disable();
3114 207 : if (waitqueue_active(&log_wait)) {
3115 69 : this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
3116 69 : irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
3117 : }
3118 207 : preempt_enable();
3119 : }
3120 :
3121 0 : void defer_console_output(void)
3122 : {
3123 0 : if (!printk_percpu_data_ready())
3124 : return;
3125 :
3126 0 : preempt_disable();
3127 0 : __this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
3128 0 : irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
3129 0 : preempt_enable();
3130 : }
3131 :
3132 0 : int vprintk_deferred(const char *fmt, va_list args)
3133 : {
3134 0 : int r;
3135 :
3136 0 : r = vprintk_emit(0, LOGLEVEL_SCHED, NULL, fmt, args);
3137 0 : defer_console_output();
3138 :
3139 0 : return r;
3140 : }
3141 :
3142 0 : int printk_deferred(const char *fmt, ...)
3143 : {
3144 0 : va_list args;
3145 0 : int r;
3146 :
3147 0 : va_start(args, fmt);
3148 0 : r = vprintk_deferred(fmt, args);
3149 0 : va_end(args);
3150 :
3151 0 : return r;
3152 : }
3153 :
3154 : /*
3155 : * printk rate limiting, lifted from the networking subsystem.
3156 : *
3157 : * This enforces a rate limit: not more than 10 kernel messages
3158 : * every 5s to make a denial-of-service attack impossible.
3159 : */
3160 : DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
3161 :
3162 0 : int __printk_ratelimit(const char *func)
3163 : {
3164 0 : return ___ratelimit(&printk_ratelimit_state, func);
3165 : }
3166 : EXPORT_SYMBOL(__printk_ratelimit);
3167 :
3168 : /**
3169 : * printk_timed_ratelimit - caller-controlled printk ratelimiting
3170 : * @caller_jiffies: pointer to caller's state
3171 : * @interval_msecs: minimum interval between prints
3172 : *
3173 : * printk_timed_ratelimit() returns true if more than @interval_msecs
3174 : * milliseconds have elapsed since the last time printk_timed_ratelimit()
3175 : * returned true.
3176 : */
3177 0 : bool printk_timed_ratelimit(unsigned long *caller_jiffies,
3178 : unsigned int interval_msecs)
3179 : {
3180 0 : unsigned long elapsed = jiffies - *caller_jiffies;
3181 :
3182 0 : if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs))
3183 : return false;
3184 :
3185 0 : *caller_jiffies = jiffies;
3186 0 : return true;
3187 : }
3188 : EXPORT_SYMBOL(printk_timed_ratelimit);
3189 :
3190 : static DEFINE_SPINLOCK(dump_list_lock);
3191 : static LIST_HEAD(dump_list);
3192 :
3193 : /**
3194 : * kmsg_dump_register - register a kernel log dumper.
3195 : * @dumper: pointer to the kmsg_dumper structure
3196 : *
3197 : * Adds a kernel log dumper to the system. The dump callback in the
3198 : * structure will be called when the kernel oopses or panics and must be
3199 : * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
3200 : */
3201 0 : int kmsg_dump_register(struct kmsg_dumper *dumper)
3202 : {
3203 0 : unsigned long flags;
3204 0 : int err = -EBUSY;
3205 :
3206 : /* The dump callback needs to be set */
3207 0 : if (!dumper->dump)
3208 : return -EINVAL;
3209 :
3210 0 : spin_lock_irqsave(&dump_list_lock, flags);
3211 : /* Don't allow registering multiple times */
3212 0 : if (!dumper->registered) {
3213 0 : dumper->registered = 1;
3214 0 : list_add_tail_rcu(&dumper->list, &dump_list);
3215 0 : err = 0;
3216 : }
3217 0 : spin_unlock_irqrestore(&dump_list_lock, flags);
3218 :
3219 0 : return err;
3220 : }
3221 : EXPORT_SYMBOL_GPL(kmsg_dump_register);
3222 :
3223 : /**
3224 : * kmsg_dump_unregister - unregister a kmsg dumper.
3225 : * @dumper: pointer to the kmsg_dumper structure
3226 : *
3227 : * Removes a dump device from the system. Returns zero on success and
3228 : * %-EINVAL otherwise.
3229 : */
3230 0 : int kmsg_dump_unregister(struct kmsg_dumper *dumper)
3231 : {
3232 0 : unsigned long flags;
3233 0 : int err = -EINVAL;
3234 :
3235 0 : spin_lock_irqsave(&dump_list_lock, flags);
3236 0 : if (dumper->registered) {
3237 0 : dumper->registered = 0;
3238 0 : list_del_rcu(&dumper->list);
3239 0 : err = 0;
3240 : }
3241 0 : spin_unlock_irqrestore(&dump_list_lock, flags);
3242 0 : synchronize_rcu();
3243 :
3244 0 : return err;
3245 : }
3246 : EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
3247 :
3248 : static bool always_kmsg_dump;
3249 : module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
3250 :
3251 0 : const char *kmsg_dump_reason_str(enum kmsg_dump_reason reason)
3252 : {
3253 0 : switch (reason) {
3254 : case KMSG_DUMP_PANIC:
3255 : return "Panic";
3256 : case KMSG_DUMP_OOPS:
3257 : return "Oops";
3258 : case KMSG_DUMP_EMERG:
3259 : return "Emergency";
3260 : case KMSG_DUMP_SHUTDOWN:
3261 : return "Shutdown";
3262 : default:
3263 : return "Unknown";
3264 : }
3265 : }
3266 : EXPORT_SYMBOL_GPL(kmsg_dump_reason_str);
3267 :
3268 : /**
3269 : * kmsg_dump - dump kernel log to kernel message dumpers.
3270 : * @reason: the reason (oops, panic etc) for dumping
3271 : *
3272 : * Call each of the registered dumper's dump() callback, which can
3273 : * retrieve the kmsg records with kmsg_dump_get_line() or
3274 : * kmsg_dump_get_buffer().
3275 : */
3276 0 : void kmsg_dump(enum kmsg_dump_reason reason)
3277 : {
3278 0 : struct kmsg_dumper *dumper;
3279 0 : unsigned long flags;
3280 :
3281 0 : rcu_read_lock();
3282 0 : list_for_each_entry_rcu(dumper, &dump_list, list) {
3283 0 : enum kmsg_dump_reason max_reason = dumper->max_reason;
3284 :
3285 : /*
3286 : * If client has not provided a specific max_reason, default
3287 : * to KMSG_DUMP_OOPS, unless always_kmsg_dump was set.
3288 : */
3289 0 : if (max_reason == KMSG_DUMP_UNDEF) {
3290 0 : max_reason = always_kmsg_dump ? KMSG_DUMP_MAX :
3291 : KMSG_DUMP_OOPS;
3292 : }
3293 0 : if (reason > max_reason)
3294 0 : continue;
3295 :
3296 : /* initialize iterator with data about the stored records */
3297 0 : dumper->active = true;
3298 :
3299 0 : logbuf_lock_irqsave(flags);
3300 0 : dumper->cur_seq = clear_seq;
3301 0 : dumper->next_seq = prb_next_seq(prb);
3302 0 : logbuf_unlock_irqrestore(flags);
3303 :
3304 : /* invoke dumper which will iterate over records */
3305 0 : dumper->dump(dumper, reason);
3306 :
3307 : /* reset iterator */
3308 0 : dumper->active = false;
3309 : }
3310 0 : rcu_read_unlock();
3311 0 : }
3312 :
3313 : /**
3314 : * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3315 : * @dumper: registered kmsg dumper
3316 : * @syslog: include the "<4>" prefixes
3317 : * @line: buffer to copy the line to
3318 : * @size: maximum size of the buffer
3319 : * @len: length of line placed into buffer
3320 : *
3321 : * Start at the beginning of the kmsg buffer, with the oldest kmsg
3322 : * record, and copy one record into the provided buffer.
3323 : *
3324 : * Consecutive calls will return the next available record moving
3325 : * towards the end of the buffer with the youngest messages.
3326 : *
3327 : * A return value of FALSE indicates that there are no more records to
3328 : * read.
3329 : *
3330 : * The function is similar to kmsg_dump_get_line(), but grabs no locks.
3331 : */
3332 0 : bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
3333 : char *line, size_t size, size_t *len)
3334 : {
3335 0 : struct printk_info info;
3336 0 : unsigned int line_count;
3337 0 : struct printk_record r;
3338 0 : size_t l = 0;
3339 0 : bool ret = false;
3340 :
3341 0 : prb_rec_init_rd(&r, &info, line, size);
3342 :
3343 0 : if (!dumper->active)
3344 0 : goto out;
3345 :
3346 : /* Read text or count text lines? */
3347 0 : if (line) {
3348 0 : if (!prb_read_valid(prb, dumper->cur_seq, &r))
3349 0 : goto out;
3350 0 : l = record_print_text(&r, syslog, printk_time);
3351 : } else {
3352 0 : if (!prb_read_valid_info(prb, dumper->cur_seq,
3353 : &info, &line_count)) {
3354 0 : goto out;
3355 : }
3356 0 : l = get_record_print_text_size(&info, line_count, syslog,
3357 : printk_time);
3358 :
3359 : }
3360 :
3361 0 : dumper->cur_seq = r.info->seq + 1;
3362 0 : ret = true;
3363 0 : out:
3364 0 : if (len)
3365 0 : *len = l;
3366 0 : return ret;
3367 : }
3368 :
3369 : /**
3370 : * kmsg_dump_get_line - retrieve one kmsg log line
3371 : * @dumper: registered kmsg dumper
3372 : * @syslog: include the "<4>" prefixes
3373 : * @line: buffer to copy the line to
3374 : * @size: maximum size of the buffer
3375 : * @len: length of line placed into buffer
3376 : *
3377 : * Start at the beginning of the kmsg buffer, with the oldest kmsg
3378 : * record, and copy one record into the provided buffer.
3379 : *
3380 : * Consecutive calls will return the next available record moving
3381 : * towards the end of the buffer with the youngest messages.
3382 : *
3383 : * A return value of FALSE indicates that there are no more records to
3384 : * read.
3385 : */
3386 0 : bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
3387 : char *line, size_t size, size_t *len)
3388 : {
3389 0 : unsigned long flags;
3390 0 : bool ret;
3391 :
3392 0 : logbuf_lock_irqsave(flags);
3393 0 : ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
3394 0 : logbuf_unlock_irqrestore(flags);
3395 :
3396 0 : return ret;
3397 : }
3398 : EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
3399 :
3400 : /**
3401 : * kmsg_dump_get_buffer - copy kmsg log lines
3402 : * @dumper: registered kmsg dumper
3403 : * @syslog: include the "<4>" prefixes
3404 : * @buf: buffer to copy the line to
3405 : * @size: maximum size of the buffer
3406 : * @len: length of line placed into buffer
3407 : *
3408 : * Start at the end of the kmsg buffer and fill the provided buffer
3409 : * with as many of the *youngest* kmsg records that fit into it.
3410 : * If the buffer is large enough, all available kmsg records will be
3411 : * copied with a single call.
3412 : *
3413 : * Consecutive calls will fill the buffer with the next block of
3414 : * available older records, not including the earlier retrieved ones.
3415 : *
3416 : * A return value of FALSE indicates that there are no more records to
3417 : * read.
3418 : */
3419 0 : bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
3420 : char *buf, size_t size, size_t *len)
3421 : {
3422 0 : struct printk_info info;
3423 0 : unsigned int line_count;
3424 0 : struct printk_record r;
3425 0 : unsigned long flags;
3426 0 : u64 seq;
3427 0 : u64 next_seq;
3428 0 : size_t l = 0;
3429 0 : bool ret = false;
3430 0 : bool time = printk_time;
3431 :
3432 0 : prb_rec_init_rd(&r, &info, buf, size);
3433 :
3434 0 : if (!dumper->active || !buf || !size)
3435 0 : goto out;
3436 :
3437 0 : logbuf_lock_irqsave(flags);
3438 0 : if (prb_read_valid_info(prb, dumper->cur_seq, &info, NULL)) {
3439 0 : if (info.seq != dumper->cur_seq) {
3440 : /* messages are gone, move to first available one */
3441 0 : dumper->cur_seq = info.seq;
3442 : }
3443 : }
3444 :
3445 : /* last entry */
3446 0 : if (dumper->cur_seq >= dumper->next_seq) {
3447 0 : logbuf_unlock_irqrestore(flags);
3448 0 : goto out;
3449 : }
3450 :
3451 : /* calculate length of entire buffer */
3452 : seq = dumper->cur_seq;
3453 0 : while (prb_read_valid_info(prb, seq, &info, &line_count)) {
3454 0 : if (r.info->seq >= dumper->next_seq)
3455 : break;
3456 0 : l += get_record_print_text_size(&info, line_count, syslog, time);
3457 0 : seq = r.info->seq + 1;
3458 : }
3459 :
3460 : /* move first record forward until length fits into the buffer */
3461 0 : seq = dumper->cur_seq;
3462 0 : while (l >= size && prb_read_valid_info(prb, seq,
3463 : &info, &line_count)) {
3464 0 : if (r.info->seq >= dumper->next_seq)
3465 : break;
3466 0 : l -= get_record_print_text_size(&info, line_count, syslog, time);
3467 0 : seq = r.info->seq + 1;
3468 : }
3469 :
3470 : /* last message in next interation */
3471 0 : next_seq = seq;
3472 :
3473 : /* actually read text into the buffer now */
3474 0 : l = 0;
3475 0 : while (prb_read_valid(prb, seq, &r)) {
3476 0 : if (r.info->seq >= dumper->next_seq)
3477 : break;
3478 :
3479 0 : l += record_print_text(&r, syslog, time);
3480 :
3481 : /* adjust record to store to remaining buffer space */
3482 0 : prb_rec_init_rd(&r, &info, buf + l, size - l);
3483 :
3484 0 : seq = r.info->seq + 1;
3485 : }
3486 :
3487 0 : dumper->next_seq = next_seq;
3488 0 : ret = true;
3489 0 : logbuf_unlock_irqrestore(flags);
3490 0 : out:
3491 0 : if (len)
3492 0 : *len = l;
3493 0 : return ret;
3494 : }
3495 : EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
3496 :
3497 : /**
3498 : * kmsg_dump_rewind_nolock - reset the iterator (unlocked version)
3499 : * @dumper: registered kmsg dumper
3500 : *
3501 : * Reset the dumper's iterator so that kmsg_dump_get_line() and
3502 : * kmsg_dump_get_buffer() can be called again and used multiple
3503 : * times within the same dumper.dump() callback.
3504 : *
3505 : * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3506 : */
3507 0 : void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
3508 : {
3509 0 : dumper->cur_seq = clear_seq;
3510 0 : dumper->next_seq = prb_next_seq(prb);
3511 0 : }
3512 :
3513 : /**
3514 : * kmsg_dump_rewind - reset the iterator
3515 : * @dumper: registered kmsg dumper
3516 : *
3517 : * Reset the dumper's iterator so that kmsg_dump_get_line() and
3518 : * kmsg_dump_get_buffer() can be called again and used multiple
3519 : * times within the same dumper.dump() callback.
3520 : */
3521 0 : void kmsg_dump_rewind(struct kmsg_dumper *dumper)
3522 : {
3523 0 : unsigned long flags;
3524 :
3525 0 : logbuf_lock_irqsave(flags);
3526 0 : kmsg_dump_rewind_nolock(dumper);
3527 0 : logbuf_unlock_irqrestore(flags);
3528 0 : }
3529 : EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
3530 :
3531 : #endif
|
