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

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

Version: ~ [ linux-4.19-rc7 ] ~ [ linux-4.18.12 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.74 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.131 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.159 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.123 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.59 ] ~ [ 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-2.6.31.14 ] ~ [ linux-2.6.30.10 ] ~ [ linux-2.6.29.6 ] ~ [ linux-2.6.28.10 ] ~ [ linux-2.6.27.62 ] ~ [ 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 #include <linux/kernel.h>
 20 #include <linux/mm.h>
 21 #include <linux/tty.h>
 22 #include <linux/tty_driver.h>
 23 #include <linux/console.h>
 24 #include <linux/init.h>
 25 #include <linux/jiffies.h>
 26 #include <linux/nmi.h>
 27 #include <linux/module.h>
 28 #include <linux/moduleparam.h>
 29 #include <linux/delay.h>
 30 #include <linux/smp.h>
 31 #include <linux/security.h>
 32 #include <linux/bootmem.h>
 33 #include <linux/memblock.h>
 34 #include <linux/syscalls.h>
 35 #include <linux/crash_core.h>
 36 #include <linux/kdb.h>
 37 #include <linux/ratelimit.h>
 38 #include <linux/kmsg_dump.h>
 39 #include <linux/syslog.h>
 40 #include <linux/cpu.h>
 41 #include <linux/rculist.h>
 42 #include <linux/poll.h>
 43 #include <linux/irq_work.h>
 44 #include <linux/ctype.h>
 45 #include <linux/uio.h>
 46 #include <linux/sched/clock.h>
 47 #include <linux/sched/debug.h>
 48 #include <linux/sched/task_stack.h>
 49 
 50 #include <linux/uaccess.h>
 51 #include <asm/sections.h>
 52 
 53 #include <trace/events/initcall.h>
 54 #define CREATE_TRACE_POINTS
 55 #include <trace/events/printk.h>
 56 
 57 #include "console_cmdline.h"
 58 #include "braille.h"
 59 #include "internal.h"
 60 
 61 int console_printk[4] = {
 62         CONSOLE_LOGLEVEL_DEFAULT,       /* console_loglevel */
 63         MESSAGE_LOGLEVEL_DEFAULT,       /* default_message_loglevel */
 64         CONSOLE_LOGLEVEL_MIN,           /* minimum_console_loglevel */
 65         CONSOLE_LOGLEVEL_DEFAULT,       /* default_console_loglevel */
 66 };
 67 
 68 atomic_t ignore_console_lock_warning __read_mostly = ATOMIC_INIT(0);
 69 EXPORT_SYMBOL(ignore_console_lock_warning);
 70 
 71 /*
 72  * Low level drivers may need that to know if they can schedule in
 73  * their unblank() callback or not. So let's export it.
 74  */
 75 int oops_in_progress;
 76 EXPORT_SYMBOL(oops_in_progress);
 77 
 78 /*
 79  * console_sem protects the console_drivers list, and also
 80  * provides serialisation for access to the entire console
 81  * driver system.
 82  */
 83 static DEFINE_SEMAPHORE(console_sem);
 84 struct console *console_drivers;
 85 EXPORT_SYMBOL_GPL(console_drivers);
 86 
 87 #ifdef CONFIG_LOCKDEP
 88 static struct lockdep_map console_lock_dep_map = {
 89         .name = "console_lock"
 90 };
 91 #endif
 92 
 93 enum devkmsg_log_bits {
 94         __DEVKMSG_LOG_BIT_ON = 0,
 95         __DEVKMSG_LOG_BIT_OFF,
 96         __DEVKMSG_LOG_BIT_LOCK,
 97 };
 98 
 99 enum devkmsg_log_masks {
100         DEVKMSG_LOG_MASK_ON             = BIT(__DEVKMSG_LOG_BIT_ON),
101         DEVKMSG_LOG_MASK_OFF            = BIT(__DEVKMSG_LOG_BIT_OFF),
102         DEVKMSG_LOG_MASK_LOCK           = BIT(__DEVKMSG_LOG_BIT_LOCK),
103 };
104 
105 /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */
106 #define DEVKMSG_LOG_MASK_DEFAULT        0
107 
108 static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
109 
110 static int __control_devkmsg(char *str)
111 {
112         if (!str)
113                 return -EINVAL;
114 
115         if (!strncmp(str, "on", 2)) {
116                 devkmsg_log = DEVKMSG_LOG_MASK_ON;
117                 return 2;
118         } else if (!strncmp(str, "off", 3)) {
119                 devkmsg_log = DEVKMSG_LOG_MASK_OFF;
120                 return 3;
121         } else if (!strncmp(str, "ratelimit", 9)) {
122                 devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT;
123                 return 9;
124         }
125         return -EINVAL;
126 }
127 
128 static int __init control_devkmsg(char *str)
129 {
130         if (__control_devkmsg(str) < 0)
131                 return 1;
132 
133         /*
134          * Set sysctl string accordingly:
135          */
136         if (devkmsg_log == DEVKMSG_LOG_MASK_ON)
137                 strcpy(devkmsg_log_str, "on");
138         else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF)
139                 strcpy(devkmsg_log_str, "off");
140         /* else "ratelimit" which is set by default. */
141 
142         /*
143          * Sysctl cannot change it anymore. The kernel command line setting of
144          * this parameter is to force the setting to be permanent throughout the
145          * runtime of the system. This is a precation measure against userspace
146          * trying to be a smarta** and attempting to change it up on us.
147          */
148         devkmsg_log |= DEVKMSG_LOG_MASK_LOCK;
149 
150         return 0;
151 }
152 __setup("printk.devkmsg=", control_devkmsg);
153 
154 char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit";
155 
156 int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write,
157                               void __user *buffer, size_t *lenp, loff_t *ppos)
158 {
159         char old_str[DEVKMSG_STR_MAX_SIZE];
160         unsigned int old;
161         int err;
162 
163         if (write) {
164                 if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK)
165                         return -EINVAL;
166 
167                 old = devkmsg_log;
168                 strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE);
169         }
170 
171         err = proc_dostring(table, write, buffer, lenp, ppos);
172         if (err)
173                 return err;
174 
175         if (write) {
176                 err = __control_devkmsg(devkmsg_log_str);
177 
178                 /*
179                  * Do not accept an unknown string OR a known string with
180                  * trailing crap...
181                  */
182                 if (err < 0 || (err + 1 != *lenp)) {
183 
184                         /* ... and restore old setting. */
185                         devkmsg_log = old;
186                         strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE);
187 
188                         return -EINVAL;
189                 }
190         }
191 
192         return 0;
193 }
194 
195 /*
196  * Number of registered extended console drivers.
197  *
198  * If extended consoles are present, in-kernel cont reassembly is disabled
199  * and each fragment is stored as a separate log entry with proper
200  * continuation flag so that every emitted message has full metadata.  This
201  * doesn't change the result for regular consoles or /proc/kmsg.  For
202  * /dev/kmsg, as long as the reader concatenates messages according to
203  * consecutive continuation flags, the end result should be the same too.
204  */
205 static int nr_ext_console_drivers;
206 
207 /*
208  * Helper macros to handle lockdep when locking/unlocking console_sem. We use
209  * macros instead of functions so that _RET_IP_ contains useful information.
210  */
211 #define down_console_sem() do { \
212         down(&console_sem);\
213         mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\
214 } while (0)
215 
216 static int __down_trylock_console_sem(unsigned long ip)
217 {
218         int lock_failed;
219         unsigned long flags;
220 
221         /*
222          * Here and in __up_console_sem() we need to be in safe mode,
223          * because spindump/WARN/etc from under console ->lock will
224          * deadlock in printk()->down_trylock_console_sem() otherwise.
225          */
226         printk_safe_enter_irqsave(flags);
227         lock_failed = down_trylock(&console_sem);
228         printk_safe_exit_irqrestore(flags);
229 
230         if (lock_failed)
231                 return 1;
232         mutex_acquire(&console_lock_dep_map, 0, 1, ip);
233         return 0;
234 }
235 #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_)
236 
237 static void __up_console_sem(unsigned long ip)
238 {
239         unsigned long flags;
240 
241         mutex_release(&console_lock_dep_map, 1, ip);
242 
243         printk_safe_enter_irqsave(flags);
244         up(&console_sem);
245         printk_safe_exit_irqrestore(flags);
246 }
247 #define up_console_sem() __up_console_sem(_RET_IP_)
248 
249 /*
250  * This is used for debugging the mess that is the VT code by
251  * keeping track if we have the console semaphore held. It's
252  * definitely not the perfect debug tool (we don't know if _WE_
253  * hold it and are racing, but it helps tracking those weird code
254  * paths in the console code where we end up in places I want
255  * locked without the console sempahore held).
256  */
257 static int console_locked, console_suspended;
258 
259 /*
260  * If exclusive_console is non-NULL then only this console is to be printed to.
261  */
262 static struct console *exclusive_console;
263 
264 /*
265  *      Array of consoles built from command line options (console=)
266  */
267 
268 #define MAX_CMDLINECONSOLES 8
269 
270 static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
271 
272 static int preferred_console = -1;
273 int console_set_on_cmdline;
274 EXPORT_SYMBOL(console_set_on_cmdline);
275 
276 /* Flag: console code may call schedule() */
277 static int console_may_schedule;
278 
279 enum con_msg_format_flags {
280         MSG_FORMAT_DEFAULT      = 0,
281         MSG_FORMAT_SYSLOG       = (1 << 0),
282 };
283 
284 static int console_msg_format = MSG_FORMAT_DEFAULT;
285 
286 /*
287  * The printk log buffer consists of a chain of concatenated variable
288  * length records. Every record starts with a record header, containing
289  * the overall length of the record.
290  *
291  * The heads to the first and last entry in the buffer, as well as the
292  * sequence numbers of these entries are maintained when messages are
293  * stored.
294  *
295  * If the heads indicate available messages, the length in the header
296  * tells the start next message. A length == 0 for the next message
297  * indicates a wrap-around to the beginning of the buffer.
298  *
299  * Every record carries the monotonic timestamp in microseconds, as well as
300  * the standard userspace syslog level and syslog facility. The usual
301  * kernel messages use LOG_KERN; userspace-injected messages always carry
302  * a matching syslog facility, by default LOG_USER. The origin of every
303  * message can be reliably determined that way.
304  *
305  * The human readable log message directly follows the message header. The
306  * length of the message text is stored in the header, the stored message
307  * is not terminated.
308  *
309  * Optionally, a message can carry a dictionary of properties (key/value pairs),
310  * to provide userspace with a machine-readable message context.
311  *
312  * Examples for well-defined, commonly used property names are:
313  *   DEVICE=b12:8               device identifier
314  *                                b12:8         block dev_t
315  *                                c127:3        char dev_t
316  *                                n8            netdev ifindex
317  *                                +sound:card0  subsystem:devname
318  *   SUBSYSTEM=pci              driver-core subsystem name
319  *
320  * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
321  * follows directly after a '=' character. Every property is terminated by
322  * a '\0' character. The last property is not terminated.
323  *
324  * Example of a message structure:
325  *   0000  ff 8f 00 00 00 00 00 00      monotonic time in nsec
326  *   0008  34 00                        record is 52 bytes long
327  *   000a        0b 00                  text is 11 bytes long
328  *   000c              1f 00            dictionary is 23 bytes long
329  *   000e                    03 00      LOG_KERN (facility) LOG_ERR (level)
330  *   0010  69 74 27 73 20 61 20 6c      "it's a l"
331  *         69 6e 65                     "ine"
332  *   001b           44 45 56 49 43      "DEVIC"
333  *         45 3d 62 38 3a 32 00 44      "E=b8:2\0D"
334  *         52 49 56 45 52 3d 62 75      "RIVER=bu"
335  *         67                           "g"
336  *   0032     00 00 00                  padding to next message header
337  *
338  * The 'struct printk_log' buffer header must never be directly exported to
339  * userspace, it is a kernel-private implementation detail that might
340  * need to be changed in the future, when the requirements change.
341  *
342  * /dev/kmsg exports the structured data in the following line format:
343  *   "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n"
344  *
345  * Users of the export format should ignore possible additional values
346  * separated by ',', and find the message after the ';' character.
347  *
348  * The optional key/value pairs are attached as continuation lines starting
349  * with a space character and terminated by a newline. All possible
350  * non-prinatable characters are escaped in the "\xff" notation.
351  */
352 
353 enum log_flags {
354         LOG_NEWLINE     = 2,    /* text ended with a newline */
355         LOG_PREFIX      = 4,    /* text started with a prefix */
356         LOG_CONT        = 8,    /* text is a fragment of a continuation line */
357 };
358 
359 struct printk_log {
360         u64 ts_nsec;            /* timestamp in nanoseconds */
361         u16 len;                /* length of entire record */
362         u16 text_len;           /* length of text buffer */
363         u16 dict_len;           /* length of dictionary buffer */
364         u8 facility;            /* syslog facility */
365         u8 flags:5;             /* internal record flags */
366         u8 level:3;             /* syslog level */
367 }
368 #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
369 __packed __aligned(4)
370 #endif
371 ;
372 
373 /*
374  * The logbuf_lock protects kmsg buffer, indices, counters.  This can be taken
375  * within the scheduler's rq lock. It must be released before calling
376  * console_unlock() or anything else that might wake up a process.
377  */
378 DEFINE_RAW_SPINLOCK(logbuf_lock);
379 
380 /*
381  * Helper macros to lock/unlock logbuf_lock and switch between
382  * printk-safe/unsafe modes.
383  */
384 #define logbuf_lock_irq()                               \
385         do {                                            \
386                 printk_safe_enter_irq();                \
387                 raw_spin_lock(&logbuf_lock);            \
388         } while (0)
389 
390 #define logbuf_unlock_irq()                             \
391         do {                                            \
392                 raw_spin_unlock(&logbuf_lock);          \
393                 printk_safe_exit_irq();                 \
394         } while (0)
395 
396 #define logbuf_lock_irqsave(flags)                      \
397         do {                                            \
398                 printk_safe_enter_irqsave(flags);       \
399                 raw_spin_lock(&logbuf_lock);            \
400         } while (0)
401 
402 #define logbuf_unlock_irqrestore(flags)         \
403         do {                                            \
404                 raw_spin_unlock(&logbuf_lock);          \
405                 printk_safe_exit_irqrestore(flags);     \
406         } while (0)
407 
408 #ifdef CONFIG_PRINTK
409 DECLARE_WAIT_QUEUE_HEAD(log_wait);
410 /* the next printk record to read by syslog(READ) or /proc/kmsg */
411 static u64 syslog_seq;
412 static u32 syslog_idx;
413 static size_t syslog_partial;
414 
415 /* index and sequence number of the first record stored in the buffer */
416 static u64 log_first_seq;
417 static u32 log_first_idx;
418 
419 /* index and sequence number of the next record to store in the buffer */
420 static u64 log_next_seq;
421 static u32 log_next_idx;
422 
423 /* the next printk record to write to the console */
424 static u64 console_seq;
425 static u32 console_idx;
426 
427 /* the next printk record to read after the last 'clear' command */
428 static u64 clear_seq;
429 static u32 clear_idx;
430 
431 #define PREFIX_MAX              32
432 #define LOG_LINE_MAX            (1024 - PREFIX_MAX)
433 
434 #define LOG_LEVEL(v)            ((v) & 0x07)
435 #define LOG_FACILITY(v)         ((v) >> 3 & 0xff)
436 
437 /* record buffer */
438 #define LOG_ALIGN __alignof__(struct printk_log)
439 #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
440 static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
441 static char *log_buf = __log_buf;
442 static u32 log_buf_len = __LOG_BUF_LEN;
443 
444 /* Return log buffer address */
445 char *log_buf_addr_get(void)
446 {
447         return log_buf;
448 }
449 
450 /* Return log buffer size */
451 u32 log_buf_len_get(void)
452 {
453         return log_buf_len;
454 }
455 
456 /* human readable text of the record */
457 static char *log_text(const struct printk_log *msg)
458 {
459         return (char *)msg + sizeof(struct printk_log);
460 }
461 
462 /* optional key/value pair dictionary attached to the record */
463 static char *log_dict(const struct printk_log *msg)
464 {
465         return (char *)msg + sizeof(struct printk_log) + msg->text_len;
466 }
467 
468 /* get record by index; idx must point to valid msg */
469 static struct printk_log *log_from_idx(u32 idx)
470 {
471         struct printk_log *msg = (struct printk_log *)(log_buf + idx);
472 
473         /*
474          * A length == 0 record is the end of buffer marker. Wrap around and
475          * read the message at the start of the buffer.
476          */
477         if (!msg->len)
478                 return (struct printk_log *)log_buf;
479         return msg;
480 }
481 
482 /* get next record; idx must point to valid msg */
483 static u32 log_next(u32 idx)
484 {
485         struct printk_log *msg = (struct printk_log *)(log_buf + idx);
486 
487         /* length == 0 indicates the end of the buffer; wrap */
488         /*
489          * A length == 0 record is the end of buffer marker. Wrap around and
490          * read the message at the start of the buffer as *this* one, and
491          * return the one after that.
492          */
493         if (!msg->len) {
494                 msg = (struct printk_log *)log_buf;
495                 return msg->len;
496         }
497         return idx + msg->len;
498 }
499 
500 /*
501  * Check whether there is enough free space for the given message.
502  *
503  * The same values of first_idx and next_idx mean that the buffer
504  * is either empty or full.
505  *
506  * If the buffer is empty, we must respect the position of the indexes.
507  * They cannot be reset to the beginning of the buffer.
508  */
509 static int logbuf_has_space(u32 msg_size, bool empty)
510 {
511         u32 free;
512 
513         if (log_next_idx > log_first_idx || empty)
514                 free = max(log_buf_len - log_next_idx, log_first_idx);
515         else
516                 free = log_first_idx - log_next_idx;
517 
518         /*
519          * We need space also for an empty header that signalizes wrapping
520          * of the buffer.
521          */
522         return free >= msg_size + sizeof(struct printk_log);
523 }
524 
525 static int log_make_free_space(u32 msg_size)
526 {
527         while (log_first_seq < log_next_seq &&
528                !logbuf_has_space(msg_size, false)) {
529                 /* drop old messages until we have enough contiguous space */
530                 log_first_idx = log_next(log_first_idx);
531                 log_first_seq++;
532         }
533 
534         if (clear_seq < log_first_seq) {
535                 clear_seq = log_first_seq;
536                 clear_idx = log_first_idx;
537         }
538 
539         /* sequence numbers are equal, so the log buffer is empty */
540         if (logbuf_has_space(msg_size, log_first_seq == log_next_seq))
541                 return 0;
542 
543         return -ENOMEM;
544 }
545 
546 /* compute the message size including the padding bytes */
547 static u32 msg_used_size(u16 text_len, u16 dict_len, u32 *pad_len)
548 {
549         u32 size;
550 
551         size = sizeof(struct printk_log) + text_len + dict_len;
552         *pad_len = (-size) & (LOG_ALIGN - 1);
553         size += *pad_len;
554 
555         return size;
556 }
557 
558 /*
559  * Define how much of the log buffer we could take at maximum. The value
560  * must be greater than two. Note that only half of the buffer is available
561  * when the index points to the middle.
562  */
563 #define MAX_LOG_TAKE_PART 4
564 static const char trunc_msg[] = "<truncated>";
565 
566 static u32 truncate_msg(u16 *text_len, u16 *trunc_msg_len,
567                         u16 *dict_len, u32 *pad_len)
568 {
569         /*
570          * The message should not take the whole buffer. Otherwise, it might
571          * get removed too soon.
572          */
573         u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART;
574         if (*text_len > max_text_len)
575                 *text_len = max_text_len;
576         /* enable the warning message */
577         *trunc_msg_len = strlen(trunc_msg);
578         /* disable the "dict" completely */
579         *dict_len = 0;
580         /* compute the size again, count also the warning message */
581         return msg_used_size(*text_len + *trunc_msg_len, 0, pad_len);
582 }
583 
584 /* insert record into the buffer, discard old ones, update heads */
585 static int log_store(int facility, int level,
586                      enum log_flags flags, u64 ts_nsec,
587                      const char *dict, u16 dict_len,
588                      const char *text, u16 text_len)
589 {
590         struct printk_log *msg;
591         u32 size, pad_len;
592         u16 trunc_msg_len = 0;
593 
594         /* number of '\0' padding bytes to next message */
595         size = msg_used_size(text_len, dict_len, &pad_len);
596 
597         if (log_make_free_space(size)) {
598                 /* truncate the message if it is too long for empty buffer */
599                 size = truncate_msg(&text_len, &trunc_msg_len,
600                                     &dict_len, &pad_len);
601                 /* survive when the log buffer is too small for trunc_msg */
602                 if (log_make_free_space(size))
603                         return 0;
604         }
605 
606         if (log_next_idx + size + sizeof(struct printk_log) > log_buf_len) {
607                 /*
608                  * This message + an additional empty header does not fit
609                  * at the end of the buffer. Add an empty header with len == 0
610                  * to signify a wrap around.
611                  */
612                 memset(log_buf + log_next_idx, 0, sizeof(struct printk_log));
613                 log_next_idx = 0;
614         }
615 
616         /* fill message */
617         msg = (struct printk_log *)(log_buf + log_next_idx);
618         memcpy(log_text(msg), text, text_len);
619         msg->text_len = text_len;
620         if (trunc_msg_len) {
621                 memcpy(log_text(msg) + text_len, trunc_msg, trunc_msg_len);
622                 msg->text_len += trunc_msg_len;
623         }
624         memcpy(log_dict(msg), dict, dict_len);
625         msg->dict_len = dict_len;
626         msg->facility = facility;
627         msg->level = level & 7;
628         msg->flags = flags & 0x1f;
629         if (ts_nsec > 0)
630                 msg->ts_nsec = ts_nsec;
631         else
632                 msg->ts_nsec = local_clock();
633         memset(log_dict(msg) + dict_len, 0, pad_len);
634         msg->len = size;
635 
636         /* insert message */
637         log_next_idx += msg->len;
638         log_next_seq++;
639 
640         return msg->text_len;
641 }
642 
643 int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT);
644 
645 static int syslog_action_restricted(int type)
646 {
647         if (dmesg_restrict)
648                 return 1;
649         /*
650          * Unless restricted, we allow "read all" and "get buffer size"
651          * for everybody.
652          */
653         return type != SYSLOG_ACTION_READ_ALL &&
654                type != SYSLOG_ACTION_SIZE_BUFFER;
655 }
656 
657 static int check_syslog_permissions(int type, int source)
658 {
659         /*
660          * If this is from /proc/kmsg and we've already opened it, then we've
661          * already done the capabilities checks at open time.
662          */
663         if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN)
664                 goto ok;
665 
666         if (syslog_action_restricted(type)) {
667                 if (capable(CAP_SYSLOG))
668                         goto ok;
669                 /*
670                  * For historical reasons, accept CAP_SYS_ADMIN too, with
671                  * a warning.
672                  */
673                 if (capable(CAP_SYS_ADMIN)) {
674                         pr_warn_once("%s (%d): Attempt to access syslog with "
675                                      "CAP_SYS_ADMIN but no CAP_SYSLOG "
676                                      "(deprecated).\n",
677                                  current->comm, task_pid_nr(current));
678                         goto ok;
679                 }
680                 return -EPERM;
681         }
682 ok:
683         return security_syslog(type);
684 }
685 
686 static void append_char(char **pp, char *e, char c)
687 {
688         if (*pp < e)
689                 *(*pp)++ = c;
690 }
691 
692 static ssize_t msg_print_ext_header(char *buf, size_t size,
693                                     struct printk_log *msg, u64 seq)
694 {
695         u64 ts_usec = msg->ts_nsec;
696 
697         do_div(ts_usec, 1000);
698 
699         return scnprintf(buf, size, "%u,%llu,%llu,%c;",
700                        (msg->facility << 3) | msg->level, seq, ts_usec,
701                        msg->flags & LOG_CONT ? 'c' : '-');
702 }
703 
704 static ssize_t msg_print_ext_body(char *buf, size_t size,
705                                   char *dict, size_t dict_len,
706                                   char *text, size_t text_len)
707 {
708         char *p = buf, *e = buf + size;
709         size_t i;
710 
711         /* escape non-printable characters */
712         for (i = 0; i < text_len; i++) {
713                 unsigned char c = text[i];
714 
715                 if (c < ' ' || c >= 127 || c == '\\')
716                         p += scnprintf(p, e - p, "\\x%02x", c);
717                 else
718                         append_char(&p, e, c);
719         }
720         append_char(&p, e, '\n');
721 
722         if (dict_len) {
723                 bool line = true;
724 
725                 for (i = 0; i < dict_len; i++) {
726                         unsigned char c = dict[i];
727 
728                         if (line) {
729                                 append_char(&p, e, ' ');
730                                 line = false;
731                         }
732 
733                         if (c == '\0') {
734                                 append_char(&p, e, '\n');
735                                 line = true;
736                                 continue;
737                         }
738 
739                         if (c < ' ' || c >= 127 || c == '\\') {
740                                 p += scnprintf(p, e - p, "\\x%02x", c);
741                                 continue;
742                         }
743 
744                         append_char(&p, e, c);
745                 }
746                 append_char(&p, e, '\n');
747         }
748 
749         return p - buf;
750 }
751 
752 /* /dev/kmsg - userspace message inject/listen interface */
753 struct devkmsg_user {
754         u64 seq;
755         u32 idx;
756         struct ratelimit_state rs;
757         struct mutex lock;
758         char buf[CONSOLE_EXT_LOG_MAX];
759 };
760 
761 static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from)
762 {
763         char *buf, *line;
764         int level = default_message_loglevel;
765         int facility = 1;       /* LOG_USER */
766         struct file *file = iocb->ki_filp;
767         struct devkmsg_user *user = file->private_data;
768         size_t len = iov_iter_count(from);
769         ssize_t ret = len;
770 
771         if (!user || len > LOG_LINE_MAX)
772                 return -EINVAL;
773 
774         /* Ignore when user logging is disabled. */
775         if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
776                 return len;
777 
778         /* Ratelimit when not explicitly enabled. */
779         if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) {
780                 if (!___ratelimit(&user->rs, current->comm))
781                         return ret;
782         }
783 
784         buf = kmalloc(len+1, GFP_KERNEL);
785         if (buf == NULL)
786                 return -ENOMEM;
787 
788         buf[len] = '\0';
789         if (!copy_from_iter_full(buf, len, from)) {
790                 kfree(buf);
791                 return -EFAULT;
792         }
793 
794         /*
795          * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
796          * the decimal value represents 32bit, the lower 3 bit are the log
797          * level, the rest are the log facility.
798          *
799          * If no prefix or no userspace facility is specified, we
800          * enforce LOG_USER, to be able to reliably distinguish
801          * kernel-generated messages from userspace-injected ones.
802          */
803         line = buf;
804         if (line[0] == '<') {
805                 char *endp = NULL;
806                 unsigned int u;
807 
808                 u = simple_strtoul(line + 1, &endp, 10);
809                 if (endp && endp[0] == '>') {
810                         level = LOG_LEVEL(u);
811                         if (LOG_FACILITY(u) != 0)
812                                 facility = LOG_FACILITY(u);
813                         endp++;
814                         len -= endp - line;
815                         line = endp;
816                 }
817         }
818 
819         printk_emit(facility, level, NULL, 0, "%s", line);
820         kfree(buf);
821         return ret;
822 }
823 
824 static ssize_t devkmsg_read(struct file *file, char __user *buf,
825                             size_t count, loff_t *ppos)
826 {
827         struct devkmsg_user *user = file->private_data;
828         struct printk_log *msg;
829         size_t len;
830         ssize_t ret;
831 
832         if (!user)
833                 return -EBADF;
834 
835         ret = mutex_lock_interruptible(&user->lock);
836         if (ret)
837                 return ret;
838 
839         logbuf_lock_irq();
840         while (user->seq == log_next_seq) {
841                 if (file->f_flags & O_NONBLOCK) {
842                         ret = -EAGAIN;
843                         logbuf_unlock_irq();
844                         goto out;
845                 }
846 
847                 logbuf_unlock_irq();
848                 ret = wait_event_interruptible(log_wait,
849                                                user->seq != log_next_seq);
850                 if (ret)
851                         goto out;
852                 logbuf_lock_irq();
853         }
854 
855         if (user->seq < log_first_seq) {
856                 /* our last seen message is gone, return error and reset */
857                 user->idx = log_first_idx;
858                 user->seq = log_first_seq;
859                 ret = -EPIPE;
860                 logbuf_unlock_irq();
861                 goto out;
862         }
863 
864         msg = log_from_idx(user->idx);
865         len = msg_print_ext_header(user->buf, sizeof(user->buf),
866                                    msg, user->seq);
867         len += msg_print_ext_body(user->buf + len, sizeof(user->buf) - len,
868                                   log_dict(msg), msg->dict_len,
869                                   log_text(msg), msg->text_len);
870 
871         user->idx = log_next(user->idx);
872         user->seq++;
873         logbuf_unlock_irq();
874 
875         if (len > count) {
876                 ret = -EINVAL;
877                 goto out;
878         }
879 
880         if (copy_to_user(buf, user->buf, len)) {
881                 ret = -EFAULT;
882                 goto out;
883         }
884         ret = len;
885 out:
886         mutex_unlock(&user->lock);
887         return ret;
888 }
889 
890 static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
891 {
892         struct devkmsg_user *user = file->private_data;
893         loff_t ret = 0;
894 
895         if (!user)
896                 return -EBADF;
897         if (offset)
898                 return -ESPIPE;
899 
900         logbuf_lock_irq();
901         switch (whence) {
902         case SEEK_SET:
903                 /* the first record */
904                 user->idx = log_first_idx;
905                 user->seq = log_first_seq;
906                 break;
907         case SEEK_DATA:
908                 /*
909                  * The first record after the last SYSLOG_ACTION_CLEAR,
910                  * like issued by 'dmesg -c'. Reading /dev/kmsg itself
911                  * changes no global state, and does not clear anything.
912                  */
913                 user->idx = clear_idx;
914                 user->seq = clear_seq;
915                 break;
916         case SEEK_END:
917                 /* after the last record */
918                 user->idx = log_next_idx;
919                 user->seq = log_next_seq;
920                 break;
921         default:
922                 ret = -EINVAL;
923         }
924         logbuf_unlock_irq();
925         return ret;
926 }
927 
928 static __poll_t devkmsg_poll(struct file *file, poll_table *wait)
929 {
930         struct devkmsg_user *user = file->private_data;
931         __poll_t ret = 0;
932 
933         if (!user)
934                 return EPOLLERR|EPOLLNVAL;
935 
936         poll_wait(file, &log_wait, wait);
937 
938         logbuf_lock_irq();
939         if (user->seq < log_next_seq) {
940                 /* return error when data has vanished underneath us */
941                 if (user->seq < log_first_seq)
942                         ret = EPOLLIN|EPOLLRDNORM|EPOLLERR|EPOLLPRI;
943                 else
944                         ret = EPOLLIN|EPOLLRDNORM;
945         }
946         logbuf_unlock_irq();
947 
948         return ret;
949 }
950 
951 static int devkmsg_open(struct inode *inode, struct file *file)
952 {
953         struct devkmsg_user *user;
954         int err;
955 
956         if (devkmsg_log & DEVKMSG_LOG_MASK_OFF)
957                 return -EPERM;
958 
959         /* write-only does not need any file context */
960         if ((file->f_flags & O_ACCMODE) != O_WRONLY) {
961                 err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
962                                                SYSLOG_FROM_READER);
963                 if (err)
964                         return err;
965         }
966 
967         user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
968         if (!user)
969                 return -ENOMEM;
970 
971         ratelimit_default_init(&user->rs);
972         ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE);
973 
974         mutex_init(&user->lock);
975 
976         logbuf_lock_irq();
977         user->idx = log_first_idx;
978         user->seq = log_first_seq;
979         logbuf_unlock_irq();
980 
981         file->private_data = user;
982         return 0;
983 }
984 
985 static int devkmsg_release(struct inode *inode, struct file *file)
986 {
987         struct devkmsg_user *user = file->private_data;
988 
989         if (!user)
990                 return 0;
991 
992         ratelimit_state_exit(&user->rs);
993 
994         mutex_destroy(&user->lock);
995         kfree(user);
996         return 0;
997 }
998 
999 const struct file_operations kmsg_fops = {
1000         .open = devkmsg_open,
1001         .read = devkmsg_read,
1002         .write_iter = devkmsg_write,
1003         .llseek = devkmsg_llseek,
1004         .poll = devkmsg_poll,
1005         .release = devkmsg_release,
1006 };
1007 
1008 #ifdef CONFIG_CRASH_CORE
1009 /*
1010  * This appends the listed symbols to /proc/vmcore
1011  *
1012  * /proc/vmcore is used by various utilities, like crash and makedumpfile to
1013  * obtain access to symbols that are otherwise very difficult to locate.  These
1014  * symbols are specifically used so that utilities can access and extract the
1015  * dmesg log from a vmcore file after a crash.
1016  */
1017 void log_buf_vmcoreinfo_setup(void)
1018 {
1019         VMCOREINFO_SYMBOL(log_buf);
1020         VMCOREINFO_SYMBOL(log_buf_len);
1021         VMCOREINFO_SYMBOL(log_first_idx);
1022         VMCOREINFO_SYMBOL(clear_idx);
1023         VMCOREINFO_SYMBOL(log_next_idx);
1024         /*
1025          * Export struct printk_log size and field offsets. User space tools can
1026          * parse it and detect any changes to structure down the line.
1027          */
1028         VMCOREINFO_STRUCT_SIZE(printk_log);
1029         VMCOREINFO_OFFSET(printk_log, ts_nsec);
1030         VMCOREINFO_OFFSET(printk_log, len);
1031         VMCOREINFO_OFFSET(printk_log, text_len);
1032         VMCOREINFO_OFFSET(printk_log, dict_len);
1033 }
1034 #endif
1035 
1036 /* requested log_buf_len from kernel cmdline */
1037 static unsigned long __initdata new_log_buf_len;
1038 
1039 /* we practice scaling the ring buffer by powers of 2 */
1040 static void __init log_buf_len_update(unsigned size)
1041 {
1042         if (size)
1043                 size = roundup_pow_of_two(size);
1044         if (size > log_buf_len)
1045                 new_log_buf_len = size;
1046 }
1047 
1048 /* save requested log_buf_len since it's too early to process it */
1049 static int __init log_buf_len_setup(char *str)
1050 {
1051         unsigned size = memparse(str, &str);
1052 
1053         log_buf_len_update(size);
1054 
1055         return 0;
1056 }
1057 early_param("log_buf_len", log_buf_len_setup);
1058 
1059 #ifdef CONFIG_SMP
1060 #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT)
1061 
1062 static void __init log_buf_add_cpu(void)
1063 {
1064         unsigned int cpu_extra;
1065 
1066         /*
1067          * archs should set up cpu_possible_bits properly with
1068          * set_cpu_possible() after setup_arch() but just in
1069          * case lets ensure this is valid.
1070          */
1071         if (num_possible_cpus() == 1)
1072                 return;
1073 
1074         cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
1075 
1076         /* by default this will only continue through for large > 64 CPUs */
1077         if (cpu_extra <= __LOG_BUF_LEN / 2)
1078                 return;
1079 
1080         pr_info("log_buf_len individual max cpu contribution: %d bytes\n",
1081                 __LOG_CPU_MAX_BUF_LEN);
1082         pr_info("log_buf_len total cpu_extra contributions: %d bytes\n",
1083                 cpu_extra);
1084         pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN);
1085 
1086         log_buf_len_update(cpu_extra + __LOG_BUF_LEN);
1087 }
1088 #else /* !CONFIG_SMP */
1089 static inline void log_buf_add_cpu(void) {}
1090 #endif /* CONFIG_SMP */
1091 
1092 void __init setup_log_buf(int early)
1093 {
1094         unsigned long flags;
1095         char *new_log_buf;
1096         int free;
1097 
1098         if (log_buf != __log_buf)
1099                 return;
1100 
1101         if (!early && !new_log_buf_len)
1102                 log_buf_add_cpu();
1103 
1104         if (!new_log_buf_len)
1105                 return;
1106 
1107         if (early) {
1108                 new_log_buf =
1109                         memblock_virt_alloc(new_log_buf_len, LOG_ALIGN);
1110         } else {
1111                 new_log_buf = memblock_virt_alloc_nopanic(new_log_buf_len,
1112                                                           LOG_ALIGN);
1113         }
1114 
1115         if (unlikely(!new_log_buf)) {
1116                 pr_err("log_buf_len: %ld bytes not available\n",
1117                         new_log_buf_len);
1118                 return;
1119         }
1120 
1121         logbuf_lock_irqsave(flags);
1122         log_buf_len = new_log_buf_len;
1123         log_buf = new_log_buf;
1124         new_log_buf_len = 0;
1125         free = __LOG_BUF_LEN - log_next_idx;
1126         memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
1127         logbuf_unlock_irqrestore(flags);
1128 
1129         pr_info("log_buf_len: %d bytes\n", log_buf_len);
1130         pr_info("early log buf free: %d(%d%%)\n",
1131                 free, (free * 100) / __LOG_BUF_LEN);
1132 }
1133 
1134 static bool __read_mostly ignore_loglevel;
1135 
1136 static int __init ignore_loglevel_setup(char *str)
1137 {
1138         ignore_loglevel = true;
1139         pr_info("debug: ignoring loglevel setting.\n");
1140 
1141         return 0;
1142 }
1143 
1144 early_param("ignore_loglevel", ignore_loglevel_setup);
1145 module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1146 MODULE_PARM_DESC(ignore_loglevel,
1147                  "ignore loglevel setting (prints all kernel messages to the console)");
1148 
1149 static bool suppress_message_printing(int level)
1150 {
1151         return (level >= console_loglevel && !ignore_loglevel);
1152 }
1153 
1154 #ifdef CONFIG_BOOT_PRINTK_DELAY
1155 
1156 static int boot_delay; /* msecs delay after each printk during bootup */
1157 static unsigned long long loops_per_msec;       /* based on boot_delay */
1158 
1159 static int __init boot_delay_setup(char *str)
1160 {
1161         unsigned long lpj;
1162 
1163         lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
1164         loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
1165 
1166         get_option(&str, &boot_delay);
1167         if (boot_delay > 10 * 1000)
1168                 boot_delay = 0;
1169 
1170         pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
1171                 "HZ: %d, loops_per_msec: %llu\n",
1172                 boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
1173         return 0;
1174 }
1175 early_param("boot_delay", boot_delay_setup);
1176 
1177 static void boot_delay_msec(int level)
1178 {
1179         unsigned long long k;
1180         unsigned long timeout;
1181 
1182         if ((boot_delay == 0 || system_state >= SYSTEM_RUNNING)
1183                 || suppress_message_printing(level)) {
1184                 return;
1185         }
1186 
1187         k = (unsigned long long)loops_per_msec * boot_delay;
1188 
1189         timeout = jiffies + msecs_to_jiffies(boot_delay);
1190         while (k) {
1191                 k--;
1192                 cpu_relax();
1193                 /*
1194                  * use (volatile) jiffies to prevent
1195                  * compiler reduction; loop termination via jiffies
1196                  * is secondary and may or may not happen.
1197                  */
1198                 if (time_after(jiffies, timeout))
1199                         break;
1200                 touch_nmi_watchdog();
1201         }
1202 }
1203 #else
1204 static inline void boot_delay_msec(int level)
1205 {
1206 }
1207 #endif
1208 
1209 static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME);
1210 module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
1211 
1212 static size_t print_time(u64 ts, char *buf)
1213 {
1214         unsigned long rem_nsec;
1215 
1216         if (!printk_time)
1217                 return 0;
1218 
1219         rem_nsec = do_div(ts, 1000000000);
1220 
1221         if (!buf)
1222                 return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
1223 
1224         return sprintf(buf, "[%5lu.%06lu] ",
1225                        (unsigned long)ts, rem_nsec / 1000);
1226 }
1227 
1228 static size_t print_prefix(const struct printk_log *msg, bool syslog, char *buf)
1229 {
1230         size_t len = 0;
1231         unsigned int prefix = (msg->facility << 3) | msg->level;
1232 
1233         if (syslog) {
1234                 if (buf) {
1235                         len += sprintf(buf, "<%u>", prefix);
1236                 } else {
1237                         len += 3;
1238                         if (prefix > 999)
1239                                 len += 3;
1240                         else if (prefix > 99)
1241                                 len += 2;
1242                         else if (prefix > 9)
1243                                 len++;
1244                 }
1245         }
1246 
1247         len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
1248         return len;
1249 }
1250 
1251 static size_t msg_print_text(const struct printk_log *msg, bool syslog, char *buf, size_t size)
1252 {
1253         const char *text = log_text(msg);
1254         size_t text_size = msg->text_len;
1255         size_t len = 0;
1256 
1257         do {
1258                 const char *next = memchr(text, '\n', text_size);
1259                 size_t text_len;
1260 
1261                 if (next) {
1262                         text_len = next - text;
1263                         next++;
1264                         text_size -= next - text;
1265                 } else {
1266                         text_len = text_size;
1267                 }
1268 
1269                 if (buf) {
1270                         if (print_prefix(msg, syslog, NULL) +
1271                             text_len + 1 >= size - len)
1272                                 break;
1273 
1274                         len += print_prefix(msg, syslog, buf + len);
1275                         memcpy(buf + len, text, text_len);
1276                         len += text_len;
1277                         buf[len++] = '\n';
1278                 } else {
1279                         /* SYSLOG_ACTION_* buffer size only calculation */
1280                         len += print_prefix(msg, syslog, NULL);
1281                         len += text_len;
1282                         len++;
1283                 }
1284 
1285                 text = next;
1286         } while (text);
1287 
1288         return len;
1289 }
1290 
1291 static int syslog_print(char __user *buf, int size)
1292 {
1293         char *text;
1294         struct printk_log *msg;
1295         int len = 0;
1296 
1297         text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1298         if (!text)
1299                 return -ENOMEM;
1300 
1301         while (size > 0) {
1302                 size_t n;
1303                 size_t skip;
1304 
1305                 logbuf_lock_irq();
1306                 if (syslog_seq < log_first_seq) {
1307                         /* messages are gone, move to first one */
1308                         syslog_seq = log_first_seq;
1309                         syslog_idx = log_first_idx;
1310                         syslog_partial = 0;
1311                 }
1312                 if (syslog_seq == log_next_seq) {
1313                         logbuf_unlock_irq();
1314                         break;
1315                 }
1316 
1317                 skip = syslog_partial;
1318                 msg = log_from_idx(syslog_idx);
1319                 n = msg_print_text(msg, true, text, LOG_LINE_MAX + PREFIX_MAX);
1320                 if (n - syslog_partial <= size) {
1321                         /* message fits into buffer, move forward */
1322                         syslog_idx = log_next(syslog_idx);
1323                         syslog_seq++;
1324                         n -= syslog_partial;
1325                         syslog_partial = 0;
1326                 } else if (!len){
1327                         /* partial read(), remember position */
1328                         n = size;
1329                         syslog_partial += n;
1330                 } else
1331                         n = 0;
1332                 logbuf_unlock_irq();
1333 
1334                 if (!n)
1335                         break;
1336 
1337                 if (copy_to_user(buf, text + skip, n)) {
1338                         if (!len)
1339                                 len = -EFAULT;
1340                         break;
1341                 }
1342 
1343                 len += n;
1344                 size -= n;
1345                 buf += n;
1346         }
1347 
1348         kfree(text);
1349         return len;
1350 }
1351 
1352 static int syslog_print_all(char __user *buf, int size, bool clear)
1353 {
1354         char *text;
1355         int len = 0;
1356         u64 next_seq;
1357         u64 seq;
1358         u32 idx;
1359 
1360         text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1361         if (!text)
1362                 return -ENOMEM;
1363 
1364         logbuf_lock_irq();
1365         /*
1366          * Find first record that fits, including all following records,
1367          * into the user-provided buffer for this dump.
1368          */
1369         seq = clear_seq;
1370         idx = clear_idx;
1371         while (seq < log_next_seq) {
1372                 struct printk_log *msg = log_from_idx(idx);
1373 
1374                 len += msg_print_text(msg, true, NULL, 0);
1375                 idx = log_next(idx);
1376                 seq++;
1377         }
1378 
1379         /* move first record forward until length fits into the buffer */
1380         seq = clear_seq;
1381         idx = clear_idx;
1382         while (len > size && seq < log_next_seq) {
1383                 struct printk_log *msg = log_from_idx(idx);
1384 
1385                 len -= msg_print_text(msg, true, NULL, 0);
1386                 idx = log_next(idx);
1387                 seq++;
1388         }
1389 
1390         /* last message fitting into this dump */
1391         next_seq = log_next_seq;
1392 
1393         len = 0;
1394         while (len >= 0 && seq < next_seq) {
1395                 struct printk_log *msg = log_from_idx(idx);
1396                 int textlen;
1397 
1398                 textlen = msg_print_text(msg, true, text,
1399                                          LOG_LINE_MAX + PREFIX_MAX);
1400                 if (textlen < 0) {
1401                         len = textlen;
1402                         break;
1403                 }
1404                 idx = log_next(idx);
1405                 seq++;
1406 
1407                 logbuf_unlock_irq();
1408                 if (copy_to_user(buf + len, text, textlen))
1409                         len = -EFAULT;
1410                 else
1411                         len += textlen;
1412                 logbuf_lock_irq();
1413 
1414                 if (seq < log_first_seq) {
1415                         /* messages are gone, move to next one */
1416                         seq = log_first_seq;
1417                         idx = log_first_idx;
1418                 }
1419         }
1420 
1421         if (clear) {
1422                 clear_seq = log_next_seq;
1423                 clear_idx = log_next_idx;
1424         }
1425         logbuf_unlock_irq();
1426 
1427         kfree(text);
1428         return len;
1429 }
1430 
1431 static void syslog_clear(void)
1432 {
1433         logbuf_lock_irq();
1434         clear_seq = log_next_seq;
1435         clear_idx = log_next_idx;
1436         logbuf_unlock_irq();
1437 }
1438 
1439 int do_syslog(int type, char __user *buf, int len, int source)
1440 {
1441         bool clear = false;
1442         static int saved_console_loglevel = LOGLEVEL_DEFAULT;
1443         int error;
1444 
1445         error = check_syslog_permissions(type, source);
1446         if (error)
1447                 return error;
1448 
1449         switch (type) {
1450         case SYSLOG_ACTION_CLOSE:       /* Close log */
1451                 break;
1452         case SYSLOG_ACTION_OPEN:        /* Open log */
1453                 break;
1454         case SYSLOG_ACTION_READ:        /* Read from log */
1455                 if (!buf || len < 0)
1456                         return -EINVAL;
1457                 if (!len)
1458                         return 0;
1459                 if (!access_ok(VERIFY_WRITE, buf, len))
1460                         return -EFAULT;
1461                 error = wait_event_interruptible(log_wait,
1462                                                  syslog_seq != log_next_seq);
1463                 if (error)
1464                         return error;
1465                 error = syslog_print(buf, len);
1466                 break;
1467         /* Read/clear last kernel messages */
1468         case SYSLOG_ACTION_READ_CLEAR:
1469                 clear = true;
1470                 /* FALL THRU */
1471         /* Read last kernel messages */
1472         case SYSLOG_ACTION_READ_ALL:
1473                 if (!buf || len < 0)
1474                         return -EINVAL;
1475                 if (!len)
1476                         return 0;
1477                 if (!access_ok(VERIFY_WRITE, buf, len))
1478                         return -EFAULT;
1479                 error = syslog_print_all(buf, len, clear);
1480                 break;
1481         /* Clear ring buffer */
1482         case SYSLOG_ACTION_CLEAR:
1483                 syslog_clear();
1484                 break;
1485         /* Disable logging to console */
1486         case SYSLOG_ACTION_CONSOLE_OFF:
1487                 if (saved_console_loglevel == LOGLEVEL_DEFAULT)
1488                         saved_console_loglevel = console_loglevel;
1489                 console_loglevel = minimum_console_loglevel;
1490                 break;
1491         /* Enable logging to console */
1492         case SYSLOG_ACTION_CONSOLE_ON:
1493                 if (saved_console_loglevel != LOGLEVEL_DEFAULT) {
1494                         console_loglevel = saved_console_loglevel;
1495                         saved_console_loglevel = LOGLEVEL_DEFAULT;
1496                 }
1497                 break;
1498         /* Set level of messages printed to console */
1499         case SYSLOG_ACTION_CONSOLE_LEVEL:
1500                 if (len < 1 || len > 8)
1501                         return -EINVAL;
1502                 if (len < minimum_console_loglevel)
1503                         len = minimum_console_loglevel;
1504                 console_loglevel = len;
1505                 /* Implicitly re-enable logging to console */
1506                 saved_console_loglevel = LOGLEVEL_DEFAULT;
1507                 break;
1508         /* Number of chars in the log buffer */
1509         case SYSLOG_ACTION_SIZE_UNREAD:
1510                 logbuf_lock_irq();
1511                 if (syslog_seq < log_first_seq) {
1512                         /* messages are gone, move to first one */
1513                         syslog_seq = log_first_seq;
1514                         syslog_idx = log_first_idx;
1515                         syslog_partial = 0;
1516                 }
1517                 if (source == SYSLOG_FROM_PROC) {
1518                         /*
1519                          * Short-cut for poll(/"proc/kmsg") which simply checks
1520                          * for pending data, not the size; return the count of
1521                          * records, not the length.
1522                          */
1523                         error = log_next_seq - syslog_seq;
1524                 } else {
1525                         u64 seq = syslog_seq;
1526                         u32 idx = syslog_idx;
1527 
1528                         while (seq < log_next_seq) {
1529                                 struct printk_log *msg = log_from_idx(idx);
1530 
1531                                 error += msg_print_text(msg, true, NULL, 0);
1532                                 idx = log_next(idx);
1533                                 seq++;
1534                         }
1535                         error -= syslog_partial;
1536                 }
1537                 logbuf_unlock_irq();
1538                 break;
1539         /* Size of the log buffer */
1540         case SYSLOG_ACTION_SIZE_BUFFER:
1541                 error = log_buf_len;
1542                 break;
1543         default:
1544                 error = -EINVAL;
1545                 break;
1546         }
1547 
1548         return error;
1549 }
1550 
1551 SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1552 {
1553         return do_syslog(type, buf, len, SYSLOG_FROM_READER);
1554 }
1555 
1556 /*
1557  * Special console_lock variants that help to reduce the risk of soft-lockups.
1558  * They allow to pass console_lock to another printk() call using a busy wait.
1559  */
1560 
1561 #ifdef CONFIG_LOCKDEP
1562 static struct lockdep_map console_owner_dep_map = {
1563         .name = "console_owner"
1564 };
1565 #endif
1566 
1567 static DEFINE_RAW_SPINLOCK(console_owner_lock);
1568 static struct task_struct *console_owner;
1569 static bool console_waiter;
1570 
1571 /**
1572  * console_lock_spinning_enable - mark beginning of code where another
1573  *      thread might safely busy wait
1574  *
1575  * This basically converts console_lock into a spinlock. This marks
1576  * the section where the console_lock owner can not sleep, because
1577  * there may be a waiter spinning (like a spinlock). Also it must be
1578  * ready to hand over the lock at the end of the section.
1579  */
1580 static void console_lock_spinning_enable(void)
1581 {
1582         raw_spin_lock(&console_owner_lock);
1583         console_owner = current;
1584         raw_spin_unlock(&console_owner_lock);
1585 
1586         /* The waiter may spin on us after setting console_owner */
1587         spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1588 }
1589 
1590 /**
1591  * console_lock_spinning_disable_and_check - mark end of code where another
1592  *      thread was able to busy wait and check if there is a waiter
1593  *
1594  * This is called at the end of the section where spinning is allowed.
1595  * It has two functions. First, it is a signal that it is no longer
1596  * safe to start busy waiting for the lock. Second, it checks if
1597  * there is a busy waiter and passes the lock rights to her.
1598  *
1599  * Important: Callers lose the lock if there was a busy waiter.
1600  *      They must not touch items synchronized by console_lock
1601  *      in this case.
1602  *
1603  * Return: 1 if the lock rights were passed, 0 otherwise.
1604  */
1605 static int console_lock_spinning_disable_and_check(void)
1606 {
1607         int waiter;
1608 
1609         raw_spin_lock(&console_owner_lock);
1610         waiter = READ_ONCE(console_waiter);
1611         console_owner = NULL;
1612         raw_spin_unlock(&console_owner_lock);
1613 
1614         if (!waiter) {
1615                 spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1616                 return 0;
1617         }
1618 
1619         /* The waiter is now free to continue */
1620         WRITE_ONCE(console_waiter, false);
1621 
1622         spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1623 
1624         /*
1625          * Hand off console_lock to waiter. The waiter will perform
1626          * the up(). After this, the waiter is the console_lock owner.
1627          */
1628         mutex_release(&console_lock_dep_map, 1, _THIS_IP_);
1629         return 1;
1630 }
1631 
1632 /**
1633  * console_trylock_spinning - try to get console_lock by busy waiting
1634  *
1635  * This allows to busy wait for the console_lock when the current
1636  * owner is running in specially marked sections. It means that
1637  * the current owner is running and cannot reschedule until it
1638  * is ready to lose the lock.
1639  *
1640  * Return: 1 if we got the lock, 0 othrewise
1641  */
1642 static int console_trylock_spinning(void)
1643 {
1644         struct task_struct *owner = NULL;
1645         bool waiter;
1646         bool spin = false;
1647         unsigned long flags;
1648 
1649         if (console_trylock())
1650                 return 1;
1651 
1652         printk_safe_enter_irqsave(flags);
1653 
1654         raw_spin_lock(&console_owner_lock);
1655         owner = READ_ONCE(console_owner);
1656         waiter = READ_ONCE(console_waiter);
1657         if (!waiter && owner && owner != current) {
1658                 WRITE_ONCE(console_waiter, true);
1659                 spin = true;
1660         }
1661         raw_spin_unlock(&console_owner_lock);
1662 
1663         /*
1664          * If there is an active printk() writing to the
1665          * consoles, instead of having it write our data too,
1666          * see if we can offload that load from the active
1667          * printer, and do some printing ourselves.
1668          * Go into a spin only if there isn't already a waiter
1669          * spinning, and there is an active printer, and
1670          * that active printer isn't us (recursive printk?).
1671          */
1672         if (!spin) {
1673                 printk_safe_exit_irqrestore(flags);
1674                 return 0;
1675         }
1676 
1677         /* We spin waiting for the owner to release us */
1678         spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_);
1679         /* Owner will clear console_waiter on hand off */
1680         while (READ_ONCE(console_waiter))
1681                 cpu_relax();
1682         spin_release(&console_owner_dep_map, 1, _THIS_IP_);
1683 
1684         printk_safe_exit_irqrestore(flags);
1685         /*
1686          * The owner passed the console lock to us.
1687          * Since we did not spin on console lock, annotate
1688          * this as a trylock. Otherwise lockdep will
1689          * complain.
1690          */
1691         mutex_acquire(&console_lock_dep_map, 0, 1, _THIS_IP_);
1692 
1693         return 1;
1694 }
1695 
1696 /*
1697  * Call the console drivers, asking them to write out
1698  * log_buf[start] to log_buf[end - 1].
1699  * The console_lock must be held.
1700  */
1701 static void call_console_drivers(const char *ext_text, size_t ext_len,
1702                                  const char *text, size_t len)
1703 {
1704         struct console *con;
1705 
1706         trace_console_rcuidle(text, len);
1707 
1708         if (!console_drivers)
1709                 return;
1710 
1711         for_each_console(con) {
1712                 if (exclusive_console && con != exclusive_console)
1713                         continue;
1714                 if (!(con->flags & CON_ENABLED))
1715                         continue;
1716                 if (!con->write)
1717                         continue;
1718                 if (!cpu_online(smp_processor_id()) &&
1719                     !(con->flags & CON_ANYTIME))
1720                         continue;
1721                 if (con->flags & CON_EXTENDED)
1722                         con->write(con, ext_text, ext_len);
1723                 else
1724                         con->write(con, text, len);
1725         }
1726 }
1727 
1728 int printk_delay_msec __read_mostly;
1729 
1730 static inline void printk_delay(void)
1731 {
1732         if (unlikely(printk_delay_msec)) {
1733                 int m = printk_delay_msec;
1734 
1735                 while (m--) {
1736                         mdelay(1);
1737                         touch_nmi_watchdog();
1738                 }
1739         }
1740 }
1741 
1742 /*
1743  * Continuation lines are buffered, and not committed to the record buffer
1744  * until the line is complete, or a race forces it. The line fragments
1745  * though, are printed immediately to the consoles to ensure everything has
1746  * reached the console in case of a kernel crash.
1747  */
1748 static struct cont {
1749         char buf[LOG_LINE_MAX];
1750         size_t len;                     /* length == 0 means unused buffer */
1751         struct task_struct *owner;      /* task of first print*/
1752         u64 ts_nsec;                    /* time of first print */
1753         u8 level;                       /* log level of first message */
1754         u8 facility;                    /* log facility of first message */
1755         enum log_flags flags;           /* prefix, newline flags */
1756 } cont;
1757 
1758 static void cont_flush(void)
1759 {
1760         if (cont.len == 0)
1761                 return;
1762 
1763         log_store(cont.facility, cont.level, cont.flags, cont.ts_nsec,
1764                   NULL, 0, cont.buf, cont.len);
1765         cont.len = 0;
1766 }
1767 
1768 static bool cont_add(int facility, int level, enum log_flags flags, const char *text, size_t len)
1769 {
1770         /*
1771          * If ext consoles are present, flush and skip in-kernel
1772          * continuation.  See nr_ext_console_drivers definition.  Also, if
1773          * the line gets too long, split it up in separate records.
1774          */
1775         if (nr_ext_console_drivers || cont.len + len > sizeof(cont.buf)) {
1776                 cont_flush();
1777                 return false;
1778         }
1779 
1780         if (!cont.len) {
1781                 cont.facility = facility;
1782                 cont.level = level;
1783                 cont.owner = current;
1784                 cont.ts_nsec = local_clock();
1785                 cont.flags = flags;
1786         }
1787 
1788         memcpy(cont.buf + cont.len, text, len);
1789         cont.len += len;
1790 
1791         // The original flags come from the first line,
1792         // but later continuations can add a newline.
1793         if (flags & LOG_NEWLINE) {
1794                 cont.flags |= LOG_NEWLINE;
1795                 cont_flush();
1796         }
1797 
1798         if (cont.len > (sizeof(cont.buf) * 80) / 100)
1799                 cont_flush();
1800 
1801         return true;
1802 }
1803 
1804 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)
1805 {
1806         /*
1807          * If an earlier line was buffered, and we're a continuation
1808          * write from the same process, try to add it to the buffer.
1809          */
1810         if (cont.len) {
1811                 if (cont.owner == current && (lflags & LOG_CONT)) {
1812                         if (cont_add(facility, level, lflags, text, text_len))
1813                                 return text_len;
1814                 }
1815                 /* Otherwise, make sure it's flushed */
1816                 cont_flush();
1817         }
1818 
1819         /* Skip empty continuation lines that couldn't be added - they just flush */
1820         if (!text_len && (lflags & LOG_CONT))
1821                 return 0;
1822 
1823         /* If it doesn't end in a newline, try to buffer the current line */
1824         if (!(lflags & LOG_NEWLINE)) {
1825                 if (cont_add(facility, level, lflags, text, text_len))
1826                         return text_len;
1827         }
1828 
1829         /* Store it in the record log */
1830         return log_store(facility, level, lflags, 0, dict, dictlen, text, text_len);
1831 }
1832 
1833 /* Must be called under logbuf_lock. */
1834 int vprintk_store(int facility, int level,
1835                   const char *dict, size_t dictlen,
1836                   const char *fmt, va_list args)
1837 {
1838         static char textbuf[LOG_LINE_MAX];
1839         char *text = textbuf;
1840         size_t text_len;
1841         enum log_flags lflags = 0;
1842 
1843         /*
1844          * The printf needs to come first; we need the syslog
1845          * prefix which might be passed-in as a parameter.
1846          */
1847         text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1848 
1849         /* mark and strip a trailing newline */
1850         if (text_len && text[text_len-1] == '\n') {
1851                 text_len--;
1852                 lflags |= LOG_NEWLINE;
1853         }
1854 
1855         /* strip kernel syslog prefix and extract log level or control flags */
1856         if (facility == 0) {
1857                 int kern_level;
1858 
1859                 while ((kern_level = printk_get_level(text)) != 0) {
1860                         switch (kern_level) {
1861                         case '' ... '7':
1862                                 if (level == LOGLEVEL_DEFAULT)
1863                                         level = kern_level - '';
1864                                 /* fallthrough */
1865                         case 'd':       /* KERN_DEFAULT */
1866                                 lflags |= LOG_PREFIX;
1867                                 break;
1868                         case 'c':       /* KERN_CONT */
1869                                 lflags |= LOG_CONT;
1870                         }
1871 
1872                         text_len -= 2;
1873                         text += 2;
1874                 }
1875         }
1876 
1877         if (level == LOGLEVEL_DEFAULT)
1878                 level = default_message_loglevel;
1879 
1880         if (dict)
1881                 lflags |= LOG_PREFIX|LOG_NEWLINE;
1882 
1883         return log_output(facility, level, lflags,
1884                           dict, dictlen, text, text_len);
1885 }
1886 
1887 asmlinkage int vprintk_emit(int facility, int level,
1888                             const char *dict, size_t dictlen,
1889                             const char *fmt, va_list args)
1890 {
1891         int printed_len;
1892         bool in_sched = false;
1893         unsigned long flags;
1894 
1895         if (level == LOGLEVEL_SCHED) {
1896                 level = LOGLEVEL_DEFAULT;
1897                 in_sched = true;
1898         }
1899 
1900         boot_delay_msec(level);
1901         printk_delay();
1902 
1903         /* This stops the holder of console_sem just where we want him */
1904         logbuf_lock_irqsave(flags);
1905         printed_len = vprintk_store(facility, level, dict, dictlen, fmt, args);
1906         logbuf_unlock_irqrestore(flags);
1907 
1908         /* If called from the scheduler, we can not call up(). */
1909         if (!in_sched) {
1910                 /*
1911                  * Disable preemption to avoid being preempted while holding
1912                  * console_sem which would prevent anyone from printing to
1913                  * console
1914                  */
1915                 preempt_disable();
1916                 /*
1917                  * Try to acquire and then immediately release the console
1918                  * semaphore.  The release will print out buffers and wake up
1919                  * /dev/kmsg and syslog() users.
1920                  */
1921                 if (console_trylock_spinning())
1922                         console_unlock();
1923                 preempt_enable();
1924         }
1925 
1926         wake_up_klogd();
1927         return printed_len;
1928 }
1929 EXPORT_SYMBOL(vprintk_emit);
1930 
1931 asmlinkage int vprintk(const char *fmt, va_list args)
1932 {
1933         return vprintk_func(fmt, args);
1934 }
1935 EXPORT_SYMBOL(vprintk);
1936 
1937 asmlinkage int printk_emit(int facility, int level,
1938                            const char *dict, size_t dictlen,
1939                            const char *fmt, ...)
1940 {
1941         va_list args;
1942         int r;
1943 
1944         va_start(args, fmt);
1945         r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1946         va_end(args);
1947 
1948         return r;
1949 }
1950 EXPORT_SYMBOL(printk_emit);
1951 
1952 int vprintk_default(const char *fmt, va_list args)
1953 {
1954         int r;
1955 
1956 #ifdef CONFIG_KGDB_KDB
1957         /* Allow to pass printk() to kdb but avoid a recursion. */
1958         if (unlikely(kdb_trap_printk && kdb_printf_cpu < 0)) {
1959                 r = vkdb_printf(KDB_MSGSRC_PRINTK, fmt, args);
1960                 return r;
1961         }
1962 #endif
1963         r = vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
1964 
1965         return r;
1966 }
1967 EXPORT_SYMBOL_GPL(vprintk_default);
1968 
1969 /**
1970  * printk - print a kernel message
1971  * @fmt: format string
1972  *
1973  * This is printk(). It can be called from any context. We want it to work.
1974  *
1975  * We try to grab the console_lock. If we succeed, it's easy - we log the
1976  * output and call the console drivers.  If we fail to get the semaphore, we
1977  * place the output into the log buffer and return. The current holder of
1978  * the console_sem will notice the new output in console_unlock(); and will
1979  * send it to the consoles before releasing the lock.
1980  *
1981  * One effect of this deferred printing is that code which calls printk() and
1982  * then changes console_loglevel may break. This is because console_loglevel
1983  * is inspected when the actual printing occurs.
1984  *
1985  * See also:
1986  * printf(3)
1987  *
1988  * See the vsnprintf() documentation for format string extensions over C99.
1989  */
1990 asmlinkage __visible int printk(const char *fmt, ...)
1991 {
1992         va_list args;
1993         int r;
1994 
1995         va_start(args, fmt);
1996         r = vprintk_func(fmt, args);
1997         va_end(args);
1998 
1999         return r;
2000 }
2001 EXPORT_SYMBOL(printk);
2002 
2003 #else /* CONFIG_PRINTK */
2004 
2005 #define LOG_LINE_MAX            0
2006 #define PREFIX_MAX              0
2007 
2008 static u64 syslog_seq;
2009 static u32 syslog_idx;
2010 static u64 console_seq;
2011 static u32 console_idx;
2012 static u64 log_first_seq;
2013 static u32 log_first_idx;
2014 static u64 log_next_seq;
2015 static char *log_text(const struct printk_log *msg) { return NULL; }
2016 static char *log_dict(const struct printk_log *msg) { return NULL; }
2017 static struct printk_log *log_from_idx(u32 idx) { return NULL; }
2018 static u32 log_next(u32 idx) { return 0; }
2019 static ssize_t msg_print_ext_header(char *buf, size_t size,
2020                                     struct printk_log *msg,
2021                                     u64 seq) { return 0; }
2022 static ssize_t msg_print_ext_body(char *buf, size_t size,
2023                                   char *dict, size_t dict_len,
2024                                   char *text, size_t text_len) { return 0; }
2025 static void console_lock_spinning_enable(void) { }
2026 static int console_lock_spinning_disable_and_check(void) { return 0; }
2027 static void call_console_drivers(const char *ext_text, size_t ext_len,
2028                                  const char *text, size_t len) {}
2029 static size_t msg_print_text(const struct printk_log *msg,
2030                              bool syslog, char *buf, size_t size) { return 0; }
2031 static bool suppress_message_printing(int level) { return false; }
2032 
2033 #endif /* CONFIG_PRINTK */
2034 
2035 #ifdef CONFIG_EARLY_PRINTK
2036 struct console *early_console;
2037 
2038 asmlinkage __visible void early_printk(const char *fmt, ...)
2039 {
2040         va_list ap;
2041         char buf[512];
2042         int n;
2043 
2044         if (!early_console)
2045                 return;
2046 
2047         va_start(ap, fmt);
2048         n = vscnprintf(buf, sizeof(buf), fmt, ap);
2049         va_end(ap);
2050 
2051         early_console->write(early_console, buf, n);
2052 }
2053 #endif
2054 
2055 static int __add_preferred_console(char *name, int idx, char *options,
2056                                    char *brl_options)
2057 {
2058         struct console_cmdline *c;
2059         int i;
2060 
2061         /*
2062          *      See if this tty is not yet registered, and
2063          *      if we have a slot free.
2064          */
2065         for (i = 0, c = console_cmdline;
2066              i < MAX_CMDLINECONSOLES && c->name[0];
2067              i++, c++) {
2068                 if (strcmp(c->name, name) == 0 && c->index == idx) {
2069                         if (!brl_options)
2070                                 preferred_console = i;
2071                         return 0;
2072                 }
2073         }
2074         if (i == MAX_CMDLINECONSOLES)
2075                 return -E2BIG;
2076         if (!brl_options)
2077                 preferred_console = i;
2078         strlcpy(c->name, name, sizeof(c->name));
2079         c->options = options;
2080         braille_set_options(c, brl_options);
2081 
2082         c->index = idx;
2083         return 0;
2084 }
2085 
2086 static int __init console_msg_format_setup(char *str)
2087 {
2088         if (!strcmp(str, "syslog"))
2089                 console_msg_format = MSG_FORMAT_SYSLOG;
2090         if (!strcmp(str, "default"))
2091                 console_msg_format = MSG_FORMAT_DEFAULT;
2092         return 1;
2093 }
2094 __setup("console_msg_format=", console_msg_format_setup);
2095 
2096 /*
2097  * Set up a console.  Called via do_early_param() in init/main.c
2098  * for each "console=" parameter in the boot command line.
2099  */
2100 static int __init console_setup(char *str)
2101 {
2102         char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */
2103         char *s, *options, *brl_options = NULL;
2104         int idx;
2105 
2106         if (_braille_console_setup(&str, &brl_options))
2107                 return 1;
2108 
2109         /*
2110          * Decode str into name, index, options.
2111          */
2112         if (str[0] >= '' && str[0] <= '9') {
2113                 strcpy(buf, "ttyS");
2114                 strncpy(buf + 4, str, sizeof(buf) - 5);
2115         } else {
2116                 strncpy(buf, str, sizeof(buf) - 1);
2117         }
2118         buf[sizeof(buf) - 1] = 0;
2119         options = strchr(str, ',');
2120         if (options)
2121                 *(options++) = 0;
2122 #ifdef __sparc__
2123         if (!strcmp(str, "ttya"))
2124                 strcpy(buf, "ttyS0");
2125         if (!strcmp(str, "ttyb"))
2126                 strcpy(buf, "ttyS1");
2127 #endif
2128         for (s = buf; *s; s++)
2129                 if (isdigit(*s) || *s == ',')
2130                         break;
2131         idx = simple_strtoul(s, NULL, 10);
2132         *s = 0;
2133 
2134         __add_preferred_console(buf, idx, options, brl_options);
2135         console_set_on_cmdline = 1;
2136         return 1;
2137 }
2138 __setup("console=", console_setup);
2139 
2140 /**
2141  * add_preferred_console - add a device to the list of preferred consoles.
2142  * @name: device name
2143  * @idx: device index
2144  * @options: options for this console
2145  *
2146  * The last preferred console added will be used for kernel messages
2147  * and stdin/out/err for init.  Normally this is used by console_setup
2148  * above to handle user-supplied console arguments; however it can also
2149  * be used by arch-specific code either to override the user or more
2150  * commonly to provide a default console (ie from PROM variables) when
2151  * the user has not supplied one.
2152  */
2153 int add_preferred_console(char *name, int idx, char *options)
2154 {
2155         return __add_preferred_console(name, idx, options, NULL);
2156 }
2157 
2158 bool console_suspend_enabled = true;
2159 EXPORT_SYMBOL(console_suspend_enabled);
2160 
2161 static int __init console_suspend_disable(char *str)
2162 {
2163         console_suspend_enabled = false;
2164         return 1;
2165 }
2166 __setup("no_console_suspend", console_suspend_disable);
2167 module_param_named(console_suspend, console_suspend_enabled,
2168                 bool, S_IRUGO | S_IWUSR);
2169 MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
2170         " and hibernate operations");
2171 
2172 /**
2173  * suspend_console - suspend the console subsystem
2174  *
2175  * This disables printk() while we go into suspend states
2176  */
2177 void suspend_console(void)
2178 {
2179         if (!console_suspend_enabled)
2180                 return;
2181         pr_info("Suspending console(s) (use no_console_suspend to debug)\n");
2182         console_lock();
2183         console_suspended = 1;
2184         up_console_sem();
2185 }
2186 
2187 void resume_console(void)
2188 {
2189         if (!console_suspend_enabled)
2190                 return;
2191         down_console_sem();
2192         console_suspended = 0;
2193         console_unlock();
2194 }
2195 
2196 /**
2197  * console_cpu_notify - print deferred console messages after CPU hotplug
2198  * @cpu: unused
2199  *
2200  * If printk() is called from a CPU that is not online yet, the messages
2201  * will be printed on the console only if there are CON_ANYTIME consoles.
2202  * This function is called when a new CPU comes online (or fails to come
2203  * up) or goes offline.
2204  */
2205 static int console_cpu_notify(unsigned int cpu)
2206 {
2207         if (!cpuhp_tasks_frozen) {
2208                 /* If trylock fails, someone else is doing the printing */
2209                 if (console_trylock())
2210                         console_unlock();
2211         }
2212         return 0;
2213 }
2214 
2215 /**
2216  * console_lock - lock the console system for exclusive use.
2217  *
2218  * Acquires a lock which guarantees that the caller has
2219  * exclusive access to the console system and the console_drivers list.
2220  *
2221  * Can sleep, returns nothing.
2222  */
2223 void console_lock(void)
2224 {
2225         might_sleep();
2226 
2227         down_console_sem();
2228         if (console_suspended)
2229                 return;
2230         console_locked = 1;
2231         console_may_schedule = 1;
2232 }
2233 EXPORT_SYMBOL(console_lock);
2234 
2235 /**
2236  * console_trylock - try to lock the console system for exclusive use.
2237  *
2238  * Try to acquire a lock which guarantees that the caller has exclusive
2239  * access to the console system and the console_drivers list.
2240  *
2241  * returns 1 on success, and 0 on failure to acquire the lock.
2242  */
2243 int console_trylock(void)
2244 {
2245         if (down_trylock_console_sem())
2246                 return 0;
2247         if (console_suspended) {
2248                 up_console_sem();
2249                 return 0;
2250         }
2251         console_locked = 1;
2252         console_may_schedule = 0;
2253         return 1;
2254 }
2255 EXPORT_SYMBOL(console_trylock);
2256 
2257 int is_console_locked(void)
2258 {
2259         return console_locked;
2260 }
2261 EXPORT_SYMBOL(is_console_locked);
2262 
2263 /*
2264  * Check if we have any console that is capable of printing while cpu is
2265  * booting or shutting down. Requires console_sem.
2266  */
2267 static int have_callable_console(void)
2268 {
2269         struct console *con;
2270 
2271         for_each_console(con)
2272                 if ((con->flags & CON_ENABLED) &&
2273                                 (con->flags & CON_ANYTIME))
2274                         return 1;
2275 
2276         return 0;
2277 }
2278 
2279 /*
2280  * Can we actually use the console at this time on this cpu?
2281  *
2282  * Console drivers may assume that per-cpu resources have been allocated. So
2283  * unless they're explicitly marked as being able to cope (CON_ANYTIME) don't
2284  * call them until this CPU is officially up.
2285  */
2286 static inline int can_use_console(void)
2287 {
2288         return cpu_online(raw_smp_processor_id()) || have_callable_console();
2289 }
2290 
2291 /**
2292  * console_unlock - unlock the console system
2293  *
2294  * Releases the console_lock which the caller holds on the console system
2295  * and the console driver list.
2296  *
2297  * While the console_lock was held, console output may have been buffered
2298  * by printk().  If this is the case, console_unlock(); emits
2299  * the output prior to releasing the lock.
2300  *
2301  * If there is output waiting, we wake /dev/kmsg and syslog() users.
2302  *
2303  * console_unlock(); may be called from any context.
2304  */
2305 void console_unlock(void)
2306 {
2307         static char ext_text[CONSOLE_EXT_LOG_MAX];
2308         static char text[LOG_LINE_MAX + PREFIX_MAX];
2309         unsigned long flags;
2310         bool do_cond_resched, retry;
2311 
2312         if (console_suspended) {
2313                 up_console_sem();
2314                 return;
2315         }
2316 
2317         /*
2318          * Console drivers are called with interrupts disabled, so
2319          * @console_may_schedule should be cleared before; however, we may
2320          * end up dumping a lot of lines, for example, if called from
2321          * console registration path, and should invoke cond_resched()
2322          * between lines if allowable.  Not doing so can cause a very long
2323          * scheduling stall on a slow console leading to RCU stall and
2324          * softlockup warnings which exacerbate the issue with more
2325          * messages practically incapacitating the system.
2326          *
2327          * console_trylock() is not able to detect the preemptive
2328          * context reliably. Therefore the value must be stored before
2329          * and cleared after the the "again" goto label.
2330          */
2331         do_cond_resched = console_may_schedule;
2332 again:
2333         console_may_schedule = 0;
2334 
2335         /*
2336          * We released the console_sem lock, so we need to recheck if
2337          * cpu is online and (if not) is there at least one CON_ANYTIME
2338          * console.
2339          */
2340         if (!can_use_console()) {
2341                 console_locked = 0;
2342                 up_console_sem();
2343                 return;
2344         }
2345 
2346         for (;;) {
2347                 struct printk_log *msg;
2348                 size_t ext_len = 0;
2349                 size_t len;
2350 
2351                 printk_safe_enter_irqsave(flags);
2352                 raw_spin_lock(&logbuf_lock);
2353                 if (console_seq < log_first_seq) {
2354                         len = sprintf(text, "** %u printk messages dropped **\n",
2355                                       (unsigned)(log_first_seq - console_seq));
2356 
2357                         /* messages are gone, move to first one */
2358                         console_seq = log_first_seq;
2359                         console_idx = log_first_idx;
2360                 } else {
2361                         len = 0;
2362                 }
2363 skip:
2364                 if (console_seq == log_next_seq)
2365                         break;
2366 
2367                 msg = log_from_idx(console_idx);
2368                 if (suppress_message_printing(msg->level)) {
2369                         /*
2370                          * Skip record we have buffered and already printed
2371                          * directly to the console when we received it, and
2372                          * record that has level above the console loglevel.
2373                          */
2374                         console_idx = log_next(console_idx);
2375                         console_seq++;
2376                         goto skip;
2377                 }
2378 
2379                 len += msg_print_text(msg,
2380                                 console_msg_format & MSG_FORMAT_SYSLOG,
2381                                 text + len,
2382                                 sizeof(text) - len);
2383                 if (nr_ext_console_drivers) {
2384                         ext_len = msg_print_ext_header(ext_text,
2385                                                 sizeof(ext_text),
2386                                                 msg, console_seq);
2387                         ext_len += msg_print_ext_body(ext_text + ext_len,
2388                                                 sizeof(ext_text) - ext_len,
2389                                                 log_dict(msg), msg->dict_len,
2390                                                 log_text(msg), msg->text_len);
2391                 }
2392                 console_idx = log_next(console_idx);
2393                 console_seq++;
2394                 raw_spin_unlock(&logbuf_lock);
2395 
2396                 /*
2397                  * While actively printing out messages, if another printk()
2398                  * were to occur on another CPU, it may wait for this one to
2399                  * finish. This task can not be preempted if there is a
2400                  * waiter waiting to take over.
2401                  */
2402                 console_lock_spinning_enable();
2403 
2404                 stop_critical_timings();        /* don't trace print latency */
2405                 call_console_drivers(ext_text, ext_len, text, len);
2406                 start_critical_timings();
2407 
2408                 if (console_lock_spinning_disable_and_check()) {
2409                         printk_safe_exit_irqrestore(flags);
2410                         return;
2411                 }
2412 
2413                 printk_safe_exit_irqrestore(flags);
2414 
2415                 if (do_cond_resched)
2416                         cond_resched();
2417         }
2418 
2419         console_locked = 0;
2420 
2421         /* Release the exclusive_console once it is used */
2422         if (unlikely(exclusive_console))
2423                 exclusive_console = NULL;
2424 
2425         raw_spin_unlock(&logbuf_lock);
2426 
2427         up_console_sem();
2428 
2429         /*
2430          * Someone could have filled up the buffer again, so re-check if there's
2431          * something to flush. In case we cannot trylock the console_sem again,
2432          * there's a new owner and the console_unlock() from them will do the
2433          * flush, no worries.
2434          */
2435         raw_spin_lock(&logbuf_lock);
2436         retry = console_seq != log_next_seq;
2437         raw_spin_unlock(&logbuf_lock);
2438         printk_safe_exit_irqrestore(flags);
2439 
2440         if (retry && console_trylock())
2441                 goto again;
2442 }
2443 EXPORT_SYMBOL(console_unlock);
2444 
2445 /**
2446  * console_conditional_schedule - yield the CPU if required
2447  *
2448  * If the console code is currently allowed to sleep, and
2449  * if this CPU should yield the CPU to another task, do
2450  * so here.
2451  *
2452  * Must be called within console_lock();.
2453  */
2454 void __sched console_conditional_schedule(void)
2455 {
2456         if (console_may_schedule)
2457                 cond_resched();
2458 }
2459 EXPORT_SYMBOL(console_conditional_schedule);
2460 
2461 void console_unblank(void)
2462 {
2463         struct console *c;
2464 
2465         /*
2466          * console_unblank can no longer be called in interrupt context unless
2467          * oops_in_progress is set to 1..
2468          */
2469         if (oops_in_progress) {
2470                 if (down_trylock_console_sem() != 0)
2471                         return;
2472         } else
2473                 console_lock();
2474 
2475         console_locked = 1;
2476         console_may_schedule = 0;
2477         for_each_console(c)
2478                 if ((c->flags & CON_ENABLED) && c->unblank)
2479                         c->unblank();
2480         console_unlock();
2481 }
2482 
2483 /**
2484  * console_flush_on_panic - flush console content on panic
2485  *
2486  * Immediately output all pending messages no matter what.
2487  */
2488 void console_flush_on_panic(void)
2489 {
2490         /*
2491          * If someone else is holding the console lock, trylock will fail
2492          * and may_schedule may be set.  Ignore and proceed to unlock so
2493          * that messages are flushed out.  As this can be called from any
2494          * context and we don't want to get preempted while flushing,
2495          * ensure may_schedule is cleared.
2496          */
2497         console_trylock();
2498         console_may_schedule = 0;
2499         console_unlock();
2500 }
2501 
2502 /*
2503  * Return the console tty driver structure and its associated index
2504  */
2505 struct tty_driver *console_device(int *index)
2506 {
2507         struct console *c;
2508         struct tty_driver *driver = NULL;
2509 
2510         console_lock();
2511         for_each_console(c) {
2512                 if (!c->device)
2513                         continue;
2514                 driver = c->device(c, index);
2515                 if (driver)
2516                         break;
2517         }
2518         console_unlock();
2519         return driver;
2520 }
2521 
2522 /*
2523  * Prevent further output on the passed console device so that (for example)
2524  * serial drivers can disable console output before suspending a port, and can
2525  * re-enable output afterwards.
2526  */
2527 void console_stop(struct console *console)
2528 {
2529         console_lock();
2530         console->flags &= ~CON_ENABLED;
2531         console_unlock();
2532 }
2533 EXPORT_SYMBOL(console_stop);
2534 
2535 void console_start(struct console *console)
2536 {
2537         console_lock();
2538         console->flags |= CON_ENABLED;
2539         console_unlock();
2540 }
2541 EXPORT_SYMBOL(console_start);
2542 
2543 static int __read_mostly keep_bootcon;
2544 
2545 static int __init keep_bootcon_setup(char *str)
2546 {
2547         keep_bootcon = 1;
2548         pr_info("debug: skip boot console de-registration.\n");
2549 
2550         return 0;
2551 }
2552 
2553 early_param("keep_bootcon", keep_bootcon_setup);
2554 
2555 /*
2556  * The console driver calls this routine during kernel initialization
2557  * to register the console printing procedure with printk() and to
2558  * print any messages that were printed by the kernel before the
2559  * console driver was initialized.
2560  *
2561  * This can happen pretty early during the boot process (because of
2562  * early_printk) - sometimes before setup_arch() completes - be careful
2563  * of what kernel features are used - they may not be initialised yet.
2564  *
2565  * There are two types of consoles - bootconsoles (early_printk) and
2566  * "real" consoles (everything which is not a bootconsole) which are
2567  * handled differently.
2568  *  - Any number of bootconsoles can be registered at any time.
2569  *  - As soon as a "real" console is registered, all bootconsoles
2570  *    will be unregistered automatically.
2571  *  - Once a "real" console is registered, any attempt to register a
2572  *    bootconsoles will be rejected
2573  */
2574 void register_console(struct console *newcon)
2575 {
2576         int i;
2577         unsigned long flags;
2578         struct console *bcon = NULL;
2579         struct console_cmdline *c;
2580         static bool has_preferred;
2581 
2582         if (console_drivers)
2583                 for_each_console(bcon)
2584                         if (WARN(bcon == newcon,
2585                                         "console '%s%d' already registered\n",
2586                                         bcon->name, bcon->index))
2587                                 return;
2588 
2589         /*
2590          * before we register a new CON_BOOT console, make sure we don't
2591          * already have a valid console
2592          */
2593         if (console_drivers && newcon->flags & CON_BOOT) {
2594                 /* find the last or real console */
2595                 for_each_console(bcon) {
2596                         if (!(bcon->flags & CON_BOOT)) {
2597                                 pr_info("Too late to register bootconsole %s%d\n",
2598                                         newcon->name, newcon->index);
2599                                 return;
2600                         }
2601                 }
2602         }
2603 
2604         if (console_drivers && console_drivers->flags & CON_BOOT)
2605                 bcon = console_drivers;
2606 
2607         if (!has_preferred || bcon || !console_drivers)
2608                 has_preferred = preferred_console >= 0;
2609 
2610         /*
2611          *      See if we want to use this console driver. If we
2612          *      didn't select a console we take the first one
2613          *      that registers here.
2614          */
2615         if (!has_preferred) {
2616                 if (newcon->index < 0)
2617                         newcon->index = 0;
2618                 if (newcon->setup == NULL ||
2619                     newcon->setup(newcon, NULL) == 0) {
2620                         newcon->flags |= CON_ENABLED;
2621                         if (newcon->device) {
2622                                 newcon->flags |= CON_CONSDEV;
2623                                 has_preferred = true;
2624                         }
2625                 }
2626         }
2627 
2628         /*
2629          *      See if this console matches one we selected on
2630          *      the command line.
2631          */
2632         for (i = 0, c = console_cmdline;
2633              i < MAX_CMDLINECONSOLES && c->name[0];
2634              i++, c++) {
2635                 if (!newcon->match ||
2636                     newcon->match(newcon, c->name, c->index, c->options) != 0) {
2637                         /* default matching */
2638                         BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name));
2639                         if (strcmp(c->name, newcon->name) != 0)
2640                                 continue;
2641                         if (newcon->index >= 0 &&
2642                             newcon->index != c->index)
2643                                 continue;
2644                         if (newcon->index < 0)
2645                                 newcon->index = c->index;
2646 
2647                         if (_braille_register_console(newcon, c))
2648                                 return;
2649 
2650                         if (newcon->setup &&
2651                             newcon->setup(newcon, c->options) != 0)
2652                                 break;
2653                 }
2654 
2655                 newcon->flags |= CON_ENABLED;
2656                 if (i == preferred_console) {
2657                         newcon->flags |= CON_CONSDEV;
2658                         has_preferred = true;
2659                 }
2660                 break;
2661         }
2662 
2663         if (!(newcon->flags & CON_ENABLED))
2664                 return;
2665 
2666         /*
2667          * If we have a bootconsole, and are switching to a real console,
2668          * don't print everything out again, since when the boot console, and
2669          * the real console are the same physical device, it's annoying to
2670          * see the beginning boot messages twice
2671          */
2672         if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2673                 newcon->flags &= ~CON_PRINTBUFFER;
2674 
2675         /*
2676          *      Put this console in the list - keep the
2677          *      preferred driver at the head of the list.
2678          */
2679         console_lock();
2680         if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2681                 newcon->next = console_drivers;
2682                 console_drivers = newcon;
2683                 if (newcon->next)
2684                         newcon->next->flags &= ~CON_CONSDEV;
2685         } else {
2686                 newcon->next = console_drivers->next;
2687                 console_drivers->next = newcon;
2688         }
2689 
2690         if (newcon->flags & CON_EXTENDED)
2691                 if (!nr_ext_console_drivers++)
2692                         pr_info("printk: continuation disabled due to ext consoles, expect more fragments in /dev/kmsg\n");
2693 
2694         if (newcon->flags & CON_PRINTBUFFER) {
2695                 /*
2696                  * console_unlock(); will print out the buffered messages
2697                  * for us.
2698                  */
2699                 logbuf_lock_irqsave(flags);
2700                 console_seq = syslog_seq;
2701                 console_idx = syslog_idx;
2702                 logbuf_unlock_irqrestore(flags);
2703                 /*
2704                  * We're about to replay the log buffer.  Only do this to the
2705                  * just-registered console to avoid excessive message spam to
2706                  * the already-registered consoles.
2707                  */
2708                 exclusive_console = newcon;
2709         }
2710         console_unlock();
2711         console_sysfs_notify();
2712 
2713         /*
2714          * By unregistering the bootconsoles after we enable the real console
2715          * we get the "console xxx enabled" message on all the consoles -
2716          * boot consoles, real consoles, etc - this is to ensure that end
2717          * users know there might be something in the kernel's log buffer that
2718          * went to the bootconsole (that they do not see on the real console)
2719          */
2720         pr_info("%sconsole [%s%d] enabled\n",
2721                 (newcon->flags & CON_BOOT) ? "boot" : "" ,
2722                 newcon->name, newcon->index);
2723         if (bcon &&
2724             ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2725             !keep_bootcon) {
2726                 /* We need to iterate through all boot consoles, to make
2727                  * sure we print everything out, before we unregister them.
2728                  */
2729                 for_each_console(bcon)
2730                         if (bcon->flags & CON_BOOT)
2731                                 unregister_console(bcon);
2732         }
2733 }
2734 EXPORT_SYMBOL(register_console);
2735 
2736 int unregister_console(struct console *console)
2737 {
2738         struct console *a, *b;
2739         int res;
2740 
2741         pr_info("%sconsole [%s%d] disabled\n",
2742                 (console->flags & CON_BOOT) ? "boot" : "" ,
2743                 console->name, console->index);
2744 
2745         res = _braille_unregister_console(console);
2746         if (res)
2747                 return res;
2748 
2749         res = 1;
2750         console_lock();
2751         if (console_drivers == console) {
2752                 console_drivers=console->next;
2753                 res = 0;
2754         } else if (console_drivers) {
2755                 for (a=console_drivers->next, b=console_drivers ;
2756                      a; b=a, a=b->next) {
2757                         if (a == console) {
2758                                 b->next = a->next;
2759                                 res = 0;
2760                                 break;
2761                         }
2762                 }
2763         }
2764 
2765         if (!res && (console->flags & CON_EXTENDED))
2766                 nr_ext_console_drivers--;
2767 
2768         /*
2769          * If this isn't the last console and it has CON_CONSDEV set, we
2770          * need to set it on the next preferred console.
2771          */
2772         if (console_drivers != NULL && console->flags & CON_CONSDEV)
2773                 console_drivers->flags |= CON_CONSDEV;
2774 
2775         console->flags &= ~CON_ENABLED;
2776         console_unlock();
2777         console_sysfs_notify();
2778         return res;
2779 }
2780 EXPORT_SYMBOL(unregister_console);
2781 
2782 /*
2783  * Initialize the console device. This is called *early*, so
2784  * we can't necessarily depend on lots of kernel help here.
2785  * Just do some early initializations, and do the complex setup
2786  * later.
2787  */
2788 void __init console_init(void)
2789 {
2790         int ret;
2791         initcall_t call;
2792         initcall_entry_t *ce;
2793 
2794         /* Setup the default TTY line discipline. */
2795         n_tty_init();
2796 
2797         /*
2798          * set up the console device so that later boot sequences can
2799          * inform about problems etc..
2800          */
2801         ce = __con_initcall_start;
2802         trace_initcall_level("console");
2803         while (ce < __con_initcall_end) {
2804                 call = initcall_from_entry(ce);
2805                 trace_initcall_start(call);
2806                 ret = call();
2807                 trace_initcall_finish(call, ret);
2808                 ce++;
2809         }
2810 }
2811 
2812 /*
2813  * Some boot consoles access data that is in the init section and which will
2814  * be discarded after the initcalls have been run. To make sure that no code
2815  * will access this data, unregister the boot consoles in a late initcall.
2816  *
2817  * If for some reason, such as deferred probe or the driver being a loadable
2818  * module, the real console hasn't registered yet at this point, there will
2819  * be a brief interval in which no messages are logged to the console, which
2820  * makes it difficult to diagnose problems that occur during this time.
2821  *
2822  * To mitigate this problem somewhat, only unregister consoles whose memory
2823  * intersects with the init section. Note that all other boot consoles will
2824  * get unregistred when the real preferred console is registered.
2825  */
2826 static int __init printk_late_init(void)
2827 {
2828         struct console *con;
2829         int ret;
2830 
2831         for_each_console(con) {
2832                 if (!(con->flags & CON_BOOT))
2833                         continue;
2834 
2835                 /* Check addresses that might be used for enabled consoles. */
2836                 if (init_section_intersects(con, sizeof(*con)) ||
2837                     init_section_contains(con->write, 0) ||
2838                     init_section_contains(con->read, 0) ||
2839                     init_section_contains(con->device, 0) ||
2840                     init_section_contains(con->unblank, 0) ||
2841                     init_section_contains(con->data, 0)) {
2842                         /*
2843                          * Please, consider moving the reported consoles out
2844                          * of the init section.
2845                          */
2846                         pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n",
2847                                 con->name, con->index);
2848                         unregister_console(con);
2849                 }
2850         }
2851         ret = cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD, "printk:dead", NULL,
2852                                         console_cpu_notify);
2853         WARN_ON(ret < 0);
2854         ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "printk:online",
2855                                         console_cpu_notify, NULL);
2856         WARN_ON(ret < 0);
2857         return 0;
2858 }
2859 late_initcall(printk_late_init);
2860 
2861 #if defined CONFIG_PRINTK
2862 /*
2863  * Delayed printk version, for scheduler-internal messages:
2864  */
2865 #define PRINTK_PENDING_WAKEUP   0x01
2866 #define PRINTK_PENDING_OUTPUT   0x02
2867 
2868 static DEFINE_PER_CPU(int, printk_pending);
2869 
2870 static void wake_up_klogd_work_func(struct irq_work *irq_work)
2871 {
2872         int pending = __this_cpu_xchg(printk_pending, 0);
2873 
2874         if (pending & PRINTK_PENDING_OUTPUT) {
2875                 /* If trylock fails, someone else is doing the printing */
2876                 if (console_trylock())
2877                         console_unlock();
2878         }
2879 
2880         if (pending & PRINTK_PENDING_WAKEUP)
2881                 wake_up_interruptible(&log_wait);
2882 }
2883 
2884 static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
2885         .func = wake_up_klogd_work_func,
2886         .flags = IRQ_WORK_LAZY,
2887 };
2888 
2889 void wake_up_klogd(void)
2890 {
2891         preempt_disable();
2892         if (waitqueue_active(&log_wait)) {
2893                 this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
2894                 irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2895         }
2896         preempt_enable();
2897 }
2898 
2899 void defer_console_output(void)
2900 {
2901         preempt_disable();
2902         __this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
2903         irq_work_queue(this_cpu_ptr(&wake_up_klogd_work));
2904         preempt_enable();
2905 }
2906 
2907 int vprintk_deferred(const char *fmt, va_list args)
2908 {
2909         int r;
2910 
2911         r = vprintk_emit(0, LOGLEVEL_SCHED, NULL, 0, fmt, args);
2912         defer_console_output();
2913 
2914         return r;
2915 }
2916 
2917 int printk_deferred(const char *fmt, ...)
2918 {
2919         va_list args;
2920         int r;
2921 
2922         va_start(args, fmt);
2923         r = vprintk_deferred(fmt, args);
2924         va_end(args);
2925 
2926         return r;
2927 }
2928 
2929 /*
2930  * printk rate limiting, lifted from the networking subsystem.
2931  *
2932  * This enforces a rate limit: not more than 10 kernel messages
2933  * every 5s to make a denial-of-service attack impossible.
2934  */
2935 DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2936 
2937 int __printk_ratelimit(const char *func)
2938 {
2939         return ___ratelimit(&printk_ratelimit_state, func);
2940 }
2941 EXPORT_SYMBOL(__printk_ratelimit);
2942 
2943 /**
2944  * printk_timed_ratelimit - caller-controlled printk ratelimiting
2945  * @caller_jiffies: pointer to caller's state
2946  * @interval_msecs: minimum interval between prints
2947  *
2948  * printk_timed_ratelimit() returns true if more than @interval_msecs
2949  * milliseconds have elapsed since the last time printk_timed_ratelimit()
2950  * returned true.
2951  */
2952 bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2953                         unsigned int interval_msecs)
2954 {
2955         unsigned long elapsed = jiffies - *caller_jiffies;
2956 
2957         if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs))
2958                 return false;
2959 
2960         *caller_jiffies = jiffies;
2961         return true;
2962 }
2963 EXPORT_SYMBOL(printk_timed_ratelimit);
2964 
2965 static DEFINE_SPINLOCK(dump_list_lock);
2966 static LIST_HEAD(dump_list);
2967 
2968 /**
2969  * kmsg_dump_register - register a kernel log dumper.
2970  * @dumper: pointer to the kmsg_dumper structure
2971  *
2972  * Adds a kernel log dumper to the system. The dump callback in the
2973  * structure will be called when the kernel oopses or panics and must be
2974  * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2975  */
2976 int kmsg_dump_register(struct kmsg_dumper *dumper)
2977 {
2978         unsigned long flags;
2979         int err = -EBUSY;
2980 
2981         /* The dump callback needs to be set */
2982         if (!dumper->dump)
2983                 return -EINVAL;
2984 
2985         spin_lock_irqsave(&dump_list_lock, flags);
2986         /* Don't allow registering multiple times */
2987         if (!dumper->registered) {
2988                 dumper->registered = 1;
2989                 list_add_tail_rcu(&dumper->list, &dump_list);
2990                 err = 0;
2991         }
2992         spin_unlock_irqrestore(&dump_list_lock, flags);
2993 
2994         return err;
2995 }
2996 EXPORT_SYMBOL_GPL(kmsg_dump_register);
2997 
2998 /**
2999  * kmsg_dump_unregister - unregister a kmsg dumper.
3000  * @dumper: pointer to the kmsg_dumper structure
3001  *
3002  * Removes a dump device from the system. Returns zero on success and
3003  * %-EINVAL otherwise.
3004  */
3005 int kmsg_dump_unregister(struct kmsg_dumper *dumper)
3006 {
3007         unsigned long flags;
3008         int err = -EINVAL;
3009 
3010         spin_lock_irqsave(&dump_list_lock, flags);
3011         if (dumper->registered) {
3012                 dumper->registered = 0;
3013                 list_del_rcu(&dumper->list);
3014                 err = 0;
3015         }
3016         spin_unlock_irqrestore(&dump_list_lock, flags);
3017         synchronize_rcu();
3018 
3019         return err;
3020 }
3021 EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
3022 
3023 static bool always_kmsg_dump;
3024 module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
3025 
3026 /**
3027  * kmsg_dump - dump kernel log to kernel message dumpers.
3028  * @reason: the reason (oops, panic etc) for dumping
3029  *
3030  * Call each of the registered dumper's dump() callback, which can
3031  * retrieve the kmsg records with kmsg_dump_get_line() or
3032  * kmsg_dump_get_buffer().
3033  */
3034 void kmsg_dump(enum kmsg_dump_reason reason)
3035 {
3036         struct kmsg_dumper *dumper;
3037         unsigned long flags;
3038 
3039         if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
3040                 return;
3041 
3042         rcu_read_lock();
3043         list_for_each_entry_rcu(dumper, &dump_list, list) {
3044                 if (dumper->max_reason && reason > dumper->max_reason)
3045                         continue;
3046 
3047                 /* initialize iterator with data about the stored records */
3048                 dumper->active = true;
3049 
3050                 logbuf_lock_irqsave(flags);
3051                 dumper->cur_seq = clear_seq;
3052                 dumper->cur_idx = clear_idx;
3053                 dumper->next_seq = log_next_seq;
3054                 dumper->next_idx = log_next_idx;
3055                 logbuf_unlock_irqrestore(flags);
3056 
3057                 /* invoke dumper which will iterate over records */
3058                 dumper->dump(dumper, reason);
3059 
3060                 /* reset iterator */
3061                 dumper->active = false;
3062         }
3063         rcu_read_unlock();
3064 }
3065 
3066 /**
3067  * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
3068  * @dumper: registered kmsg dumper
3069  * @syslog: include the "<4>" prefixes
3070  * @line: buffer to copy the line to
3071  * @size: maximum size of the buffer
3072  * @len: length of line placed into buffer
3073  *
3074  * Start at the beginning of the kmsg buffer, with the oldest kmsg
3075  * record, and copy one record into the provided buffer.
3076  *
3077  * Consecutive calls will return the next available record moving
3078  * towards the end of the buffer with the youngest messages.
3079  *
3080  * A return value of FALSE indicates that there are no more records to
3081  * read.
3082  *
3083  * The function is similar to kmsg_dump_get_line(), but grabs no locks.
3084  */
3085 bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
3086                                char *line, size_t size, size_t *len)
3087 {
3088         struct printk_log *msg;
3089         size_t l = 0;
3090         bool ret = false;
3091 
3092         if (!dumper->active)
3093                 goto out;
3094 
3095         if (dumper->cur_seq < log_first_seq) {
3096                 /* messages are gone, move to first available one */
3097                 dumper->cur_seq = log_first_seq;
3098                 dumper->cur_idx = log_first_idx;
3099         }
3100 
3101         /* last entry */
3102         if (dumper->cur_seq >= log_next_seq)
3103                 goto out;
3104 
3105         msg = log_from_idx(dumper->cur_idx);
3106         l = msg_print_text(msg, syslog, line, size);
3107 
3108         dumper->cur_idx = log_next(dumper->cur_idx);
3109         dumper->cur_seq++;
3110         ret = true;
3111 out:
3112         if (len)
3113                 *len = l;
3114         return ret;
3115 }
3116 
3117 /**
3118  * kmsg_dump_get_line - retrieve one kmsg log line
3119  * @dumper: registered kmsg dumper
3120  * @syslog: include the "<4>" prefixes
3121  * @line: buffer to copy the line to
3122  * @size: maximum size of the buffer
3123  * @len: length of line placed into buffer
3124  *
3125  * Start at the beginning of the kmsg buffer, with the oldest kmsg
3126  * record, and copy one record into the provided buffer.
3127  *
3128  * Consecutive calls will return the next available record moving
3129  * towards the end of the buffer with the youngest messages.
3130  *
3131  * A return value of FALSE indicates that there are no more records to
3132  * read.
3133  */
3134 bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
3135                         char *line, size_t size, size_t *len)
3136 {
3137         unsigned long flags;
3138         bool ret;
3139 
3140         logbuf_lock_irqsave(flags);
3141         ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
3142         logbuf_unlock_irqrestore(flags);
3143 
3144         return ret;
3145 }
3146 EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
3147 
3148 /**
3149  * kmsg_dump_get_buffer - copy kmsg log lines
3150  * @dumper: registered kmsg dumper
3151  * @syslog: include the "<4>" prefixes
3152  * @buf: buffer to copy the line to
3153  * @size: maximum size of the buffer
3154  * @len: length of line placed into buffer
3155  *
3156  * Start at the end of the kmsg buffer and fill the provided buffer
3157  * with as many of the the *youngest* kmsg records that fit into it.
3158  * If the buffer is large enough, all available kmsg records will be
3159  * copied with a single call.
3160  *
3161  * Consecutive calls will fill the buffer with the next block of
3162  * available older records, not including the earlier retrieved ones.
3163  *
3164  * A return value of FALSE indicates that there are no more records to
3165  * read.
3166  */
3167 bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
3168                           char *buf, size_t size, size_t *len)
3169 {
3170         unsigned long flags;
3171         u64 seq;
3172         u32 idx;
3173         u64 next_seq;
3174         u32 next_idx;
3175         size_t l = 0;
3176         bool ret = false;
3177 
3178         if (!dumper->active)
3179                 goto out;
3180 
3181         logbuf_lock_irqsave(flags);
3182         if (dumper->cur_seq < log_first_seq) {
3183                 /* messages are gone, move to first available one */
3184                 dumper->cur_seq = log_first_seq;
3185                 dumper->cur_idx = log_first_idx;
3186         }
3187 
3188         /* last entry */
3189         if (dumper->cur_seq >= dumper->next_seq) {
3190                 logbuf_unlock_irqrestore(flags);
3191                 goto out;
3192         }
3193 
3194         /* calculate length of entire buffer */
3195         seq = dumper->cur_seq;
3196         idx = dumper->cur_idx;
3197         while (seq < dumper->next_seq) {
3198                 struct printk_log *msg = log_from_idx(idx);
3199 
3200                 l += msg_print_text(msg, true, NULL, 0);
3201                 idx = log_next(idx);
3202                 seq++;
3203         }
3204 
3205         /* move first record forward until length fits into the buffer */
3206         seq = dumper->cur_seq;
3207         idx = dumper->cur_idx;
3208         while (l > size && seq < dumper->next_seq) {
3209                 struct printk_log *msg = log_from_idx(idx);
3210 
3211                 l -= msg_print_text(msg, true, NULL, 0);
3212                 idx = log_next(idx);
3213                 seq++;
3214         }
3215 
3216         /* last message in next interation */
3217         next_seq = seq;
3218         next_idx = idx;
3219 
3220         l = 0;
3221         while (seq < dumper->next_seq) {
3222                 struct printk_log *msg = log_from_idx(idx);
3223 
3224                 l += msg_print_text(msg, syslog, buf + l, size - l);
3225                 idx = log_next(idx);
3226                 seq++;
3227         }
3228 
3229         dumper->next_seq = next_seq;
3230         dumper->next_idx = next_idx;
3231         ret = true;
3232         logbuf_unlock_irqrestore(flags);
3233 out:
3234         if (len)
3235                 *len = l;
3236         return ret;
3237 }
3238 EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
3239 
3240 /**
3241  * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
3242  * @dumper: registered kmsg dumper
3243  *
3244  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3245  * kmsg_dump_get_buffer() can be called again and used multiple
3246  * times within the same dumper.dump() callback.
3247  *
3248  * The function is similar to kmsg_dump_rewind(), but grabs no locks.
3249  */
3250 void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
3251 {
3252         dumper->cur_seq = clear_seq;
3253         dumper->cur_idx = clear_idx;
3254         dumper->next_seq = log_next_seq;
3255         dumper->next_idx = log_next_idx;
3256 }
3257 
3258 /**
3259  * kmsg_dump_rewind - reset the interator
3260  * @dumper: registered kmsg dumper
3261  *
3262  * Reset the dumper's iterator so that kmsg_dump_get_line() and
3263  * kmsg_dump_get_buffer() can be called again and used multiple
3264  * times within the same dumper.dump() callback.
3265  */
3266 void kmsg_dump_rewind(struct kmsg_dumper *dumper)
3267 {
3268         unsigned long flags;
3269 
3270         logbuf_lock_irqsave(flags);
3271         kmsg_dump_rewind_nolock(dumper);
3272         logbuf_unlock_irqrestore(flags);
3273 }
3274 EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
3275 
3276 #endif
3277 

~ [ 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