~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/printk/printk.c

Version: ~ [ linux-5.0-rc6 ] ~ [ linux-4.20.10 ] ~ [ linux-4.19.23 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.101 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.158 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.174 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.134 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.63 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.39.4 ] ~ [ linux-2.6.38.8 ] ~ [ linux-2.6.37.6 ] ~ [ linux-2.6.36.4 ] ~ [ linux-2.6.35.14 ] ~ [ linux-2.6.34.15 ] ~ [ linux-2.6.33.20 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *  linux/kernel/printk.c
  3  *
  4  *  Copyright (C) 1991, 1992  Linus Torvalds
  5  *
  6  * Modified to make sys_syslog() more flexible: added commands to
  7  * return the last 4k of kernel messages, regardless of whether
  8  * they've been read or not.  Added option to suppress kernel printk's
  9  * to the console.  Added hook for sending the console messages
 10  * elsewhere, in preparation for a serial line console (someday).
 11  * Ted Ts'o, 2/11/93.
 12  * Modified for sysctl support, 1/8/97, Chris Horn.
 13  * Fixed SMP synchronization, 08/08/99, Manfred Spraul
 14  *     manfred@colorfullife.com
 15  * Rewrote bits to get rid of console_lock
 16  *      01Mar01 Andrew Morton
 17  */
 18 
 19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 20 
 21 #include <linux/kernel.h>
 22 #include <linux/mm.h>
 23 #include <linux/tty.h>
 24 #include <linux/tty_driver.h>
 25 #include <linux/console.h>
 26 #include <linux/init.h>
 27 #include <linux/jiffies.h>
 28 #include <linux/nmi.h>
 29 #include <linux/module.h>
 30 #include <linux/moduleparam.h>
 31 #include <linux/delay.h>
 32 #include <linux/smp.h>
 33 #include <linux/security.h>
 34 #include <linux/memblock.h>
 35 #include <linux/syscalls.h>
 36 #include <linux/crash_core.h>
 37 #include <linux/kdb.h>
 38 #include <linux/ratelimit.h>
 39 #include <linux/kmsg_dump.h>
 40 #include <linux/syslog.h>
 41 #include <linux/cpu.h>
 42 #include <linux/rculist.h>
 43 #include <linux/poll.h>
 44 #include <linux/irq_work.h>
 45 #include <linux/ctype.h>
 46 #include <linux/uio.h>
 47 #include <linux/sched/clock.h>
 48 #include <linux/sched/debug.h>
 49 #include <linux/sched/task_stack.h>
 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 "console_cmdline.h"
 59 #include "braille.h"
 60 #include "internal.h"
 61 
 62 int console_printk[4] = {
 63         CONSOLE_LOGLEVEL_DEFAULT,       /* console_loglevel */
 64         MESSAGE_LOGLEVEL_DEFAULT,       /* default_message_loglevel */
 65         CONSOLE_LOGLEVEL_MIN,           /* minimum_console_loglevel */
 66         CONSOLE_LOGLEVEL_DEFAULT,       /* default_console_loglevel */
 67 };
 68 
 69 atomic_t ignore_console_lock_warning __read_mostly = ATOMIC_INIT(0);
 70 EXPORT_SYMBOL(ignore_console_lock_warning);
 71 
 72 /*
 73  * Low level drivers may need that to know if they can schedule in
 74  * their unblank() callback or not. So let's export it.
 75  */
 76 int oops_in_progress;
 77 EXPORT_SYMBOL(oops_in_progress);
 78 
 79 /*
 80  * console_sem protects the console_drivers list, and also
 81  * provides serialisation for access to the entire console
 82  * driver system.
 83  */
 84 static DEFINE_SEMAPHORE(console_sem);
 85 struct console *console_drivers;
 86 EXPORT_SYMBOL_GPL(console_drivers);
 87 
 88 #ifdef CONFIG_LOCKDEP
 89 static struct lockdep_map console_lock_dep_map = {
 90         .name = "console_lock"
 91 };
 92 #endif
 93 
 94 enum devkmsg_log_bits {
 95         __DEVKMSG_LOG_BIT_ON = 0,
 96         __DEVKMSG_LOG_BIT_OFF,
 97         __DEVKMSG_LOG_BIT_LOCK,
 98 };
 99 
100 enum devkmsg_log_masks {
101         DEVKMSG_LOG_MASK_ON             = BIT(__DEVKMSG_LOG_BIT_ON),
102         DEVKMSG_LOG_MASK_OFF            = BIT(__DEVKMSG_LOG_BIT_OFF),
103         DEVKMSG_LOG_MASK_LOCK           = BIT(__DEVKMSG_LOG_BIT_LOCK),
104 };
105 
106 /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
107 #define DEVKMSG_LOG_MASK_DEFAULT        0
108 
109 static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
110 
111 static int __control_devkmsg(char *str)
112 {
113         if (!str)
114                 return -EINVAL;
115 
116         if (!strncmp(str, "on", 2)) {
117                 devkmsg_log = DEVKMSG_LOG_MASK_ON;
118                 return 2;
119         } else if (!strncmp(str, "off", 3)) {
120                 devkmsg_log = DEVKMSG_LOG_MASK_OFF;
121                 return 3;
122         } else if (!strncmp(str, "ratelimit", 9)) {
123                 devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
124                 return 9;
125         }
126         return -EINVAL;
127 }
128 
129 static int __init control_devkmsg(char *str)
130 {
131         if (__control_devkmsg(str) < 0)
132                 return 1;
133 
134         /*
135          * Set sysctl string accordingly:
136          */
137         if (devkmsg_log == DEVKMSG_LOG_MASK_ON)
138                 strcpy(devkmsg_log_str, "on");
139         else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF)
140                 strcpy(devkmsg_log_str, "off");
141         /* else "ratelimit" which is set by default. */
142 
143         /*
144          * Sysctl cannot change it anymore. The kernel command line setting of
145          * this parameter is to force the setting to be permanent throughout the
146          * runtime of the system. This is a precation measure against userspace
147          * trying to be a smarta** and attempting to change it up on us.
148          */
149         devkmsg_log |= DEVKMSG_LOG_MASK_LOCK;
150 
151         return 0;
152 }
153 __setup("printk.devkmsg=", control_devkmsg);
154 
155 char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit";
156 
157 int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
158                               void __user *buffer, size_t *lenp, loff_t *ppos)
159 {
160         char old_str[DEVKMSG_STR_MAX_SIZE];
161         unsigned int old;
162         int err;
163 
164         if (write) {
165                 if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK)
166                         return -EINVAL;
167 
168                 old = devkmsg_log;
169                 strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE);
170         }
171 
172         err = proc_dostring(table, write, buffer, lenp, ppos);
173         if (err)
174                 return err;
175 
176         if (write) {
177                 err = __control_devkmsg(devkmsg_log_str);
178 
179                 /*
180                  * Do not accept an unknown string OR a known string with
181                  * trailing crap...
182                  */
183                 if (err < 0 || (err + 1 != *lenp)) {
184 
185                         /* ... and restore old setting. */
186                         devkmsg_log = old;
187                         strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE);
188 
189                         return -EINVAL;
190                 }
191         }
192 
193         return 0;
194 }
195 
196 /* Number of registered extended console drivers. */
197 static int nr_ext_console_drivers;
198 
199 /*
200  * Helper macros to handle lockdep when locking/unlocking console_sem. We use
201  * macros instead of functions so that _RET_IP_ contains useful information.
202  */
203 #define down_console_sem() do { \
204         down(&console_sem);\
205         mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
206 } while (0)
207 
208 static int __down_trylock_console_sem(unsigned long ip)
209 {
210         int lock_failed;
211         unsigned long flags;
212 
213         /*
214          * Here and in __up_console_sem() we need to be in safe mode,
215          * because spindump/WARN/etc from under console ->lock will
216          * deadlock in printk()->down_trylock_console_sem() otherwise.
217          */
218         printk_safe_enter_irqsave(flags);
219         lock_failed = down_trylock(&console_sem);
220         printk_safe_exit_irqrestore(flags);
221 
222         if (lock_failed)
223                 return 1;
224         mutex_acquire(&console_lock_dep_map, 0, 1, ip);
225         return 0;
226 }
227 #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
228 
229 static void __up_console_sem(unsigned long ip)
230 {
231         unsigned long flags;
232 
233         mutex_release(&console_lock_dep_map, 1, ip);
234 
235         printk_safe_enter_irqsave(flags);
236         up(&console_sem);
237         printk_safe_exit_irqrestore(flags);
238 }
239 #define up_console_sem() __up_console_sem(_RET_IP_)
240 
241 /*
242  * This is used for debugging the mess that is the VT code by
243  * keeping track if we have the console semaphore held. It's
244  * definitely not the perfect debug tool (we don't know if _WE_
245  * hold it and are racing, but it helps tracking those weird code
246  * paths in the console code where we end up in places I want
247  * locked without the console sempahore held).
248  */
249 static int console_locked, console_suspended;
250 
251 /*
252  * If exclusive_console is non-NULL then only this console is to be printed to.
253  */
254 static struct console *exclusive_console;
255 
256 /*
257  *      Array of consoles built from command line options (console=)
258  */
259 
260 #define MAX_CMDLINECONSOLES 8
261 
262 static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
263 
264 static int preferred_console = -1;
265 int console_set_on_cmdline;
266 EXPORT_SYMBOL(console_set_on_cmdline);
267 
268 /* Flag: console code may call schedule() */
269 static int console_may_schedule;
270 
271 enum con_msg_format_flags {
272         MSG_FORMAT_DEFAULT      = 0,
273         MSG_FORMAT_SYSLOG       = (1 << 0),
274 };
275 
276 static int console_msg_format = MSG_FORMAT_DEFAULT;
277 
278 /*
279  * The printk log buffer consists of a chain of concatenated variable
280  * length records. Every record starts with a record header, containing
281  * the overall length of the record.
282  *
283  * The heads to the first and last entry in the buffer, as well as the
284  * sequence numbers of these entries are maintained when messages are
285  * stored.
286  *
287  * If the heads indicate available messages, the length in the header
288  * tells the start next message. A length == 0 for the next message
289  * indicates a wrap-around to the beginning of the buffer.
290  *
291  * Every record carries the monotonic timestamp in microseconds, as well as
292  * the standard userspace syslog level and syslog facility. The usual
293  * kernel messages use LOG_KERN; userspace-injected messages always carry
294  * a matching syslog facility, by default LOG_USER. The origin of every
295  * message can be reliably determined that way.
296  *
297  * The human readable log message directly follows the message header. The
298  * length of the message text is stored in the header, the stored message
299  * is not terminated.
300  *
301  * Optionally, a message can carry a dictionary of properties (key/value pairs),
302  * to provide userspace with a machine-readable message context.
303  *
304  * Examples for well-defined, commonly used property names are:
305  *   DEVICE=b12:8               device identifier
306  *                                b12:8         block dev_t
307  *                                c127:3        char dev_t
308  *                                n8            netdev ifindex
309  *                                +sound:card0  subsystem:devname
310  *   SUBSYSTEM=pci              driver-core subsystem name
311  *
312  * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
313  * follows directly after a '=' character. Every property is terminated by
314  * a '\0' character. The last property is not terminated.
315  *
316  * Example of a message structure:
317  *   0000  ff 8f 00 00 00 00 00 00      monotonic time in nsec
318  *   0008  34 00                        record is 52 bytes long
319  *   000a        0b 00                  text is 11 bytes long
320  *   000c              1f 00            dictionary is 23 bytes long
321  *   000e                    03 00      LOG_KERN (facility) LOG_ERR (level)
322  *   0010  69 74 27 73 20 61 20 6c      "it's a l"
323  *         69 6e 65                     "ine"
324  *   001b           44 45 56 49 43      "DEVIC"
325  *         45 3d 62 38 3a 32 00 44      "E=b8:2\0D"
326  *         52 49 56 45 52 3d 62 75      "RIVER=bu"
327  *         67                           "g"
328  *   0032     00 00 00                  padding to next message header
329  *
330  * The 'struct printk_log' buffer header must never be directly exported to
331  * userspace, it is a kernel-private implementation detail that might
332  * need to be changed in the future, when the requirements change.
333  *
334  * /dev/kmsg exports the structured data in the following line format:
335  *   "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
336  *
337  * Users of the export format should ignore possible additional values
338  * separated by ',', and find the message after the ';' character.
339  *
340  * The optional key/value pairs are attached as continuation lines starting
341  * with a space character and terminated by a newline. All possible
342  * non-prinatable characters are escaped in the "\xff" notation.
343  */
344 
345 enum log_flags {
346         LOG_NEWLINE     = 2,    /* text ended with a newline */
347         LOG_PREFIX      = 4,    /* text started with a prefix */
348         LOG_CONT        = 8,    /* text is a fragment of a continuation line */
349 };
350 
351 struct printk_log {
352         u64 ts_nsec;            /* timestamp in nanoseconds */
353         u16 len;                /* length of entire record */
354         u16 text_len;           /* length of text buffer */
355         u16 dict_len;           /* length of dictionary buffer */
356         u8 facility;            /* syslog facility */
357         u8 flags:5;             /* internal record flags */
358         u8 level:3;             /* syslog level */
359 }
360 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
361 __packed __aligned(4)
362 #endif
363 ;
364 
365 /*
366  * The logbuf_lock protects kmsg buffer, indices, counters.  This can be taken
367  * within the scheduler's rq lock. It must be released before calling
368  * console_unlock() or anything else that might wake up a process.
369  */
370 DEFINE_RAW_SPINLOCK(logbuf_lock);
371 
372 /*
373  * Helper macros to lock/unlock logbuf_lock and switch between
374  * printk-safe/unsafe modes.
375  */
376 #define logbuf_lock_irq()                               \
377         do {                                            \
378                 printk_safe_enter_irq();                \
379                 raw_spin_lock(&logbuf_lock);            \
380         } while (0)
381 
382 #define logbuf_unlock_irq()                             \
383         do {                                            \
384                 raw_spin_unlock(&logbuf_lock);          \
385                 printk_safe_exit_irq();                 \
386         } while (0)
387 
388 #define logbuf_lock_irqsave(flags)                      \
389         do {                                            \
390                 printk_safe_enter_irqsave(flags);       \
391                 raw_spin_lock(&logbuf_lock);            \
392         } while (0)
393 
394 #define logbuf_unlock_irqrestore(flags)         \
395         do {                                            \
396                 raw_spin_unlock(&logbuf_lock);          \
397                 printk_safe_exit_irqrestore(flags);     \
398         } while (0)
399 
400 #ifdef CONFIG_PRINTK
401 DECLARE_WAIT_QUEUE_HEAD(log_wait);
402 /* the next printk record to read by syslog(READ) or /proc/kmsg */
403 static u64 syslog_seq;
404 static u32 syslog_idx;
405 static size_t syslog_partial;
406 static bool syslog_time;
407 
408 /* index and sequence number of the first record stored in the buffer */
409 static u64 log_first_seq;
410 static u32 log_first_idx;
411 
412 /* index and sequence number of the next record to store in the buffer */
413 static u64 log_next_seq;
414 static u32 log_next_idx;
415 
416 /* the next printk record to write to the console */
417 static u64 console_seq;
418 static u32 console_idx;
419 static u64 exclusive_console_stop_seq;
420 
421 /* the next printk record to read after the last 'clear' command */
422 static u64 clear_seq;
423 static u32 clear_idx;
424 
425 #define PREFIX_MAX              32
426 #define LOG_LINE_MAX            (1024 - PREFIX_MAX)
427 
428 #define LOG_LEVEL(v)            ((v) & 0x07)
429 #define LOG_FACILITY(v)         ((v) >> 3 & 0xff)
430 
431 /* record buffer */
432 #define LOG_ALIGN __alignof__(struct printk_log)
433 #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
434 #define LOG_BUF_LEN_MAX (u32)(1 << 31)
435 static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
436 static char *log_buf = __log_buf;
437 static u32 log_buf_len = __LOG_BUF_LEN;
438 
439 /* Return log buffer address */
440 char *log_buf_addr_get(void)
441 {
442         return log_buf;
443 }
444 
445 /* Return log buffer size */
446 u32 log_buf_len_get(void)
447 {
448         return log_buf_len;
449 }
450 
451 /* human readable text of the record */
452 static char *log_text(const struct printk_log *msg)
453 {
454         return (char *)msg + sizeof(struct printk_log);
455 }
456 
457 /* optional key/value pair dictionary attached to the record */
458 static char *log_dict(const struct printk_log *msg)
459 {
460         return (char *)msg + sizeof(struct printk_log) + msg->text_len;
461 }
462 
463 /* get record by index; idx must point to valid msg */
464 static struct printk_log *log_from_idx(u32 idx)
465 {
466         struct printk_log *msg = (struct printk_log *)(log_buf + idx);
467 
468         /*
469          * A length == 0 record is the end of buffer marker. Wrap around and
470          * read the message at the start of the buffer.
471          */
472         if (!msg->len)
473                 return (struct printk_log *)log_buf;
474         return msg;
475 }
476 
477 /* get next record; idx must point to valid msg */
478 static u32 log_next(u32 idx)
479 {
480         struct printk_log *msg = (struct printk_log *)(log_buf + idx);
481 
482         /* length == 0 indicates the end of the buffer; wrap */
483         /*
484          * A length == 0 record is the end of buffer marker. Wrap around and
485          * read the message at the start of the buffer as *this* one, and
486          * return the one after that.
487          */
488         if (!msg->len) {
489                 msg = (struct printk_log *)log_buf;
490                 return msg->len;
491         }
492         return idx + msg->len;
493 }
494 
495 /*
496  * Check whether there is enough free space for the given message.
497  *
498  * The same values of first_idx and next_idx mean that the buffer
499  * is either empty or full.
500  *
501  * If the buffer is empty, we must respect the position of the indexes.
502  * They cannot be reset to the beginning of the buffer.
503  */
504 static int logbuf_has_space(u32 msg_size, bool empty)
505 {
506         u32 free;
507 
508         if (log_next_idx > log_first_idx || empty)
509                 free = max(log_buf_len - log_next_idx, log_first_idx);
510         else
511                 free = log_first_idx - log_next_idx;
512 
513         /*
514          * We need space also for an empty header that signalizes wrapping
515          * of the buffer.
516          */
517         return free >= msg_size + sizeof(struct printk_log);
518 }
519 
520 static int log_make_free_space(u32 msg_size)
521 {
522         while (log_first_seq < log_next_seq &&
523                !logbuf_has_space(msg_size, false)) {
524                 /* drop old messages until we have enough contiguous space */
525                 log_first_idx = log_next(log_first_idx);
526                 log_first_seq++;
527         }
528 
529         if (clear_seq < log_first_seq) {
530                 clear_seq = log_first_seq;
531                 clear_idx = log_first_idx;
532         }
533 
534         /* sequence numbers are equal, so the log buffer is empty */
535         if (logbuf_has_space(msg_size, log_first_seq == log_next_seq))
536                 return 0;
537 
538         return -ENOMEM;
539 }
540 
541 /* compute the message size including the padding bytes */
542 static u32 msg_used_size(u16 text_len, u16 dict_len, u32 *pad_len)
543 {
544         u32 size;
545 
546         size = sizeof(struct printk_log) + text_len + dict_len;
547         *pad_len = (-size) & (LOG_ALIGN - 1);
548         size += *pad_len;
549 
550         return size;
551 }
552 
553 /*
554  * Define how much of the log buffer we could take at maximum. The value
555  * must be greater than two. Note that only half of the buffer is available
556  * when the index points to the middle.
557  */
558 #define MAX_LOG_TAKE_PART 4
559 static const char trunc_msg[] = "<truncated>";
560 
561 static u32 truncate_msg(u16 *text_len, u16 *trunc_msg_len,
562                         u16 *dict_len, u32 *pad_len)
563 {
564         /*
565          * The message should not take the whole buffer. Otherwise, it might
566          * get removed too soon.
567          */
568         u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART;
569         if (*text_len > max_text_len)
570                 *text_len = max_text_len;
571         /* enable the warning message */
572         *trunc_msg_len = strlen(trunc_msg);
573         /* disable the "dict" completely */
574         *dict_len = 0;
575         /* compute the size again, count also the warning message */
576         return msg_used_size(*text_len + *trunc_msg_len, 0, pad_len);
577 }
578 
579 /* insert record into the buffer, discard old ones, update heads */
580 static int log_store(int facility, int level,
581                      enum log_flags flags, u64 ts_nsec,
582                      const char *dict, u16 dict_len,
583                      const char *text, u16 text_len)
584 {
585         struct printk_log *msg;
586         u32 size, pad_len;
587         u16 trunc_msg_len = 0;
588 
589         /* number of '\0' padding bytes to next message */
590         size = msg_used_size(text_len, dict_len, &pad_len);
591 
592         if (log_make_free_space(size)) {
593                 /* truncate the message if it is too long for empty buffer */
594                 size = truncate_msg(&text_len, &trunc_msg_len,
595                                     &dict_len, &pad_len);
596                 /* survive when the log buffer is too small for trunc_msg */
597                 if (log_make_free_space(size))
598                         return 0;
599         }
600 
601         if (log_next_idx + size + sizeof(struct printk_log) > log_buf_len) {
602                 /*
603                  * This message + an additional empty header does not fit
604                  * at the end of the buffer. Add an empty header with len == 0
605                  * to signify a wrap around.
606                  */
607                 memset(log_buf + log_next_idx, 0, sizeof(struct printk_log));
608                 log_next_idx = 0;
609         }
610 
611         /* fill message */
612         msg = (struct printk_log *)(log_buf + log_next_idx);
613         memcpy(log_text(msg), text, text_len);
614         msg->text_len = text_len;
615         if (trunc_msg_len) {
616                 memcpy(log_text(msg) + text_len, trunc_msg, trunc_msg_len);
617                 msg->text_len += trunc_msg_len;
618         }
619         memcpy(log_dict(msg), dict, dict_len);
620         msg->dict_len = dict_len;
621         msg->facility = facility;
622         msg->level = level & 7;
623         msg->flags = flags & 0x1f;
624         if (ts_nsec > 0)
625                 msg->ts_nsec = ts_nsec;
626         else
627                 msg->ts_nsec = local_clock();
628         memset(log_dict(msg) + dict_len, 0, pad_len);
629         msg->len = size;
630 
631         /* insert message */
632         log_next_idx += msg->len;
633         log_next_seq++;
634 
635         return msg->text_len;
636 }
637 
638 int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT);
639 
640 static int syslog_action_restricted(int type)
641 {
642         if (dmesg_restrict)
643                 return 1;
644         /*
645          * Unless restricted, we allow "read all" and "get buffer size"
646          * for everybody.
647          */
648         return type != SYSLOG_ACTION_READ_ALL &&
649                type != SYSLOG_ACTION_SIZE_BUFFER;
650 }
651 
652 static int check_syslog_permissions(int type, int source)
653 {
654         /*
655          * If this is from /proc/kmsg and we've already opened it, then we've
656          * already done the capabilities checks at open time.
657          */
658         if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN)
659                 goto ok;
660 
661         if (syslog_action_restricted(type)) {
662                 if (capable(CAP_SYSLOG))
663                         goto ok;
664                 /*
665                  * For historical reasons, accept CAP_SYS_ADMIN too, with
666                  * a warning.
667                  */
668                 if (capable(CAP_SYS_ADMIN)) {
669                         pr_warn_once("%s (%d): Attempt to access syslog with "
670                                      "CAP_SYS_ADMIN but no CAP_SYSLOG "
671                                      "(deprecated).\n",
672                                  current->comm, task_pid_nr(current));
673                         goto ok;
674                 }
675                 return -EPERM;
676         }
677 ok:
678         return security_syslog(type);
679 }
680 
681 static void append_char(char **pp, char *e, char c)
682 {
683         if (*pp < e)
684                 *(*pp)++ = c;
685 }
686 
687 static ssize_t msg_print_ext_header(char *buf, size_t size,
688                                     struct printk_log *msg, u64 seq)
689 {
690         u64 ts_usec = msg->ts_nsec;
691 
692         do_div(ts_usec, 1000);
693 
694         return scnprintf(buf, size, "%u,%llu,%llu,%c;",
695                        (msg->facility << 3) | msg->level, seq, ts_usec,
696                        msg->flags & LOG_CONT ? 'c' : '-');
697 }
698 
699 static ssize_t msg_print_ext_body(char *buf, size_t size,
700                                   char *dict, size_t dict_len,
701                                   char *text, size_t text_len)
702 {
703         char *p = buf, *e = buf + size;
704         size_t i;
705 
706         /* escape non-printable characters */
707         for (i = 0; i < text_len; i++) {
708                 unsigned char c = text[i];
709 
710                 if (c < ' ' || c >= 127 || c == '\\')
711                         p += scnprintf(p, e - p, "\\x%02x", c);
712                 else
713                         append_char(&p, e, c);
714         }
715         append_char(&p, e, '\n');
716 
717         if (dict_len) {
718                 bool line = true;
719 
720                 for (i = 0; i < dict_len; i++) {
721                         unsigned char c = dict[i];
722 
723                         if (line) {
724                                 append_char(&p, e, ' ');
725                                 line = false;
726                         }
727 
728                         if (c == '\0') {
729                                 append_char(&p, e, '\n');
730                                 line = true;
731                                 continue;
732                         }
733 
734                         if (c < ' ' || c >= 127 || c == '\\') {
735                                 p += scnprintf(p, e - p, "\\x%02x", c);
736                                 continue;
737                         }
738 
739                         append_char(&p, e, c);
740                 }
741                 append_char(&p, e, '\n');
742         }
743 
744         return p - buf;
745 }
746 
747 /* /dev/kmsg - userspace message inject/listen interface */
748 struct devkmsg_user {
749         u64 seq;
750         u32 idx;
751         struct ratelimit_state rs;
752         struct mutex lock;
753         char buf[CONSOLE_EXT_LOG_MAX];
754 };
755 
756 static __printf(3, 4) __cold
757 int devkmsg_emit(int facility, int level, const char *fmt, ...)
758 {
759         va_list args;
760         int r;
761 
762         va_start(args, fmt);
763         r = vprintk_emit(facility, level, NULL, 0, fmt, args);
764         va_end(args);
765 
766         return r;
767 }
768 
769 static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from)
770 {
771         char *buf, *line;
772         int level = default_message_loglevel;
773         int facility = 1;       /* LOG_USER */
774         struct file *file = iocb->ki_filp;
775         struct devkmsg_user *user = file->private_data;
776         size_t len = iov_iter_count(from);
777         ssize_t ret = len;
778 
779         if (!user || len > LOG_LINE_MAX)
780                 return -EINVAL;
781 
782         /* Ignore when user logging is disabled. */
783         if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
784                 return len;
785 
786         /* Ratelimit when not explicitly enabled. */
787         if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) {
788                 if (!___ratelimit(&user->rs, current->comm))
789                         return ret;
790         }
791 
792         buf = kmalloc(len+1, GFP_KERNEL);
793         if (buf == NULL)
794                 return -ENOMEM;
795 
796         buf[len] = '\0';
797         if (!copy_from_iter_full(buf, len, from)) {
798                 kfree(buf);
799                 return -EFAULT;
800         }
801 
802         /*
803          * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
804          * the decimal value represents 32bit, the lower 3 bit are the log
805          * level, the rest are the log facility.
806          *
807          * If no prefix or no userspace facility is specified, we
808          * enforce LOG_USER, to be able to reliably distinguish
809          * kernel-generated messages from userspace-injected ones.
810          */
811         line = buf;
812         if (line[0] == '<') {
813                 char *endp = NULL;
814                 unsigned int u;
815 
816                 u = simple_strtoul(line + 1, &endp, 10);
817                 if (endp && endp[0] == '>') {
818                         level = LOG_LEVEL(u);
819                         if (LOG_FACILITY(u) != 0)
820                                 facility = LOG_FACILITY(u);
821                         endp++;
822                         len -= endp - line;
823                         line = endp;
824                 }
825         }
826 
827         devkmsg_emit(facility, level, "%s", line);
828         kfree(buf);
829         return ret;
830 }
831 
832 static ssize_t devkmsg_read(struct file *file, char __user *buf,
833                             size_t count, loff_t *ppos)
834 {
835         struct devkmsg_user *user = file->private_data;
836         struct printk_log *msg;
837         size_t len;
838         ssize_t ret;
839 
840         if (!user)
841                 return -EBADF;
842 
843         ret = mutex_lock_interruptible(&user->lock);
844         if (ret)
845                 return ret;
846 
847         logbuf_lock_irq();
848         while (user->seq == log_next_seq) {
849                 if (file->f_flags & O_NONBLOCK) {
850                         ret = -EAGAIN;
851                         logbuf_unlock_irq();
852                         goto out;
853                 }
854 
855                 logbuf_unlock_irq();
856                 ret = wait_event_interruptible(log_wait,
857                                                user->seq != log_next_seq);
858                 if (ret)
859                         goto out;
860                 logbuf_lock_irq();
861         }
862 
863         if (user->seq < log_first_seq) {
864                 /* our last seen message is gone, return error and reset */
865                 user->idx = log_first_idx;
866                 user->seq = log_first_seq;
867                 ret = -EPIPE;
868                 logbuf_unlock_irq();
869                 goto out;
870         }
871 
872         msg = log_from_idx(user->idx);
873         len = msg_print_ext_header(user->buf, sizeof(user->buf),
874                                    msg, user->seq);
875         len += msg_print_ext_body(user->buf + len, sizeof(user->buf) - len,
876                                   log_dict(msg), msg->dict_len,
877                                   log_text(msg), msg->text_len);
878 
879         user->idx = log_next(user->idx);
880         user->seq++;
881         logbuf_unlock_irq();
882 
883         if (len > count) {
884                 ret = -EINVAL;
885                 goto out;
886         }
887 
888         if (copy_to_user(buf, user->buf, len)) {
889                 ret = -EFAULT;
890                 goto out;
891         }
892         ret = len;
893 out:
894         mutex_unlock(&user->lock);
895         return ret;
896 }
897 
898 static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
899 {
900         struct devkmsg_user *user = file->private_data;
901         loff_t ret = 0;
902 
903         if (!user)
904                 return -EBADF;
905         if (offset)
906                 return -ESPIPE;
907 
908         logbuf_lock_irq();
909         switch (whence) {
910         case SEEK_SET:
911                 /* the first record */
912                 user->idx = log_first_idx;
913                 user->seq = log_first_seq;
914                 break;
915         case SEEK_DATA:
916                 /*
917                  * The first record after the last SYSLOG_ACTION_CLEAR,
918                  * like issued by 'dmesg -c'. Reading /dev/kmsg itself
919                  * changes no global state, and does not clear anything.
920                  */
921                 user->idx = clear_idx;
922                 user->seq = clear_seq;
923                 break;
924         case SEEK_END:
925                 /* after the last record */
926                 user->idx = log_next_idx;
927                 user->seq = log_next_seq;
928                 break;
929         default:
930                 ret = -EINVAL;
931         }
932         logbuf_unlock_irq();
933         return ret;
934 }
935 
936 static __poll_t devkmsg_poll(struct file *file, poll_table *wait)
937 {
938         struct devkmsg_user *user = file->private_data;
939         __poll_t ret = 0;
940 
941         if (!user)
942                 return EPOLLERR|EPOLLNVAL;
943 
944         poll_wait(file, &log_wait, wait);
945 
946         logbuf_lock_irq();
947         if (user->seq < log_next_seq) {
948                 /* return error when data has vanished underneath us */
949                 if (user->seq < log_first_seq)
950                         ret = EPOLLIN|EPOLLRDNORM|EPOLLERR|EPOLLPRI;
951                 else
952                         ret = EPOLLIN|EPOLLRDNORM;
953         }
954         logbuf_unlock_irq();
955 
956         return ret;
957 }
958 
959 static int devkmsg_open(struct inode *inode, struct file *file)
960 {
961         struct devkmsg_user *user;
962         int err;
963 
964         if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
965                 return -EPERM;
966 
967         /* write-only does not need any file context */
968         if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
969                 err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
970                                                SYSLOG_FROM_READER);
971                 if (err)
972                         return err;
973         }
974 
975         user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
976         if (!user)
977                 return -ENOMEM;
978 
979         ratelimit_default_init(&user->rs);
980         ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE);
981 
982         mutex_init(&user->lock);
983 
984         logbuf_lock_irq();
985         user->idx = log_first_idx;
986         user->seq = log_first_seq;
987         logbuf_unlock_irq();
988 
989         file->private_data = user;
990         return 0;
991 }
992 
993 static int devkmsg_release(struct inode *inode, struct file *file)
994 {
995         struct devkmsg_user *user = file->private_data;
996 
997         if (!user)
998                 return 0;
999 
1000         ratelimit_state_exit(&user->rs);
1001 
1002         mutex_destroy(&user->lock);
1003         kfree(user);
1004         return 0;
1005 }
1006 
1007 const struct file_operations kmsg_fops = {
1008         .open = devkmsg_open,
1009         .read = devkmsg_read,
1010         .write_iter = devkmsg_write,
1011         .llseek = devkmsg_llseek,
1012         .poll = devkmsg_poll,
1013         .release = devkmsg_release,
1014 };
1015 
1016 #ifdef CONFIG_CRASH_CORE
1017 /*
1018  * This appends the listed symbols to /proc/vmcore
1019  *
1020  * /proc/vmcore is used by various utilities, like crash and makedumpfile to
1021  * obtain access to symbols that are otherwise very difficult to locate.  These
1022  * symbols are specifically used so that utilities can access and extract the
1023  * dmesg log from a vmcore file after a crash.
1024  */
1025 void log_buf_vmcoreinfo_setup(void)
1026 {
1027         VMCOREINFO_SYMBOL(log_buf);
1028         VMCOREINFO_SYMBOL(log_buf_len);
1029         VMCOREINFO_SYMBOL(log_first_idx);
1030         VMCOREINFO_SYMBOL(clear_idx);
1031         VMCOREINFO_SYMBOL(log_next_idx);
1032         /*
1033          * Export struct printk_log size and field offsets. User space tools can
1034          * parse it and detect any changes to structure down the line.
1035          */
1036         VMCOREINFO_STRUCT_SIZE(printk_log);
1037         VMCOREINFO_OFFSET(printk_log, ts_nsec);
1038         VMCOREINFO_OFFSET(printk_log, len);
1039         VMCOREINFO_OFFSET(printk_log, text_len);
1040         VMCOREINFO_OFFSET(printk_log, dict_len);
1041 }
1042 #endif
1043 
1044 /* requested log_buf_len from kernel cmdline */
1045 static unsigned long __initdata new_log_buf_len;
1046 
1047 /* we practice scaling the ring buffer by powers of 2 */
1048 static void __init log_buf_len_update(u64 size)
1049 {
1050         if (size > (u64)LOG_BUF_LEN_MAX) {
1051                 size = (u64)LOG_BUF_LEN_MAX;
1052                 pr_err("log_buf over 2G is not supported.\n");
1053         }
1054 
1055         if (size)
1056                 size = roundup_pow_of_two(size);
1057         if (size > log_buf_len)
1058                 new_log_buf_len = (unsigned long)size;
1059 }
1060 
1061 /* save requested log_buf_len since it's too early to process it */
1062 static int __init log_buf_len_setup(char *str)
1063 {
1064         u64 size;
1065 
1066         if (!str)
1067                 return -EINVAL;
1068 
1069         size = memparse(str, &str);
1070 
1071         log_buf_len_update(size);
1072 
1073         return 0;
1074 }
1075 early_param("log_buf_len", log_buf_len_setup);
1076 
1077 #ifdef CONFIG_SMP
1078 #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1079 
1080 static void __init log_buf_add_cpu(void)
1081 {
1082         unsigned int cpu_extra;
1083 
1084         /*
1085          * archs should set up cpu_possible_bits properly with
1086          * set_cpu_possible() after setup_arch() but just in
1087          * case lets ensure this is valid.
1088          */
1089         if (num_possible_cpus() == 1)
1090                 return;
1091 
1092         cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
1093 
1094         /* by default this will only continue through for large > 64 CPUs */
1095         if (cpu_extra <= __LOG_BUF_LEN / 2)
1096                 return;
1097 
1098         pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1099                 __LOG_CPU_MAX_BUF_LEN);
1100         pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1101                 cpu_extra);
1102         pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN);
1103 
1104         log_buf_len_update(cpu_extra + __LOG_BUF_LEN);
1105 }
1106 #else /* !CONFIG_SMP */
1107 static inline void log_buf_add_cpu(void) {}
1108 #endif /* CONFIG_SMP */
1109 
1110 void __init setup_log_buf(int early)
1111 {
1112         unsigned long flags;
1113         char *new_log_buf;
1114         unsigned int free;
1115 
1116         if (log_buf != __log_buf)
1117                 return;
1118 
1119         if (!early && !new_log_buf_len)
1120                 log_buf_add_cpu();
1121 
1122         if (!new_log_buf_len)
1123                 return;
1124 
1125         if (early) {
1126                 new_log_buf =
1127                         memblock_alloc(new_log_buf_len, LOG_ALIGN);
1128         } else {
1129                 new_log_buf = memblock_alloc_nopanic(new_log_buf_len,
1130                                                           LOG_ALIGN);
1131         }
1132 
1133         if (unlikely(!new_log_buf)) {
1134                 pr_err("log_buf_len: %lu bytes not available\n",
1135                         new_log_buf_len);
1136                 return;
1137         }
1138 
1139         logbuf_lock_irqsave(flags);
1140         log_buf_len = new_log_buf_len;
1141         log_buf = new_log_buf;
1142         new_log_buf_len = 0;
1143         free = __LOG_BUF_LEN - log_next_idx;
1144         memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
1145         logbuf_unlock_irqrestore(flags);
1146 
1147         pr_info("log_buf_len: %u bytes\n", log_buf_len);
1148         pr_info("early log buf free: %u(%u%%)\n",
1149                 free, (free * 100) / __LOG_BUF_LEN);
1150 }
1151 
1152 static bool __read_mostly ignore_loglevel;
1153 
1154 static int __init ignore_loglevel_setup(char *str)
1155 {
1156         ignore_loglevel = true;
1157         pr_info("debug: ignoring loglevel setting.\n");
1158 
1159         return 0;
1160 }
1161 
1162 early_param("ignore_loglevel", ignore_loglevel_setup);
1163 module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1164 MODULE_PARM_DESC(ignore_loglevel,
1165                  "ignore loglevel setting (prints all kernel messages to the console)");
1166 
1167 static bool suppress_message_printing(int level)
1168 {
1169         return (level >= console_loglevel && !ignore_loglevel);
1170 }
1171 
1172 #ifdef CONFIG_BOOT_PRINTK_DELAY
1173 
1174 static int boot_delay; /* msecs delay after each printk during bootup */
1175 static unsigned long long loops_per_msec;       /* based on boot_delay */
1176 
1177 static int __init boot_delay_setup(char *str)
1178 {
1179         unsigned long lpj;
1180 
1181         lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
1182         loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
1183 
1184         get_option(&str, &boot_delay);
1185         if (boot_delay > 10 * 1000)
1186                 boot_delay = 0;
1187 
1188         pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1189                 "HZ: %d, loops_per_msec: %llu\n",
1190                 boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
1191         return 0;
1192 }
1193 early_param("boot_delay", boot_delay_setup);
1194 
1195 static void boot_delay_msec(int level)
1196 {
1197         unsigned long long k;
1198         unsigned long timeout;
1199 
1200         if ((boot_delay == 0 || system_state >= SYSTEM_RUNNING)
1201                 || suppress_message_printing(level)) {
1202                 return;
1203         }
1204 
1205         k = (unsigned long long)loops_per_msec * boot_delay;
1206 
1207         timeout = jiffies + msecs_to_jiffies(boot_delay);
1208         while (k) {
1209                 k--;
1210                 cpu_relax();
1211                 /*
1212                  * use (volatile) jiffies to prevent
1213                  * compiler reduction; loop termination via jiffies
1214                  * is secondary and may or may not happen.
1215                  */
1216                 if (time_after(jiffies, timeout))
1217                         break;
1218                 touch_nmi_watchdog();
1219         }
1220 }
1221 #else
1222 static inline void boot_delay_msec(int level)
1223 {
1224 }
1225 #endif
1226 
1227 static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME);
1228 module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
1229 
1230 static size_t print_syslog(unsigned int level, char *buf)
1231 {
1232         return sprintf(buf, "<%u>", level);
1233 }
1234 
1235 static size_t print_time(u64 ts, char *buf)
1236 {
1237         unsigned long rem_nsec = do_div(ts, 1000000000);
1238 
1239         return sprintf(buf, "[%5lu.%06lu] ",
1240                        (unsigned long)ts, rem_nsec / 1000);
1241 }
1242 
1243 static size_t print_prefix(const struct printk_log *msg, bool syslog,
1244                            bool time, char *buf)
1245 {
1246         size_t len = 0;
1247 
1248         if (syslog)
1249                 len = print_syslog((msg->facility << 3) | msg->level, buf);
1250         if (time)
1251                 len += print_time(msg->ts_nsec, buf + len);
1252         return len;
1253 }
1254 
1255 static size_t msg_print_text(const struct printk_log *msg, bool syslog,
1256                              bool time, char *buf, size_t size)
1257 {
1258         const char *text = log_text(msg);
1259         size_t text_size = msg->text_len;
1260         size_t len = 0;
1261         char prefix[PREFIX_MAX];
1262         const size_t prefix_len = print_prefix(msg, syslog, time, prefix);
1263 
1264         do {
1265                 const char *next = memchr(text, '\n', text_size);
1266                 size_t text_len;
1267 
1268                 if (next) {
1269                         text_len = next - text;
1270                         next++;
1271                         text_size -= next - text;
1272                 } else {
1273                         text_len = text_size;
1274                 }
1275 
1276                 if (buf) {
1277                         if (prefix_len + text_len + 1 >= size - len)
1278                                 break;
1279 
1280                         memcpy(buf + len, prefix, prefix_len);
1281                         len += prefix_len;
1282                         memcpy(buf + len, text, text_len);
1283                         len += text_len;
1284                         buf[len++] = '\n';
1285                 } else {
1286                         /* SYSLOG_ACTION_* buffer size only calculation */
1287                         len += prefix_len + text_len + 1;
1288                 }
1289 
1290                 text = next;
1291         } while (text);
1292 
1293         return len;
1294 }
1295 
1296 static int syslog_print(char __user *buf, int size)
1297 {
1298         char *text;
1299         struct printk_log *msg;
1300         int len = 0;
1301 
1302         text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1303         if (!text)
1304                 return -ENOMEM;
1305 
1306         while (size > 0) {
1307                 size_t n;
1308                 size_t skip;
1309 
1310                 logbuf_lock_irq();
1311                 if (syslog_seq < log_first_seq) {
1312                         /* messages are gone, move to first one */
1313                         syslog_seq = log_first_seq;
1314                         syslog_idx = log_first_idx;
1315                         syslog_partial = 0;
1316                 }
1317                 if (syslog_seq == log_next_seq) {
1318                         logbuf_unlock_irq();
1319                         break;
1320                 }
1321 
1322                 /*
1323                  * To keep reading/counting partial line consistent,
1324                  * use printk_time value as of the beginning of a line.
1325                  */
1326                 if (!syslog_partial)
1327                         syslog_time = printk_time;
1328 
1329                 skip = syslog_partial;
1330                 msg = log_from_idx(syslog_idx);
1331                 n = msg_print_text(msg, true, syslog_time, text,
1332                                    LOG_LINE_MAX + PREFIX_MAX);
1333                 if (n - syslog_partial <= size) {
1334                         /* message fits into buffer, move forward */
1335                         syslog_idx = log_next(syslog_idx);
1336                         syslog_seq++;
1337                         n -= syslog_partial;
1338                         syslog_partial = 0;
1339                 } else if (!len){
1340                         /* partial read(), remember position */
1341                         n = size;
1342                         syslog_partial += n;
1343                 } else
1344                         n = 0;
1345                 logbuf_unlock_irq();
1346 
1347                 if (!n)
1348                         break;
1349 
1350                 if (copy_to_user(buf, text + skip, n)) {
1351                         if (!len)
1352                                 len = -EFAULT;
1353                         break;
1354                 }
1355 
1356                 len += n;
1357                 size -= n;
1358                 buf += n;
1359         }
1360 
1361         kfree(text);
1362         return len;
1363 }
1364 
1365 static int syslog_print_all(char __user *buf, int size, bool clear)
1366 {
1367         char *text;
1368         int len = 0;
1369         u64 next_seq;
1370         u64 seq;
1371         u32 idx;
1372         bool time;
1373 
1374         text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1375         if (!text)
1376                 return -ENOMEM;
1377 
1378         time = printk_time;
1379         logbuf_lock_irq();
1380         /*
1381          * Find first record that fits, including all following records,
1382          * into the user-provided buffer for this dump.
1383          */
1384         seq = clear_seq;
1385         idx = clear_idx;
1386         while (seq < log_next_seq) {
1387                 struct printk_log *msg = log_from_idx(idx);
1388 
1389                 len += msg_print_text(msg, true, time, NULL, 0);
1390                 idx = log_next(idx);
1391                 seq++;
1392         }
1393 
1394         /* move first record forward until length fits into the buffer */
1395         seq = clear_seq;
1396         idx = clear_idx;
1397         while (len > size && seq < log_next_seq) {
1398                 struct printk_log *msg = log_from_idx(idx);
1399 
1400                 len -= msg_print_text(msg, true, time, NULL, 0);
1401                 idx = log_next(idx);
1402                 seq++;
1403         }
1404 
1405         /* last message fitting into this dump */
1406         next_seq = log_next_seq;
1407 
1408         len = 0;
1409         while (len >= 0 && seq < next_seq) {
1410                 struct printk_log *msg = log_from_idx(idx);
1411                 int textlen = msg_print_text(msg, true, time, text,
1412                                              LOG_LINE_MAX + PREFIX_MAX);
1413 
1414                 idx = log_next(idx);
1415                 seq++;
1416 
1417                 logbuf_unlock_irq();
1418                 if (copy_to_user(buf + len, text, textlen))
1419                         len = -EFAULT;
1420                 else
1421                         len += textlen;
1422                 logbuf_lock_irq();
1423 
1424                 if (seq < log_first_seq) {
1425                         /* messages are gone, move to next one */
1426                         seq = log_first_seq;
1427                         idx = log_first_idx;
1428                 }
1429         }
1430 
1431         if (clear) {
1432                 clear_seq = log_next_seq;
1433                 clear_idx = log_next_idx;
1434         }
1435         logbuf_unlock_irq();
1436 
1437         kfree(text);
1438         return len;
1439 }
1440 
1441 static void syslog_clear(void)
1442 {
1443         logbuf_lock_irq();
1444         clear_seq = log_next_seq;
1445         clear_idx = log_next_idx;
1446         logbuf_unlock_irq();
1447 }
1448 
1449 int do_syslog(int type, char __user *buf, int len, int source)
1450 {
1451         bool clear = false;
1452         static int saved_console_loglevel = LOGLEVEL_DEFAULT;
1453         int error;
1454 
1455         error = check_syslog_permissions(type, source);
1456         if (error)
1457                 return error;
1458 
1459         switch (type) {
1460         case SYSLOG_ACTION_CLOSE:       /* Close log */
1461                 break;
1462         case SYSLOG_ACTION_OPEN:        /* Open log */
1463                 break;
1464         case SYSLOG_ACTION_READ:        /* Read from log */
1465                 if (!buf || len < 0)
1466                         return -EINVAL;
1467                 if (!len)
1468                         return 0;
1469                 if (!access_ok(buf, len))
1470                         return -EFAULT;
1471                 error = wait_event_interruptible(log_wait,
1472                                                  syslog_seq != log_next_seq);
1473                 if (error)
1474                         return error;
1475                 error = syslog_print(buf, len);
1476                 break;
1477         /* Read/clear last kernel messages */
1478         case SYSLOG_ACTION_READ_CLEAR:
1479                 clear = true;
1480                 /* FALL THRU */
1481         /* Read last kernel messages */
1482         case SYSLOG_ACTION_READ_ALL:
1483                 if (!buf || len < 0)
1484                         return -EINVAL;
1485                 if (!len)
1486                         return 0;
1487                 if (!access_ok(buf, len))
1488                         return -EFAULT;
1489                 error = syslog_print_all(buf, len, clear);
1490                 break;
1491         /* Clear ring buffer */
1492         case SYSLOG_ACTION_CLEAR:
1493                 syslog_clear();
1494                 break;
1495         /* Disable logging to console */
1496         case SYSLOG_ACTION_CONSOLE_OFF:
1497                 if (saved_console_loglevel == LOGLEVEL_DEFAULT)
1498                         saved_console_loglevel = console_loglevel;
1499                 console_loglevel = minimum_console_loglevel;
1500                 break;
1501         /* Enable logging to console */
1502         case SYSLOG_ACTION_CONSOLE_ON:
1503                 if (saved_console_loglevel != LOGLEVEL_DEFAULT) {
1504                         console_loglevel = saved_console_loglevel;
1505                         saved_console_loglevel = LOGLEVEL_DEFAULT;
1506                 }
1507                 break;
1508         /* Set level of messages printed to console */
1509         case SYSLOG_ACTION_CONSOLE_LEVEL:
1510                 if (len < 1 || len > 8)
1511                         return -EINVAL;
1512                 if (len < minimum_console_loglevel)
1513                         len = minimum_console_loglevel;
1514                 console_loglevel = len;
1515                 /* Implicitly re-enable logging to console */
1516                 saved_console_loglevel = LOGLEVEL_DEFAULT;
1517                 break;
1518         /* Number of chars in the log buffer */
1519         case SYSLOG_ACTION_SIZE_UNREAD:
1520                 logbuf_lock_irq();
1521                 if (syslog_seq < log_first_seq) {
1522                         /* messages are gone, move to first one */
1523                         syslog_seq = log_first_seq;
1524                         syslog_idx = log_first_idx;
1525                         syslog_partial = 0;
1526                 }
1527                 if (source == SYSLOG_FROM_PROC) {
1528                         /*
1529                          * Short-cut for poll(/"proc/kmsg") which simply checks
1530                          * for pending data, not the size; return the count of
1531                          * records, not the length.
1532                          */
1533                         error = log_next_seq - syslog_seq;
1534                 } else {
1535                         u64 seq = syslog_seq;
1536                         u32 idx = syslog_idx;
1537                         bool time = syslog_partial ? syslog_time : printk_time;
1538 
1539                         while (seq < log_next_seq) {
1540                                 struct printk_log *msg = log_from_idx(idx);
1541 
1542                                 error += msg_print_text(msg, true, time, NULL,
1543                                                         0);
1544                                 time = printk_time;
1545                                 idx = log_next(idx);
1546                                 seq++;
1547                         }
1548                         error -= syslog_partial;
1549                 }
1550                 logbuf_unlock_irq();
1551                 break;
1552         /* Size of the log buffer */
1553         case SYSLOG_ACTION_SIZE_BUFFER:
1554                 error = log_buf_len;
1555                 break;
1556         default:
1557                 error = -EINVAL;
1558                 break;
1559         }
1560 
1561         return error;
1562 }
1563 
1564 SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1565 {
1566         return do_syslog(type, buf, len, SYSLOG_FROM_READER);
1567 }
1568 
1569 /*
1570  * Special console_lock variants that help to reduce the risk of soft-lockups.
1571  * They allow to pass console_lock to another printk() call using a busy wait.
1572  */
1573 
1574 #ifdef CONFIG_LOCKDEP
1575 static struct lockdep_map console_owner_dep_map = {
1576         .name = "console_owner"
1577 };
1578 #endif
1579 
1580 static DEFINE_RAW_SPINLOCK(console_owner_lock);
1581 static struct task_struct *console_owner;
1582 static bool console_waiter;
1583 
1584 /**
1585  * console_lock_spinning_enable - mark beginning of code where another
1586  *      thread might safely busy wait
1587  *
1588  * This basically converts console_lock into a spinlock. This marks
1589  * the section where the console_lock owner can not sleep, because
1590  * there may be a waiter spinning (like a spinlock). Also it must be
1591  * ready to hand over the lock at the end of the section.
1592  */
1593 static void console_lock_spinning_enable(void)
1594 {
1595         raw_spin_lock(&console_owner_lock);
1596         console_owner = current;
1597         raw_spin_unlock(&console_owner_lock);
1598 
1599         /* The waiter may spin on us after setting console_owner */
1600         spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1601 }
1602 
1603 /**
1604  * console_lock_spinning_disable_and_check - mark end of code where another
1605  *      thread was able to busy wait and check if there is a waiter
1606  *
1607  * This is called at the end of the section where spinning is allowed.
1608  * It has two functions. First, it is a signal that it is no longer
1609  * safe to start busy waiting for the lock. Second, it checks if
1610  * there is a busy waiter and passes the lock rights to her.
1611  *
1612  * Important: Callers lose the lock if there was a busy waiter.
1613  *      They must not touch items synchronized by console_lock
1614  *      in this case.
1615  *
1616  * Return: 1 if the lock rights were passed, 0 otherwise.
1617  */
1618 static int console_lock_spinning_disable_and_check(void)
1619 {
1620         int waiter;
1621 
1622         raw_spin_lock(&console_owner_lock);
1623         waiter = READ_ONCE(console_waiter);
1624         console_owner = NULL;
1625         raw_spin_unlock(&console_owner_lock);
1626 
1627         if (!waiter) {
1628                 spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1629                 return 0;
1630         }
1631 
1632         /* The waiter is now free to continue */
1633         WRITE_ONCE(console_waiter, false);
1634 
1635         spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1636 
1637         /*
1638          * Hand off console_lock to waiter. The waiter will perform
1639          * the up(). After this, the waiter is the console_lock owner.
1640          */
1641         mutex_release(&console_lock_dep_map, 1, _THIS_IP_);
1642         return 1;
1643 }
1644 
1645 /**
1646  * console_trylock_spinning - try to get console_lock by busy waiting
1647  *
1648  * This allows to busy wait for the console_lock when the current
1649  * owner is running in specially marked sections. It means that
1650  * the current owner is running and cannot reschedule until it
1651  * is ready to lose the lock.
1652  *
1653  * Return: 1 if we got the lock, 0 othrewise
1654  */
1655 static int console_trylock_spinning(void)
1656 {
1657         struct task_struct *owner = NULL;
1658         bool waiter;
1659         bool spin = false;
1660         unsigned long flags;
1661 
1662         if (console_trylock())
1663                 return 1;
1664 
1665         printk_safe_enter_irqsave(flags);
1666 
1667         raw_spin_lock(&console_owner_lock);
1668         owner = READ_ONCE(console_owner);
1669         waiter = READ_ONCE(console_waiter);
1670         if (!waiter && owner && owner != current) {
1671                 WRITE_ONCE(console_waiter, true);
1672                 spin = true;
1673         }
1674         raw_spin_unlock(&console_owner_lock);
1675 
1676         /*
1677          * If there is an active printk() writing to the
1678          * consoles, instead of having it write our data too,
1679          * see if we can offload that load from the active
1680          * printer, and do some printing ourselves.
1681          * Go into a spin only if there isn't already a waiter
1682          * spinning, and there is an active printer, and
1683          * that active printer isn't us (recursive printk?).
1684          */
1685         if (!spin) {
1686                 printk_safe_exit_irqrestore(flags);
1687                 return 0;
1688         }
1689 
1690         /* We spin waiting for the owner to release us */
1691         spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1692         /* Owner will clear console_waiter on hand off */
1693         while (READ_ONCE(console_waiter))
1694                 cpu_relax();
1695         spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1696 
1697         printk_safe_exit_irqrestore(flags);
1698         /*
1699          * The owner passed the console lock to us.
1700          * Since we did not spin on console lock, annotate
1701          * this as a trylock. Otherwise lockdep will
1702          * complain.
1703          */
1704         mutex_acquire(&console_lock_dep_map, 0, 1, _THIS_IP_);
1705 
1706         return 1;
1707 }
1708 
1709 /*
1710  * Call the console drivers, asking them to write out
1711  * log_buf[start] to log_buf[end - 1].
1712  * The console_lock must be held.
1713  */
1714 static void call_console_drivers(const char *ext_text, size_t ext_len,
1715                                  const char *text, size_t len)
1716 {
1717         struct console *con;
1718 
1719         trace_console_rcuidle(text, len);
1720 
1721         if (!console_drivers)
1722                 return;
1723 
1724         for_each_console(con) {
1725                 if (exclusive_console && con != exclusive_console)
1726                         continue;
1727                 if (!(con->flags & CON_ENABLED))
1728                         continue;
1729                 if (!con->write)
1730                         continue;
1731                 if (!cpu_online(smp_processor_id()) &&
1732                     !(con->flags & CON_ANYTIME))
1733                         continue;
1734                 if (con->flags & CON_EXTENDED)
1735                         con->write(con, ext_text, ext_len);
1736                 else
1737                         con->write(con, text, len);
1738         }
1739 }
1740 
1741 int printk_delay_msec __read_mostly;
1742 
1743 static inline void printk_delay(void)
1744 {
1745         if (unlikely(printk_delay_msec)) {
1746                 int m = printk_delay_msec;
1747 
1748                 while (m--) {
1749                         mdelay(1);
1750                         touch_nmi_watchdog();
1751                 }
1752         }
1753 }
1754 
1755 /*
1756  * Continuation lines are buffered, and not committed to the record buffer
1757  * until the line is complete, or a race forces it. The line fragments
1758  * though, are printed immediately to the consoles to ensure everything has
1759  * reached the console in case of a kernel crash.
1760  */
1761 static struct cont {
1762         char buf[LOG_LINE_MAX];
1763         size_t len;                     /* length == 0 means unused buffer */
1764         struct task_struct *owner;      /* task of first print*/
1765         u64 ts_nsec;                    /* time of first print */
1766         u8 level;                       /* log level of first message */
1767         u8 facility;                    /* log facility of first message */
1768         enum log_flags flags;           /* prefix, newline flags */
1769 } cont;
1770 
1771 static void cont_flush(void)
1772 {
1773         if (cont.len == 0)
1774                 return;
1775 
1776         log_store(cont.facility, cont.level, cont.flags, cont.ts_nsec,
1777                   NULL, 0, cont.buf, cont.len);
1778         cont.len = 0;
1779 }
1780 
1781 static bool cont_add(int facility, int level, enum log_flags flags, const char *text, size_t len)
1782 {
1783         /* If the line gets too long, split it up in separate records. */
1784         if (cont.len + len > sizeof(cont.buf)) {
1785                 cont_flush();
1786                 return false;
1787         }
1788 
1789         if (!cont.len) {
1790                 cont.facility = facility;
1791                 cont.level = level;
1792                 cont.owner = current;
1793                 cont.ts_nsec = local_clock();
1794                 cont.flags = flags;
1795         }
1796 
1797         memcpy(cont.buf + cont.len, text, len);
1798         cont.len += len;
1799 
1800         // The original flags come from the first line,
1801         // but later continuations can add a newline.
1802         if (flags & LOG_NEWLINE) {
1803                 cont.flags |= LOG_NEWLINE;
1804                 cont_flush();
1805         }
1806 
1807         return true;
1808 }
1809 
1810 static size_t log_output(int facility, int level, enum log_flags lflags, const char *dict, size_t dictlen, char *text, size_t text_len)
1811 {
1812         /*
1813          * If an earlier line was buffered, and we're a continuation
1814          * write from the same process, try to add it to the buffer.
1815          */
1816         if (cont.len) {
1817                 if (cont.owner == current && (lflags & LOG_CONT)) {
1818                         if (cont_add(facility, level, lflags, text, text_len))
1819                                 return text_len;
1820                 }
1821                 /* Otherwise, make sure it's flushed */
1822                 cont_flush();
1823         }
1824 
1825         /* Skip empty continuation lines that couldn't be added - they just flush */
1826         if (!text_len && (lflags & LOG_CONT))
1827                 return 0;
1828 
1829         /* If it doesn't end in a newline, try to buffer the current line */
1830         if (!(lflags & LOG_NEWLINE)) {
1831                 if (cont_add(facility, level, lflags, text, text_len))
1832                         return text_len;
1833         }
1834 
1835         /* Store it in the record log */
1836         return log_store(facility, level, lflags, 0, dict, dictlen, text, text_len);
1837 }
1838 
1839 /* Must be called under logbuf_lock. */
1840 int vprintk_store(int facility, int level,
1841                   const char *dict, size_t dictlen,
1842                   const char *fmt, va_list args)
1843 {
1844         static char textbuf[LOG_LINE_MAX];
1845         char *text = textbuf;
1846         size_t text_len;
1847         enum log_flags lflags = 0;
1848 
1849         /*
1850          * The printf needs to come first; we need the syslog
1851          * prefix which might be passed-in as a parameter.
1852          */
1853         text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1854 
1855         /* mark and strip a trailing newline */
1856         if (text_len && text[text_len-1] == '\n') {
1857                 text_len--;
1858                 lflags |= LOG_NEWLINE;
1859         }
1860 
1861         /* strip kernel syslog prefix and extract log level or control flags */
1862         if (facility == 0) {
1863                 int kern_level;
1864 
1865                 while ((kern_level = printk_get_level(text)) != 0) {
1866                         switch (kern_level) {
1867                         case '' ... '7':
1868                                 if (level == LOGLEVEL_DEFAULT)
1869                                         level = kern_level - '';
1870                                 /* fallthrough */
1871                         case 'd':       /* KERN_DEFAULT */
1872                                 lflags |= LOG_PREFIX;
1873                                 break;
1874                         case 'c':       /* KERN_CONT */
1875                                 lflags |= LOG_CONT;
1876                         }
1877 
1878                         text_len -= 2;
1879                         text += 2;
1880                 }
1881         }
1882 
1883         if (level == LOGLEVEL_DEFAULT)
1884                 level = default_message_loglevel;
1885 
1886         if (dict)
1887                 lflags |= LOG_PREFIX|LOG_NEWLINE;
1888 
1889         return log_output(facility, level, lflags,
1890                           dict, dictlen, text, text_len);
1891 }
1892 
1893 asmlinkage int vprintk_emit(int facility, int level,
1894                             const char *dict, size_t dictlen,
1895                             const char *fmt, va_list args)
1896 {
1897         int printed_len;
1898         bool in_sched = false, pending_output;
1899         unsigned long flags;
1900         u64 curr_log_seq;
1901 
1902         if (level == LOGLEVEL_SCHED) {
1903                 level = LOGLEVEL_DEFAULT;
1904                 in_sched = true;
1905         }
1906 
1907         boot_delay_msec(level);
1908         printk_delay();
1909 
1910         /* This stops the holder of console_sem just where we want him */
1911         logbuf_lock_irqsave(flags);
1912         curr_log_seq = log_next_seq;
1913         printed_len = vprintk_store(facility, level, dict, dictlen, fmt, args);
1914         pending_output = (curr_log_seq != log_next_seq);
1915         logbuf_unlock_irqrestore(flags);
1916 
1917         /* If called from the scheduler, we can not call up(). */
1918         if (!in_sched && pending_output) {
1919                 /*
1920                  * Disable preemption to avoid being preempted while holding
1921                  * console_sem which would prevent anyone from printing to
1922                  * console
1923                  */
1924                 preempt_disable();
1925                 /*
1926                  * Try to acquire and then immediately release the console
1927                  * semaphore.  The release will print out buffers and wake up
1928                  * /dev/kmsg and syslog() users.
1929                  */
1930                 if (console_trylock_spinning())
1931                         console_unlock();
1932                 preempt_enable();
1933         }
1934 
1935         if (pending_output)
1936                 wake_up_klogd();
1937         return printed_len;
1938 }
1939 EXPORT_SYMBOL(vprintk_emit);
1940 
1941 asmlinkage int vprintk(const char *fmt, va_list args)
1942 {
1943         return vprintk_func(fmt, args);
1944 }
1945 EXPORT_SYMBOL(vprintk);
1946 
1947 int vprintk_default(const char *fmt, va_list args)
1948 {
1949         int r;
1950 
1951 #ifdef CONFIG_KGDB_KDB
1952         /* Allow to pass printk() to kdb but avoid a recursion. */
1953         if (unlikely(kdb_trap_printk && kdb_printf_cpu < 0)) {
1954                 r = vkdb_printf(KDB_MSGSRC_PRINTK, fmt, args);
1955                 return r;
1956         }
1957 #endif
1958         r = vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
1959 
1960         return r;
1961 }
1962 EXPORT_SYMBOL_GPL(vprintk_default);
1963 
1964 /**
1965  * printk - print a kernel message
1966  * @fmt: format string
1967  *
1968  * This is printk(). It can be called from any context. We want it to work.
1969  *
1970  * We try to grab the console_lock. If we succeed, it's easy - we log the
1971  * output and call the console drivers.  If we fail to get the semaphore, we
1972  * place the output into the log buffer and return. The current holder of
1973  * the console_sem will notice the new output in console_unlock(); and will
1974  * send it to the consoles before releasing the lock.
1975  *
1976  * One effect of this deferred printing is that code which calls printk() and
1977  * then changes console_loglevel may break. This is because console_loglevel
1978  * is inspected when the actual printing occurs.
1979  *
1980  * See also:
1981  * printf(3)
1982  *
1983  * See the vsnprintf() documentation for format string extensions over C99.
1984  */
1985 asmlinkage __visible int printk(const char *fmt, ...)
1986 {
1987         va_list args;
1988         int r;
1989 
1990         va_start(args, fmt);
1991         r = vprintk_func(fmt, args);
1992         va_end(args);
1993 
1994         return r;
1995 }
1996 EXPORT_SYMBOL(printk);
1997 
1998 #else /* CONFIG_PRINTK */
1999 
2000 #define LOG_LINE_MAX            0
2001 #define PREFIX_MAX              0
2002 #define printk_time             false
2003 
2004 static u64 syslog_seq;
2005 static u32 syslog_idx;
2006 static u64 console_seq;
2007 static u32 console_idx;
2008 static u64 exclusive_console_stop_seq;
2009 static u64 log_first_seq;
2010 static u32 log_first_idx;
2011 static u64 log_next_seq;
2012 static char *log_text(const struct printk_log *msg) { return NULL; }
2013 static char *log_dict(const struct printk_log *msg) { return NULL; }
2014 static struct printk_log *log_from_idx(u32 idx) { return NULL; }
2015 static u32 log_next(u32 idx) { return 0; }
2016 static ssize_t msg_print_ext_header(char *buf, size_t size,
2017                                     struct printk_log *msg,
2018                                     u64 seq) { return 0; }
2019 static ssize_t msg_print_ext_body(char *buf, size_t size,
2020                                   char *dict, size_t dict_len,
2021                                   char *text, size_t text_len) { return 0; }
2022 static void console_lock_spinning_enable(void) { }
2023 static int console_lock_spinning_disable_and_check(void) { return 0; }
2024 static void call_console_drivers(const char *ext_text, size_t ext_len,
2025                                  const char *text, size_t len) {}
2026 static size_t msg_print_text(const struct printk_log *msg, bool syslog,
2027                              bool time, char *buf, size_t size) { return 0; }
2028 static bool suppress_message_printing(int level) { return false; }
2029 
2030 #endif /* CONFIG_PRINTK */
2031 
2032 #ifdef CONFIG_EARLY_PRINTK
2033 struct console *early_console;
2034 
2035 asmlinkage __visible void early_printk(const char *fmt, ...)
2036 {
2037         va_list ap;
2038         char buf[512];
2039         int n;
2040 
2041         if (!early_console)
2042                 return;
2043 
2044         va_start(ap, fmt);
2045         n = vscnprintf(buf, sizeof(buf), fmt, ap);
2046         va_end(ap);
2047 
2048         early_console->write(early_console, buf, n);
2049 }
2050 #endif
2051 
2052 static int __add_preferred_console(char *name, int idx, char *options,
2053                                    char *brl_options)
2054 {
2055         struct console_cmdline *c;
2056         int i;
2057 
2058         /*
2059          *      See if this tty is not yet registered, and
2060          *      if we have a slot free.
2061          */
2062         for (i = 0, c = console_cmdline;
2063              i < MAX_CMDLINECONSOLES && c->name[0];
2064              i++, c++) {
2065                 if (strcmp(c->name, name) == 0 && c->index == idx) {
2066                         if (!brl_options)
2067                                 preferred_console = i;
2068                         return 0;
2069                 }
2070         }
2071         if (i == MAX_CMDLINECONSOLES)
2072                 return -E2BIG;
2073         if (!brl_options)
2074                 preferred_console = i;
2075         strlcpy(c->name, name, sizeof(c->name));
2076         c->options = options;
2077         braille_set_options(c, brl_options);
2078 
2079         c->index = idx;
2080         return 0;
2081 }
2082 
2083 static int __init console_msg_format_setup(char *str)
2084 {
2085         if (!strcmp(str, "syslog"))
2086                 console_msg_format = MSG_FORMAT_SYSLOG;
2087         if (!strcmp(str, "default"))
2088                 console_msg_format = MSG_FORMAT_DEFAULT;
2089         return 1;
2090 }
2091 __setup("console_msg_format=", console_msg_format_setup);
2092 
2093 /*
2094  * Set up a console.  Called via do_early_param() in init/main.c
2095  * for each "console=" parameter in the boot command line.
2096  */
2097 static int __init console_setup(char *str)
2098 {
2099         char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */
2100         char *s, *options, *brl_options = NULL;
2101         int idx;
2102 
2103         if (_braille_console_setup(&str, &brl_options))
2104                 return 1;
2105 
2106         /*
2107          * Decode str into name, index, options.
2108          */
2109         if (str[0] >= '' && str[0] <= '9') {
2110                 strcpy(buf, "ttyS");
2111                 strncpy(buf + 4, str, sizeof(buf) - 5);
2112         } else {
2113                 strncpy(buf, str, sizeof(buf) - 1);
2114         }
2115         buf[sizeof(buf) - 1] = 0;
2116         options = strchr(str, ',');
2117         if (options)
2118                 *(options++) = 0;
2119 #ifdef __sparc__
2120         if (!strcmp(str, "ttya"))
2121                 strcpy(buf, "ttyS0");
2122         if (!strcmp(str, "ttyb"))
2123                 strcpy(buf, "ttyS1");
2124 #endif
2125         for (s = buf; *s; s++)
2126                 if (isdigit(*s) || *s == ',')
2127                         break;
2128         idx = simple_strtoul(s, NULL, 10);
2129         *s = 0;
2130 
2131         __add_preferred_console(buf, idx, options, brl_options);
2132         console_set_on_cmdline = 1;
2133         return 1;
2134 }
2135 __setup("console=", console_setup);
2136 
2137 /**
2138  * add_preferred_console - add a device to the list of preferred consoles.
2139  * @name: device name
2140  * @idx: device index
2141  * @options: options for this console
2142  *
2143  * The last preferred console added will be used for kernel messages
2144  * and stdin/out/err for init.  Normally this is used by console_setup
2145  * above to handle user-supplied console arguments; however it can also
2146  * be used by arch-specific code either to override the user or more
2147  * commonly to provide a default console (ie from PROM variables) when
2148  * the user has not supplied one.
2149  */
2150 int add_preferred_console(char *name, int idx, char *options)
2151 {
2152         return __add_preferred_console(name, idx, options, NULL);
2153 }
2154 
2155 bool console_suspend_enabled = true;
2156 EXPORT_SYMBOL(console_suspend_enabled);
2157 
2158 static int __init console_suspend_disable(char *str)
2159 {
2160         console_suspend_enabled = false;
2161         return 1;
2162 }
2163 __setup("no_console_suspend", console_suspend_disable);
2164 module_param_named(console_suspend, console_suspend_enabled,
2165                 bool, S_IRUGO | S_IWUSR);
2166 MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
2167         " and hibernate operations");
2168 
2169 /**
2170  * suspend_console - suspend the console subsystem
2171  *
2172  * This disables printk() while we go into suspend states
2173  */
2174 void suspend_console(void)
2175 {
2176         if (!console_suspend_enabled)
2177                 return;
2178         pr_info("Suspending console(s) (use no_console_suspend to debug)\n");
2179         console_lock();
2180         console_suspended = 1;
2181         up_console_sem();
2182 }
2183 
2184 void resume_console(void)
2185 {
2186         if (!console_suspend_enabled)
2187                 return;
2188         down_console_sem();
2189         console_suspended = 0;
2190         console_unlock();
2191 }
2192 
2193 /**
2194  * console_cpu_notify - print deferred console messages after CPU hotplug
2195  * @cpu: unused
2196  *
2197  * If printk() is called from a CPU that is not online yet, the messages
2198  * will be printed on the console only if there are CON_ANYTIME consoles.
2199  * This function is called when a new CPU comes online (or fails to come
2200  * up) or goes offline.
2201  */
2202 static int console_cpu_notify(unsigned int cpu)
2203 {
2204         if (!cpuhp_tasks_frozen) {
2205                 /* If trylock fails, someone else is doing the printing */
2206                 if (console_trylock())
2207                         console_unlock();
2208         }
2209         return 0;
2210 }
2211 
2212 /**
2213  * console_lock - lock the console system for exclusive use.
2214  *
2215  * Acquires a lock which guarantees that the caller has
2216  * exclusive access to the console system and the console_drivers list.
2217  *
2218  * Can sleep, returns nothing.
2219  */
2220 void console_lock(void)
2221 {
2222         might_sleep();
2223 
2224         down_console_sem();
2225         if (console_suspended)
2226                 return;
2227         console_locked = 1;
2228         console_may_schedule = 1;
2229 }
2230 EXPORT_SYMBOL(console_lock);
2231 
2232 /**
2233  * console_trylock - try to lock the console system for exclusive use.
2234  *
2235  * Try to acquire a lock which guarantees that the caller has exclusive
2236  * access to the console system and the console_drivers list.
2237  *
2238  * returns 1 on success, and 0 on failure to acquire the lock.
2239  */
2240 int console_trylock(void)
2241 {
2242         if (down_trylock_console_sem())
2243                 return 0;
2244         if (console_suspended) {
2245                 up_console_sem();
2246                 return 0;
2247         }
2248         console_locked = 1;
2249         console_may_schedule = 0;
2250         return 1;
2251 }
2252 EXPORT_SYMBOL(console_trylock);
2253 
2254 int is_console_locked(void)
2255 {
2256         return console_locked;
2257 }
2258 EXPORT_SYMBOL(is_console_locked);
2259 
2260 /*
2261  * Check if we have any console that is capable of printing while cpu is
2262  * booting or shutting down. Requires console_sem.
2263  */
2264 static int have_callable_console(void)
2265 {
2266         struct console *con;
2267 
2268         for_each_console(con)
2269                 if ((con->flags & CON_ENABLED) &&
2270                                 (con->flags & CON_ANYTIME))
2271                         return 1;
2272 
2273         return 0;
2274 }
2275 
2276 /*
2277  * Can we actually use the console at this time on this cpu?
2278  *
2279  * Console drivers may assume that per-cpu resources have been allocated. So
2280  * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2281  * call them until this CPU is officially up.
2282  */
2283 static inline int can_use_console(void)
2284 {
2285         return cpu_online(raw_smp_processor_id()) || have_callable_console();
2286 }
2287 
2288 /**
2289  * console_unlock - unlock the console system
2290  *
2291  * Releases the console_lock which the caller holds on the console system
2292  * and the console driver list.
2293  *
2294  * While the console_lock was held, console output may have been buffered
2295  * by printk().  If this is the case, console_unlock(); emits
2296  * the output prior to releasing the lock.
2297  *
2298  * If there is output waiting, we wake /dev/kmsg and syslog() users.
2299  *
2300  * console_unlock(); may be called from any context.
2301  */
2302 void console_unlock(void)
2303 {
2304         static char ext_text[CONSOLE_EXT_LOG_MAX];
2305         static char text[LOG_LINE_MAX + PREFIX_MAX];
2306         unsigned long flags;
2307         bool do_cond_resched, retry;
2308 
2309         if (console_suspended) {
2310                 up_console_sem();
2311                 return;
2312         }
2313 
2314         /*
2315          * Console drivers are called with interrupts disabled, so
2316          * @console_may_schedule should be cleared before; however, we may
2317          * end up dumping a lot of lines, for example, if called from
2318          * console registration path, and should invoke cond_resched()
2319          * between lines if allowable.  Not doing so can cause a very long
2320          * scheduling stall on a slow console leading to RCU stall and
2321          * softlockup warnings which exacerbate the issue with more
2322          * messages practically incapacitating the system.
2323          *
2324          * console_trylock() is not able to detect the preemptive
2325          * context reliably. Therefore the value must be stored before
2326          * and cleared after the the "again" goto label.
2327          */
2328         do_cond_resched = console_may_schedule;
2329 again:
2330         console_may_schedule = 0;
2331 
2332         /*
2333          * We released the console_sem lock, so we need to recheck if
2334          * cpu is online and (if not) is there at least one CON_ANYTIME
2335          * console.
2336          */
2337         if (!can_use_console()) {
2338                 console_locked = 0;
2339                 up_console_sem();
2340                 return;
2341         }
2342 
2343         for (;;) {
2344                 struct printk_log *msg;
2345                 size_t ext_len = 0;
2346                 size_t len;
2347 
2348                 printk_safe_enter_irqsave(flags);
2349                 raw_spin_lock(&logbuf_lock);
2350                 if (console_seq < log_first_seq) {
2351                         len = sprintf(text,
2352                                       "** %llu printk messages dropped **\n",
2353                                       log_first_seq - console_seq);
2354 
2355                         /* messages are gone, move to first one */
2356                         console_seq = log_first_seq;
2357                         console_idx = log_first_idx;
2358                 } else {
2359                         len = 0;
2360                 }
2361 skip:
2362                 if (console_seq == log_next_seq)
2363                         break;
2364 
2365                 msg = log_from_idx(console_idx);
2366                 if (suppress_message_printing(msg->level)) {
2367                         /*
2368                          * Skip record we have buffered and already printed
2369                          * directly to the console when we received it, and
2370                          * record that has level above the console loglevel.
2371                          */
2372                         console_idx = log_next(console_idx);
2373                         console_seq++;
2374                         goto skip;
2375                 }
2376 
2377                 /* Output to all consoles once old messages replayed. */
2378                 if (unlikely(exclusive_console &&
2379                              console_seq >= exclusive_console_stop_seq)) {
2380                         exclusive_console = NULL;
2381                 }
2382 
2383                 len += msg_print_text(msg,
2384                                 console_msg_format & MSG_FORMAT_SYSLOG,
2385                                 printk_time, text + len, sizeof(text) - len);
2386                 if (nr_ext_console_drivers) {
2387                         ext_len = msg_print_ext_header(ext_text,
2388                                                 sizeof(ext_text),
2389                                                 msg, console_seq);
2390                         ext_len += msg_print_ext_body(ext_text + ext_len,
2391                                                 sizeof(ext_text) - ext_len,
2392                                                 log_dict(msg), msg->dict_len,
2393                                                 log_text(msg), msg->text_len);
2394                 }
2395                 console_idx = log_next(console_idx);
2396                 console_seq++;
2397                 raw_spin_unlock(&logbuf_lock);
2398 
2399                 /*
2400                  * While actively printing out messages, if another printk()
2401                  * were to occur on another CPU, it may wait for this one to
2402                  * finish. This task can not be preempted if there is a
2403                  * waiter waiting to take over.
2404                  */
2405                 console_lock_spinning_enable();
2406 
2407                 stop_critical_timings();        /* don't trace print latency */
2408                 call_console_drivers(ext_text, ext_len, text, len);
2409                 start_critical_timings();
2410 
2411                 if (console_lock_spinning_disable_and_check()) {
2412                         printk_safe_exit_irqrestore(flags);
2413                         return;
2414                 }
2415 
2416                 printk_safe_exit_irqrestore(flags);
2417 
2418                 if (do_cond_resched)
2419                         cond_resched();
2420         }
2421 
2422         console_locked = 0;
2423 
2424         raw_spin_unlock(&logbuf_lock);
2425 
2426         up_console_sem();
2427 
2428         /*
2429          * Someone could have filled up the buffer again, so re-check if there's
2430          * something to flush. In case we cannot trylock the console_sem again,
2431          * there's a new owner and the console_unlock() from them will do the
2432          * flush, no worries.
2433          */
2434         raw_spin_lock(&logbuf_lock);
2435         retry = console_seq != log_next_seq;
2436         raw_spin_unlock(&logbuf_lock);
2437         printk_safe_exit_irqrestore(flags);
2438 
2439         if (retry && console_trylock())
2440                 goto again;
2441 }
2442 EXPORT_SYMBOL(console_unlock);
2443 
2444 /**
2445  * console_conditional_schedule - yield the CPU if required
2446  *
2447  * If the console code is currently allowed to sleep, and
2448  * if this CPU should yield the CPU to another task, do
2449  * so here.
2450  *
2451  * Must be called within console_lock();.
2452  */
2453 void __sched console_conditional_schedule(void)
2454 {
2455         if (console_may_schedule)
2456                 cond_resched();
2457 }
2458 EXPORT_SYMBOL(console_conditional_schedule);
2459 
2460 void console_unblank(void)
2461 {
2462         struct console *c;
2463 
2464         /*
2465          * console_unblank can no longer be called in interrupt context unless
2466          * oops_in_progress is set to 1..
2467          */
2468         if (oops_in_progress) {
2469                 if (down_trylock_console_sem() != 0)
2470                         return;
2471         } else
2472                 console_lock();
2473 
2474         console_locked = 1;
2475         console_may_schedule = 0;
2476         for_each_console(c)
2477                 if ((c->flags & CON_ENABLED) && c->unblank)
2478                         c->unblank();
2479         console_unlock();
2480 }
2481 
2482 /**
2483  * console_flush_on_panic - flush console content on panic
2484  *
2485  * Immediately output all pending messages no matter what.
2486  */
2487 void console_flush_on_panic(void)
2488 {
2489         /*
2490          * If someone else is holding the console lock, trylock will fail
2491          * and may_schedule may be set.  Ignore and proceed to unlock so
2492          * that messages are flushed out.  As this can be called from any
2493          * context and we don't want to get preempted while flushing,
2494          * ensure may_schedule is cleared.
2495          */
2496         console_trylock();
2497         console_may_schedule = 0;
2498         console_unlock();
2499 }
2500 
2501 /*
2502  * Return the console tty driver structure and its associated index
2503  */
2504 struct tty_driver *console_device(int *index)
2505 {
2506         struct console *c;
2507         struct tty_driver *driver = NULL;
2508 
2509         console_lock();
2510         for_each_console(c) {
2511                 if (!c->device)
2512                         continue;
2513                 driver = c->device(c, index);
2514                 if (driver)
2515                         break;
2516         }
2517         console_unlock();
2518         return driver;
2519 }
2520 
2521 /*
2522  * Prevent further output on the passed console device so that (for example)
2523  * serial drivers can disable console output before suspending a port, and can
2524  * re-enable output afterwards.
2525  */
2526 void console_stop(struct console *console)
2527 {
2528         console_lock();
2529         console->flags &= ~CON_ENABLED;
2530         console_unlock();
2531 }
2532 EXPORT_SYMBOL(console_stop);
2533 
2534 void console_start(struct console *console)
2535 {
2536         console_lock();
2537         console->flags |= CON_ENABLED;
2538         console_unlock();
2539 }
2540 EXPORT_SYMBOL(console_start);
2541 
2542 static int __read_mostly keep_bootcon;
2543 
2544 static int __init keep_bootcon_setup(char *str)
2545 {
2546         keep_bootcon = 1;
2547         pr_info("debug: skip boot console de-registration.\n");
2548 
2549         return 0;
2550 }
2551 
2552 early_param("keep_bootcon", keep_bootcon_setup);
2553 
2554 /*
2555  * The console driver calls this routine during kernel initialization
2556  * to register the console printing procedure with printk() and to
2557  * print any messages that were printed by the kernel before the
2558  * console driver was initialized.
2559  *
2560  * This can happen pretty early during the boot process (because of
2561  * early_printk) - sometimes before setup_arch() completes - be careful
2562  * of what kernel features are used - they may not be initialised yet.
2563  *
2564  * There are two types of consoles - bootconsoles (early_printk) and
2565  * "real" consoles (everything which is not a bootconsole) which are
2566  * handled differently.
2567  *  - Any number of bootconsoles can be registered at any time.
2568  *  - As soon as a "real" console is registered, all bootconsoles
2569  *    will be unregistered automatically.
2570  *  - Once a "real" console is registered, any attempt to register a
2571  *    bootconsoles will be rejected
2572  */
2573 void register_console(struct console *newcon)
2574 {
2575         int i;
2576         unsigned long flags;
2577         struct console *bcon = NULL;
2578         struct console_cmdline *c;
2579         static bool has_preferred;
2580 
2581         if (console_drivers)
2582                 for_each_console(bcon)
2583                         if (WARN(bcon == newcon,
2584                                         "console '%s%d' already registered\n",
2585                                         bcon->name, bcon->index))
2586                                 return;
2587 
2588         /*
2589          * before we register a new CON_BOOT console, make sure we don't
2590          * already have a valid console
2591          */
2592         if (console_drivers && newcon->flags & CON_BOOT) {
2593                 /* find the last or real console */
2594                 for_each_console(bcon) {
2595                         if (!(bcon->flags & CON_BOOT)) {
2596                                 pr_info("Too late to register bootconsole %s%d\n",
2597                                         newcon->name, newcon->index);
2598                                 return;
2599                         }
2600                 }
2601         }
2602 
2603         if (console_drivers && console_drivers->flags & CON_BOOT)
2604                 bcon = console_drivers;
2605 
2606         if (!has_preferred || bcon || !console_drivers)
2607                 has_preferred = preferred_console >= 0;
2608 
2609         /*
2610          *      See if we want to use this console driver. If we
2611          *      didn't select a console we take the first one
2612          *      that registers here.
2613          */
2614         if (!has_preferred) {
2615                 if (newcon->index < 0)
2616                         newcon->index = 0;
2617                 if (newcon->setup == NULL ||
2618                     newcon->setup(newcon, NULL) == 0) {
2619                         newcon->flags |= CON_ENABLED;
2620                         if (newcon->device) {
2621                                 newcon->flags |= CON_CONSDEV;
2622                                 has_preferred = true;
2623                         }
2624                 }
2625         }
2626 
2627         /*
2628          *      See if this console matches one we selected on
2629          *      the command line.
2630          */
2631         for (i = 0, c = console_cmdline;
2632              i < MAX_CMDLINECONSOLES && c->name[0];
2633              i++, c++) {
2634                 if (!newcon->match ||
2635                     newcon->match(newcon, c->name, c->index, c->options) != 0) {
2636                         /* default matching */
2637                         BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name));
2638                         if (strcmp(c->name, newcon->name) != 0)
2639                                 continue;
2640                         if (newcon->index >= 0 &&
2641                             newcon->index != c->index)
2642                                 continue;
2643                         if (newcon->index < 0)
2644                                 newcon->index = c->index;
2645 
2646                         if (_braille_register_console(newcon, c))
2647                                 return;
2648 
2649                         if (newcon->setup &&
2650                             newcon->setup(newcon, c->options) != 0)
2651                                 break;
2652                 }
2653 
2654                 newcon->flags |= CON_ENABLED;
2655                 if (i == preferred_console) {
2656                         newcon->flags |= CON_CONSDEV;
2657                         has_preferred = true;
2658                 }
2659                 break;
2660         }
2661 
2662         if (!(newcon->flags & CON_ENABLED))
2663                 return;
2664 
2665         /*
2666          * If we have a bootconsole, and are switching to a real console,
2667          * don't print everything out again, since when the boot console, and
2668          * the real console are the same physical device, it's annoying to
2669          * see the beginning boot messages twice
2670          */
2671         if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2672                 newcon->flags &= ~CON_PRINTBUFFER;
2673 
2674         /*
2675          *      Put this console in the list - keep the
2676          *      preferred driver at the head of the list.
2677          */
2678         console_lock();
2679         if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2680                 newcon->next = console_drivers;
2681                 console_drivers = newcon;
2682                 if (newcon->next)
2683                         newcon->next->flags &= ~CON_CONSDEV;
2684         } else {
2685                 newcon->next = console_drivers->next;
2686                 console_drivers->next = newcon;
2687         }
2688 
2689         if (newcon->flags & CON_EXTENDED)
2690                 nr_ext_console_drivers++;
2691 
2692         if (newcon->flags & CON_PRINTBUFFER) {
2693                 /*
2694                  * console_unlock(); will print out the buffered messages
2695                  * for us.
2696                  */
2697                 logbuf_lock_irqsave(flags);
2698                 console_seq = syslog_seq;
2699                 console_idx = syslog_idx;
2700                 /*
2701                  * We're about to replay the log buffer.  Only do this to the
2702                  * just-registered console to avoid excessive message spam to
2703                  * the already-registered consoles.
2704                  *
2705                  * Set exclusive_console with disabled interrupts to reduce
2706                  * race window with eventual console_flush_on_panic() that
2707                  * ignores console_lock.
2708                  */
2709                 exclusive_console = newcon;
2710                 exclusive_console_stop_seq = console_seq;
2711                 logbuf_unlock_irqrestore(flags);
2712         }
2713         console_unlock();
2714         console_sysfs_notify();
2715 
2716         /*
2717          * By unregistering the bootconsoles after we enable the real console
2718          * we get the "console xxx enabled" message on all the consoles -
2719          * boot consoles, real consoles, etc - this is to ensure that end
2720          * users know there might be something in the kernel's log buffer that
2721          * went to the bootconsole (that they do not see on the real console)
2722          */
2723         pr_info("%sconsole [%s%d] enabled\n",
2724                 (newcon->flags & CON_BOOT) ? "boot" : "" ,
2725                 newcon->name, newcon->index);
2726         if (bcon &&
2727             ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2728             !keep_bootcon) {
2729                 /* We need to iterate through all boot consoles, to make
2730                  * sure we print everything out, before we unregister them.
2731                  */
2732                 for_each_console(bcon)
2733                         if (bcon->flags & CON_BOOT)
2734                                 unregister_console(bcon);
2735         }
2736 }
2737 EXPORT_SYMBOL(register_console);
2738 
2739 int unregister_console(struct console *console)
2740 {
2741         struct console *a, *b;
2742         int res;
2743 
2744         pr_info("%sconsole [%s%d] disabled\n",
2745                 (console->flags & CON_BOOT) ? "boot" : "" ,
2746                 console->name, console->index);
2747 
2748         res = _braille_unregister_console(console);
2749         if (res)
2750                 return res;
2751 
2752         res = 1;
2753         console_lock();
2754         if (console_drivers == console) {
2755                 console_drivers=console->next;
2756                 res = 0;
2757         } else if (console_drivers) {
2758                 for (a=console_drivers->next, b=console_drivers ;
2759                      a; b=a, a=b->next) {
2760                         if (a == console) {
2761                                 b->next = a->next;
2762                                 res = 0;
2763                                 break;
2764                         }
2765                 }
2766         }
2767 
2768         if (!res && (console->flags & CON_EXTENDED))
2769                 nr_ext_console_drivers--;
2770 
2771         /*
2772          * If this isn't the last console and it has CON_CONSDEV set, we
2773          * need to set it on the next preferred console.
2774          */
2775         if (console_drivers != NULL && console->flags & CON_CONSDEV)
2776                 console_drivers->flags |= CON_CONSDEV;
2777 
2778         console->flags &= ~CON_ENABLED;
2779         console_unlock();
2780         console_sysfs_notify();
2781         return res;
2782 }
2783 EXPORT_SYMBOL(unregister_console);
2784 
2785 /*
2786  * Initialize the console device. This is called *early*, so
2787  * we can't necessarily depend on lots of kernel help here.
2788  * Just do some early initializations, and do the complex setup
2789  * later.
2790  */
2791 void __init console_init(void)
2792 {
2793         int ret;
2794         initcall_t call;
2795         initcall_entry_t *ce;
2796 
2797         /* Setup the default TTY line discipline. */
2798         n_tty_init();
2799 
2800         /*
2801          * set up the console device so that later boot sequences can
2802          * inform about problems etc..
2803          */
2804         ce = __con_initcall_start;
2805         trace_initcall_level("console");
2806         while (ce < __con_initcall_end) {
2807                 call = initcall_from_entry(ce);
2808                 trace_initcall_start(call);
2809                 ret = call();
2810                 trace_initcall_finish(call, ret);
2811                 ce++;
2812         }
2813 }
2814 
2815 /*
2816  * Some boot consoles access data that is in the init section and which will
2817  * be discarded after the initcalls have been run. To make sure that no code
2818  * will access this data, unregister the boot consoles in a late initcall.
2819  *
2820  * If for some reason, such as deferred probe or the driver being a loadable
2821  * module, the real console hasn't registered yet at this point, there will
2822  * be a brief interval in which no messages are logged to the console, which
2823  * makes it difficult to diagnose problems that occur during this time.
2824  *
2825  * To mitigate this problem somewhat, only unregister consoles whose memory
2826  * intersects with the init section. Note that all other boot consoles will
2827  * get unregistred when the real preferred console is registered.
2828  */
2829 static int __init printk_late_init(void)
2830 {
2831         struct console *con;
2832         int ret;
2833 
2834         for_each_console(con) {
2835                 if (!(con->flags & CON_BOOT))
2836                         continue;
2837 
2838                 /* Check addresses that might be used for enabled consoles. */
2839                 if (init_section_intersects(con, sizeof(*con)) ||
2840                     init_section_contains(con->write, 0) ||
2841                     init_section_contains(con->read, 0) ||
2842                     init_section_contains(con->device, 0) ||
2843                     init_section_contains(con->unblank, 0) ||
2844                     init_section_contains(con->data, 0)) {
2845                         /*
2846                          * Please, consider moving the reported consoles out
2847                          * of the init section.
2848                          */
2849                         pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n",
2850                                 con->name, con->index);
2851                         unregister_console(con);
2852                 }
2853         }
2854         ret = cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD, "printk:dead", NULL,
2855                                         console_cpu_notify);
2856         WARN_ON(ret < 0);
2857         ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "printk:online",
2858                                         console_cpu_notify, NULL);
2859         WARN_ON(ret < 0);
2860         return 0;
2861 }
2862 late_initcall(printk_late_init);
2863 
2864 #if defined CONFIG_PRINTK
2865 /*
2866  * Delayed printk version, for scheduler-internal messages:
2867  */
2868 #define PRINTK_PENDING_WAKEUP   0x01
2869 #define PRINTK_PENDING_OUTPUT   0x02
2870 
2871 static DEFINE_PER_CPU(int, printk_pending);
2872 
2873 static void wake_up_klogd_work_func(struct irq_work *irq_work)
2874 {
2875         int pending = __this_cpu_xchg(printk_pending, 0);
2876 
2877         if (pending & PRINTK_PENDING_OUTPUT) {
2878                 /* If trylock fails, someone else is doing the printing */
2879                 if (console_trylock())
2880                         console_unlock();
2881         }
2882 
2883         if (pending & PRINTK_PENDING_WAKEUP)
2884                 wake_up_interruptible(&log_wait);
2885 }
2886 
2887 static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
2888         .func = wake_up_klogd_work_func,
2889         .flags = IRQ_WORK_LAZY,
2890 };
2891 
2892 void wake_up_klogd(void)
2893 {
2894         preempt_disable();
2895         if (waitqueue_active(&log_wait)) {
2896                 this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
2897                 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2898         }
2899         preempt_enable();
2900 }
2901 
2902 void defer_console_output(void)
2903 {
2904         preempt_disable();
2905         __this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
2906         irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2907         preempt_enable();
2908 }
2909 
2910 int vprintk_deferred(const char *fmt, va_list args)
2911 {
2912         int r;
2913 
2914         r = vprintk_emit(0, LOGLEVEL_SCHED, NULL, 0, fmt, args);
2915         defer_console_output();
2916 
2917         return r;
2918 }
2919 
2920 int printk_deferred(const char *fmt, ...)
2921 {
2922         va_list args;
2923         int r;
2924 
2925         va_start(args, fmt);
2926         r = vprintk_deferred(fmt, args);
2927         va_end(args);
2928 
2929         return r;
2930 }
2931 
2932 /*
2933  * printk rate limiting, lifted from the networking subsystem.
2934  *
2935  * This enforces a rate limit: not more than 10 kernel messages
2936  * every 5s to make a denial-of-service attack impossible.
2937  */
2938 DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2939 
2940 int __printk_ratelimit(const char *func)
2941 {
2942         return ___ratelimit(&printk_ratelimit_state, func);
2943 }
2944 EXPORT_SYMBOL(__printk_ratelimit);
2945 
2946 /**
2947  * printk_timed_ratelimit - caller-controlled printk ratelimiting
2948  * @caller_jiffies: pointer to caller's state
2949  * @interval_msecs: minimum interval between prints
2950  *
2951  * printk_timed_ratelimit() returns true if more than @interval_msecs
2952  * milliseconds have elapsed since the last time printk_timed_ratelimit()
2953  * returned true.
2954  */
2955 bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2956                         unsigned int interval_msecs)
2957 {
2958         unsigned long elapsed = jiffies - *caller_jiffies;
2959 
2960         if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs))
2961                 return false;
2962 
2963         *caller_jiffies = jiffies;
2964         return true;
2965 }
2966 EXPORT_SYMBOL(printk_timed_ratelimit);
2967 
2968 static DEFINE_SPINLOCK(dump_list_lock);
2969 static LIST_HEAD(dump_list);
2970 
2971 /**
2972  * kmsg_dump_register - register a kernel log dumper.
2973  * @dumper: pointer to the kmsg_dumper structure
2974  *
2975  * Adds a kernel log dumper to the system. The dump callback in the
2976  * structure will be called when the kernel oopses or panics and must be
2977  * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2978  */
2979 int kmsg_dump_register(struct kmsg_dumper *dumper)
2980 {
2981         unsigned long flags;
2982         int err = -EBUSY;
2983 
2984         /* The dump callback needs to be set */
2985         if (!dumper->dump)
2986                 return -EINVAL;
2987 
2988         spin_lock_irqsave(&dump_list_lock, flags);
2989         /* Don't allow registering multiple times */
2990         if (!dumper->registered) {
2991                 dumper->registered = 1;
2992                 list_add_tail_rcu(&dumper->list, &dump_list);
2993                 err = 0;
2994         }
2995         spin_unlock_irqrestore(&dump_list_lock, flags);
2996 
2997         return err;
2998 }
2999 EXPORT_SYMBOL_GPL(kmsg_dump_register);
3000 
3001 /**
3002  * kmsg_dump_unregister - unregister a kmsg dumper.
3003  * @dumper: pointer to the kmsg_dumper structure
3004  *
3005  * Removes a dump device from the system. Returns zero on success and
3006  * %-EINVAL otherwise.
3007  */
3008 int kmsg_dump_unregister(struct kmsg_dumper *dumper)
3009 {
3010         unsigned long flags;
3011         int err = -EINVAL;
3012 
3013         spin_lock_irqsave(&dump_list_lock, flags);
3014         if (dumper->registered) {
3015                 dumper->registered = 0;
3016                 list_del_rcu(&dumper->list);
3017                 err = 0;
3018         }
3019         spin_unlock_irqrestore(&dump_list_lock, flags);
3020         synchronize_rcu();
3021 
3022         return err;
3023 }
3024 EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
3025 
3026 static bool always_kmsg_dump;
3027 module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
3028 
3029 /**
3030  * kmsg_dump - dump kernel log to kernel message dumpers.
3031  * @reason: the reason (oops, panic etc) for dumping
3032  *
3033  * Call each of the registered dumper's dump() callback, which can
3034  * retrieve the kmsg records with kmsg_dump_get_line() or
3035  * kmsg_dump_get_buffer().
3036  */
3037 void kmsg_dump(enum kmsg_dump_reason reason)
3038 {
3039         struct kmsg_dumper *dumper;
3040         unsigned long flags;
3041 
3042         if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
3043                 return;
3044 
3045         rcu_read_lock();
3046         list_for_each_entry_rcu(dumper, &dump_list, list) {
3047                 if (dumper->max_reason && reason > dumper->max_reason)
3048                         continue;
3049 
3050                 /* initialize iterator with data about the stored records */
3051                 dumper->active = true;
3052 
3053                 logbuf_lock_irqsave(flags);
3054                 dumper->cur_seq = clear_seq;
3055                 dumper->cur_idx = clear_idx;
3056                 dumper->next_seq = log_next_seq;
3057                 dumper->next_idx = log_next_idx;
3058                 logbuf_unlock_irqrestore(flags);
3059 
3060                 /* invoke dumper which will iterate over records */
3061                 dumper->dump(dumper, reason);
3062 
3063                 /* reset iterator */
3064                 dumper->active = false;
3065         }
3066         rcu_read_unlock();
3067 }
3068 
3069 /**
3070  * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3071  * @dumper: registered kmsg dumper
3072  * @syslog: include the "<4>" prefixes
3073  * @line: buffer to copy the line to
3074  * @size: maximum size of the buffer
3075  * @len: length of line placed into buffer
3076  *
3077  * Start at the beginning of the kmsg buffer, with the oldest kmsg
3078  * record, and copy one record into the provided buffer.
3079  *
3080  * Consecutive calls will return the next available record moving
3081  * towards the end of the buffer with the youngest messages.
3082  *
3083  * A return value of FALSE indicates that there are no more records to
3084  * read.
3085  *
3086  * The function is similar to kmsg_dump_get_line(), but grabs no locks.
3087  */
3088 bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
3089                                char *line, size_t size, size_t *len)
3090 {
3091         struct printk_log *msg;
3092         size_t l = 0;
3093         bool ret = false;
3094 
3095         if (!dumper->active)
3096                 goto out;
3097 
3098         if (dumper->cur_seq < log_first_seq) {
3099                 /* messages are gone, move to first available one */
3100                 dumper->cur_seq = log_first_seq;
3101                 dumper->cur_idx = log_first_idx;
3102         }
3103 
3104         /* last entry */
3105         if (dumper->cur_seq >= log_next_seq)
3106                 goto out;
3107 
3108         msg = log_from_idx(dumper->cur_idx);
3109         l = msg_print_text(msg, syslog, printk_time, line, size);
3110 
3111         dumper->cur_idx = log_next(dumper->cur_idx);
3112         dumper->cur_seq++;
3113         ret = true;
3114 out:
3115         if (len)
3116                 *len = l;
3117         return ret;
3118 }
3119 
3120 /**
3121  * kmsg_dump_get_line - retrieve one kmsg log line
3122  * @dumper: registered kmsg dumper
3123  * @syslog: include the "<4>" prefixes
3124  * @line: buffer to copy the line to
3125  * @size: maximum size of the buffer
3126  * @len: length of line placed into buffer
3127  *
3128  * Start at the beginning of the kmsg buffer, with the oldest kmsg
3129  * record, and copy one record into the provided buffer.
3130  *
3131  * Consecutive calls will return the next available record moving
3132  * towards the end of the buffer with the youngest messages.
3133  *
3134  * A return value of FALSE indicates that there are no more records to
3135  * read.
3136  */
3137 bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
3138                         char *line, size_t size, size_t *len)
3139 {
3140         unsigned long flags;
3141         bool ret;
3142 
3143         logbuf_lock_irqsave(flags);
3144         ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
3145         logbuf_unlock_irqrestore(flags);
3146 
3147         return ret;
3148 }
3149 EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
3150 
3151 /**
3152  * kmsg_dump_get_buffer - copy kmsg log lines
3153  * @dumper: registered kmsg dumper
3154  * @syslog: include the "<4>" prefixes
3155  * @buf: buffer to copy the line to
3156  * @size: maximum size of the buffer
3157  * @len: length of line placed into buffer
3158  *
3159  * Start at the end of the kmsg buffer and fill the provided buffer
3160  * with as many of the the *youngest* kmsg records that fit into it.
3161  * If the buffer is large enough, all available kmsg records will be
3162  * copied with a single call.
3163  *
3164  * Consecutive calls will fill the buffer with the next block of
3165  * available older records, not including the earlier retrieved ones.
3166  *
3167  * A return value of FALSE indicates that there are no more records to
3168  * read.
3169  */
3170 bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
3171                           char *buf, size_t size, size_t *len)
3172 {
3173         unsigned long flags;
3174         u64 seq;
3175         u32 idx;
3176         u64 next_seq;
3177         u32 next_idx;
3178         size_t l = 0;
3179         bool ret = false;
3180         bool time = printk_time;
3181 
3182         if (!dumper->active)
3183                 goto out;
3184 
3185         logbuf_lock_irqsave(flags);
3186         if (dumper->cur_seq < log_first_seq) {
3187                 /* messages are gone, move to first available one */
3188                 dumper->cur_seq = log_first_seq;
3189                 dumper->cur_idx = log_first_idx;
3190         }
3191 
3192         /* last entry */
3193         if (dumper->cur_seq >= dumper->next_seq) {
3194                 logbuf_unlock_irqrestore(flags);
3195                 goto out;
3196         }
3197 
3198         /* calculate length of entire buffer */
3199         seq = dumper->cur_seq;
3200         idx = dumper->cur_idx;
3201         while (seq < dumper->next_seq) {
3202                 struct printk_log *msg = log_from_idx(idx);
3203 
3204                 l += msg_print_text(msg, true, time, NULL, 0);
3205                 idx = log_next(idx);
3206                 seq++;
3207         }
3208 
3209         /* move first record forward until length fits into the buffer */
3210         seq = dumper->cur_seq;
3211         idx = dumper->cur_idx;
3212         while (l > size && seq < dumper->next_seq) {
3213                 struct printk_log *msg = log_from_idx(idx);
3214 
3215                 l -= msg_print_text(msg, true, time, NULL, 0);
3216                 idx = log_next(idx);
3217                 seq++;
3218         }
3219 
3220         /* last message in next interation */
3221         next_seq = seq;
3222         next_idx = idx;
3223 
3224         l = 0;
3225         while (seq < dumper->next_seq) {
3226                 struct printk_log *msg = log_from_idx(idx);
3227 
3228                 l += msg_print_text(msg, syslog, time, buf + l, size - l);
3229                 idx = log_next(idx);
3230                 seq++;
3231         }
3232 
3233         dumper->next_seq = next_seq;
3234         dumper->next_idx = next_idx;
3235         ret = true;
3236         logbuf_unlock_irqrestore(flags);
3237 out:
3238         if (len)
3239                 *len = l;
3240         return ret;
3241 }
3242 EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
3243 
3244 /**
3245  * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3246  * @dumper: registered kmsg dumper
3247  *
3248  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3249  * kmsg_dump_get_buffer() can be called again and used multiple
3250  * times within the same dumper.dump() callback.
3251  *
3252  * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3253  */
3254 void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
3255 {
3256         dumper->cur_seq = clear_seq;
3257         dumper->cur_idx = clear_idx;
3258         dumper->next_seq = log_next_seq;
3259         dumper->next_idx = log_next_idx;
3260 }
3261 
3262 /**
3263  * kmsg_dump_rewind - reset the interator
3264  * @dumper: registered kmsg dumper
3265  *
3266  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3267  * kmsg_dump_get_buffer() can be called again and used multiple
3268  * times within the same dumper.dump() callback.
3269  */
3270 void kmsg_dump_rewind(struct kmsg_dumper *dumper)
3271 {
3272         unsigned long flags;
3273 
3274         logbuf_lock_irqsave(flags);
3275         kmsg_dump_rewind_nolock(dumper);
3276         logbuf_unlock_irqrestore(flags);
3277 }
3278 EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
3279 
3280 #endif
3281 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp