]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/printk.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * Modified to make sys_syslog() more flexible: added commands to | |
8 | * return the last 4k of kernel messages, regardless of whether | |
9 | * they've been read or not. Added option to suppress kernel printk's | |
10 | * to the console. Added hook for sending the console messages | |
11 | * elsewhere, in preparation for a serial line console (someday). | |
12 | * Ted Ts'o, 2/11/93. | |
13 | * Modified for sysctl support, 1/8/97, Chris Horn. | |
40dc5651 | 14 | * Fixed SMP synchronization, 08/08/99, Manfred Spraul |
624dffcb | 15 | * manfred@colorfullife.com |
1da177e4 | 16 | * Rewrote bits to get rid of console_lock |
e1f8e874 | 17 | * 01Mar01 Andrew Morton |
1da177e4 LT |
18 | */ |
19 | ||
dd5adbfb HZ |
20 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
21 | ||
1da177e4 LT |
22 | #include <linux/kernel.h> |
23 | #include <linux/mm.h> | |
24 | #include <linux/tty.h> | |
25 | #include <linux/tty_driver.h> | |
1da177e4 LT |
26 | #include <linux/console.h> |
27 | #include <linux/init.h> | |
bfe8df3d RD |
28 | #include <linux/jiffies.h> |
29 | #include <linux/nmi.h> | |
1da177e4 | 30 | #include <linux/module.h> |
3b9c0410 | 31 | #include <linux/moduleparam.h> |
1da177e4 LT |
32 | #include <linux/delay.h> |
33 | #include <linux/smp.h> | |
34 | #include <linux/security.h> | |
162a7e75 | 35 | #include <linux/memblock.h> |
1da177e4 | 36 | #include <linux/syscalls.h> |
692f66f2 | 37 | #include <linux/crash_core.h> |
3fff4c42 | 38 | #include <linux/ratelimit.h> |
456b565c | 39 | #include <linux/kmsg_dump.h> |
00234592 | 40 | #include <linux/syslog.h> |
034260d6 | 41 | #include <linux/cpu.h> |
fb842b00 | 42 | #include <linux/rculist.h> |
e11fea92 | 43 | #include <linux/poll.h> |
74876a98 | 44 | #include <linux/irq_work.h> |
249771b8 | 45 | #include <linux/ctype.h> |
e2e40f2c | 46 | #include <linux/uio.h> |
e6017571 | 47 | #include <linux/sched/clock.h> |
b17b0153 | 48 | #include <linux/sched/debug.h> |
68db0cf1 | 49 | #include <linux/sched/task_stack.h> |
1da177e4 | 50 | |
7c0f6ba6 | 51 | #include <linux/uaccess.h> |
40a7d9f5 | 52 | #include <asm/sections.h> |
1da177e4 | 53 | |
58eacfff | 54 | #include <trace/events/initcall.h> |
95100358 JB |
55 | #define CREATE_TRACE_POINTS |
56 | #include <trace/events/printk.h> | |
57 | ||
896fbe20 | 58 | #include "printk_ringbuffer.h" |
d197c43d | 59 | #include "console_cmdline.h" |
bbeddf52 | 60 | #include "braille.h" |
42a0bb3f | 61 | #include "internal.h" |
d197c43d | 62 | |
1da177e4 | 63 | int console_printk[4] = { |
a8fe19eb | 64 | CONSOLE_LOGLEVEL_DEFAULT, /* console_loglevel */ |
42a9dc0b | 65 | MESSAGE_LOGLEVEL_DEFAULT, /* default_message_loglevel */ |
a8fe19eb BP |
66 | CONSOLE_LOGLEVEL_MIN, /* minimum_console_loglevel */ |
67 | CONSOLE_LOGLEVEL_DEFAULT, /* default_console_loglevel */ | |
1da177e4 | 68 | }; |
a1939185 | 69 | EXPORT_SYMBOL_GPL(console_printk); |
1da177e4 | 70 | |
56e6c104 TZ |
71 | atomic_t ignore_console_lock_warning __read_mostly = ATOMIC_INIT(0); |
72 | EXPORT_SYMBOL(ignore_console_lock_warning); | |
73 | ||
1f6ab566 PK |
74 | EXPORT_TRACEPOINT_SYMBOL_GPL(console); |
75 | ||
1da177e4 | 76 | /* |
0bbfb7c2 | 77 | * Low level drivers may need that to know if they can schedule in |
1da177e4 LT |
78 | * their unblank() callback or not. So let's export it. |
79 | */ | |
80 | int oops_in_progress; | |
81 | EXPORT_SYMBOL(oops_in_progress); | |
82 | ||
83 | /* | |
4dc64682 JO |
84 | * console_mutex protects console_list updates and console->flags updates. |
85 | * The flags are synchronized only for consoles that are registered, i.e. | |
86 | * accessible via the console list. | |
87 | */ | |
88 | static DEFINE_MUTEX(console_mutex); | |
89 | ||
1da177e4 | 90 | /* |
9e70a5e1 | 91 | * console_sem protects updates to console->seq |
848a9c10 | 92 | * and also provides serialization for console printing. |
1da177e4 | 93 | */ |
48380368 | 94 | static DEFINE_SEMAPHORE(console_sem, 1); |
d9a4af56 TG |
95 | HLIST_HEAD(console_list); |
96 | EXPORT_SYMBOL_GPL(console_list); | |
6c4afa79 | 97 | DEFINE_STATIC_SRCU(console_srcu); |
a29d1cfe | 98 | |
c39ea0b9 FT |
99 | /* |
100 | * System may need to suppress printk message under certain | |
101 | * circumstances, like after kernel panic happens. | |
102 | */ | |
103 | int __read_mostly suppress_printk; | |
104 | ||
daee7797 DV |
105 | #ifdef CONFIG_LOCKDEP |
106 | static struct lockdep_map console_lock_dep_map = { | |
107 | .name = "console_lock" | |
108 | }; | |
4dc64682 JO |
109 | |
110 | void lockdep_assert_console_list_lock_held(void) | |
111 | { | |
112 | lockdep_assert_held(&console_mutex); | |
113 | } | |
114 | EXPORT_SYMBOL(lockdep_assert_console_list_lock_held); | |
daee7797 DV |
115 | #endif |
116 | ||
6c4afa79 JO |
117 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
118 | bool console_srcu_read_lock_is_held(void) | |
119 | { | |
120 | return srcu_read_lock_held(&console_srcu); | |
121 | } | |
3ef5abd9 | 122 | EXPORT_SYMBOL(console_srcu_read_lock_is_held); |
daee7797 DV |
123 | #endif |
124 | ||
750afe7b BP |
125 | enum devkmsg_log_bits { |
126 | __DEVKMSG_LOG_BIT_ON = 0, | |
127 | __DEVKMSG_LOG_BIT_OFF, | |
128 | __DEVKMSG_LOG_BIT_LOCK, | |
129 | }; | |
130 | ||
131 | enum devkmsg_log_masks { | |
132 | DEVKMSG_LOG_MASK_ON = BIT(__DEVKMSG_LOG_BIT_ON), | |
133 | DEVKMSG_LOG_MASK_OFF = BIT(__DEVKMSG_LOG_BIT_OFF), | |
134 | DEVKMSG_LOG_MASK_LOCK = BIT(__DEVKMSG_LOG_BIT_LOCK), | |
135 | }; | |
136 | ||
137 | /* Keep both the 'on' and 'off' bits clear, i.e. ratelimit by default: */ | |
138 | #define DEVKMSG_LOG_MASK_DEFAULT 0 | |
139 | ||
140 | static unsigned int __read_mostly devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT; | |
141 | ||
142 | static int __control_devkmsg(char *str) | |
143 | { | |
35c35493 CY |
144 | size_t len; |
145 | ||
750afe7b BP |
146 | if (!str) |
147 | return -EINVAL; | |
148 | ||
35c35493 CY |
149 | len = str_has_prefix(str, "on"); |
150 | if (len) { | |
750afe7b | 151 | devkmsg_log = DEVKMSG_LOG_MASK_ON; |
35c35493 CY |
152 | return len; |
153 | } | |
154 | ||
155 | len = str_has_prefix(str, "off"); | |
156 | if (len) { | |
750afe7b | 157 | devkmsg_log = DEVKMSG_LOG_MASK_OFF; |
35c35493 CY |
158 | return len; |
159 | } | |
160 | ||
161 | len = str_has_prefix(str, "ratelimit"); | |
162 | if (len) { | |
750afe7b | 163 | devkmsg_log = DEVKMSG_LOG_MASK_DEFAULT; |
35c35493 | 164 | return len; |
750afe7b | 165 | } |
35c35493 | 166 | |
750afe7b BP |
167 | return -EINVAL; |
168 | } | |
169 | ||
170 | static int __init control_devkmsg(char *str) | |
171 | { | |
b665eae7 RD |
172 | if (__control_devkmsg(str) < 0) { |
173 | pr_warn("printk.devkmsg: bad option string '%s'\n", str); | |
750afe7b | 174 | return 1; |
b665eae7 | 175 | } |
750afe7b BP |
176 | |
177 | /* | |
178 | * Set sysctl string accordingly: | |
179 | */ | |
6fd78a1a SS |
180 | if (devkmsg_log == DEVKMSG_LOG_MASK_ON) |
181 | strcpy(devkmsg_log_str, "on"); | |
182 | else if (devkmsg_log == DEVKMSG_LOG_MASK_OFF) | |
183 | strcpy(devkmsg_log_str, "off"); | |
750afe7b BP |
184 | /* else "ratelimit" which is set by default. */ |
185 | ||
186 | /* | |
187 | * Sysctl cannot change it anymore. The kernel command line setting of | |
188 | * this parameter is to force the setting to be permanent throughout the | |
189 | * runtime of the system. This is a precation measure against userspace | |
190 | * trying to be a smarta** and attempting to change it up on us. | |
191 | */ | |
192 | devkmsg_log |= DEVKMSG_LOG_MASK_LOCK; | |
193 | ||
b665eae7 | 194 | return 1; |
750afe7b BP |
195 | } |
196 | __setup("printk.devkmsg=", control_devkmsg); | |
197 | ||
198 | char devkmsg_log_str[DEVKMSG_STR_MAX_SIZE] = "ratelimit"; | |
fdcd4073 | 199 | #if defined(CONFIG_PRINTK) && defined(CONFIG_SYSCTL) |
750afe7b | 200 | int devkmsg_sysctl_set_loglvl(struct ctl_table *table, int write, |
32927393 | 201 | void *buffer, size_t *lenp, loff_t *ppos) |
750afe7b BP |
202 | { |
203 | char old_str[DEVKMSG_STR_MAX_SIZE]; | |
204 | unsigned int old; | |
205 | int err; | |
206 | ||
207 | if (write) { | |
208 | if (devkmsg_log & DEVKMSG_LOG_MASK_LOCK) | |
209 | return -EINVAL; | |
210 | ||
211 | old = devkmsg_log; | |
212 | strncpy(old_str, devkmsg_log_str, DEVKMSG_STR_MAX_SIZE); | |
213 | } | |
214 | ||
215 | err = proc_dostring(table, write, buffer, lenp, ppos); | |
216 | if (err) | |
217 | return err; | |
218 | ||
219 | if (write) { | |
220 | err = __control_devkmsg(devkmsg_log_str); | |
221 | ||
222 | /* | |
223 | * Do not accept an unknown string OR a known string with | |
224 | * trailing crap... | |
225 | */ | |
226 | if (err < 0 || (err + 1 != *lenp)) { | |
227 | ||
228 | /* ... and restore old setting. */ | |
229 | devkmsg_log = old; | |
230 | strncpy(devkmsg_log_str, old_str, DEVKMSG_STR_MAX_SIZE); | |
231 | ||
232 | return -EINVAL; | |
233 | } | |
234 | } | |
235 | ||
236 | return 0; | |
237 | } | |
fdcd4073 | 238 | #endif /* CONFIG_PRINTK && CONFIG_SYSCTL */ |
750afe7b | 239 | |
4dc64682 JO |
240 | /** |
241 | * console_list_lock - Lock the console list | |
242 | * | |
243 | * For console list or console->flags updates | |
244 | */ | |
245 | void console_list_lock(void) | |
246 | { | |
247 | /* | |
6f883675 JO |
248 | * In unregister_console() and console_force_preferred_locked(), |
249 | * synchronize_srcu() is called with the console_list_lock held. | |
250 | * Therefore it is not allowed that the console_list_lock is taken | |
251 | * with the srcu_lock held. | |
4dc64682 JO |
252 | * |
253 | * Detecting if this context is really in the read-side critical | |
254 | * section is only possible if the appropriate debug options are | |
255 | * enabled. | |
256 | */ | |
257 | WARN_ON_ONCE(debug_lockdep_rcu_enabled() && | |
258 | srcu_read_lock_held(&console_srcu)); | |
259 | ||
260 | mutex_lock(&console_mutex); | |
261 | } | |
262 | EXPORT_SYMBOL(console_list_lock); | |
263 | ||
264 | /** | |
265 | * console_list_unlock - Unlock the console list | |
266 | * | |
267 | * Counterpart to console_list_lock() | |
268 | */ | |
269 | void console_list_unlock(void) | |
270 | { | |
271 | mutex_unlock(&console_mutex); | |
272 | } | |
273 | EXPORT_SYMBOL(console_list_unlock); | |
274 | ||
6c4afa79 JO |
275 | /** |
276 | * console_srcu_read_lock - Register a new reader for the | |
277 | * SRCU-protected console list | |
278 | * | |
279 | * Use for_each_console_srcu() to iterate the console list | |
280 | * | |
281 | * Context: Any context. | |
5074ffbe | 282 | * Return: A cookie to pass to console_srcu_read_unlock(). |
6c4afa79 JO |
283 | */ |
284 | int console_srcu_read_lock(void) | |
285 | { | |
286 | return srcu_read_lock_nmisafe(&console_srcu); | |
287 | } | |
288 | EXPORT_SYMBOL(console_srcu_read_lock); | |
289 | ||
290 | /** | |
291 | * console_srcu_read_unlock - Unregister an old reader from | |
292 | * the SRCU-protected console list | |
5074ffbe | 293 | * @cookie: cookie returned from console_srcu_read_lock() |
6c4afa79 JO |
294 | * |
295 | * Counterpart to console_srcu_read_lock() | |
296 | */ | |
297 | void console_srcu_read_unlock(int cookie) | |
298 | { | |
299 | srcu_read_unlock_nmisafe(&console_srcu, cookie); | |
300 | } | |
301 | EXPORT_SYMBOL(console_srcu_read_unlock); | |
302 | ||
bd8d7cf5 JK |
303 | /* |
304 | * Helper macros to handle lockdep when locking/unlocking console_sem. We use | |
305 | * macros instead of functions so that _RET_IP_ contains useful information. | |
306 | */ | |
307 | #define down_console_sem() do { \ | |
308 | down(&console_sem);\ | |
309 | mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);\ | |
310 | } while (0) | |
311 | ||
312 | static int __down_trylock_console_sem(unsigned long ip) | |
313 | { | |
f975237b SS |
314 | int lock_failed; |
315 | unsigned long flags; | |
316 | ||
317 | /* | |
318 | * Here and in __up_console_sem() we need to be in safe mode, | |
319 | * because spindump/WARN/etc from under console ->lock will | |
320 | * deadlock in printk()->down_trylock_console_sem() otherwise. | |
321 | */ | |
322 | printk_safe_enter_irqsave(flags); | |
323 | lock_failed = down_trylock(&console_sem); | |
324 | printk_safe_exit_irqrestore(flags); | |
325 | ||
326 | if (lock_failed) | |
bd8d7cf5 JK |
327 | return 1; |
328 | mutex_acquire(&console_lock_dep_map, 0, 1, ip); | |
329 | return 0; | |
330 | } | |
331 | #define down_trylock_console_sem() __down_trylock_console_sem(_RET_IP_) | |
332 | ||
f975237b SS |
333 | static void __up_console_sem(unsigned long ip) |
334 | { | |
335 | unsigned long flags; | |
336 | ||
5facae4f | 337 | mutex_release(&console_lock_dep_map, ip); |
f975237b SS |
338 | |
339 | printk_safe_enter_irqsave(flags); | |
340 | up(&console_sem); | |
341 | printk_safe_exit_irqrestore(flags); | |
342 | } | |
343 | #define up_console_sem() __up_console_sem(_RET_IP_) | |
bd8d7cf5 | 344 | |
77498617 SB |
345 | static bool panic_in_progress(void) |
346 | { | |
347 | return unlikely(atomic_read(&panic_cpu) != PANIC_CPU_INVALID); | |
348 | } | |
349 | ||
36652d0f JO |
350 | /* Return true if a panic is in progress on the current CPU. */ |
351 | bool this_cpu_in_panic(void) | |
352 | { | |
353 | /* | |
354 | * We can use raw_smp_processor_id() here because it is impossible for | |
355 | * the task to be migrated to the panic_cpu, or away from it. If | |
356 | * panic_cpu has already been set, and we're not currently executing on | |
357 | * that CPU, then we never will be. | |
358 | */ | |
359 | return unlikely(atomic_read(&panic_cpu) == raw_smp_processor_id()); | |
360 | } | |
361 | ||
362 | /* | |
363 | * Return true if a panic is in progress on a remote CPU. | |
364 | * | |
365 | * On true, the local CPU should immediately release any printing resources | |
366 | * that may be needed by the panic CPU. | |
367 | */ | |
368 | bool other_cpu_in_panic(void) | |
369 | { | |
370 | return (panic_in_progress() && !this_cpu_in_panic()); | |
371 | } | |
372 | ||
007eeab7 PM |
373 | /* |
374 | * This is used for debugging the mess that is the VT code by | |
375 | * keeping track if we have the console semaphore held. It's | |
376 | * definitely not the perfect debug tool (we don't know if _WE_ | |
377 | * hold it and are racing, but it helps tracking those weird code | |
378 | * paths in the console code where we end up in places I want | |
379 | * locked without the console semaphore held). | |
380 | */ | |
9e70a5e1 | 381 | static int console_locked; |
1da177e4 | 382 | |
1da177e4 LT |
383 | /* |
384 | * Array of consoles built from command line options (console=) | |
385 | */ | |
1da177e4 LT |
386 | |
387 | #define MAX_CMDLINECONSOLES 8 | |
388 | ||
389 | static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES]; | |
d197c43d | 390 | |
1da177e4 | 391 | static int preferred_console = -1; |
9e124fe1 MA |
392 | int console_set_on_cmdline; |
393 | EXPORT_SYMBOL(console_set_on_cmdline); | |
1da177e4 LT |
394 | |
395 | /* Flag: console code may call schedule() */ | |
396 | static int console_may_schedule; | |
397 | ||
cca10d58 SS |
398 | enum con_msg_format_flags { |
399 | MSG_FORMAT_DEFAULT = 0, | |
400 | MSG_FORMAT_SYSLOG = (1 << 0), | |
401 | }; | |
402 | ||
403 | static int console_msg_format = MSG_FORMAT_DEFAULT; | |
404 | ||
7ff9554b | 405 | /* |
896fbe20 | 406 | * The printk log buffer consists of a sequenced collection of records, each |
74caba7f JO |
407 | * containing variable length message text. Every record also contains its |
408 | * own meta-data (@info). | |
7ff9554b | 409 | * |
896fbe20 JO |
410 | * Every record meta-data carries the timestamp in microseconds, as well as |
411 | * the standard userspace syslog level and syslog facility. The usual kernel | |
412 | * messages use LOG_KERN; userspace-injected messages always carry a matching | |
413 | * syslog facility, by default LOG_USER. The origin of every message can be | |
414 | * reliably determined that way. | |
7ff9554b | 415 | * |
896fbe20 JO |
416 | * The human readable log message of a record is available in @text, the |
417 | * length of the message text in @text_len. The stored message is not | |
418 | * terminated. | |
7ff9554b | 419 | * |
896fbe20 | 420 | * Optionally, a record can carry a dictionary of properties (key/value |
74caba7f | 421 | * pairs), to provide userspace with a machine-readable message context. |
e11fea92 KS |
422 | * |
423 | * Examples for well-defined, commonly used property names are: | |
424 | * DEVICE=b12:8 device identifier | |
425 | * b12:8 block dev_t | |
426 | * c127:3 char dev_t | |
427 | * n8 netdev ifindex | |
428 | * +sound:card0 subsystem:devname | |
429 | * SUBSYSTEM=pci driver-core subsystem name | |
430 | * | |
74caba7f JO |
431 | * Valid characters in property names are [a-zA-Z0-9.-_]. Property names |
432 | * and values are terminated by a '\0' character. | |
e11fea92 | 433 | * |
896fbe20 | 434 | * Example of record values: |
74caba7f JO |
435 | * record.text_buf = "it's a line" (unterminated) |
436 | * record.info.seq = 56 | |
437 | * record.info.ts_nsec = 36863 | |
438 | * record.info.text_len = 11 | |
439 | * record.info.facility = 0 (LOG_KERN) | |
440 | * record.info.flags = 0 | |
441 | * record.info.level = 3 (LOG_ERR) | |
442 | * record.info.caller_id = 299 (task 299) | |
443 | * record.info.dev_info.subsystem = "pci" (terminated) | |
444 | * record.info.dev_info.device = "+pci:0000:00:01.0" (terminated) | |
896fbe20 JO |
445 | * |
446 | * The 'struct printk_info' buffer must never be directly exported to | |
e11fea92 KS |
447 | * userspace, it is a kernel-private implementation detail that might |
448 | * need to be changed in the future, when the requirements change. | |
449 | * | |
450 | * /dev/kmsg exports the structured data in the following line format: | |
b389645f AO |
451 | * "<level>,<sequnum>,<timestamp>,<contflag>[,additional_values, ... ];<message text>\n" |
452 | * | |
453 | * Users of the export format should ignore possible additional values | |
454 | * separated by ',', and find the message after the ';' character. | |
e11fea92 KS |
455 | * |
456 | * The optional key/value pairs are attached as continuation lines starting | |
457 | * with a space character and terminated by a newline. All possible | |
458 | * non-prinatable characters are escaped in the "\xff" notation. | |
7ff9554b KS |
459 | */ |
460 | ||
636babdc | 461 | /* syslog_lock protects syslog_* variables and write access to clear_seq. */ |
b371cbb5 | 462 | static DEFINE_MUTEX(syslog_lock); |
636babdc | 463 | |
96efedf1 | 464 | #ifdef CONFIG_PRINTK |
98a04655 JO |
465 | /* |
466 | * During panic, heavy printk by other CPUs can delay the | |
467 | * panic and risk deadlock on console resources. | |
468 | */ | |
469 | static int __read_mostly suppress_panic_printk; | |
470 | ||
dc72c32e | 471 | DECLARE_WAIT_QUEUE_HEAD(log_wait); |
636babdc | 472 | /* All 3 protected by @syslog_lock. */ |
7f3a781d KS |
473 | /* the next printk record to read by syslog(READ) or /proc/kmsg */ |
474 | static u64 syslog_seq; | |
eb02dac9 | 475 | static size_t syslog_partial; |
e80c1a9d | 476 | static bool syslog_time; |
7ff9554b | 477 | |
7d7a23a9 JO |
478 | struct latched_seq { |
479 | seqcount_latch_t latch; | |
480 | u64 val[2]; | |
481 | }; | |
482 | ||
483 | /* | |
484 | * The next printk record to read after the last 'clear' command. There are | |
485 | * two copies (updated with seqcount_latch) so that reads can locklessly | |
636babdc | 486 | * access a valid value. Writers are synchronized by @syslog_lock. |
7d7a23a9 JO |
487 | */ |
488 | static struct latched_seq clear_seq = { | |
489 | .latch = SEQCNT_LATCH_ZERO(clear_seq.latch), | |
490 | .val[0] = 0, | |
491 | .val[1] = 0, | |
492 | }; | |
7ff9554b | 493 | |
3824657c MK |
494 | #define LOG_LEVEL(v) ((v) & 0x07) |
495 | #define LOG_FACILITY(v) ((v) >> 3 & 0xff) | |
496 | ||
7f3a781d | 497 | /* record buffer */ |
896fbe20 | 498 | #define LOG_ALIGN __alignof__(unsigned long) |
7f3a781d | 499 | #define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT) |
e6fe3e5b | 500 | #define LOG_BUF_LEN_MAX (u32)(1 << 31) |
f8450fca | 501 | static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN); |
7f3a781d KS |
502 | static char *log_buf = __log_buf; |
503 | static u32 log_buf_len = __LOG_BUF_LEN; | |
504 | ||
896fbe20 JO |
505 | /* |
506 | * Define the average message size. This only affects the number of | |
507 | * descriptors that will be available. Underestimating is better than | |
508 | * overestimating (too many available descriptors is better than not enough). | |
896fbe20 JO |
509 | */ |
510 | #define PRB_AVGBITS 5 /* 32 character average length */ | |
511 | ||
512 | #if CONFIG_LOG_BUF_SHIFT <= PRB_AVGBITS | |
513 | #error CONFIG_LOG_BUF_SHIFT value too small. | |
514 | #endif | |
515 | _DEFINE_PRINTKRB(printk_rb_static, CONFIG_LOG_BUF_SHIFT - PRB_AVGBITS, | |
f35efc78 | 516 | PRB_AVGBITS, &__log_buf[0]); |
896fbe20 JO |
517 | |
518 | static struct printk_ringbuffer printk_rb_dynamic; | |
519 | ||
ad56ebd1 | 520 | struct printk_ringbuffer *prb = &printk_rb_static; |
896fbe20 | 521 | |
ab6f762f SS |
522 | /* |
523 | * We cannot access per-CPU data (e.g. per-CPU flush irq_work) before | |
524 | * per_cpu_areas are initialised. This variable is set to true when | |
525 | * it's safe to access per-CPU data. | |
526 | */ | |
78ba392c | 527 | static bool __printk_percpu_data_ready __ro_after_init; |
ab6f762f SS |
528 | |
529 | bool printk_percpu_data_ready(void) | |
530 | { | |
531 | return __printk_percpu_data_ready; | |
532 | } | |
533 | ||
636babdc | 534 | /* Must be called under syslog_lock. */ |
7d7a23a9 JO |
535 | static void latched_seq_write(struct latched_seq *ls, u64 val) |
536 | { | |
537 | raw_write_seqcount_latch(&ls->latch); | |
538 | ls->val[0] = val; | |
539 | raw_write_seqcount_latch(&ls->latch); | |
540 | ls->val[1] = val; | |
541 | } | |
542 | ||
543 | /* Can be called from any context. */ | |
544 | static u64 latched_seq_read_nolock(struct latched_seq *ls) | |
545 | { | |
546 | unsigned int seq; | |
547 | unsigned int idx; | |
548 | u64 val; | |
549 | ||
550 | do { | |
551 | seq = raw_read_seqcount_latch(&ls->latch); | |
552 | idx = seq & 0x1; | |
553 | val = ls->val[idx]; | |
d16317de | 554 | } while (raw_read_seqcount_latch_retry(&ls->latch, seq)); |
7d7a23a9 JO |
555 | |
556 | return val; | |
557 | } | |
558 | ||
14c4000a VH |
559 | /* Return log buffer address */ |
560 | char *log_buf_addr_get(void) | |
561 | { | |
562 | return log_buf; | |
563 | } | |
564 | ||
565 | /* Return log buffer size */ | |
566 | u32 log_buf_len_get(void) | |
567 | { | |
568 | return log_buf_len; | |
569 | } | |
570 | ||
55bd53a4 PM |
571 | /* |
572 | * Define how much of the log buffer we could take at maximum. The value | |
573 | * must be greater than two. Note that only half of the buffer is available | |
574 | * when the index points to the middle. | |
575 | */ | |
576 | #define MAX_LOG_TAKE_PART 4 | |
577 | static const char trunc_msg[] = "<truncated>"; | |
578 | ||
896fbe20 | 579 | static void truncate_msg(u16 *text_len, u16 *trunc_msg_len) |
55bd53a4 PM |
580 | { |
581 | /* | |
582 | * The message should not take the whole buffer. Otherwise, it might | |
583 | * get removed too soon. | |
584 | */ | |
585 | u32 max_text_len = log_buf_len / MAX_LOG_TAKE_PART; | |
896fbe20 | 586 | |
55bd53a4 PM |
587 | if (*text_len > max_text_len) |
588 | *text_len = max_text_len; | |
896fbe20 JO |
589 | |
590 | /* enable the warning message (if there is room) */ | |
55bd53a4 | 591 | *trunc_msg_len = strlen(trunc_msg); |
896fbe20 JO |
592 | if (*text_len >= *trunc_msg_len) |
593 | *text_len -= *trunc_msg_len; | |
594 | else | |
595 | *trunc_msg_len = 0; | |
55bd53a4 PM |
596 | } |
597 | ||
e99aa461 | 598 | int dmesg_restrict = IS_ENABLED(CONFIG_SECURITY_DMESG_RESTRICT); |
637241a9 KC |
599 | |
600 | static int syslog_action_restricted(int type) | |
601 | { | |
602 | if (dmesg_restrict) | |
603 | return 1; | |
604 | /* | |
605 | * Unless restricted, we allow "read all" and "get buffer size" | |
606 | * for everybody. | |
607 | */ | |
608 | return type != SYSLOG_ACTION_READ_ALL && | |
609 | type != SYSLOG_ACTION_SIZE_BUFFER; | |
610 | } | |
611 | ||
c71b02e4 | 612 | static int check_syslog_permissions(int type, int source) |
637241a9 KC |
613 | { |
614 | /* | |
615 | * If this is from /proc/kmsg and we've already opened it, then we've | |
616 | * already done the capabilities checks at open time. | |
617 | */ | |
3ea4331c | 618 | if (source == SYSLOG_FROM_PROC && type != SYSLOG_ACTION_OPEN) |
d194e5d6 | 619 | goto ok; |
637241a9 KC |
620 | |
621 | if (syslog_action_restricted(type)) { | |
622 | if (capable(CAP_SYSLOG)) | |
d194e5d6 | 623 | goto ok; |
637241a9 KC |
624 | /* |
625 | * For historical reasons, accept CAP_SYS_ADMIN too, with | |
626 | * a warning. | |
627 | */ | |
628 | if (capable(CAP_SYS_ADMIN)) { | |
629 | pr_warn_once("%s (%d): Attempt to access syslog with " | |
630 | "CAP_SYS_ADMIN but no CAP_SYSLOG " | |
631 | "(deprecated).\n", | |
632 | current->comm, task_pid_nr(current)); | |
d194e5d6 | 633 | goto ok; |
637241a9 KC |
634 | } |
635 | return -EPERM; | |
636 | } | |
d194e5d6 | 637 | ok: |
637241a9 KC |
638 | return security_syslog(type); |
639 | } | |
640 | ||
d43ff430 TH |
641 | static void append_char(char **pp, char *e, char c) |
642 | { | |
643 | if (*pp < e) | |
644 | *(*pp)++ = c; | |
645 | } | |
637241a9 | 646 | |
896fbe20 JO |
647 | static ssize_t info_print_ext_header(char *buf, size_t size, |
648 | struct printk_info *info) | |
0a295e67 | 649 | { |
896fbe20 | 650 | u64 ts_usec = info->ts_nsec; |
15ff2069 TH |
651 | char caller[20]; |
652 | #ifdef CONFIG_PRINTK_CALLER | |
896fbe20 | 653 | u32 id = info->caller_id; |
15ff2069 TH |
654 | |
655 | snprintf(caller, sizeof(caller), ",caller=%c%u", | |
656 | id & 0x80000000 ? 'C' : 'T', id & ~0x80000000); | |
657 | #else | |
658 | caller[0] = '\0'; | |
659 | #endif | |
0a295e67 TH |
660 | |
661 | do_div(ts_usec, 1000); | |
662 | ||
15ff2069 | 663 | return scnprintf(buf, size, "%u,%llu,%llu,%c%s;", |
896fbe20 JO |
664 | (info->facility << 3) | info->level, info->seq, |
665 | ts_usec, info->flags & LOG_CONT ? 'c' : '-', caller); | |
0a295e67 TH |
666 | } |
667 | ||
74caba7f JO |
668 | static ssize_t msg_add_ext_text(char *buf, size_t size, |
669 | const char *text, size_t text_len, | |
670 | unsigned char endc) | |
0a295e67 TH |
671 | { |
672 | char *p = buf, *e = buf + size; | |
673 | size_t i; | |
674 | ||
675 | /* escape non-printable characters */ | |
676 | for (i = 0; i < text_len; i++) { | |
677 | unsigned char c = text[i]; | |
678 | ||
679 | if (c < ' ' || c >= 127 || c == '\\') | |
680 | p += scnprintf(p, e - p, "\\x%02x", c); | |
681 | else | |
682 | append_char(&p, e, c); | |
683 | } | |
74caba7f | 684 | append_char(&p, e, endc); |
0a295e67 | 685 | |
74caba7f JO |
686 | return p - buf; |
687 | } | |
0a295e67 | 688 | |
74caba7f JO |
689 | static ssize_t msg_add_dict_text(char *buf, size_t size, |
690 | const char *key, const char *val) | |
691 | { | |
692 | size_t val_len = strlen(val); | |
693 | ssize_t len; | |
0a295e67 | 694 | |
74caba7f JO |
695 | if (!val_len) |
696 | return 0; | |
0a295e67 | 697 | |
74caba7f JO |
698 | len = msg_add_ext_text(buf, size, "", 0, ' '); /* dict prefix */ |
699 | len += msg_add_ext_text(buf + len, size - len, key, strlen(key), '='); | |
700 | len += msg_add_ext_text(buf + len, size - len, val, val_len, '\n'); | |
0a295e67 | 701 | |
74caba7f JO |
702 | return len; |
703 | } | |
0a295e67 | 704 | |
74caba7f JO |
705 | static ssize_t msg_print_ext_body(char *buf, size_t size, |
706 | char *text, size_t text_len, | |
707 | struct dev_printk_info *dev_info) | |
708 | { | |
709 | ssize_t len; | |
0a295e67 | 710 | |
74caba7f JO |
711 | len = msg_add_ext_text(buf, size, text, text_len, '\n'); |
712 | ||
713 | if (!dev_info) | |
714 | goto out; | |
715 | ||
716 | len += msg_add_dict_text(buf + len, size - len, "SUBSYSTEM", | |
717 | dev_info->subsystem); | |
718 | len += msg_add_dict_text(buf + len, size - len, "DEVICE", | |
719 | dev_info->device); | |
720 | out: | |
721 | return len; | |
0a295e67 TH |
722 | } |
723 | ||
e11fea92 KS |
724 | /* /dev/kmsg - userspace message inject/listen interface */ |
725 | struct devkmsg_user { | |
35b2b163 | 726 | atomic64_t seq; |
750afe7b | 727 | struct ratelimit_state rs; |
e11fea92 | 728 | struct mutex lock; |
ea308da1 | 729 | struct printk_buffers pbufs; |
e11fea92 KS |
730 | }; |
731 | ||
9adcfaff TH |
732 | static __printf(3, 4) __cold |
733 | int devkmsg_emit(int facility, int level, const char *fmt, ...) | |
734 | { | |
735 | va_list args; | |
736 | int r; | |
737 | ||
738 | va_start(args, fmt); | |
74caba7f | 739 | r = vprintk_emit(facility, level, NULL, fmt, args); |
9adcfaff TH |
740 | va_end(args); |
741 | ||
742 | return r; | |
743 | } | |
744 | ||
849f3127 | 745 | static ssize_t devkmsg_write(struct kiocb *iocb, struct iov_iter *from) |
e11fea92 KS |
746 | { |
747 | char *buf, *line; | |
e11fea92 KS |
748 | int level = default_message_loglevel; |
749 | int facility = 1; /* LOG_USER */ | |
750afe7b BP |
750 | struct file *file = iocb->ki_filp; |
751 | struct devkmsg_user *user = file->private_data; | |
66ee59af | 752 | size_t len = iov_iter_count(from); |
e11fea92 KS |
753 | ssize_t ret = len; |
754 | ||
bee43904 | 755 | if (len > PRINTKRB_RECORD_MAX) |
e11fea92 | 756 | return -EINVAL; |
750afe7b BP |
757 | |
758 | /* Ignore when user logging is disabled. */ | |
759 | if (devkmsg_log & DEVKMSG_LOG_MASK_OFF) | |
760 | return len; | |
761 | ||
762 | /* Ratelimit when not explicitly enabled. */ | |
763 | if (!(devkmsg_log & DEVKMSG_LOG_MASK_ON)) { | |
764 | if (!___ratelimit(&user->rs, current->comm)) | |
765 | return ret; | |
766 | } | |
767 | ||
e11fea92 KS |
768 | buf = kmalloc(len+1, GFP_KERNEL); |
769 | if (buf == NULL) | |
770 | return -ENOMEM; | |
771 | ||
849f3127 | 772 | buf[len] = '\0'; |
cbbd26b8 | 773 | if (!copy_from_iter_full(buf, len, from)) { |
849f3127 AV |
774 | kfree(buf); |
775 | return -EFAULT; | |
e11fea92 KS |
776 | } |
777 | ||
778 | /* | |
779 | * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace | |
780 | * the decimal value represents 32bit, the lower 3 bit are the log | |
781 | * level, the rest are the log facility. | |
782 | * | |
783 | * If no prefix or no userspace facility is specified, we | |
784 | * enforce LOG_USER, to be able to reliably distinguish | |
785 | * kernel-generated messages from userspace-injected ones. | |
786 | */ | |
787 | line = buf; | |
788 | if (line[0] == '<') { | |
789 | char *endp = NULL; | |
3824657c | 790 | unsigned int u; |
e11fea92 | 791 | |
3824657c | 792 | u = simple_strtoul(line + 1, &endp, 10); |
e11fea92 | 793 | if (endp && endp[0] == '>') { |
3824657c MK |
794 | level = LOG_LEVEL(u); |
795 | if (LOG_FACILITY(u) != 0) | |
796 | facility = LOG_FACILITY(u); | |
e11fea92 | 797 | endp++; |
e11fea92 KS |
798 | line = endp; |
799 | } | |
800 | } | |
e11fea92 | 801 | |
9adcfaff | 802 | devkmsg_emit(facility, level, "%s", line); |
e11fea92 KS |
803 | kfree(buf); |
804 | return ret; | |
805 | } | |
806 | ||
807 | static ssize_t devkmsg_read(struct file *file, char __user *buf, | |
808 | size_t count, loff_t *ppos) | |
809 | { | |
810 | struct devkmsg_user *user = file->private_data; | |
ea308da1 JO |
811 | char *outbuf = &user->pbufs.outbuf[0]; |
812 | struct printk_message pmsg = { | |
813 | .pbufs = &user->pbufs, | |
814 | }; | |
e11fea92 KS |
815 | ssize_t ret; |
816 | ||
4a77a5a0 YL |
817 | ret = mutex_lock_interruptible(&user->lock); |
818 | if (ret) | |
819 | return ret; | |
de6fcbdb | 820 | |
ea308da1 | 821 | if (!printk_get_next_message(&pmsg, atomic64_read(&user->seq), true, false)) { |
e11fea92 KS |
822 | if (file->f_flags & O_NONBLOCK) { |
823 | ret = -EAGAIN; | |
e11fea92 KS |
824 | goto out; |
825 | } | |
826 | ||
1f5d7830 JO |
827 | /* |
828 | * Guarantee this task is visible on the waitqueue before | |
829 | * checking the wake condition. | |
830 | * | |
831 | * The full memory barrier within set_current_state() of | |
832 | * prepare_to_wait_event() pairs with the full memory barrier | |
833 | * within wq_has_sleeper(). | |
834 | * | |
5341b93d | 835 | * This pairs with __wake_up_klogd:A. |
1f5d7830 | 836 | */ |
e11fea92 | 837 | ret = wait_event_interruptible(log_wait, |
ea308da1 JO |
838 | printk_get_next_message(&pmsg, atomic64_read(&user->seq), true, |
839 | false)); /* LMM(devkmsg_read:A) */ | |
e11fea92 KS |
840 | if (ret) |
841 | goto out; | |
e11fea92 KS |
842 | } |
843 | ||
ea308da1 | 844 | if (pmsg.dropped) { |
e11fea92 | 845 | /* our last seen message is gone, return error and reset */ |
ea308da1 | 846 | atomic64_set(&user->seq, pmsg.seq); |
e11fea92 | 847 | ret = -EPIPE; |
e11fea92 KS |
848 | goto out; |
849 | } | |
850 | ||
ea308da1 | 851 | atomic64_set(&user->seq, pmsg.seq + 1); |
e11fea92 | 852 | |
ea308da1 | 853 | if (pmsg.outbuf_len > count) { |
e11fea92 KS |
854 | ret = -EINVAL; |
855 | goto out; | |
856 | } | |
857 | ||
ea308da1 | 858 | if (copy_to_user(buf, outbuf, pmsg.outbuf_len)) { |
e11fea92 KS |
859 | ret = -EFAULT; |
860 | goto out; | |
861 | } | |
ea308da1 | 862 | ret = pmsg.outbuf_len; |
e11fea92 KS |
863 | out: |
864 | mutex_unlock(&user->lock); | |
865 | return ret; | |
866 | } | |
867 | ||
bc885f1a BM |
868 | /* |
869 | * Be careful when modifying this function!!! | |
870 | * | |
871 | * Only few operations are supported because the device works only with the | |
872 | * entire variable length messages (records). Non-standard values are | |
873 | * returned in the other cases and has been this way for quite some time. | |
874 | * User space applications might depend on this behavior. | |
875 | */ | |
e11fea92 KS |
876 | static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence) |
877 | { | |
878 | struct devkmsg_user *user = file->private_data; | |
879 | loff_t ret = 0; | |
880 | ||
e11fea92 KS |
881 | if (offset) |
882 | return -ESPIPE; | |
883 | ||
e11fea92 KS |
884 | switch (whence) { |
885 | case SEEK_SET: | |
886 | /* the first record */ | |
35b2b163 | 887 | atomic64_set(&user->seq, prb_first_valid_seq(prb)); |
e11fea92 KS |
888 | break; |
889 | case SEEK_DATA: | |
890 | /* | |
891 | * The first record after the last SYSLOG_ACTION_CLEAR, | |
892 | * like issued by 'dmesg -c'. Reading /dev/kmsg itself | |
893 | * changes no global state, and does not clear anything. | |
894 | */ | |
35b2b163 | 895 | atomic64_set(&user->seq, latched_seq_read_nolock(&clear_seq)); |
e11fea92 KS |
896 | break; |
897 | case SEEK_END: | |
898 | /* after the last record */ | |
35b2b163 | 899 | atomic64_set(&user->seq, prb_next_seq(prb)); |
e11fea92 KS |
900 | break; |
901 | default: | |
902 | ret = -EINVAL; | |
903 | } | |
e11fea92 KS |
904 | return ret; |
905 | } | |
906 | ||
9dd95748 | 907 | static __poll_t devkmsg_poll(struct file *file, poll_table *wait) |
e11fea92 KS |
908 | { |
909 | struct devkmsg_user *user = file->private_data; | |
13791c80 | 910 | struct printk_info info; |
9dd95748 | 911 | __poll_t ret = 0; |
e11fea92 | 912 | |
e11fea92 KS |
913 | poll_wait(file, &log_wait, wait); |
914 | ||
35b2b163 | 915 | if (prb_read_valid_info(prb, atomic64_read(&user->seq), &info, NULL)) { |
e11fea92 | 916 | /* return error when data has vanished underneath us */ |
35b2b163 | 917 | if (info.seq != atomic64_read(&user->seq)) |
a9a08845 | 918 | ret = EPOLLIN|EPOLLRDNORM|EPOLLERR|EPOLLPRI; |
0a285317 | 919 | else |
a9a08845 | 920 | ret = EPOLLIN|EPOLLRDNORM; |
e11fea92 | 921 | } |
e11fea92 KS |
922 | |
923 | return ret; | |
924 | } | |
925 | ||
926 | static int devkmsg_open(struct inode *inode, struct file *file) | |
927 | { | |
928 | struct devkmsg_user *user; | |
929 | int err; | |
930 | ||
750afe7b BP |
931 | if (devkmsg_log & DEVKMSG_LOG_MASK_OFF) |
932 | return -EPERM; | |
e11fea92 | 933 | |
750afe7b BP |
934 | /* write-only does not need any file context */ |
935 | if ((file->f_flags & O_ACCMODE) != O_WRONLY) { | |
936 | err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL, | |
937 | SYSLOG_FROM_READER); | |
938 | if (err) | |
939 | return err; | |
940 | } | |
e11fea92 | 941 | |
9980c425 | 942 | user = kvmalloc(sizeof(struct devkmsg_user), GFP_KERNEL); |
e11fea92 KS |
943 | if (!user) |
944 | return -ENOMEM; | |
945 | ||
750afe7b BP |
946 | ratelimit_default_init(&user->rs); |
947 | ratelimit_set_flags(&user->rs, RATELIMIT_MSG_ON_RELEASE); | |
948 | ||
e11fea92 KS |
949 | mutex_init(&user->lock); |
950 | ||
35b2b163 | 951 | atomic64_set(&user->seq, prb_first_valid_seq(prb)); |
e11fea92 KS |
952 | |
953 | file->private_data = user; | |
954 | return 0; | |
955 | } | |
956 | ||
957 | static int devkmsg_release(struct inode *inode, struct file *file) | |
958 | { | |
959 | struct devkmsg_user *user = file->private_data; | |
960 | ||
750afe7b BP |
961 | ratelimit_state_exit(&user->rs); |
962 | ||
e11fea92 | 963 | mutex_destroy(&user->lock); |
9980c425 | 964 | kvfree(user); |
e11fea92 KS |
965 | return 0; |
966 | } | |
967 | ||
968 | const struct file_operations kmsg_fops = { | |
969 | .open = devkmsg_open, | |
970 | .read = devkmsg_read, | |
849f3127 | 971 | .write_iter = devkmsg_write, |
e11fea92 KS |
972 | .llseek = devkmsg_llseek, |
973 | .poll = devkmsg_poll, | |
974 | .release = devkmsg_release, | |
975 | }; | |
976 | ||
692f66f2 | 977 | #ifdef CONFIG_CRASH_CORE |
04d491ab | 978 | /* |
4c1ace64 | 979 | * This appends the listed symbols to /proc/vmcore |
04d491ab | 980 | * |
4c1ace64 | 981 | * /proc/vmcore is used by various utilities, like crash and makedumpfile to |
04d491ab NH |
982 | * obtain access to symbols that are otherwise very difficult to locate. These |
983 | * symbols are specifically used so that utilities can access and extract the | |
984 | * dmesg log from a vmcore file after a crash. | |
985 | */ | |
692f66f2 | 986 | void log_buf_vmcoreinfo_setup(void) |
04d491ab | 987 | { |
74caba7f JO |
988 | struct dev_printk_info *dev_info = NULL; |
989 | ||
896fbe20 JO |
990 | VMCOREINFO_SYMBOL(prb); |
991 | VMCOREINFO_SYMBOL(printk_rb_static); | |
992 | VMCOREINFO_SYMBOL(clear_seq); | |
993 | ||
6791457a | 994 | /* |
896fbe20 | 995 | * Export struct size and field offsets. User space tools can |
6791457a VG |
996 | * parse it and detect any changes to structure down the line. |
997 | */ | |
896fbe20 JO |
998 | |
999 | VMCOREINFO_STRUCT_SIZE(printk_ringbuffer); | |
1000 | VMCOREINFO_OFFSET(printk_ringbuffer, desc_ring); | |
1001 | VMCOREINFO_OFFSET(printk_ringbuffer, text_data_ring); | |
896fbe20 JO |
1002 | VMCOREINFO_OFFSET(printk_ringbuffer, fail); |
1003 | ||
1004 | VMCOREINFO_STRUCT_SIZE(prb_desc_ring); | |
1005 | VMCOREINFO_OFFSET(prb_desc_ring, count_bits); | |
1006 | VMCOREINFO_OFFSET(prb_desc_ring, descs); | |
cfe2790b | 1007 | VMCOREINFO_OFFSET(prb_desc_ring, infos); |
896fbe20 JO |
1008 | VMCOREINFO_OFFSET(prb_desc_ring, head_id); |
1009 | VMCOREINFO_OFFSET(prb_desc_ring, tail_id); | |
1010 | ||
1011 | VMCOREINFO_STRUCT_SIZE(prb_desc); | |
896fbe20 JO |
1012 | VMCOREINFO_OFFSET(prb_desc, state_var); |
1013 | VMCOREINFO_OFFSET(prb_desc, text_blk_lpos); | |
896fbe20 JO |
1014 | |
1015 | VMCOREINFO_STRUCT_SIZE(prb_data_blk_lpos); | |
1016 | VMCOREINFO_OFFSET(prb_data_blk_lpos, begin); | |
1017 | VMCOREINFO_OFFSET(prb_data_blk_lpos, next); | |
1018 | ||
1019 | VMCOREINFO_STRUCT_SIZE(printk_info); | |
1020 | VMCOREINFO_OFFSET(printk_info, seq); | |
1021 | VMCOREINFO_OFFSET(printk_info, ts_nsec); | |
1022 | VMCOREINFO_OFFSET(printk_info, text_len); | |
896fbe20 | 1023 | VMCOREINFO_OFFSET(printk_info, caller_id); |
74caba7f JO |
1024 | VMCOREINFO_OFFSET(printk_info, dev_info); |
1025 | ||
1026 | VMCOREINFO_STRUCT_SIZE(dev_printk_info); | |
1027 | VMCOREINFO_OFFSET(dev_printk_info, subsystem); | |
1028 | VMCOREINFO_LENGTH(printk_info_subsystem, sizeof(dev_info->subsystem)); | |
1029 | VMCOREINFO_OFFSET(dev_printk_info, device); | |
1030 | VMCOREINFO_LENGTH(printk_info_device, sizeof(dev_info->device)); | |
896fbe20 JO |
1031 | |
1032 | VMCOREINFO_STRUCT_SIZE(prb_data_ring); | |
1033 | VMCOREINFO_OFFSET(prb_data_ring, size_bits); | |
1034 | VMCOREINFO_OFFSET(prb_data_ring, data); | |
1035 | VMCOREINFO_OFFSET(prb_data_ring, head_lpos); | |
1036 | VMCOREINFO_OFFSET(prb_data_ring, tail_lpos); | |
1037 | ||
1038 | VMCOREINFO_SIZE(atomic_long_t); | |
1039 | VMCOREINFO_TYPE_OFFSET(atomic_long_t, counter); | |
7d7a23a9 JO |
1040 | |
1041 | VMCOREINFO_STRUCT_SIZE(latched_seq); | |
1042 | VMCOREINFO_OFFSET(latched_seq, val); | |
04d491ab NH |
1043 | } |
1044 | #endif | |
1045 | ||
162a7e75 MT |
1046 | /* requested log_buf_len from kernel cmdline */ |
1047 | static unsigned long __initdata new_log_buf_len; | |
1048 | ||
c0a318a3 | 1049 | /* we practice scaling the ring buffer by powers of 2 */ |
e6fe3e5b | 1050 | static void __init log_buf_len_update(u64 size) |
1da177e4 | 1051 | { |
e6fe3e5b HZ |
1052 | if (size > (u64)LOG_BUF_LEN_MAX) { |
1053 | size = (u64)LOG_BUF_LEN_MAX; | |
1054 | pr_err("log_buf over 2G is not supported.\n"); | |
1055 | } | |
1056 | ||
1da177e4 LT |
1057 | if (size) |
1058 | size = roundup_pow_of_two(size); | |
162a7e75 | 1059 | if (size > log_buf_len) |
e6fe3e5b | 1060 | new_log_buf_len = (unsigned long)size; |
c0a318a3 LR |
1061 | } |
1062 | ||
1063 | /* save requested log_buf_len since it's too early to process it */ | |
1064 | static int __init log_buf_len_setup(char *str) | |
1065 | { | |
e6fe3e5b | 1066 | u64 size; |
277fcdb2 HZ |
1067 | |
1068 | if (!str) | |
1069 | return -EINVAL; | |
1070 | ||
1071 | size = memparse(str, &str); | |
c0a318a3 LR |
1072 | |
1073 | log_buf_len_update(size); | |
162a7e75 MT |
1074 | |
1075 | return 0; | |
1da177e4 | 1076 | } |
162a7e75 MT |
1077 | early_param("log_buf_len", log_buf_len_setup); |
1078 | ||
2240a31d GU |
1079 | #ifdef CONFIG_SMP |
1080 | #define __LOG_CPU_MAX_BUF_LEN (1 << CONFIG_LOG_CPU_MAX_BUF_SHIFT) | |
1081 | ||
23b2899f LR |
1082 | static void __init log_buf_add_cpu(void) |
1083 | { | |
1084 | unsigned int cpu_extra; | |
1085 | ||
1086 | /* | |
1087 | * archs should set up cpu_possible_bits properly with | |
1088 | * set_cpu_possible() after setup_arch() but just in | |
1089 | * case lets ensure this is valid. | |
1090 | */ | |
1091 | if (num_possible_cpus() == 1) | |
1092 | return; | |
1093 | ||
1094 | cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN; | |
1095 | ||
1096 | /* by default this will only continue through for large > 64 CPUs */ | |
1097 | if (cpu_extra <= __LOG_BUF_LEN / 2) | |
1098 | return; | |
1099 | ||
1100 | pr_info("log_buf_len individual max cpu contribution: %d bytes\n", | |
1101 | __LOG_CPU_MAX_BUF_LEN); | |
1102 | pr_info("log_buf_len total cpu_extra contributions: %d bytes\n", | |
1103 | cpu_extra); | |
1104 | pr_info("log_buf_len min size: %d bytes\n", __LOG_BUF_LEN); | |
1105 | ||
1106 | log_buf_len_update(cpu_extra + __LOG_BUF_LEN); | |
1107 | } | |
2240a31d GU |
1108 | #else /* !CONFIG_SMP */ |
1109 | static inline void log_buf_add_cpu(void) {} | |
1110 | #endif /* CONFIG_SMP */ | |
23b2899f | 1111 | |
ab6f762f SS |
1112 | static void __init set_percpu_data_ready(void) |
1113 | { | |
ab6f762f SS |
1114 | __printk_percpu_data_ready = true; |
1115 | } | |
1116 | ||
896fbe20 JO |
1117 | static unsigned int __init add_to_rb(struct printk_ringbuffer *rb, |
1118 | struct printk_record *r) | |
1119 | { | |
1120 | struct prb_reserved_entry e; | |
1121 | struct printk_record dest_r; | |
1122 | ||
f35efc78 | 1123 | prb_rec_init_wr(&dest_r, r->info->text_len); |
896fbe20 JO |
1124 | |
1125 | if (!prb_reserve(&e, rb, &dest_r)) | |
1126 | return 0; | |
1127 | ||
cc5c7041 JO |
1128 | memcpy(&dest_r.text_buf[0], &r->text_buf[0], r->info->text_len); |
1129 | dest_r.info->text_len = r->info->text_len; | |
896fbe20 JO |
1130 | dest_r.info->facility = r->info->facility; |
1131 | dest_r.info->level = r->info->level; | |
1132 | dest_r.info->flags = r->info->flags; | |
1133 | dest_r.info->ts_nsec = r->info->ts_nsec; | |
1134 | dest_r.info->caller_id = r->info->caller_id; | |
74caba7f | 1135 | memcpy(&dest_r.info->dev_info, &r->info->dev_info, sizeof(dest_r.info->dev_info)); |
896fbe20 | 1136 | |
f5f022e5 | 1137 | prb_final_commit(&e); |
896fbe20 JO |
1138 | |
1139 | return prb_record_text_space(&e); | |
1140 | } | |
1141 | ||
b0975c47 | 1142 | static char setup_text_buf[PRINTKRB_RECORD_MAX] __initdata; |
896fbe20 | 1143 | |
162a7e75 MT |
1144 | void __init setup_log_buf(int early) |
1145 | { | |
cfe2790b | 1146 | struct printk_info *new_infos; |
896fbe20 JO |
1147 | unsigned int new_descs_count; |
1148 | struct prb_desc *new_descs; | |
1149 | struct printk_info info; | |
1150 | struct printk_record r; | |
93d102f0 | 1151 | unsigned int text_size; |
896fbe20 | 1152 | size_t new_descs_size; |
cfe2790b | 1153 | size_t new_infos_size; |
162a7e75 | 1154 | unsigned long flags; |
162a7e75 | 1155 | char *new_log_buf; |
d2130e82 | 1156 | unsigned int free; |
896fbe20 | 1157 | u64 seq; |
162a7e75 | 1158 | |
ab6f762f SS |
1159 | /* |
1160 | * Some archs call setup_log_buf() multiple times - first is very | |
1161 | * early, e.g. from setup_arch(), and second - when percpu_areas | |
1162 | * are initialised. | |
1163 | */ | |
1164 | if (!early) | |
1165 | set_percpu_data_ready(); | |
1166 | ||
23b2899f LR |
1167 | if (log_buf != __log_buf) |
1168 | return; | |
1169 | ||
1170 | if (!early && !new_log_buf_len) | |
1171 | log_buf_add_cpu(); | |
1172 | ||
162a7e75 MT |
1173 | if (!new_log_buf_len) |
1174 | return; | |
1da177e4 | 1175 | |
896fbe20 JO |
1176 | new_descs_count = new_log_buf_len >> PRB_AVGBITS; |
1177 | if (new_descs_count == 0) { | |
1178 | pr_err("new_log_buf_len: %lu too small\n", new_log_buf_len); | |
1179 | return; | |
1180 | } | |
1181 | ||
26fb3dae | 1182 | new_log_buf = memblock_alloc(new_log_buf_len, LOG_ALIGN); |
162a7e75 | 1183 | if (unlikely(!new_log_buf)) { |
896fbe20 JO |
1184 | pr_err("log_buf_len: %lu text bytes not available\n", |
1185 | new_log_buf_len); | |
162a7e75 MT |
1186 | return; |
1187 | } | |
1188 | ||
896fbe20 JO |
1189 | new_descs_size = new_descs_count * sizeof(struct prb_desc); |
1190 | new_descs = memblock_alloc(new_descs_size, LOG_ALIGN); | |
1191 | if (unlikely(!new_descs)) { | |
1192 | pr_err("log_buf_len: %zu desc bytes not available\n", | |
1193 | new_descs_size); | |
f35efc78 | 1194 | goto err_free_log_buf; |
cfe2790b JO |
1195 | } |
1196 | ||
1197 | new_infos_size = new_descs_count * sizeof(struct printk_info); | |
1198 | new_infos = memblock_alloc(new_infos_size, LOG_ALIGN); | |
1199 | if (unlikely(!new_infos)) { | |
1200 | pr_err("log_buf_len: %zu info bytes not available\n", | |
1201 | new_infos_size); | |
1202 | goto err_free_descs; | |
896fbe20 JO |
1203 | } |
1204 | ||
f35efc78 | 1205 | prb_rec_init_rd(&r, &info, &setup_text_buf[0], sizeof(setup_text_buf)); |
896fbe20 JO |
1206 | |
1207 | prb_init(&printk_rb_dynamic, | |
1208 | new_log_buf, ilog2(new_log_buf_len), | |
cfe2790b JO |
1209 | new_descs, ilog2(new_descs_count), |
1210 | new_infos); | |
896fbe20 | 1211 | |
93d102f0 | 1212 | local_irq_save(flags); |
896fbe20 | 1213 | |
162a7e75 MT |
1214 | log_buf_len = new_log_buf_len; |
1215 | log_buf = new_log_buf; | |
1216 | new_log_buf_len = 0; | |
896fbe20 JO |
1217 | |
1218 | free = __LOG_BUF_LEN; | |
93d102f0 JO |
1219 | prb_for_each_record(0, &printk_rb_static, seq, &r) { |
1220 | text_size = add_to_rb(&printk_rb_dynamic, &r); | |
1221 | if (text_size > free) | |
1222 | free = 0; | |
1223 | else | |
1224 | free -= text_size; | |
1225 | } | |
896fbe20 | 1226 | |
896fbe20 JO |
1227 | prb = &printk_rb_dynamic; |
1228 | ||
93d102f0 JO |
1229 | local_irq_restore(flags); |
1230 | ||
1231 | /* | |
1232 | * Copy any remaining messages that might have appeared from | |
1233 | * NMI context after copying but before switching to the | |
1234 | * dynamic buffer. | |
1235 | */ | |
1236 | prb_for_each_record(seq, &printk_rb_static, seq, &r) { | |
1237 | text_size = add_to_rb(&printk_rb_dynamic, &r); | |
1238 | if (text_size > free) | |
1239 | free = 0; | |
1240 | else | |
1241 | free -= text_size; | |
1242 | } | |
162a7e75 | 1243 | |
896fbe20 JO |
1244 | if (seq != prb_next_seq(&printk_rb_static)) { |
1245 | pr_err("dropped %llu messages\n", | |
1246 | prb_next_seq(&printk_rb_static) - seq); | |
1247 | } | |
1248 | ||
e6fe3e5b HZ |
1249 | pr_info("log_buf_len: %u bytes\n", log_buf_len); |
1250 | pr_info("early log buf free: %u(%u%%)\n", | |
162a7e75 | 1251 | free, (free * 100) / __LOG_BUF_LEN); |
cfe2790b JO |
1252 | return; |
1253 | ||
1254 | err_free_descs: | |
4421cca0 | 1255 | memblock_free(new_descs, new_descs_size); |
cfe2790b | 1256 | err_free_log_buf: |
4421cca0 | 1257 | memblock_free(new_log_buf, new_log_buf_len); |
162a7e75 | 1258 | } |
1da177e4 | 1259 | |
2fa72c8f AC |
1260 | static bool __read_mostly ignore_loglevel; |
1261 | ||
1262 | static int __init ignore_loglevel_setup(char *str) | |
1263 | { | |
d25d9fec | 1264 | ignore_loglevel = true; |
27083bac | 1265 | pr_info("debug: ignoring loglevel setting.\n"); |
2fa72c8f AC |
1266 | |
1267 | return 0; | |
1268 | } | |
1269 | ||
1270 | early_param("ignore_loglevel", ignore_loglevel_setup); | |
1271 | module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR); | |
205bd3d2 JP |
1272 | MODULE_PARM_DESC(ignore_loglevel, |
1273 | "ignore loglevel setting (prints all kernel messages to the console)"); | |
2fa72c8f | 1274 | |
cf775444 SS |
1275 | static bool suppress_message_printing(int level) |
1276 | { | |
1277 | return (level >= console_loglevel && !ignore_loglevel); | |
1278 | } | |
1279 | ||
bfe8df3d RD |
1280 | #ifdef CONFIG_BOOT_PRINTK_DELAY |
1281 | ||
674dff65 | 1282 | static int boot_delay; /* msecs delay after each printk during bootup */ |
3a3b6ed2 | 1283 | static unsigned long long loops_per_msec; /* based on boot_delay */ |
bfe8df3d RD |
1284 | |
1285 | static int __init boot_delay_setup(char *str) | |
1286 | { | |
1287 | unsigned long lpj; | |
bfe8df3d RD |
1288 | |
1289 | lpj = preset_lpj ? preset_lpj : 1000000; /* some guess */ | |
1290 | loops_per_msec = (unsigned long long)lpj / 1000 * HZ; | |
1291 | ||
1292 | get_option(&str, &boot_delay); | |
1293 | if (boot_delay > 10 * 1000) | |
1294 | boot_delay = 0; | |
1295 | ||
3a3b6ed2 DY |
1296 | pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, " |
1297 | "HZ: %d, loops_per_msec: %llu\n", | |
1298 | boot_delay, preset_lpj, lpj, HZ, loops_per_msec); | |
29e9d225 | 1299 | return 0; |
bfe8df3d | 1300 | } |
29e9d225 | 1301 | early_param("boot_delay", boot_delay_setup); |
bfe8df3d | 1302 | |
2fa72c8f | 1303 | static void boot_delay_msec(int level) |
bfe8df3d RD |
1304 | { |
1305 | unsigned long long k; | |
1306 | unsigned long timeout; | |
1307 | ||
ff48cd26 | 1308 | if ((boot_delay == 0 || system_state >= SYSTEM_RUNNING) |
cf775444 | 1309 | || suppress_message_printing(level)) { |
bfe8df3d | 1310 | return; |
2fa72c8f | 1311 | } |
bfe8df3d | 1312 | |
3a3b6ed2 | 1313 | k = (unsigned long long)loops_per_msec * boot_delay; |
bfe8df3d RD |
1314 | |
1315 | timeout = jiffies + msecs_to_jiffies(boot_delay); | |
1316 | while (k) { | |
1317 | k--; | |
1318 | cpu_relax(); | |
1319 | /* | |
1320 | * use (volatile) jiffies to prevent | |
1321 | * compiler reduction; loop termination via jiffies | |
1322 | * is secondary and may or may not happen. | |
1323 | */ | |
1324 | if (time_after(jiffies, timeout)) | |
1325 | break; | |
1326 | touch_nmi_watchdog(); | |
1327 | } | |
1328 | } | |
1329 | #else | |
2fa72c8f | 1330 | static inline void boot_delay_msec(int level) |
bfe8df3d RD |
1331 | { |
1332 | } | |
1333 | #endif | |
1334 | ||
e99aa461 | 1335 | static bool printk_time = IS_ENABLED(CONFIG_PRINTK_TIME); |
7ff9554b KS |
1336 | module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR); |
1337 | ||
07c17732 | 1338 | static size_t print_syslog(unsigned int level, char *buf) |
084681d1 | 1339 | { |
07c17732 TH |
1340 | return sprintf(buf, "<%u>", level); |
1341 | } | |
35dac27c | 1342 | |
084681d1 KS |
1343 | static size_t print_time(u64 ts, char *buf) |
1344 | { | |
e80c1a9d | 1345 | unsigned long rem_nsec = do_div(ts, 1000000000); |
084681d1 | 1346 | |
15ff2069 | 1347 | return sprintf(buf, "[%5lu.%06lu]", |
084681d1 KS |
1348 | (unsigned long)ts, rem_nsec / 1000); |
1349 | } | |
1350 | ||
15ff2069 TH |
1351 | #ifdef CONFIG_PRINTK_CALLER |
1352 | static size_t print_caller(u32 id, char *buf) | |
1353 | { | |
1354 | char caller[12]; | |
1355 | ||
1356 | snprintf(caller, sizeof(caller), "%c%u", | |
1357 | id & 0x80000000 ? 'C' : 'T', id & ~0x80000000); | |
1358 | return sprintf(buf, "[%6s]", caller); | |
1359 | } | |
1360 | #else | |
1361 | #define print_caller(id, buf) 0 | |
1362 | #endif | |
1363 | ||
896fbe20 JO |
1364 | static size_t info_print_prefix(const struct printk_info *info, bool syslog, |
1365 | bool time, char *buf) | |
649e6ee3 | 1366 | { |
3ce9a7c0 | 1367 | size_t len = 0; |
649e6ee3 | 1368 | |
07c17732 | 1369 | if (syslog) |
896fbe20 | 1370 | len = print_syslog((info->facility << 3) | info->level, buf); |
15ff2069 | 1371 | |
e80c1a9d | 1372 | if (time) |
896fbe20 | 1373 | len += print_time(info->ts_nsec, buf + len); |
15ff2069 | 1374 | |
896fbe20 | 1375 | len += print_caller(info->caller_id, buf + len); |
15ff2069 TH |
1376 | |
1377 | if (IS_ENABLED(CONFIG_PRINTK_CALLER) || time) { | |
1378 | buf[len++] = ' '; | |
1379 | buf[len] = '\0'; | |
1380 | } | |
1381 | ||
3ce9a7c0 | 1382 | return len; |
649e6ee3 KS |
1383 | } |
1384 | ||
896fbe20 JO |
1385 | /* |
1386 | * Prepare the record for printing. The text is shifted within the given | |
1387 | * buffer to avoid a need for another one. The following operations are | |
1388 | * done: | |
1389 | * | |
1390 | * - Add prefix for each line. | |
f0e386ee | 1391 | * - Drop truncated lines that no longer fit into the buffer. |
896fbe20 | 1392 | * - Add the trailing newline that has been removed in vprintk_store(). |
f0e386ee JO |
1393 | * - Add a string terminator. |
1394 | * | |
1395 | * Since the produced string is always terminated, the maximum possible | |
1396 | * return value is @r->text_buf_size - 1; | |
896fbe20 JO |
1397 | * |
1398 | * Return: The length of the updated/prepared text, including the added | |
f0e386ee JO |
1399 | * prefixes and the newline. The terminator is not counted. The dropped |
1400 | * line(s) are not counted. | |
896fbe20 JO |
1401 | */ |
1402 | static size_t record_print_text(struct printk_record *r, bool syslog, | |
1403 | bool time) | |
7ff9554b | 1404 | { |
896fbe20 JO |
1405 | size_t text_len = r->info->text_len; |
1406 | size_t buf_size = r->text_buf_size; | |
1407 | char *text = r->text_buf; | |
b0975c47 | 1408 | char prefix[PRINTK_PREFIX_MAX]; |
896fbe20 JO |
1409 | bool truncated = false; |
1410 | size_t prefix_len; | |
1411 | size_t line_len; | |
1412 | size_t len = 0; | |
1413 | char *next; | |
3ce9a7c0 | 1414 | |
59f8bcca JO |
1415 | /* |
1416 | * If the message was truncated because the buffer was not large | |
1417 | * enough, treat the available text as if it were the full text. | |
1418 | */ | |
1419 | if (text_len > buf_size) | |
1420 | text_len = buf_size; | |
3ce9a7c0 | 1421 | |
896fbe20 | 1422 | prefix_len = info_print_prefix(r->info, syslog, time, prefix); |
3ce9a7c0 | 1423 | |
896fbe20 JO |
1424 | /* |
1425 | * @text_len: bytes of unprocessed text | |
1426 | * @line_len: bytes of current line _without_ newline | |
1427 | * @text: pointer to beginning of current line | |
1428 | * @len: number of bytes prepared in r->text_buf | |
1429 | */ | |
1430 | for (;;) { | |
1431 | next = memchr(text, '\n', text_len); | |
3ce9a7c0 | 1432 | if (next) { |
896fbe20 | 1433 | line_len = next - text; |
3ce9a7c0 | 1434 | } else { |
896fbe20 JO |
1435 | /* Drop truncated line(s). */ |
1436 | if (truncated) | |
1437 | break; | |
1438 | line_len = text_len; | |
3ce9a7c0 | 1439 | } |
7ff9554b | 1440 | |
896fbe20 JO |
1441 | /* |
1442 | * Truncate the text if there is not enough space to add the | |
f0e386ee | 1443 | * prefix and a trailing newline and a terminator. |
896fbe20 | 1444 | */ |
f0e386ee | 1445 | if (len + prefix_len + text_len + 1 + 1 > buf_size) { |
896fbe20 | 1446 | /* Drop even the current line if no space. */ |
f0e386ee | 1447 | if (len + prefix_len + line_len + 1 + 1 > buf_size) |
3ce9a7c0 | 1448 | break; |
7ff9554b | 1449 | |
f0e386ee | 1450 | text_len = buf_size - len - prefix_len - 1 - 1; |
896fbe20 | 1451 | truncated = true; |
3ce9a7c0 | 1452 | } |
7ff9554b | 1453 | |
896fbe20 JO |
1454 | memmove(text + prefix_len, text, text_len); |
1455 | memcpy(text, prefix, prefix_len); | |
1456 | ||
f0e386ee JO |
1457 | /* |
1458 | * Increment the prepared length to include the text and | |
1459 | * prefix that were just moved+copied. Also increment for the | |
1460 | * newline at the end of this line. If this is the last line, | |
1461 | * there is no newline, but it will be added immediately below. | |
1462 | */ | |
896fbe20 | 1463 | len += prefix_len + line_len + 1; |
896fbe20 JO |
1464 | if (text_len == line_len) { |
1465 | /* | |
f0e386ee JO |
1466 | * This is the last line. Add the trailing newline |
1467 | * removed in vprintk_store(). | |
896fbe20 JO |
1468 | */ |
1469 | text[prefix_len + line_len] = '\n'; | |
1470 | break; | |
1471 | } | |
1472 | ||
1473 | /* | |
1474 | * Advance beyond the added prefix and the related line with | |
1475 | * its newline. | |
1476 | */ | |
1477 | text += prefix_len + line_len + 1; | |
1478 | ||
1479 | /* | |
1480 | * The remaining text has only decreased by the line with its | |
1481 | * newline. | |
1482 | * | |
1483 | * Note that @text_len can become zero. It happens when @text | |
1484 | * ended with a newline (either due to truncation or the | |
1485 | * original string ending with "\n\n"). The loop is correctly | |
1486 | * repeated and (if not truncated) an empty line with a prefix | |
1487 | * will be prepared. | |
1488 | */ | |
1489 | text_len -= line_len + 1; | |
1490 | } | |
7ff9554b | 1491 | |
f0e386ee JO |
1492 | /* |
1493 | * If a buffer was provided, it will be terminated. Space for the | |
1494 | * string terminator is guaranteed to be available. The terminator is | |
1495 | * not counted in the return value. | |
1496 | */ | |
1497 | if (buf_size > 0) | |
08d60e59 | 1498 | r->text_buf[len] = 0; |
f0e386ee | 1499 | |
7ff9554b KS |
1500 | return len; |
1501 | } | |
1502 | ||
896fbe20 JO |
1503 | static size_t get_record_print_text_size(struct printk_info *info, |
1504 | unsigned int line_count, | |
1505 | bool syslog, bool time) | |
1506 | { | |
b0975c47 | 1507 | char prefix[PRINTK_PREFIX_MAX]; |
896fbe20 JO |
1508 | size_t prefix_len; |
1509 | ||
1510 | prefix_len = info_print_prefix(info, syslog, time, prefix); | |
1511 | ||
1512 | /* | |
1513 | * Each line will be preceded with a prefix. The intermediate | |
1514 | * newlines are already within the text, but a final trailing | |
1515 | * newline will be added. | |
1516 | */ | |
1517 | return ((prefix_len * line_count) + info->text_len + 1); | |
1518 | } | |
1519 | ||
4260e0e5 JO |
1520 | /* |
1521 | * Beginning with @start_seq, find the first record where it and all following | |
1522 | * records up to (but not including) @max_seq fit into @size. | |
1523 | * | |
1524 | * @max_seq is simply an upper bound and does not need to exist. If the caller | |
1525 | * does not require an upper bound, -1 can be used for @max_seq. | |
1526 | */ | |
1527 | static u64 find_first_fitting_seq(u64 start_seq, u64 max_seq, size_t size, | |
1528 | bool syslog, bool time) | |
1529 | { | |
1530 | struct printk_info info; | |
1531 | unsigned int line_count; | |
1532 | size_t len = 0; | |
1533 | u64 seq; | |
1534 | ||
1535 | /* Determine the size of the records up to @max_seq. */ | |
1536 | prb_for_each_info(start_seq, prb, seq, &info, &line_count) { | |
1537 | if (info.seq >= max_seq) | |
1538 | break; | |
1539 | len += get_record_print_text_size(&info, line_count, syslog, time); | |
1540 | } | |
1541 | ||
1542 | /* | |
1543 | * Adjust the upper bound for the next loop to avoid subtracting | |
1544 | * lengths that were never added. | |
1545 | */ | |
1546 | if (seq < max_seq) | |
1547 | max_seq = seq; | |
1548 | ||
1549 | /* | |
1550 | * Move first record forward until length fits into the buffer. Ignore | |
1551 | * newest messages that were not counted in the above cycle. Messages | |
1552 | * might appear and get lost in the meantime. This is a best effort | |
1553 | * that prevents an infinite loop that could occur with a retry. | |
1554 | */ | |
1555 | prb_for_each_info(start_seq, prb, seq, &info, &line_count) { | |
1556 | if (len <= size || info.seq >= max_seq) | |
1557 | break; | |
1558 | len -= get_record_print_text_size(&info, line_count, syslog, time); | |
1559 | } | |
1560 | ||
1561 | return seq; | |
1562 | } | |
1563 | ||
8d909b23 | 1564 | /* The caller is responsible for making sure @size is greater than 0. */ |
7ff9554b KS |
1565 | static int syslog_print(char __user *buf, int size) |
1566 | { | |
896fbe20 JO |
1567 | struct printk_info info; |
1568 | struct printk_record r; | |
7ff9554b | 1569 | char *text; |
116e90b2 | 1570 | int len = 0; |
8d909b23 | 1571 | u64 seq; |
7ff9554b | 1572 | |
b0975c47 | 1573 | text = kmalloc(PRINTK_MESSAGE_MAX, GFP_KERNEL); |
7ff9554b KS |
1574 | if (!text) |
1575 | return -ENOMEM; | |
1576 | ||
b0975c47 | 1577 | prb_rec_init_rd(&r, &info, text, PRINTK_MESSAGE_MAX); |
896fbe20 | 1578 | |
8d909b23 JO |
1579 | mutex_lock(&syslog_lock); |
1580 | ||
1581 | /* | |
1582 | * Wait for the @syslog_seq record to be available. @syslog_seq may | |
1583 | * change while waiting. | |
1584 | */ | |
1585 | do { | |
1586 | seq = syslog_seq; | |
1587 | ||
1588 | mutex_unlock(&syslog_lock); | |
1f5d7830 JO |
1589 | /* |
1590 | * Guarantee this task is visible on the waitqueue before | |
1591 | * checking the wake condition. | |
1592 | * | |
1593 | * The full memory barrier within set_current_state() of | |
1594 | * prepare_to_wait_event() pairs with the full memory barrier | |
1595 | * within wq_has_sleeper(). | |
1596 | * | |
5341b93d | 1597 | * This pairs with __wake_up_klogd:A. |
1f5d7830 JO |
1598 | */ |
1599 | len = wait_event_interruptible(log_wait, | |
1600 | prb_read_valid(prb, seq, NULL)); /* LMM(syslog_print:A) */ | |
8d909b23 JO |
1601 | mutex_lock(&syslog_lock); |
1602 | ||
1603 | if (len) | |
1604 | goto out; | |
1605 | } while (syslog_seq != seq); | |
1606 | ||
1607 | /* | |
1608 | * Copy records that fit into the buffer. The above cycle makes sure | |
1609 | * that the first record is always available. | |
1610 | */ | |
1611 | do { | |
116e90b2 | 1612 | size_t n; |
eb02dac9 | 1613 | size_t skip; |
8d909b23 | 1614 | int err; |
116e90b2 | 1615 | |
8d909b23 | 1616 | if (!prb_read_valid(prb, syslog_seq, &r)) |
116e90b2 | 1617 | break; |
8d909b23 | 1618 | |
896fbe20 JO |
1619 | if (r.info->seq != syslog_seq) { |
1620 | /* message is gone, move to next valid one */ | |
1621 | syslog_seq = r.info->seq; | |
1622 | syslog_partial = 0; | |
1623 | } | |
eb02dac9 | 1624 | |
e80c1a9d TH |
1625 | /* |
1626 | * To keep reading/counting partial line consistent, | |
1627 | * use printk_time value as of the beginning of a line. | |
1628 | */ | |
1629 | if (!syslog_partial) | |
1630 | syslog_time = printk_time; | |
1631 | ||
eb02dac9 | 1632 | skip = syslog_partial; |
896fbe20 | 1633 | n = record_print_text(&r, true, syslog_time); |
eb02dac9 KS |
1634 | if (n - syslog_partial <= size) { |
1635 | /* message fits into buffer, move forward */ | |
896fbe20 | 1636 | syslog_seq = r.info->seq + 1; |
eb02dac9 KS |
1637 | n -= syslog_partial; |
1638 | syslog_partial = 0; | |
1639 | } else if (!len){ | |
1640 | /* partial read(), remember position */ | |
1641 | n = size; | |
1642 | syslog_partial += n; | |
116e90b2 JB |
1643 | } else |
1644 | n = 0; | |
116e90b2 JB |
1645 | |
1646 | if (!n) | |
1647 | break; | |
1648 | ||
8d909b23 JO |
1649 | mutex_unlock(&syslog_lock); |
1650 | err = copy_to_user(buf, text + skip, n); | |
1651 | mutex_lock(&syslog_lock); | |
1652 | ||
1653 | if (err) { | |
116e90b2 JB |
1654 | if (!len) |
1655 | len = -EFAULT; | |
1656 | break; | |
1657 | } | |
eb02dac9 KS |
1658 | |
1659 | len += n; | |
1660 | size -= n; | |
1661 | buf += n; | |
8d909b23 JO |
1662 | } while (size); |
1663 | out: | |
1664 | mutex_unlock(&syslog_lock); | |
7ff9554b KS |
1665 | kfree(text); |
1666 | return len; | |
1667 | } | |
1668 | ||
1669 | static int syslog_print_all(char __user *buf, int size, bool clear) | |
1670 | { | |
896fbe20 | 1671 | struct printk_info info; |
896fbe20 | 1672 | struct printk_record r; |
7ff9554b KS |
1673 | char *text; |
1674 | int len = 0; | |
63842c21 | 1675 | u64 seq; |
e80c1a9d | 1676 | bool time; |
63842c21 | 1677 | |
b0975c47 | 1678 | text = kmalloc(PRINTK_MESSAGE_MAX, GFP_KERNEL); |
7ff9554b KS |
1679 | if (!text) |
1680 | return -ENOMEM; | |
1681 | ||
e80c1a9d | 1682 | time = printk_time; |
63842c21 NG |
1683 | /* |
1684 | * Find first record that fits, including all following records, | |
1685 | * into the user-provided buffer for this dump. | |
1686 | */ | |
7d7a23a9 JO |
1687 | seq = find_first_fitting_seq(latched_seq_read_nolock(&clear_seq), -1, |
1688 | size, true, time); | |
7ff9554b | 1689 | |
b0975c47 | 1690 | prb_rec_init_rd(&r, &info, text, PRINTK_MESSAGE_MAX); |
7ff9554b | 1691 | |
896fbe20 JO |
1692 | prb_for_each_record(seq, prb, seq, &r) { |
1693 | int textlen; | |
7ff9554b | 1694 | |
896fbe20 JO |
1695 | textlen = record_print_text(&r, true, time); |
1696 | ||
1697 | if (len + textlen > size) { | |
1698 | seq--; | |
1699 | break; | |
1700 | } | |
7ff9554b | 1701 | |
63842c21 NG |
1702 | if (copy_to_user(buf + len, text, textlen)) |
1703 | len = -EFAULT; | |
1704 | else | |
1705 | len += textlen; | |
7ff9554b | 1706 | |
896fbe20 JO |
1707 | if (len < 0) |
1708 | break; | |
7ff9554b KS |
1709 | } |
1710 | ||
636babdc | 1711 | if (clear) { |
b371cbb5 | 1712 | mutex_lock(&syslog_lock); |
7d7a23a9 | 1713 | latched_seq_write(&clear_seq, seq); |
b371cbb5 | 1714 | mutex_unlock(&syslog_lock); |
636babdc | 1715 | } |
7ff9554b KS |
1716 | |
1717 | kfree(text); | |
1718 | return len; | |
1719 | } | |
1720 | ||
8599dc7d PM |
1721 | static void syslog_clear(void) |
1722 | { | |
b371cbb5 | 1723 | mutex_lock(&syslog_lock); |
7d7a23a9 | 1724 | latched_seq_write(&clear_seq, prb_next_seq(prb)); |
b371cbb5 | 1725 | mutex_unlock(&syslog_lock); |
636babdc JO |
1726 | } |
1727 | ||
3ea4331c | 1728 | int do_syslog(int type, char __user *buf, int len, int source) |
1da177e4 | 1729 | { |
13791c80 | 1730 | struct printk_info info; |
7ff9554b | 1731 | bool clear = false; |
a39d4a85 | 1732 | static int saved_console_loglevel = LOGLEVEL_DEFAULT; |
ee24aebf | 1733 | int error; |
1da177e4 | 1734 | |
3ea4331c | 1735 | error = check_syslog_permissions(type, source); |
ee24aebf | 1736 | if (error) |
077a1cc0 | 1737 | return error; |
12b3052c | 1738 | |
1da177e4 | 1739 | switch (type) { |
d78ca3cd | 1740 | case SYSLOG_ACTION_CLOSE: /* Close log */ |
1da177e4 | 1741 | break; |
d78ca3cd | 1742 | case SYSLOG_ACTION_OPEN: /* Open log */ |
1da177e4 | 1743 | break; |
d78ca3cd | 1744 | case SYSLOG_ACTION_READ: /* Read from log */ |
1da177e4 | 1745 | if (!buf || len < 0) |
077a1cc0 | 1746 | return -EINVAL; |
1da177e4 | 1747 | if (!len) |
077a1cc0 | 1748 | return 0; |
96d4f267 | 1749 | if (!access_ok(buf, len)) |
077a1cc0 | 1750 | return -EFAULT; |
7ff9554b | 1751 | error = syslog_print(buf, len); |
1da177e4 | 1752 | break; |
d78ca3cd KC |
1753 | /* Read/clear last kernel messages */ |
1754 | case SYSLOG_ACTION_READ_CLEAR: | |
7ff9554b | 1755 | clear = true; |
4e797e6e | 1756 | fallthrough; |
d78ca3cd KC |
1757 | /* Read last kernel messages */ |
1758 | case SYSLOG_ACTION_READ_ALL: | |
1da177e4 | 1759 | if (!buf || len < 0) |
077a1cc0 | 1760 | return -EINVAL; |
1da177e4 | 1761 | if (!len) |
077a1cc0 | 1762 | return 0; |
96d4f267 | 1763 | if (!access_ok(buf, len)) |
077a1cc0 | 1764 | return -EFAULT; |
7ff9554b | 1765 | error = syslog_print_all(buf, len, clear); |
1da177e4 | 1766 | break; |
d78ca3cd KC |
1767 | /* Clear ring buffer */ |
1768 | case SYSLOG_ACTION_CLEAR: | |
8599dc7d | 1769 | syslog_clear(); |
4661e356 | 1770 | break; |
d78ca3cd KC |
1771 | /* Disable logging to console */ |
1772 | case SYSLOG_ACTION_CONSOLE_OFF: | |
a39d4a85 | 1773 | if (saved_console_loglevel == LOGLEVEL_DEFAULT) |
1aaad49e | 1774 | saved_console_loglevel = console_loglevel; |
1da177e4 LT |
1775 | console_loglevel = minimum_console_loglevel; |
1776 | break; | |
d78ca3cd KC |
1777 | /* Enable logging to console */ |
1778 | case SYSLOG_ACTION_CONSOLE_ON: | |
a39d4a85 | 1779 | if (saved_console_loglevel != LOGLEVEL_DEFAULT) { |
1aaad49e | 1780 | console_loglevel = saved_console_loglevel; |
a39d4a85 | 1781 | saved_console_loglevel = LOGLEVEL_DEFAULT; |
1aaad49e | 1782 | } |
1da177e4 | 1783 | break; |
d78ca3cd KC |
1784 | /* Set level of messages printed to console */ |
1785 | case SYSLOG_ACTION_CONSOLE_LEVEL: | |
1da177e4 | 1786 | if (len < 1 || len > 8) |
077a1cc0 | 1787 | return -EINVAL; |
1da177e4 LT |
1788 | if (len < minimum_console_loglevel) |
1789 | len = minimum_console_loglevel; | |
1790 | console_loglevel = len; | |
1aaad49e | 1791 | /* Implicitly re-enable logging to console */ |
a39d4a85 | 1792 | saved_console_loglevel = LOGLEVEL_DEFAULT; |
1da177e4 | 1793 | break; |
d78ca3cd KC |
1794 | /* Number of chars in the log buffer */ |
1795 | case SYSLOG_ACTION_SIZE_UNREAD: | |
b371cbb5 | 1796 | mutex_lock(&syslog_lock); |
13791c80 JO |
1797 | if (!prb_read_valid_info(prb, syslog_seq, &info, NULL)) { |
1798 | /* No unread messages. */ | |
b371cbb5 | 1799 | mutex_unlock(&syslog_lock); |
13791c80 JO |
1800 | return 0; |
1801 | } | |
1802 | if (info.seq != syslog_seq) { | |
7ff9554b | 1803 | /* messages are gone, move to first one */ |
13791c80 | 1804 | syslog_seq = info.seq; |
eb02dac9 | 1805 | syslog_partial = 0; |
7ff9554b | 1806 | } |
3ea4331c | 1807 | if (source == SYSLOG_FROM_PROC) { |
7ff9554b KS |
1808 | /* |
1809 | * Short-cut for poll(/"proc/kmsg") which simply checks | |
1810 | * for pending data, not the size; return the count of | |
1811 | * records, not the length. | |
1812 | */ | |
896fbe20 | 1813 | error = prb_next_seq(prb) - syslog_seq; |
7ff9554b | 1814 | } else { |
e80c1a9d | 1815 | bool time = syslog_partial ? syslog_time : printk_time; |
896fbe20 JO |
1816 | unsigned int line_count; |
1817 | u64 seq; | |
1818 | ||
1819 | prb_for_each_info(syslog_seq, prb, seq, &info, | |
1820 | &line_count) { | |
1821 | error += get_record_print_text_size(&info, line_count, | |
1822 | true, time); | |
e80c1a9d | 1823 | time = printk_time; |
7ff9554b | 1824 | } |
eb02dac9 | 1825 | error -= syslog_partial; |
7ff9554b | 1826 | } |
b371cbb5 | 1827 | mutex_unlock(&syslog_lock); |
1da177e4 | 1828 | break; |
d78ca3cd KC |
1829 | /* Size of the log buffer */ |
1830 | case SYSLOG_ACTION_SIZE_BUFFER: | |
1da177e4 LT |
1831 | error = log_buf_len; |
1832 | break; | |
1833 | default: | |
1834 | error = -EINVAL; | |
1835 | break; | |
1836 | } | |
077a1cc0 | 1837 | |
1da177e4 LT |
1838 | return error; |
1839 | } | |
1840 | ||
1e7bfb21 | 1841 | SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len) |
1da177e4 | 1842 | { |
637241a9 | 1843 | return do_syslog(type, buf, len, SYSLOG_FROM_READER); |
1da177e4 LT |
1844 | } |
1845 | ||
c162d5b4 PM |
1846 | /* |
1847 | * Special console_lock variants that help to reduce the risk of soft-lockups. | |
1848 | * They allow to pass console_lock to another printk() call using a busy wait. | |
1849 | */ | |
1850 | ||
1851 | #ifdef CONFIG_LOCKDEP | |
1852 | static struct lockdep_map console_owner_dep_map = { | |
1853 | .name = "console_owner" | |
1854 | }; | |
1855 | #endif | |
1856 | ||
1857 | static DEFINE_RAW_SPINLOCK(console_owner_lock); | |
1858 | static struct task_struct *console_owner; | |
1859 | static bool console_waiter; | |
1860 | ||
1861 | /** | |
1862 | * console_lock_spinning_enable - mark beginning of code where another | |
1863 | * thread might safely busy wait | |
1864 | * | |
1865 | * This basically converts console_lock into a spinlock. This marks | |
1866 | * the section where the console_lock owner can not sleep, because | |
1867 | * there may be a waiter spinning (like a spinlock). Also it must be | |
1868 | * ready to hand over the lock at the end of the section. | |
1869 | */ | |
1870 | static void console_lock_spinning_enable(void) | |
1871 | { | |
d04d5882 PM |
1872 | /* |
1873 | * Do not use spinning in panic(). The panic CPU wants to keep the lock. | |
1874 | * Non-panic CPUs abandon the flush anyway. | |
1875 | * | |
1876 | * Just keep the lockdep annotation. The panic-CPU should avoid | |
1877 | * taking console_owner_lock because it might cause a deadlock. | |
1878 | * This looks like the easiest way how to prevent false lockdep | |
1879 | * reports without handling races a lockless way. | |
1880 | */ | |
1881 | if (panic_in_progress()) | |
1882 | goto lockdep; | |
1883 | ||
c162d5b4 PM |
1884 | raw_spin_lock(&console_owner_lock); |
1885 | console_owner = current; | |
1886 | raw_spin_unlock(&console_owner_lock); | |
1887 | ||
d04d5882 | 1888 | lockdep: |
c162d5b4 PM |
1889 | /* The waiter may spin on us after setting console_owner */ |
1890 | spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_); | |
1891 | } | |
1892 | ||
1893 | /** | |
1894 | * console_lock_spinning_disable_and_check - mark end of code where another | |
1895 | * thread was able to busy wait and check if there is a waiter | |
4fe59a13 | 1896 | * @cookie: cookie returned from console_srcu_read_lock() |
c162d5b4 PM |
1897 | * |
1898 | * This is called at the end of the section where spinning is allowed. | |
1899 | * It has two functions. First, it is a signal that it is no longer | |
1900 | * safe to start busy waiting for the lock. Second, it checks if | |
1901 | * there is a busy waiter and passes the lock rights to her. | |
1902 | * | |
fc956ae0 JO |
1903 | * Important: Callers lose both the console_lock and the SRCU read lock if |
1904 | * there was a busy waiter. They must not touch items synchronized by | |
1905 | * console_lock or SRCU read lock in this case. | |
c162d5b4 PM |
1906 | * |
1907 | * Return: 1 if the lock rights were passed, 0 otherwise. | |
1908 | */ | |
fc956ae0 | 1909 | static int console_lock_spinning_disable_and_check(int cookie) |
c162d5b4 PM |
1910 | { |
1911 | int waiter; | |
1912 | ||
d04d5882 PM |
1913 | /* |
1914 | * Ignore spinning waiters during panic() because they might get stopped | |
1915 | * or blocked at any time, | |
1916 | * | |
1917 | * It is safe because nobody is allowed to start spinning during panic | |
1918 | * in the first place. If there has been a waiter then non panic CPUs | |
1919 | * might stay spinning. They would get stopped anyway. The panic context | |
1920 | * will never start spinning and an interrupted spin on panic CPU will | |
1921 | * never continue. | |
1922 | */ | |
1923 | if (panic_in_progress()) { | |
1924 | /* Keep lockdep happy. */ | |
1925 | spin_release(&console_owner_dep_map, _THIS_IP_); | |
1926 | return 0; | |
1927 | } | |
1928 | ||
c162d5b4 PM |
1929 | raw_spin_lock(&console_owner_lock); |
1930 | waiter = READ_ONCE(console_waiter); | |
1931 | console_owner = NULL; | |
1932 | raw_spin_unlock(&console_owner_lock); | |
1933 | ||
1934 | if (!waiter) { | |
5facae4f | 1935 | spin_release(&console_owner_dep_map, _THIS_IP_); |
c162d5b4 PM |
1936 | return 0; |
1937 | } | |
1938 | ||
1939 | /* The waiter is now free to continue */ | |
1940 | WRITE_ONCE(console_waiter, false); | |
1941 | ||
5facae4f | 1942 | spin_release(&console_owner_dep_map, _THIS_IP_); |
c162d5b4 | 1943 | |
fc956ae0 JO |
1944 | /* |
1945 | * Preserve lockdep lock ordering. Release the SRCU read lock before | |
1946 | * releasing the console_lock. | |
1947 | */ | |
1948 | console_srcu_read_unlock(cookie); | |
1949 | ||
c162d5b4 PM |
1950 | /* |
1951 | * Hand off console_lock to waiter. The waiter will perform | |
1952 | * the up(). After this, the waiter is the console_lock owner. | |
1953 | */ | |
5facae4f | 1954 | mutex_release(&console_lock_dep_map, _THIS_IP_); |
c162d5b4 PM |
1955 | return 1; |
1956 | } | |
1957 | ||
1958 | /** | |
1959 | * console_trylock_spinning - try to get console_lock by busy waiting | |
1960 | * | |
1961 | * This allows to busy wait for the console_lock when the current | |
1962 | * owner is running in specially marked sections. It means that | |
1963 | * the current owner is running and cannot reschedule until it | |
1964 | * is ready to lose the lock. | |
1965 | * | |
1966 | * Return: 1 if we got the lock, 0 othrewise | |
1967 | */ | |
1968 | static int console_trylock_spinning(void) | |
1969 | { | |
1970 | struct task_struct *owner = NULL; | |
1971 | bool waiter; | |
1972 | bool spin = false; | |
1973 | unsigned long flags; | |
1974 | ||
1975 | if (console_trylock()) | |
1976 | return 1; | |
1977 | ||
d5150709 SB |
1978 | /* |
1979 | * It's unsafe to spin once a panic has begun. If we are the | |
1980 | * panic CPU, we may have already halted the owner of the | |
1981 | * console_sem. If we are not the panic CPU, then we should | |
1982 | * avoid taking console_sem, so the panic CPU has a better | |
1983 | * chance of cleanly acquiring it later. | |
1984 | */ | |
1985 | if (panic_in_progress()) | |
1986 | return 0; | |
1987 | ||
c162d5b4 PM |
1988 | printk_safe_enter_irqsave(flags); |
1989 | ||
1990 | raw_spin_lock(&console_owner_lock); | |
1991 | owner = READ_ONCE(console_owner); | |
1992 | waiter = READ_ONCE(console_waiter); | |
1993 | if (!waiter && owner && owner != current) { | |
1994 | WRITE_ONCE(console_waiter, true); | |
1995 | spin = true; | |
1996 | } | |
1997 | raw_spin_unlock(&console_owner_lock); | |
1998 | ||
1999 | /* | |
2000 | * If there is an active printk() writing to the | |
2001 | * consoles, instead of having it write our data too, | |
2002 | * see if we can offload that load from the active | |
2003 | * printer, and do some printing ourselves. | |
2004 | * Go into a spin only if there isn't already a waiter | |
2005 | * spinning, and there is an active printer, and | |
2006 | * that active printer isn't us (recursive printk?). | |
2007 | */ | |
2008 | if (!spin) { | |
2009 | printk_safe_exit_irqrestore(flags); | |
2010 | return 0; | |
2011 | } | |
2012 | ||
2013 | /* We spin waiting for the owner to release us */ | |
2014 | spin_acquire(&console_owner_dep_map, 0, 0, _THIS_IP_); | |
2015 | /* Owner will clear console_waiter on hand off */ | |
2016 | while (READ_ONCE(console_waiter)) | |
2017 | cpu_relax(); | |
5facae4f | 2018 | spin_release(&console_owner_dep_map, _THIS_IP_); |
c162d5b4 PM |
2019 | |
2020 | printk_safe_exit_irqrestore(flags); | |
2021 | /* | |
2022 | * The owner passed the console lock to us. | |
2023 | * Since we did not spin on console lock, annotate | |
2024 | * this as a trylock. Otherwise lockdep will | |
2025 | * complain. | |
2026 | */ | |
2027 | mutex_acquire(&console_lock_dep_map, 0, 1, _THIS_IP_); | |
2028 | ||
2029 | return 1; | |
2030 | } | |
2031 | ||
002eb6ad JO |
2032 | /* |
2033 | * Recursion is tracked separately on each CPU. If NMIs are supported, an | |
2034 | * additional NMI context per CPU is also separately tracked. Until per-CPU | |
2035 | * is available, a separate "early tracking" is performed. | |
2036 | */ | |
2037 | static DEFINE_PER_CPU(u8, printk_count); | |
2038 | static u8 printk_count_early; | |
2039 | #ifdef CONFIG_HAVE_NMI | |
2040 | static DEFINE_PER_CPU(u8, printk_count_nmi); | |
2041 | static u8 printk_count_nmi_early; | |
2042 | #endif | |
2043 | ||
2044 | /* | |
2045 | * Recursion is limited to keep the output sane. printk() should not require | |
2046 | * more than 1 level of recursion (allowing, for example, printk() to trigger | |
2047 | * a WARN), but a higher value is used in case some printk-internal errors | |
2048 | * exist, such as the ringbuffer validation checks failing. | |
2049 | */ | |
2050 | #define PRINTK_MAX_RECURSION 3 | |
2051 | ||
2052 | /* | |
2053 | * Return a pointer to the dedicated counter for the CPU+context of the | |
2054 | * caller. | |
2055 | */ | |
2056 | static u8 *__printk_recursion_counter(void) | |
2057 | { | |
2058 | #ifdef CONFIG_HAVE_NMI | |
2059 | if (in_nmi()) { | |
2060 | if (printk_percpu_data_ready()) | |
2061 | return this_cpu_ptr(&printk_count_nmi); | |
2062 | return &printk_count_nmi_early; | |
2063 | } | |
2064 | #endif | |
2065 | if (printk_percpu_data_ready()) | |
2066 | return this_cpu_ptr(&printk_count); | |
2067 | return &printk_count_early; | |
2068 | } | |
2069 | ||
2070 | /* | |
2071 | * Enter recursion tracking. Interrupts are disabled to simplify tracking. | |
2072 | * The caller must check the boolean return value to see if the recursion is | |
2073 | * allowed. On failure, interrupts are not disabled. | |
2074 | * | |
2075 | * @recursion_ptr must be a variable of type (u8 *) and is the same variable | |
2076 | * that is passed to printk_exit_irqrestore(). | |
2077 | */ | |
2078 | #define printk_enter_irqsave(recursion_ptr, flags) \ | |
2079 | ({ \ | |
2080 | bool success = true; \ | |
2081 | \ | |
2082 | typecheck(u8 *, recursion_ptr); \ | |
2083 | local_irq_save(flags); \ | |
2084 | (recursion_ptr) = __printk_recursion_counter(); \ | |
2085 | if (*(recursion_ptr) > PRINTK_MAX_RECURSION) { \ | |
2086 | local_irq_restore(flags); \ | |
2087 | success = false; \ | |
2088 | } else { \ | |
2089 | (*(recursion_ptr))++; \ | |
2090 | } \ | |
2091 | success; \ | |
2092 | }) | |
2093 | ||
2094 | /* Exit recursion tracking, restoring interrupts. */ | |
2095 | #define printk_exit_irqrestore(recursion_ptr, flags) \ | |
2096 | do { \ | |
2097 | typecheck(u8 *, recursion_ptr); \ | |
2098 | (*(recursion_ptr))--; \ | |
2099 | local_irq_restore(flags); \ | |
2100 | } while (0) | |
2101 | ||
af91322e DY |
2102 | int printk_delay_msec __read_mostly; |
2103 | ||
1f47e8af | 2104 | static inline void printk_delay(int level) |
af91322e | 2105 | { |
1f47e8af JO |
2106 | boot_delay_msec(level); |
2107 | ||
af91322e DY |
2108 | if (unlikely(printk_delay_msec)) { |
2109 | int m = printk_delay_msec; | |
2110 | ||
2111 | while (m--) { | |
2112 | mdelay(1); | |
2113 | touch_nmi_watchdog(); | |
2114 | } | |
2115 | } | |
2116 | } | |
2117 | ||
cbae05d3 TH |
2118 | static inline u32 printk_caller_id(void) |
2119 | { | |
2120 | return in_task() ? task_pid_nr(current) : | |
9f0844de | 2121 | 0x80000000 + smp_processor_id(); |
cbae05d3 TH |
2122 | } |
2123 | ||
b031a684 | 2124 | /** |
f3d75cf5 | 2125 | * printk_parse_prefix - Parse level and control flags. |
b031a684 JO |
2126 | * |
2127 | * @text: The terminated text message. | |
2128 | * @level: A pointer to the current level value, will be updated. | |
a1ad4b8a | 2129 | * @flags: A pointer to the current printk_info flags, will be updated. |
b031a684 JO |
2130 | * |
2131 | * @level may be NULL if the caller is not interested in the parsed value. | |
2132 | * Otherwise the variable pointed to by @level must be set to | |
2133 | * LOGLEVEL_DEFAULT in order to be updated with the parsed value. | |
2134 | * | |
a1ad4b8a CD |
2135 | * @flags may be NULL if the caller is not interested in the parsed value. |
2136 | * Otherwise the variable pointed to by @flags will be OR'd with the parsed | |
b031a684 JO |
2137 | * value. |
2138 | * | |
2139 | * Return: The length of the parsed level and control flags. | |
2140 | */ | |
f3d75cf5 | 2141 | u16 printk_parse_prefix(const char *text, int *level, |
a1ad4b8a | 2142 | enum printk_info_flags *flags) |
c362c7ff | 2143 | { |
b031a684 JO |
2144 | u16 prefix_len = 0; |
2145 | int kern_level; | |
cbae05d3 | 2146 | |
b031a684 JO |
2147 | while (*text) { |
2148 | kern_level = printk_get_level(text); | |
2149 | if (!kern_level) | |
2150 | break; | |
f5f022e5 | 2151 | |
b031a684 JO |
2152 | switch (kern_level) { |
2153 | case '0' ... '7': | |
2154 | if (level && *level == LOGLEVEL_DEFAULT) | |
2155 | *level = kern_level - '0'; | |
2156 | break; | |
2157 | case 'c': /* KERN_CONT */ | |
a1ad4b8a CD |
2158 | if (flags) |
2159 | *flags |= LOG_CONT; | |
b031a684 JO |
2160 | } |
2161 | ||
2162 | prefix_len += 2; | |
2163 | text += 2; | |
2164 | } | |
2165 | ||
2166 | return prefix_len; | |
2167 | } | |
2168 | ||
264a7504 | 2169 | __printf(5, 0) |
a1ad4b8a CD |
2170 | static u16 printk_sprint(char *text, u16 size, int facility, |
2171 | enum printk_info_flags *flags, const char *fmt, | |
2172 | va_list args) | |
b031a684 JO |
2173 | { |
2174 | u16 text_len; | |
2175 | ||
2176 | text_len = vscnprintf(text, size, fmt, args); | |
2177 | ||
2178 | /* Mark and strip a trailing newline. */ | |
2179 | if (text_len && text[text_len - 1] == '\n') { | |
2180 | text_len--; | |
a1ad4b8a | 2181 | *flags |= LOG_NEWLINE; |
b031a684 JO |
2182 | } |
2183 | ||
2184 | /* Strip log level and control flags. */ | |
2185 | if (facility == 0) { | |
2186 | u16 prefix_len; | |
2187 | ||
f3d75cf5 | 2188 | prefix_len = printk_parse_prefix(text, NULL, NULL); |
b031a684 JO |
2189 | if (prefix_len) { |
2190 | text_len -= prefix_len; | |
2191 | memmove(text, text + prefix_len, text_len); | |
f5f022e5 | 2192 | } |
c362c7ff LT |
2193 | } |
2194 | ||
880970b5 | 2195 | trace_console(text, text_len); |
701850dc | 2196 | |
b031a684 | 2197 | return text_len; |
c362c7ff LT |
2198 | } |
2199 | ||
b031a684 | 2200 | __printf(4, 0) |
ba552399 | 2201 | int vprintk_store(int facility, int level, |
74caba7f | 2202 | const struct dev_printk_info *dev_info, |
ba552399 | 2203 | const char *fmt, va_list args) |
1da177e4 | 2204 | { |
6b916706 | 2205 | struct prb_reserved_entry e; |
a1ad4b8a | 2206 | enum printk_info_flags flags = 0; |
6b916706 | 2207 | struct printk_record r; |
002eb6ad | 2208 | unsigned long irqflags; |
6b916706 | 2209 | u16 trunc_msg_len = 0; |
b031a684 | 2210 | char prefix_buf[8]; |
002eb6ad | 2211 | u8 *recursion_ptr; |
b031a684 JO |
2212 | u16 reserve_size; |
2213 | va_list args2; | |
9f0844de | 2214 | u32 caller_id; |
6b916706 | 2215 | u16 text_len; |
002eb6ad | 2216 | int ret = 0; |
6b916706 | 2217 | u64 ts_nsec; |
bfe8df3d | 2218 | |
9f0844de JO |
2219 | if (!printk_enter_irqsave(recursion_ptr, irqflags)) |
2220 | return 0; | |
2221 | ||
7ff9554b | 2222 | /* |
6b916706 JO |
2223 | * Since the duration of printk() can vary depending on the message |
2224 | * and state of the ringbuffer, grab the timestamp now so that it is | |
2225 | * close to the call of printk(). This provides a more deterministic | |
2226 | * timestamp with respect to the caller. | |
7ff9554b | 2227 | */ |
6b916706 | 2228 | ts_nsec = local_clock(); |
5fd29d6c | 2229 | |
9f0844de | 2230 | caller_id = printk_caller_id(); |
002eb6ad | 2231 | |
7ff9554b | 2232 | /* |
b031a684 JO |
2233 | * The sprintf needs to come first since the syslog prefix might be |
2234 | * passed in as a parameter. An extra byte must be reserved so that | |
2235 | * later the vscnprintf() into the reserved buffer has room for the | |
2236 | * terminating '\0', which is not counted by vsnprintf(). | |
7ff9554b | 2237 | */ |
b031a684 JO |
2238 | va_copy(args2, args); |
2239 | reserve_size = vsnprintf(&prefix_buf[0], sizeof(prefix_buf), fmt, args2) + 1; | |
2240 | va_end(args2); | |
9d90c8d9 | 2241 | |
b0975c47 JO |
2242 | if (reserve_size > PRINTKRB_RECORD_MAX) |
2243 | reserve_size = PRINTKRB_RECORD_MAX; | |
088a52aa | 2244 | |
b031a684 JO |
2245 | /* Extract log level or control flags. */ |
2246 | if (facility == 0) | |
f3d75cf5 | 2247 | printk_parse_prefix(&prefix_buf[0], &level, &flags); |
5fd29d6c | 2248 | |
a39d4a85 | 2249 | if (level == LOGLEVEL_DEFAULT) |
c313af14 | 2250 | level = default_message_loglevel; |
9d90c8d9 | 2251 | |
74caba7f | 2252 | if (dev_info) |
a1ad4b8a | 2253 | flags |= LOG_NEWLINE; |
9d90c8d9 | 2254 | |
a1ad4b8a | 2255 | if (flags & LOG_CONT) { |
b031a684 | 2256 | prb_rec_init_wr(&r, reserve_size); |
b0975c47 | 2257 | if (prb_reserve_in_last(&e, prb, &r, caller_id, PRINTKRB_RECORD_MAX)) { |
b031a684 | 2258 | text_len = printk_sprint(&r.text_buf[r.info->text_len], reserve_size, |
a1ad4b8a | 2259 | facility, &flags, fmt, args); |
6b916706 | 2260 | r.info->text_len += text_len; |
088a52aa | 2261 | |
a1ad4b8a | 2262 | if (flags & LOG_NEWLINE) { |
6b916706 JO |
2263 | r.info->flags |= LOG_NEWLINE; |
2264 | prb_final_commit(&e); | |
2265 | } else { | |
2266 | prb_commit(&e); | |
088a52aa | 2267 | } |
4bcc595c | 2268 | |
002eb6ad JO |
2269 | ret = text_len; |
2270 | goto out; | |
5fd29d6c LT |
2271 | } |
2272 | } | |
2273 | ||
6b916706 JO |
2274 | /* |
2275 | * Explicitly initialize the record before every prb_reserve() call. | |
2276 | * prb_reserve_in_last() and prb_reserve() purposely invalidate the | |
2277 | * structure when they fail. | |
2278 | */ | |
b031a684 | 2279 | prb_rec_init_wr(&r, reserve_size); |
6b916706 JO |
2280 | if (!prb_reserve(&e, prb, &r)) { |
2281 | /* truncate the message if it is too long for empty buffer */ | |
b031a684 | 2282 | truncate_msg(&reserve_size, &trunc_msg_len); |
9d90c8d9 | 2283 | |
b031a684 | 2284 | prb_rec_init_wr(&r, reserve_size + trunc_msg_len); |
6b916706 | 2285 | if (!prb_reserve(&e, prb, &r)) |
002eb6ad | 2286 | goto out; |
6b916706 JO |
2287 | } |
2288 | ||
2289 | /* fill message */ | |
a1ad4b8a | 2290 | text_len = printk_sprint(&r.text_buf[0], reserve_size, facility, &flags, fmt, args); |
6b916706 JO |
2291 | if (trunc_msg_len) |
2292 | memcpy(&r.text_buf[text_len], trunc_msg, trunc_msg_len); | |
2293 | r.info->text_len = text_len + trunc_msg_len; | |
2294 | r.info->facility = facility; | |
2295 | r.info->level = level & 7; | |
a1ad4b8a | 2296 | r.info->flags = flags & 0x1f; |
6b916706 JO |
2297 | r.info->ts_nsec = ts_nsec; |
2298 | r.info->caller_id = caller_id; | |
74caba7f | 2299 | if (dev_info) |
6b916706 | 2300 | memcpy(&r.info->dev_info, dev_info, sizeof(r.info->dev_info)); |
ac60ad74 | 2301 | |
6b916706 | 2302 | /* A message without a trailing newline can be continued. */ |
a1ad4b8a | 2303 | if (!(flags & LOG_NEWLINE)) |
6b916706 JO |
2304 | prb_commit(&e); |
2305 | else | |
2306 | prb_final_commit(&e); | |
2307 | ||
002eb6ad JO |
2308 | ret = text_len + trunc_msg_len; |
2309 | out: | |
2310 | printk_exit_irqrestore(recursion_ptr, irqflags); | |
2311 | return ret; | |
ba552399 | 2312 | } |
1da177e4 | 2313 | |
ba552399 | 2314 | asmlinkage int vprintk_emit(int facility, int level, |
74caba7f | 2315 | const struct dev_printk_info *dev_info, |
ba552399 PM |
2316 | const char *fmt, va_list args) |
2317 | { | |
2318 | int printed_len; | |
8749efc0 | 2319 | bool in_sched = false; |
ba552399 | 2320 | |
c39ea0b9 FT |
2321 | /* Suppress unimportant messages after panic happens */ |
2322 | if (unlikely(suppress_printk)) | |
2323 | return 0; | |
2324 | ||
0ab7cdd0 | 2325 | if (unlikely(suppress_panic_printk) && other_cpu_in_panic()) |
13fb0f74 SB |
2326 | return 0; |
2327 | ||
ba552399 PM |
2328 | if (level == LOGLEVEL_SCHED) { |
2329 | level = LOGLEVEL_DEFAULT; | |
2330 | in_sched = true; | |
2331 | } | |
2332 | ||
1f47e8af | 2333 | printk_delay(level); |
ba552399 | 2334 | |
74caba7f | 2335 | printed_len = vprintk_store(facility, level, dev_info, fmt, args); |
939f04be | 2336 | |
458df9fd | 2337 | /* If called from the scheduler, we can not call up(). */ |
5831788a | 2338 | if (!in_sched) { |
fd5f7cde | 2339 | /* |
a699449b | 2340 | * The caller may be holding system-critical or |
5831788a | 2341 | * timing-sensitive locks. Disable preemption during |
a699449b JO |
2342 | * printing of all remaining records to all consoles so that |
2343 | * this context can return as soon as possible. Hopefully | |
2344 | * another printk() caller will take over the printing. | |
fd5f7cde SS |
2345 | */ |
2346 | preempt_disable(); | |
d18bbc21 AM |
2347 | /* |
2348 | * Try to acquire and then immediately release the console | |
a699449b JO |
2349 | * semaphore. The release will print out buffers. With the |
2350 | * spinning variant, this context tries to take over the | |
2351 | * printing from another printing context. | |
d18bbc21 | 2352 | */ |
c162d5b4 | 2353 | if (console_trylock_spinning()) |
d18bbc21 | 2354 | console_unlock(); |
fd5f7cde | 2355 | preempt_enable(); |
d18bbc21 | 2356 | } |
76a8ad29 | 2357 | |
696ffaf5 JO |
2358 | if (in_sched) |
2359 | defer_console_output(); | |
2360 | else | |
2361 | wake_up_klogd(); | |
2362 | ||
1da177e4 LT |
2363 | return printed_len; |
2364 | } | |
7ff9554b KS |
2365 | EXPORT_SYMBOL(vprintk_emit); |
2366 | ||
a0cba217 | 2367 | int vprintk_default(const char *fmt, va_list args) |
afdc34a3 | 2368 | { |
74caba7f | 2369 | return vprintk_emit(0, LOGLEVEL_DEFAULT, NULL, fmt, args); |
afdc34a3 SRRH |
2370 | } |
2371 | EXPORT_SYMBOL_GPL(vprintk_default); | |
2372 | ||
33701557 | 2373 | asmlinkage __visible int _printk(const char *fmt, ...) |
7ff9554b KS |
2374 | { |
2375 | va_list args; | |
2376 | int r; | |
2377 | ||
7ff9554b | 2378 | va_start(args, fmt); |
28e1745b | 2379 | r = vprintk(fmt, args); |
7ff9554b KS |
2380 | va_end(args); |
2381 | ||
2382 | return r; | |
2383 | } | |
33701557 | 2384 | EXPORT_SYMBOL(_printk); |
7f3a781d | 2385 | |
c60ba2d3 | 2386 | static bool pr_flush(int timeout_ms, bool reset_on_progress); |
3b604ca8 JO |
2387 | static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progress); |
2388 | ||
96efedf1 | 2389 | #else /* CONFIG_PRINTK */ |
d59745ce | 2390 | |
e80c1a9d | 2391 | #define printk_time false |
249771b8 | 2392 | |
896fbe20 JO |
2393 | #define prb_read_valid(rb, seq, r) false |
2394 | #define prb_first_valid_seq(rb) 0 | |
a699449b | 2395 | #define prb_next_seq(rb) 0 |
896fbe20 | 2396 | |
96efedf1 | 2397 | static u64 syslog_seq; |
896fbe20 | 2398 | |
c60ba2d3 | 2399 | static bool pr_flush(int timeout_ms, bool reset_on_progress) { return true; } |
3b604ca8 | 2400 | static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progress) { return true; } |
d59745ce | 2401 | |
7f3a781d | 2402 | #endif /* CONFIG_PRINTK */ |
d59745ce | 2403 | |
d0380e6c TG |
2404 | #ifdef CONFIG_EARLY_PRINTK |
2405 | struct console *early_console; | |
2406 | ||
722a9f92 | 2407 | asmlinkage __visible void early_printk(const char *fmt, ...) |
d0380e6c TG |
2408 | { |
2409 | va_list ap; | |
1dc6244b JP |
2410 | char buf[512]; |
2411 | int n; | |
2412 | ||
2413 | if (!early_console) | |
2414 | return; | |
d0380e6c TG |
2415 | |
2416 | va_start(ap, fmt); | |
1dc6244b | 2417 | n = vscnprintf(buf, sizeof(buf), fmt, ap); |
d0380e6c | 2418 | va_end(ap); |
1dc6244b JP |
2419 | |
2420 | early_console->write(early_console, buf, n); | |
d0380e6c TG |
2421 | } |
2422 | #endif | |
2423 | ||
a5a763b2 AK |
2424 | static void set_user_specified(struct console_cmdline *c, bool user_specified) |
2425 | { | |
2426 | if (!user_specified) | |
2427 | return; | |
2428 | ||
2429 | /* | |
2430 | * @c console was defined by the user on the command line. | |
2431 | * Do not clear when added twice also by SPCR or the device tree. | |
2432 | */ | |
2433 | c->user_specified = true; | |
2434 | /* At least one console defined by the user on the command line. */ | |
2435 | console_set_on_cmdline = 1; | |
2436 | } | |
2437 | ||
f7511d5f | 2438 | static int __add_preferred_console(char *name, int idx, char *options, |
e369d822 | 2439 | char *brl_options, bool user_specified) |
f7511d5f ST |
2440 | { |
2441 | struct console_cmdline *c; | |
2442 | int i; | |
2443 | ||
2444 | /* | |
2445 | * See if this tty is not yet registered, and | |
2446 | * if we have a slot free. | |
2447 | */ | |
dac8bbba PM |
2448 | for (i = 0, c = console_cmdline; |
2449 | i < MAX_CMDLINECONSOLES && c->name[0]; | |
2450 | i++, c++) { | |
23475408 | 2451 | if (strcmp(c->name, name) == 0 && c->index == idx) { |
dac8bbba PM |
2452 | if (!brl_options) |
2453 | preferred_console = i; | |
a5a763b2 | 2454 | set_user_specified(c, user_specified); |
23475408 | 2455 | return 0; |
f7511d5f | 2456 | } |
23475408 | 2457 | } |
f7511d5f ST |
2458 | if (i == MAX_CMDLINECONSOLES) |
2459 | return -E2BIG; | |
2460 | if (!brl_options) | |
ad86ee2b | 2461 | preferred_console = i; |
7365df19 | 2462 | strscpy(c->name, name, sizeof(c->name)); |
f7511d5f | 2463 | c->options = options; |
a5a763b2 | 2464 | set_user_specified(c, user_specified); |
bbeddf52 JP |
2465 | braille_set_options(c, brl_options); |
2466 | ||
f7511d5f ST |
2467 | c->index = idx; |
2468 | return 0; | |
2469 | } | |
cca10d58 SS |
2470 | |
2471 | static int __init console_msg_format_setup(char *str) | |
2472 | { | |
2473 | if (!strcmp(str, "syslog")) | |
2474 | console_msg_format = MSG_FORMAT_SYSLOG; | |
2475 | if (!strcmp(str, "default")) | |
2476 | console_msg_format = MSG_FORMAT_DEFAULT; | |
2477 | return 1; | |
2478 | } | |
2479 | __setup("console_msg_format=", console_msg_format_setup); | |
2480 | ||
2ea1c539 | 2481 | /* |
0b90fec3 AE |
2482 | * Set up a console. Called via do_early_param() in init/main.c |
2483 | * for each "console=" parameter in the boot command line. | |
2ea1c539 JB |
2484 | */ |
2485 | static int __init console_setup(char *str) | |
2486 | { | |
0b90fec3 | 2487 | char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for "ttyS" */ |
f7511d5f | 2488 | char *s, *options, *brl_options = NULL; |
2ea1c539 JB |
2489 | int idx; |
2490 | ||
3cffa06a PM |
2491 | /* |
2492 | * console="" or console=null have been suggested as a way to | |
2493 | * disable console output. Use ttynull that has been created | |
acebb559 | 2494 | * for exactly this purpose. |
3cffa06a PM |
2495 | */ |
2496 | if (str[0] == 0 || strcmp(str, "null") == 0) { | |
2497 | __add_preferred_console("ttynull", 0, NULL, NULL, true); | |
48021f98 | 2498 | return 1; |
3cffa06a | 2499 | } |
48021f98 | 2500 | |
bbeddf52 JP |
2501 | if (_braille_console_setup(&str, &brl_options)) |
2502 | return 1; | |
f7511d5f | 2503 | |
2ea1c539 JB |
2504 | /* |
2505 | * Decode str into name, index, options. | |
2506 | */ | |
2507 | if (str[0] >= '0' && str[0] <= '9') { | |
eaa944af YL |
2508 | strcpy(buf, "ttyS"); |
2509 | strncpy(buf + 4, str, sizeof(buf) - 5); | |
2ea1c539 | 2510 | } else { |
eaa944af | 2511 | strncpy(buf, str, sizeof(buf) - 1); |
2ea1c539 | 2512 | } |
eaa944af | 2513 | buf[sizeof(buf) - 1] = 0; |
249771b8 AE |
2514 | options = strchr(str, ','); |
2515 | if (options) | |
2ea1c539 JB |
2516 | *(options++) = 0; |
2517 | #ifdef __sparc__ | |
2518 | if (!strcmp(str, "ttya")) | |
eaa944af | 2519 | strcpy(buf, "ttyS0"); |
2ea1c539 | 2520 | if (!strcmp(str, "ttyb")) |
eaa944af | 2521 | strcpy(buf, "ttyS1"); |
2ea1c539 | 2522 | #endif |
eaa944af | 2523 | for (s = buf; *s; s++) |
249771b8 | 2524 | if (isdigit(*s) || *s == ',') |
2ea1c539 JB |
2525 | break; |
2526 | idx = simple_strtoul(s, NULL, 10); | |
2527 | *s = 0; | |
2528 | ||
e369d822 | 2529 | __add_preferred_console(buf, idx, options, brl_options, true); |
2ea1c539 JB |
2530 | return 1; |
2531 | } | |
2532 | __setup("console=", console_setup); | |
2533 | ||
3c0547ba MM |
2534 | /** |
2535 | * add_preferred_console - add a device to the list of preferred consoles. | |
ddad86c2 MW |
2536 | * @name: device name |
2537 | * @idx: device index | |
2538 | * @options: options for this console | |
3c0547ba MM |
2539 | * |
2540 | * The last preferred console added will be used for kernel messages | |
2541 | * and stdin/out/err for init. Normally this is used by console_setup | |
2542 | * above to handle user-supplied console arguments; however it can also | |
2543 | * be used by arch-specific code either to override the user or more | |
2544 | * commonly to provide a default console (ie from PROM variables) when | |
2545 | * the user has not supplied one. | |
2546 | */ | |
fb445ee5 | 2547 | int add_preferred_console(char *name, int idx, char *options) |
3c0547ba | 2548 | { |
e369d822 | 2549 | return __add_preferred_console(name, idx, options, NULL, false); |
3c0547ba MM |
2550 | } |
2551 | ||
d25d9fec | 2552 | bool console_suspend_enabled = true; |
8f4ce8c3 AS |
2553 | EXPORT_SYMBOL(console_suspend_enabled); |
2554 | ||
2555 | static int __init console_suspend_disable(char *str) | |
2556 | { | |
d25d9fec | 2557 | console_suspend_enabled = false; |
8f4ce8c3 AS |
2558 | return 1; |
2559 | } | |
2560 | __setup("no_console_suspend", console_suspend_disable); | |
134620f7 YZ |
2561 | module_param_named(console_suspend, console_suspend_enabled, |
2562 | bool, S_IRUGO | S_IWUSR); | |
2563 | MODULE_PARM_DESC(console_suspend, "suspend console during suspend" | |
2564 | " and hibernate operations"); | |
8f4ce8c3 | 2565 | |
10102a89 DS |
2566 | static bool printk_console_no_auto_verbose; |
2567 | ||
2568 | void console_verbose(void) | |
2569 | { | |
2570 | if (console_loglevel && !printk_console_no_auto_verbose) | |
2571 | console_loglevel = CONSOLE_LOGLEVEL_MOTORMOUTH; | |
2572 | } | |
2573 | EXPORT_SYMBOL_GPL(console_verbose); | |
2574 | ||
2575 | module_param_named(console_no_auto_verbose, printk_console_no_auto_verbose, bool, 0644); | |
2576 | MODULE_PARM_DESC(console_no_auto_verbose, "Disable console loglevel raise to highest on oops/panic/etc"); | |
2577 | ||
557240b4 LT |
2578 | /** |
2579 | * suspend_console - suspend the console subsystem | |
2580 | * | |
2581 | * This disables printk() while we go into suspend states | |
2582 | */ | |
2583 | void suspend_console(void) | |
2584 | { | |
9e70a5e1 JO |
2585 | struct console *con; |
2586 | ||
8f4ce8c3 AS |
2587 | if (!console_suspend_enabled) |
2588 | return; | |
47319f71 | 2589 | pr_info("Suspending console(s) (use no_console_suspend to debug)\n"); |
3b604ca8 | 2590 | pr_flush(1000, true); |
9e70a5e1 JO |
2591 | |
2592 | console_list_lock(); | |
2593 | for_each_console(con) | |
2594 | console_srcu_write_flags(con, con->flags | CON_SUSPENDED); | |
2595 | console_list_unlock(); | |
2596 | ||
2597 | /* | |
2598 | * Ensure that all SRCU list walks have completed. All printing | |
2599 | * contexts must be able to see that they are suspended so that it | |
2600 | * is guaranteed that all printing has stopped when this function | |
2601 | * completes. | |
2602 | */ | |
2603 | synchronize_srcu(&console_srcu); | |
557240b4 LT |
2604 | } |
2605 | ||
2606 | void resume_console(void) | |
2607 | { | |
9e70a5e1 JO |
2608 | struct console *con; |
2609 | ||
8f4ce8c3 AS |
2610 | if (!console_suspend_enabled) |
2611 | return; | |
9e70a5e1 JO |
2612 | |
2613 | console_list_lock(); | |
2614 | for_each_console(con) | |
2615 | console_srcu_write_flags(con, con->flags & ~CON_SUSPENDED); | |
2616 | console_list_unlock(); | |
2617 | ||
2618 | /* | |
2619 | * Ensure that all SRCU list walks have completed. All printing | |
2620 | * contexts must be able to see they are no longer suspended so | |
2621 | * that they are guaranteed to wake up and resume printing. | |
2622 | */ | |
2623 | synchronize_srcu(&console_srcu); | |
2624 | ||
3b604ca8 | 2625 | pr_flush(1000, true); |
557240b4 LT |
2626 | } |
2627 | ||
034260d6 KC |
2628 | /** |
2629 | * console_cpu_notify - print deferred console messages after CPU hotplug | |
90b14889 | 2630 | * @cpu: unused |
034260d6 KC |
2631 | * |
2632 | * If printk() is called from a CPU that is not online yet, the messages | |
64ca752d SS |
2633 | * will be printed on the console only if there are CON_ANYTIME consoles. |
2634 | * This function is called when a new CPU comes online (or fails to come | |
2635 | * up) or goes offline. | |
034260d6 | 2636 | */ |
90b14889 SAS |
2637 | static int console_cpu_notify(unsigned int cpu) |
2638 | { | |
f97960fb | 2639 | if (!cpuhp_tasks_frozen) { |
64ca752d SS |
2640 | /* If trylock fails, someone else is doing the printing */ |
2641 | if (console_trylock()) | |
2642 | console_unlock(); | |
034260d6 | 2643 | } |
90b14889 | 2644 | return 0; |
034260d6 KC |
2645 | } |
2646 | ||
1da177e4 | 2647 | /** |
848a9c10 | 2648 | * console_lock - block the console subsystem from printing |
1da177e4 | 2649 | * |
848a9c10 JO |
2650 | * Acquires a lock which guarantees that no consoles will |
2651 | * be in or enter their write() callback. | |
1da177e4 LT |
2652 | * |
2653 | * Can sleep, returns nothing. | |
2654 | */ | |
ac751efa | 2655 | void console_lock(void) |
1da177e4 | 2656 | { |
6b898c07 DV |
2657 | might_sleep(); |
2658 | ||
51a1d258 | 2659 | /* On panic, the console_lock must be left to the panic cpu. */ |
132a90d1 | 2660 | while (other_cpu_in_panic()) |
51a1d258 JO |
2661 | msleep(1000); |
2662 | ||
bd8d7cf5 | 2663 | down_console_sem(); |
007eeab7 | 2664 | console_locked = 1; |
1da177e4 LT |
2665 | console_may_schedule = 1; |
2666 | } | |
ac751efa | 2667 | EXPORT_SYMBOL(console_lock); |
1da177e4 | 2668 | |
ac751efa | 2669 | /** |
848a9c10 | 2670 | * console_trylock - try to block the console subsystem from printing |
ac751efa | 2671 | * |
848a9c10 JO |
2672 | * Try to acquire a lock which guarantees that no consoles will |
2673 | * be in or enter their write() callback. | |
ac751efa TH |
2674 | * |
2675 | * returns 1 on success, and 0 on failure to acquire the lock. | |
2676 | */ | |
2677 | int console_trylock(void) | |
1da177e4 | 2678 | { |
51a1d258 | 2679 | /* On panic, the console_lock must be left to the panic cpu. */ |
132a90d1 | 2680 | if (other_cpu_in_panic()) |
ac751efa | 2681 | return 0; |
bd8d7cf5 | 2682 | if (down_trylock_console_sem()) |
ac751efa | 2683 | return 0; |
007eeab7 | 2684 | console_locked = 1; |
fd5f7cde | 2685 | console_may_schedule = 0; |
ac751efa | 2686 | return 1; |
1da177e4 | 2687 | } |
ac751efa | 2688 | EXPORT_SYMBOL(console_trylock); |
1da177e4 LT |
2689 | |
2690 | int is_console_locked(void) | |
2691 | { | |
2d9ef940 | 2692 | return console_locked; |
1da177e4 | 2693 | } |
d48de54a | 2694 | EXPORT_SYMBOL(is_console_locked); |
1da177e4 | 2695 | |
5831788a PM |
2696 | /* |
2697 | * Check if the given console is currently capable and allowed to print | |
2698 | * records. | |
2699 | * | |
12f1da5f | 2700 | * Requires the console_srcu_read_lock. |
5831788a PM |
2701 | */ |
2702 | static inline bool console_is_usable(struct console *con) | |
a699449b | 2703 | { |
12f1da5f JO |
2704 | short flags = console_srcu_read_flags(con); |
2705 | ||
2706 | if (!(flags & CON_ENABLED)) | |
5831788a PM |
2707 | return false; |
2708 | ||
9e70a5e1 JO |
2709 | if ((flags & CON_SUSPENDED)) |
2710 | return false; | |
2711 | ||
5831788a | 2712 | if (!con->write) |
a699449b JO |
2713 | return false; |
2714 | ||
2715 | /* | |
2716 | * Console drivers may assume that per-cpu resources have been | |
2717 | * allocated. So unless they're explicitly marked as being able to | |
2718 | * cope (CON_ANYTIME) don't call them until this CPU is officially up. | |
2719 | */ | |
12f1da5f | 2720 | if (!cpu_online(raw_smp_processor_id()) && !(flags & CON_ANYTIME)) |
a699449b JO |
2721 | return false; |
2722 | ||
2723 | return true; | |
2724 | } | |
2725 | ||
2726 | static void __console_unlock(void) | |
2727 | { | |
007eeab7 | 2728 | console_locked = 0; |
a699449b JO |
2729 | up_console_sem(); |
2730 | } | |
2731 | ||
98a04655 JO |
2732 | #ifdef CONFIG_PRINTK |
2733 | ||
a699449b | 2734 | /* |
c4fcc617 JO |
2735 | * Prepend the message in @pmsg->pbufs->outbuf with a "dropped message". This |
2736 | * is achieved by shifting the existing message over and inserting the dropped | |
2737 | * message. | |
2738 | * | |
2739 | * @pmsg is the printk message to prepend. | |
a699449b | 2740 | * |
c4fcc617 | 2741 | * @dropped is the dropped count to report in the dropped message. |
03a749e6 | 2742 | * |
c4fcc617 JO |
2743 | * If the message text in @pmsg->pbufs->outbuf does not have enough space for |
2744 | * the dropped message, the message text will be sufficiently truncated. | |
03a749e6 | 2745 | * |
c4fcc617 JO |
2746 | * If @pmsg->pbufs->outbuf is modified, @pmsg->outbuf_len is updated. |
2747 | */ | |
06653d57 | 2748 | void console_prepend_dropped(struct printk_message *pmsg, unsigned long dropped) |
c4fcc617 JO |
2749 | { |
2750 | struct printk_buffers *pbufs = pmsg->pbufs; | |
2751 | const size_t scratchbuf_sz = sizeof(pbufs->scratchbuf); | |
2752 | const size_t outbuf_sz = sizeof(pbufs->outbuf); | |
2753 | char *scratchbuf = &pbufs->scratchbuf[0]; | |
2754 | char *outbuf = &pbufs->outbuf[0]; | |
2755 | size_t len; | |
2756 | ||
d551afc2 | 2757 | len = scnprintf(scratchbuf, scratchbuf_sz, |
c4fcc617 JO |
2758 | "** %lu printk messages dropped **\n", dropped); |
2759 | ||
2760 | /* | |
2761 | * Make sure outbuf is sufficiently large before prepending. | |
2762 | * Keep at least the prefix when the message must be truncated. | |
2763 | * It is a rather theoretical problem when someone tries to | |
2764 | * use a minimalist buffer. | |
2765 | */ | |
b0975c47 | 2766 | if (WARN_ON_ONCE(len + PRINTK_PREFIX_MAX >= outbuf_sz)) |
c4fcc617 JO |
2767 | return; |
2768 | ||
2769 | if (pmsg->outbuf_len + len >= outbuf_sz) { | |
2770 | /* Truncate the message, but keep it terminated. */ | |
2771 | pmsg->outbuf_len = outbuf_sz - (len + 1); | |
2772 | outbuf[pmsg->outbuf_len] = 0; | |
2773 | } | |
2774 | ||
2775 | memmove(outbuf + len, outbuf, pmsg->outbuf_len + 1); | |
2776 | memcpy(outbuf, scratchbuf, len); | |
2777 | pmsg->outbuf_len += len; | |
2778 | } | |
c4fcc617 | 2779 | |
a699449b | 2780 | /* |
2830eec1 JO |
2781 | * Read and format the specified record (or a later record if the specified |
2782 | * record is not available). | |
a699449b | 2783 | * |
2830eec1 JO |
2784 | * @pmsg will contain the formatted result. @pmsg->pbufs must point to a |
2785 | * struct printk_buffers. | |
fc956ae0 | 2786 | * |
2830eec1 JO |
2787 | * @seq is the record to read and format. If it is not available, the next |
2788 | * valid record is read. | |
a699449b | 2789 | * |
2830eec1 JO |
2790 | * @is_extended specifies if the message should be formatted for extended |
2791 | * console output. | |
a8199371 | 2792 | * |
ea308da1 JO |
2793 | * @may_supress specifies if records may be skipped based on loglevel. |
2794 | * | |
2830eec1 JO |
2795 | * Returns false if no record is available. Otherwise true and all fields |
2796 | * of @pmsg are valid. (See the documentation of struct printk_message | |
2797 | * for information about the @pmsg fields.) | |
a8199371 | 2798 | */ |
06653d57 TG |
2799 | bool printk_get_next_message(struct printk_message *pmsg, u64 seq, |
2800 | bool is_extended, bool may_suppress) | |
a8199371 | 2801 | { |
2d9ef940 | 2802 | static int panic_console_dropped; |
daaab5b5 | 2803 | |
2830eec1 JO |
2804 | struct printk_buffers *pbufs = pmsg->pbufs; |
2805 | const size_t scratchbuf_sz = sizeof(pbufs->scratchbuf); | |
2806 | const size_t outbuf_sz = sizeof(pbufs->outbuf); | |
2807 | char *scratchbuf = &pbufs->scratchbuf[0]; | |
2808 | char *outbuf = &pbufs->outbuf[0]; | |
a699449b JO |
2809 | struct printk_info info; |
2810 | struct printk_record r; | |
2830eec1 | 2811 | size_t len = 0; |
a699449b | 2812 | |
daaab5b5 JO |
2813 | /* |
2814 | * Formatting extended messages requires a separate buffer, so use the | |
2815 | * scratch buffer to read in the ringbuffer text. | |
2816 | * | |
2817 | * Formatting normal messages is done in-place, so read the ringbuffer | |
2818 | * text directly into the output buffer. | |
2819 | */ | |
2820 | if (is_extended) | |
2821 | prb_rec_init_rd(&r, &info, scratchbuf, scratchbuf_sz); | |
2822 | else | |
2823 | prb_rec_init_rd(&r, &info, outbuf, outbuf_sz); | |
a699449b | 2824 | |
2830eec1 | 2825 | if (!prb_read_valid(prb, seq, &r)) |
a699449b JO |
2826 | return false; |
2827 | ||
2830eec1 JO |
2828 | pmsg->seq = r.info->seq; |
2829 | pmsg->dropped = r.info->seq - seq; | |
2830 | ||
2831 | /* | |
2832 | * Check for dropped messages in panic here so that printk | |
2833 | * suppression can occur as early as possible if necessary. | |
2834 | */ | |
2835 | if (pmsg->dropped && | |
2836 | panic_in_progress() && | |
2837 | panic_console_dropped++ > 10) { | |
2838 | suppress_panic_printk = 1; | |
2839 | pr_warn_once("Too many dropped messages. Suppress messages on non-panic CPUs to prevent livelock.\n"); | |
a699449b JO |
2840 | } |
2841 | ||
2842 | /* Skip record that has level above the console loglevel. */ | |
ea308da1 | 2843 | if (may_suppress && suppress_message_printing(r.info->level)) |
2830eec1 | 2844 | goto out; |
a699449b | 2845 | |
daaab5b5 JO |
2846 | if (is_extended) { |
2847 | len = info_print_ext_header(outbuf, outbuf_sz, r.info); | |
2848 | len += msg_print_ext_body(outbuf + len, outbuf_sz - len, | |
a699449b JO |
2849 | &r.text_buf[0], r.info->text_len, &r.info->dev_info); |
2850 | } else { | |
a699449b JO |
2851 | len = record_print_text(&r, console_msg_format & MSG_FORMAT_SYSLOG, printk_time); |
2852 | } | |
2830eec1 JO |
2853 | out: |
2854 | pmsg->outbuf_len = len; | |
2855 | return true; | |
2856 | } | |
2857 | ||
d818b56f JO |
2858 | /* |
2859 | * Used as the printk buffers for non-panic, serialized console printing. | |
2860 | * This is for legacy (!CON_NBCON) as well as all boot (CON_BOOT) consoles. | |
2861 | * Its usage requires the console_lock held. | |
2862 | */ | |
2863 | struct printk_buffers printk_shared_pbufs; | |
2864 | ||
2830eec1 JO |
2865 | /* |
2866 | * Print one record for the given console. The record printed is whatever | |
2867 | * record is the next available record for the given console. | |
03a749e6 | 2868 | * |
a699449b | 2869 | * @handover will be set to true if a printk waiter has taken over the |
fc956ae0 JO |
2870 | * console_lock, in which case the caller is no longer holding both the |
2871 | * console_lock and the SRCU read lock. Otherwise it is set to false. | |
2872 | * | |
2873 | * @cookie is the cookie from the SRCU read lock. | |
a699449b JO |
2874 | * |
2875 | * Returns false if the given console has no next record to print, otherwise | |
2876 | * true. | |
a8199371 | 2877 | * |
fc956ae0 | 2878 | * Requires the console_lock and the SRCU read lock. |
a8199371 | 2879 | */ |
2830eec1 | 2880 | static bool console_emit_next_record(struct console *con, bool *handover, int cookie) |
a8199371 | 2881 | { |
2830eec1 | 2882 | bool is_extended = console_srcu_read_flags(con) & CON_EXTENDED; |
d818b56f | 2883 | char *outbuf = &printk_shared_pbufs.outbuf[0]; |
2830eec1 | 2884 | struct printk_message pmsg = { |
d818b56f | 2885 | .pbufs = &printk_shared_pbufs, |
2830eec1 JO |
2886 | }; |
2887 | unsigned long flags; | |
a699449b | 2888 | |
2d9ef940 | 2889 | *handover = false; |
a699449b | 2890 | |
ea308da1 | 2891 | if (!printk_get_next_message(&pmsg, con->seq, is_extended, true)) |
a699449b JO |
2892 | return false; |
2893 | ||
2830eec1 | 2894 | con->dropped += pmsg.dropped; |
a699449b | 2895 | |
2830eec1 JO |
2896 | /* Skip messages of formatted length 0. */ |
2897 | if (pmsg.outbuf_len == 0) { | |
2898 | con->seq = pmsg.seq + 1; | |
a699449b JO |
2899 | goto skip; |
2900 | } | |
2901 | ||
c4fcc617 JO |
2902 | if (con->dropped && !is_extended) { |
2903 | console_prepend_dropped(&pmsg, con->dropped); | |
2904 | con->dropped = 0; | |
a699449b JO |
2905 | } |
2906 | ||
2d9ef940 PM |
2907 | /* |
2908 | * While actively printing out messages, if another printk() | |
2909 | * were to occur on another CPU, it may wait for this one to | |
2910 | * finish. This task can not be preempted if there is a | |
2911 | * waiter waiting to take over. | |
2912 | * | |
2913 | * Interrupts are disabled because the hand over to a waiter | |
2914 | * must not be interrupted until the hand over is completed | |
2915 | * (@console_waiter is cleared). | |
2916 | */ | |
2917 | printk_safe_enter_irqsave(flags); | |
2918 | console_lock_spinning_enable(); | |
a699449b | 2919 | |
c4fcc617 JO |
2920 | /* Do not trace print latency. */ |
2921 | stop_critical_timings(); | |
2922 | ||
2923 | /* Write everything out to the hardware. */ | |
2924 | con->write(con, outbuf, pmsg.outbuf_len); | |
2925 | ||
2d9ef940 | 2926 | start_critical_timings(); |
a699449b | 2927 | |
2830eec1 | 2928 | con->seq = pmsg.seq + 1; |
a699449b | 2929 | |
fc956ae0 | 2930 | *handover = console_lock_spinning_disable_and_check(cookie); |
2d9ef940 | 2931 | printk_safe_exit_irqrestore(flags); |
a699449b JO |
2932 | skip: |
2933 | return true; | |
2934 | } | |
2935 | ||
98a04655 JO |
2936 | #else |
2937 | ||
2938 | static bool console_emit_next_record(struct console *con, bool *handover, int cookie) | |
2939 | { | |
2940 | *handover = false; | |
2941 | return false; | |
2942 | } | |
2943 | ||
2944 | #endif /* CONFIG_PRINTK */ | |
2945 | ||
a699449b JO |
2946 | /* |
2947 | * Print out all remaining records to all consoles. | |
2948 | * | |
2949 | * @do_cond_resched is set by the caller. It can be true only in schedulable | |
2950 | * context. | |
2951 | * | |
2952 | * @next_seq is set to the sequence number after the last available record. | |
2953 | * The value is valid only when this function returns true. It means that all | |
2954 | * usable consoles are completely flushed. | |
2955 | * | |
2956 | * @handover will be set to true if a printk waiter has taken over the | |
2957 | * console_lock, in which case the caller is no longer holding the | |
2958 | * console_lock. Otherwise it is set to false. | |
2959 | * | |
2960 | * Returns true when there was at least one usable console and all messages | |
2961 | * were flushed to all usable consoles. A returned false informs the caller | |
2962 | * that everything was not flushed (either there were no usable consoles or | |
2963 | * another context has taken over printing or it is a panic situation and this | |
5831788a PM |
2964 | * is not the panic CPU). Regardless the reason, the caller should assume it |
2965 | * is not useful to immediately try again. | |
a699449b JO |
2966 | * |
2967 | * Requires the console_lock. | |
2968 | */ | |
2969 | static bool console_flush_all(bool do_cond_resched, u64 *next_seq, bool *handover) | |
2970 | { | |
2971 | bool any_usable = false; | |
2972 | struct console *con; | |
2973 | bool any_progress; | |
fc956ae0 | 2974 | int cookie; |
a699449b JO |
2975 | |
2976 | *next_seq = 0; | |
2977 | *handover = false; | |
2978 | ||
2979 | do { | |
2980 | any_progress = false; | |
2981 | ||
fc956ae0 JO |
2982 | cookie = console_srcu_read_lock(); |
2983 | for_each_console_srcu(con) { | |
a699449b JO |
2984 | bool progress; |
2985 | ||
2986 | if (!console_is_usable(con)) | |
2987 | continue; | |
2988 | any_usable = true; | |
2989 | ||
daaab5b5 | 2990 | progress = console_emit_next_record(con, handover, cookie); |
fc956ae0 JO |
2991 | |
2992 | /* | |
2993 | * If a handover has occurred, the SRCU read lock | |
2994 | * is already released. | |
2995 | */ | |
a699449b JO |
2996 | if (*handover) |
2997 | return false; | |
2998 | ||
2999 | /* Track the next of the highest seq flushed. */ | |
3000 | if (con->seq > *next_seq) | |
3001 | *next_seq = con->seq; | |
3002 | ||
3003 | if (!progress) | |
3004 | continue; | |
3005 | any_progress = true; | |
3006 | ||
3007 | /* Allow panic_cpu to take over the consoles safely. */ | |
132a90d1 | 3008 | if (other_cpu_in_panic()) |
fc956ae0 | 3009 | goto abandon; |
a699449b JO |
3010 | |
3011 | if (do_cond_resched) | |
3012 | cond_resched(); | |
3013 | } | |
fc956ae0 | 3014 | console_srcu_read_unlock(cookie); |
a699449b JO |
3015 | } while (any_progress); |
3016 | ||
3017 | return any_usable; | |
fc956ae0 JO |
3018 | |
3019 | abandon: | |
3020 | console_srcu_read_unlock(cookie); | |
3021 | return false; | |
a8199371 SS |
3022 | } |
3023 | ||
1da177e4 | 3024 | /** |
848a9c10 | 3025 | * console_unlock - unblock the console subsystem from printing |
1da177e4 | 3026 | * |
848a9c10 JO |
3027 | * Releases the console_lock which the caller holds to block printing of |
3028 | * the console subsystem. | |
1da177e4 | 3029 | * |
ac751efa TH |
3030 | * While the console_lock was held, console output may have been buffered |
3031 | * by printk(). If this is the case, console_unlock(); emits | |
3032 | * the output prior to releasing the lock. | |
1da177e4 | 3033 | * |
ac751efa | 3034 | * console_unlock(); may be called from any context. |
1da177e4 | 3035 | */ |
ac751efa | 3036 | void console_unlock(void) |
1da177e4 | 3037 | { |
a699449b JO |
3038 | bool do_cond_resched; |
3039 | bool handover; | |
3040 | bool flushed; | |
3041 | u64 next_seq; | |
1da177e4 | 3042 | |
8d91f8b1 | 3043 | /* |
257ab443 | 3044 | * Console drivers are called with interrupts disabled, so |
8d91f8b1 TH |
3045 | * @console_may_schedule should be cleared before; however, we may |
3046 | * end up dumping a lot of lines, for example, if called from | |
3047 | * console registration path, and should invoke cond_resched() | |
3048 | * between lines if allowable. Not doing so can cause a very long | |
3049 | * scheduling stall on a slow console leading to RCU stall and | |
3050 | * softlockup warnings which exacerbate the issue with more | |
a699449b JO |
3051 | * messages practically incapacitating the system. Therefore, create |
3052 | * a local to use for the printing loop. | |
8d91f8b1 TH |
3053 | */ |
3054 | do_cond_resched = console_may_schedule; | |
78944e54 | 3055 | |
a699449b JO |
3056 | do { |
3057 | console_may_schedule = 0; | |
649e6ee3 | 3058 | |
a699449b JO |
3059 | flushed = console_flush_all(do_cond_resched, &next_seq, &handover); |
3060 | if (!handover) | |
3061 | __console_unlock(); | |
f92b070f | 3062 | |
896fbe20 | 3063 | /* |
a699449b JO |
3064 | * Abort if there was a failure to flush all messages to all |
3065 | * usable consoles. Either it is not possible to flush (in | |
3066 | * which case it would be an infinite loop of retrying) or | |
3067 | * another context has taken over printing. | |
896fbe20 | 3068 | */ |
a699449b JO |
3069 | if (!flushed) |
3070 | break; | |
7ff9554b | 3071 | |
dbdda842 | 3072 | /* |
a699449b JO |
3073 | * Some context may have added new records after |
3074 | * console_flush_all() but before unlocking the console. | |
3075 | * Re-check if there is a new record to flush. If the trylock | |
3076 | * fails, another context is already handling the printing. | |
dbdda842 | 3077 | */ |
a699449b | 3078 | } while (prb_read_valid(prb, next_seq, NULL) && console_trylock()); |
1da177e4 | 3079 | } |
ac751efa | 3080 | EXPORT_SYMBOL(console_unlock); |
1da177e4 | 3081 | |
ddad86c2 MW |
3082 | /** |
3083 | * console_conditional_schedule - yield the CPU if required | |
1da177e4 LT |
3084 | * |
3085 | * If the console code is currently allowed to sleep, and | |
3086 | * if this CPU should yield the CPU to another task, do | |
3087 | * so here. | |
3088 | * | |
ac751efa | 3089 | * Must be called within console_lock();. |
1da177e4 LT |
3090 | */ |
3091 | void __sched console_conditional_schedule(void) | |
3092 | { | |
3093 | if (console_may_schedule) | |
3094 | cond_resched(); | |
3095 | } | |
3096 | EXPORT_SYMBOL(console_conditional_schedule); | |
3097 | ||
1da177e4 LT |
3098 | void console_unblank(void) |
3099 | { | |
7b23a66d | 3100 | bool found_unblank = false; |
1da177e4 | 3101 | struct console *c; |
d792db6f | 3102 | int cookie; |
1da177e4 | 3103 | |
7b23a66d JO |
3104 | /* |
3105 | * First check if there are any consoles implementing the unblank() | |
3106 | * callback. If not, there is no reason to continue and take the | |
3107 | * console lock, which in particular can be dangerous if | |
3108 | * @oops_in_progress is set. | |
3109 | */ | |
3110 | cookie = console_srcu_read_lock(); | |
3111 | for_each_console_srcu(c) { | |
3112 | if ((console_srcu_read_flags(c) & CON_ENABLED) && c->unblank) { | |
3113 | found_unblank = true; | |
3114 | break; | |
3115 | } | |
3116 | } | |
3117 | console_srcu_read_unlock(cookie); | |
3118 | if (!found_unblank) | |
3119 | return; | |
3120 | ||
1da177e4 | 3121 | /* |
d792db6f JO |
3122 | * Stop console printing because the unblank() callback may |
3123 | * assume the console is not within its write() callback. | |
3124 | * | |
3125 | * If @oops_in_progress is set, this may be an atomic context. | |
3126 | * In that case, attempt a trylock as best-effort. | |
1da177e4 LT |
3127 | */ |
3128 | if (oops_in_progress) { | |
7b23a66d JO |
3129 | /* Semaphores are not NMI-safe. */ |
3130 | if (in_nmi()) | |
3131 | return; | |
3132 | ||
3133 | /* | |
3134 | * Attempting to trylock the console lock can deadlock | |
3135 | * if another CPU was stopped while modifying the | |
3136 | * semaphore. "Hope and pray" that this is not the | |
3137 | * current situation. | |
3138 | */ | |
bd8d7cf5 | 3139 | if (down_trylock_console_sem() != 0) |
1da177e4 LT |
3140 | return; |
3141 | } else | |
ac751efa | 3142 | console_lock(); |
1da177e4 | 3143 | |
007eeab7 | 3144 | console_locked = 1; |
1da177e4 | 3145 | console_may_schedule = 0; |
d792db6f JO |
3146 | |
3147 | cookie = console_srcu_read_lock(); | |
3148 | for_each_console_srcu(c) { | |
3149 | if ((console_srcu_read_flags(c) & CON_ENABLED) && c->unblank) | |
1da177e4 | 3150 | c->unblank(); |
d792db6f JO |
3151 | } |
3152 | console_srcu_read_unlock(cookie); | |
3153 | ||
ac751efa | 3154 | console_unlock(); |
3b604ca8 JO |
3155 | |
3156 | if (!oops_in_progress) | |
3157 | pr_flush(1000, true); | |
1da177e4 | 3158 | } |
1da177e4 | 3159 | |
8d91f8b1 TH |
3160 | /** |
3161 | * console_flush_on_panic - flush console content on panic | |
de6da1e8 | 3162 | * @mode: flush all messages in buffer or just the pending ones |
8d91f8b1 TH |
3163 | * |
3164 | * Immediately output all pending messages no matter what. | |
3165 | */ | |
de6da1e8 | 3166 | void console_flush_on_panic(enum con_flush_mode mode) |
8d91f8b1 | 3167 | { |
eacb04ff JO |
3168 | bool handover; |
3169 | u64 next_seq; | |
3170 | ||
8d91f8b1 | 3171 | /* |
eacb04ff JO |
3172 | * Ignore the console lock and flush out the messages. Attempting a |
3173 | * trylock would not be useful because: | |
3174 | * | |
3175 | * - if it is contended, it must be ignored anyway | |
3176 | * - console_lock() and console_trylock() block and fail | |
3177 | * respectively in panic for non-panic CPUs | |
3178 | * - semaphores are not NMI-safe | |
3179 | */ | |
3180 | ||
3181 | /* | |
3182 | * If another context is holding the console lock, | |
3183 | * @console_may_schedule might be set. Clear it so that | |
3184 | * this context does not call cond_resched() while flushing. | |
8d91f8b1 | 3185 | */ |
8d91f8b1 | 3186 | console_may_schedule = 0; |
de6da1e8 | 3187 | |
a699449b JO |
3188 | if (mode == CONSOLE_REPLAY_ALL) { |
3189 | struct console *c; | |
ad56ebd1 | 3190 | short flags; |
87f2e4b7 | 3191 | int cookie; |
a699449b JO |
3192 | u64 seq; |
3193 | ||
3194 | seq = prb_first_valid_seq(prb); | |
87f2e4b7 JO |
3195 | |
3196 | cookie = console_srcu_read_lock(); | |
3197 | for_each_console_srcu(c) { | |
ad56ebd1 TG |
3198 | flags = console_srcu_read_flags(c); |
3199 | ||
3200 | if (flags & CON_NBCON) { | |
3201 | nbcon_seq_force(c, seq); | |
3202 | } else { | |
3203 | /* | |
3204 | * This is an unsynchronized assignment. On | |
3205 | * panic legacy consoles are only best effort. | |
3206 | */ | |
3207 | c->seq = seq; | |
3208 | } | |
87f2e4b7 JO |
3209 | } |
3210 | console_srcu_read_unlock(cookie); | |
a699449b | 3211 | } |
eacb04ff JO |
3212 | |
3213 | console_flush_all(false, &next_seq, &handover); | |
8d91f8b1 TH |
3214 | } |
3215 | ||
1da177e4 LT |
3216 | /* |
3217 | * Return the console tty driver structure and its associated index | |
3218 | */ | |
3219 | struct tty_driver *console_device(int *index) | |
3220 | { | |
3221 | struct console *c; | |
3222 | struct tty_driver *driver = NULL; | |
8cb15f7f | 3223 | int cookie; |
1da177e4 | 3224 | |
8cb15f7f JO |
3225 | /* |
3226 | * Take console_lock to serialize device() callback with | |
3227 | * other console operations. For example, fg_console is | |
3228 | * modified under console_lock when switching vt. | |
3229 | */ | |
ac751efa | 3230 | console_lock(); |
8cb15f7f JO |
3231 | |
3232 | cookie = console_srcu_read_lock(); | |
3233 | for_each_console_srcu(c) { | |
1da177e4 LT |
3234 | if (!c->device) |
3235 | continue; | |
3236 | driver = c->device(c, index); | |
3237 | if (driver) | |
3238 | break; | |
3239 | } | |
8cb15f7f JO |
3240 | console_srcu_read_unlock(cookie); |
3241 | ||
ac751efa | 3242 | console_unlock(); |
1da177e4 LT |
3243 | return driver; |
3244 | } | |
3245 | ||
3246 | /* | |
3247 | * Prevent further output on the passed console device so that (for example) | |
3248 | * serial drivers can disable console output before suspending a port, and can | |
3249 | * re-enable output afterwards. | |
3250 | */ | |
3251 | void console_stop(struct console *console) | |
3252 | { | |
3b604ca8 | 3253 | __pr_flush(console, 1000, true); |
4dc64682 | 3254 | console_list_lock(); |
100bdef2 | 3255 | console_srcu_write_flags(console, console->flags & ~CON_ENABLED); |
4dc64682 | 3256 | console_list_unlock(); |
6c4afa79 JO |
3257 | |
3258 | /* | |
3259 | * Ensure that all SRCU list walks have completed. All contexts must | |
3260 | * be able to see that this console is disabled so that (for example) | |
3261 | * the caller can suspend the port without risk of another context | |
3262 | * using the port. | |
3263 | */ | |
3264 | synchronize_srcu(&console_srcu); | |
1da177e4 LT |
3265 | } |
3266 | EXPORT_SYMBOL(console_stop); | |
3267 | ||
3268 | void console_start(struct console *console) | |
3269 | { | |
4dc64682 | 3270 | console_list_lock(); |
100bdef2 | 3271 | console_srcu_write_flags(console, console->flags | CON_ENABLED); |
4dc64682 | 3272 | console_list_unlock(); |
3b604ca8 | 3273 | __pr_flush(console, 1000, true); |
1da177e4 LT |
3274 | } |
3275 | EXPORT_SYMBOL(console_start); | |
3276 | ||
7bf69395 FDN |
3277 | static int __read_mostly keep_bootcon; |
3278 | ||
3279 | static int __init keep_bootcon_setup(char *str) | |
3280 | { | |
3281 | keep_bootcon = 1; | |
27083bac | 3282 | pr_info("debug: skip boot console de-registration.\n"); |
7bf69395 FDN |
3283 | |
3284 | return 0; | |
3285 | } | |
3286 | ||
3287 | early_param("keep_bootcon", keep_bootcon_setup); | |
3288 | ||
ad8cd1db BH |
3289 | /* |
3290 | * This is called by register_console() to try to match | |
3291 | * the newly registered console with any of the ones selected | |
3292 | * by either the command line or add_preferred_console() and | |
3293 | * setup/enable it. | |
3294 | * | |
3295 | * Care need to be taken with consoles that are statically | |
3296 | * enabled such as netconsole | |
3297 | */ | |
ed758b30 PM |
3298 | static int try_enable_preferred_console(struct console *newcon, |
3299 | bool user_specified) | |
ad8cd1db BH |
3300 | { |
3301 | struct console_cmdline *c; | |
bba18a1a | 3302 | int i, err; |
ad8cd1db BH |
3303 | |
3304 | for (i = 0, c = console_cmdline; | |
3305 | i < MAX_CMDLINECONSOLES && c->name[0]; | |
3306 | i++, c++) { | |
e369d822 BH |
3307 | if (c->user_specified != user_specified) |
3308 | continue; | |
ad8cd1db BH |
3309 | if (!newcon->match || |
3310 | newcon->match(newcon, c->name, c->index, c->options) != 0) { | |
3311 | /* default matching */ | |
3312 | BUILD_BUG_ON(sizeof(c->name) != sizeof(newcon->name)); | |
3313 | if (strcmp(c->name, newcon->name) != 0) | |
3314 | continue; | |
3315 | if (newcon->index >= 0 && | |
3316 | newcon->index != c->index) | |
3317 | continue; | |
3318 | if (newcon->index < 0) | |
3319 | newcon->index = c->index; | |
3320 | ||
3321 | if (_braille_register_console(newcon, c)) | |
3322 | return 0; | |
3323 | ||
3324 | if (newcon->setup && | |
bba18a1a AS |
3325 | (err = newcon->setup(newcon, c->options)) != 0) |
3326 | return err; | |
ad8cd1db BH |
3327 | } |
3328 | newcon->flags |= CON_ENABLED; | |
f873efe8 | 3329 | if (i == preferred_console) |
ad8cd1db | 3330 | newcon->flags |= CON_CONSDEV; |
ad8cd1db BH |
3331 | return 0; |
3332 | } | |
3333 | ||
3334 | /* | |
3335 | * Some consoles, such as pstore and netconsole, can be enabled even | |
e369d822 | 3336 | * without matching. Accept the pre-enabled consoles only when match() |
50460376 | 3337 | * and setup() had a chance to be called. |
ad8cd1db | 3338 | */ |
e369d822 | 3339 | if (newcon->flags & CON_ENABLED && c->user_specified == user_specified) |
ad8cd1db BH |
3340 | return 0; |
3341 | ||
3342 | return -ENOENT; | |
3343 | } | |
3344 | ||
ed758b30 PM |
3345 | /* Try to enable the console unconditionally */ |
3346 | static void try_enable_default_console(struct console *newcon) | |
3347 | { | |
3348 | if (newcon->index < 0) | |
3349 | newcon->index = 0; | |
3350 | ||
3351 | if (newcon->setup && newcon->setup(newcon, NULL) != 0) | |
3352 | return; | |
3353 | ||
3354 | newcon->flags |= CON_ENABLED; | |
3355 | ||
4f546939 | 3356 | if (newcon->device) |
ed758b30 | 3357 | newcon->flags |= CON_CONSDEV; |
ed758b30 PM |
3358 | } |
3359 | ||
a4242760 | 3360 | static void console_init_seq(struct console *newcon, bool bootcon_registered) |
b80ea0e8 | 3361 | { |
a4242760 JO |
3362 | struct console *con; |
3363 | bool handover; | |
3364 | ||
3365 | if (newcon->flags & (CON_PRINTBUFFER | CON_BOOT)) { | |
b80ea0e8 JO |
3366 | /* Get a consistent copy of @syslog_seq. */ |
3367 | mutex_lock(&syslog_lock); | |
3368 | newcon->seq = syslog_seq; | |
3369 | mutex_unlock(&syslog_lock); | |
3370 | } else { | |
a4242760 | 3371 | /* Begin with next message added to ringbuffer. */ |
b80ea0e8 | 3372 | newcon->seq = prb_next_seq(prb); |
a4242760 JO |
3373 | |
3374 | /* | |
3375 | * If any enabled boot consoles are due to be unregistered | |
3376 | * shortly, some may not be caught up and may be the same | |
3377 | * device as @newcon. Since it is not known which boot console | |
3378 | * is the same device, flush all consoles and, if necessary, | |
3379 | * start with the message of the enabled boot console that is | |
3380 | * the furthest behind. | |
3381 | */ | |
3382 | if (bootcon_registered && !keep_bootcon) { | |
848a9c10 JO |
3383 | /* |
3384 | * Hold the console_lock to stop console printing and | |
3385 | * guarantee safe access to console->seq. | |
3386 | */ | |
3387 | console_lock(); | |
3388 | ||
a4242760 JO |
3389 | /* |
3390 | * Flush all consoles and set the console to start at | |
3391 | * the next unprinted sequence number. | |
3392 | */ | |
3393 | if (!console_flush_all(true, &newcon->seq, &handover)) { | |
3394 | /* | |
3395 | * Flushing failed. Just choose the lowest | |
3396 | * sequence of the enabled boot consoles. | |
3397 | */ | |
3398 | ||
3399 | /* | |
3400 | * If there was a handover, this context no | |
3401 | * longer holds the console_lock. | |
3402 | */ | |
3403 | if (handover) | |
3404 | console_lock(); | |
3405 | ||
3406 | newcon->seq = prb_next_seq(prb); | |
3407 | for_each_console(con) { | |
3408 | if ((con->flags & CON_BOOT) && | |
3409 | (con->flags & CON_ENABLED) && | |
3410 | con->seq < newcon->seq) { | |
3411 | newcon->seq = con->seq; | |
3412 | } | |
3413 | } | |
3414 | } | |
848a9c10 JO |
3415 | |
3416 | console_unlock(); | |
a4242760 | 3417 | } |
b80ea0e8 JO |
3418 | } |
3419 | } | |
3420 | ||
d9a4af56 TG |
3421 | #define console_first() \ |
3422 | hlist_entry(console_list.first, struct console, node) | |
3423 | ||
4dc64682 JO |
3424 | static int unregister_console_locked(struct console *console); |
3425 | ||
1da177e4 LT |
3426 | /* |
3427 | * The console driver calls this routine during kernel initialization | |
3428 | * to register the console printing procedure with printk() and to | |
3429 | * print any messages that were printed by the kernel before the | |
3430 | * console driver was initialized. | |
4d091611 RG |
3431 | * |
3432 | * This can happen pretty early during the boot process (because of | |
3433 | * early_printk) - sometimes before setup_arch() completes - be careful | |
3434 | * of what kernel features are used - they may not be initialised yet. | |
3435 | * | |
3436 | * There are two types of consoles - bootconsoles (early_printk) and | |
3437 | * "real" consoles (everything which is not a bootconsole) which are | |
3438 | * handled differently. | |
3439 | * - Any number of bootconsoles can be registered at any time. | |
3440 | * - As soon as a "real" console is registered, all bootconsoles | |
3441 | * will be unregistered automatically. | |
3442 | * - Once a "real" console is registered, any attempt to register a | |
3443 | * bootconsoles will be rejected | |
1da177e4 | 3444 | */ |
4d091611 | 3445 | void register_console(struct console *newcon) |
1da177e4 | 3446 | { |
5e8ba485 | 3447 | struct console *con; |
11457036 JO |
3448 | bool bootcon_registered = false; |
3449 | bool realcon_registered = false; | |
ad8cd1db | 3450 | int err; |
1da177e4 | 3451 | |
4dc64682 JO |
3452 | console_list_lock(); |
3453 | ||
5e8ba485 PM |
3454 | for_each_console(con) { |
3455 | if (WARN(con == newcon, "console '%s%d' already registered\n", | |
4dc64682 JO |
3456 | con->name, con->index)) { |
3457 | goto unlock; | |
3458 | } | |
16cf48a6 | 3459 | |
5e8ba485 | 3460 | if (con->flags & CON_BOOT) |
11457036 | 3461 | bootcon_registered = true; |
5e8ba485 | 3462 | else |
11457036 | 3463 | realcon_registered = true; |
69331af7 GH |
3464 | } |
3465 | ||
5e8ba485 | 3466 | /* Do not register boot consoles when there already is a real one. */ |
11457036 | 3467 | if ((newcon->flags & CON_BOOT) && realcon_registered) { |
5e8ba485 PM |
3468 | pr_info("Too late to register bootconsole %s%d\n", |
3469 | newcon->name, newcon->index); | |
4dc64682 | 3470 | goto unlock; |
5e8ba485 | 3471 | } |
4d091611 | 3472 | |
5634c90f TG |
3473 | if (newcon->flags & CON_NBCON) { |
3474 | /* | |
3475 | * Ensure the nbcon console buffers can be allocated | |
3476 | * before modifying any global data. | |
3477 | */ | |
3478 | if (!nbcon_alloc(newcon)) | |
3479 | goto unlock; | |
3480 | } | |
3481 | ||
1da177e4 | 3482 | /* |
4f546939 PM |
3483 | * See if we want to enable this console driver by default. |
3484 | * | |
3485 | * Nope when a console is preferred by the command line, device | |
3486 | * tree, or SPCR. | |
3487 | * | |
3488 | * The first real console with tty binding (driver) wins. More | |
3489 | * consoles might get enabled before the right one is found. | |
3490 | * | |
3491 | * Note that a console with tty binding will have CON_CONSDEV | |
3492 | * flag set and will be first in the list. | |
1da177e4 | 3493 | */ |
4f546939 | 3494 | if (preferred_console < 0) { |
d9a4af56 TG |
3495 | if (hlist_empty(&console_list) || !console_first()->device || |
3496 | console_first()->flags & CON_BOOT) { | |
4f546939 PM |
3497 | try_enable_default_console(newcon); |
3498 | } | |
3499 | } | |
1da177e4 | 3500 | |
e369d822 | 3501 | /* See if this console matches one we selected on the command line */ |
ed758b30 | 3502 | err = try_enable_preferred_console(newcon, true); |
bbeddf52 | 3503 | |
e369d822 BH |
3504 | /* If not, try to match against the platform default(s) */ |
3505 | if (err == -ENOENT) | |
ed758b30 | 3506 | err = try_enable_preferred_console(newcon, false); |
1da177e4 | 3507 | |
ad8cd1db | 3508 | /* printk() messages are not printed to the Braille console. */ |
5634c90f TG |
3509 | if (err || newcon->flags & CON_BRL) { |
3510 | if (newcon->flags & CON_NBCON) | |
3511 | nbcon_free(newcon); | |
4dc64682 | 3512 | goto unlock; |
5634c90f | 3513 | } |
1da177e4 | 3514 | |
8259cf43 RG |
3515 | /* |
3516 | * If we have a bootconsole, and are switching to a real console, | |
3517 | * don't print everything out again, since when the boot console, and | |
3518 | * the real console are the same physical device, it's annoying to | |
3519 | * see the beginning boot messages twice | |
3520 | */ | |
11457036 | 3521 | if (bootcon_registered && |
5e8ba485 | 3522 | ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV)) { |
4d091611 | 3523 | newcon->flags &= ~CON_PRINTBUFFER; |
5e8ba485 | 3524 | } |
1da177e4 | 3525 | |
6c4afa79 | 3526 | newcon->dropped = 0; |
a4242760 | 3527 | console_init_seq(newcon, bootcon_registered); |
6c4afa79 | 3528 | |
6b93bb41 TG |
3529 | if (newcon->flags & CON_NBCON) |
3530 | nbcon_init(newcon); | |
3531 | ||
1da177e4 | 3532 | /* |
d9a4af56 TG |
3533 | * Put this console in the list - keep the |
3534 | * preferred driver at the head of the list. | |
1da177e4 | 3535 | */ |
d9a4af56 TG |
3536 | if (hlist_empty(&console_list)) { |
3537 | /* Ensure CON_CONSDEV is always set for the head. */ | |
33225d7b | 3538 | newcon->flags |= CON_CONSDEV; |
6c4afa79 | 3539 | hlist_add_head_rcu(&newcon->node, &console_list); |
d9a4af56 TG |
3540 | |
3541 | } else if (newcon->flags & CON_CONSDEV) { | |
3542 | /* Only the new head can have CON_CONSDEV set. */ | |
100bdef2 | 3543 | console_srcu_write_flags(console_first(), console_first()->flags & ~CON_CONSDEV); |
6c4afa79 | 3544 | hlist_add_head_rcu(&newcon->node, &console_list); |
6fe29354 | 3545 | |
a699449b | 3546 | } else { |
6c4afa79 | 3547 | hlist_add_behind_rcu(&newcon->node, console_list.first); |
1da177e4 | 3548 | } |
6c4afa79 JO |
3549 | |
3550 | /* | |
3551 | * No need to synchronize SRCU here! The caller does not rely | |
3552 | * on all contexts being able to see the new console before | |
3553 | * register_console() completes. | |
3554 | */ | |
3555 | ||
fbc92a34 | 3556 | console_sysfs_notify(); |
8259cf43 RG |
3557 | |
3558 | /* | |
3559 | * By unregistering the bootconsoles after we enable the real console | |
3560 | * we get the "console xxx enabled" message on all the consoles - | |
3561 | * boot consoles, real consoles, etc - this is to ensure that end | |
3562 | * users know there might be something in the kernel's log buffer that | |
3563 | * went to the bootconsole (that they do not see on the real console) | |
3564 | */ | |
1fc0ca9e | 3565 | con_printk(KERN_INFO, newcon, "enabled\n"); |
11457036 | 3566 | if (bootcon_registered && |
7bf69395 FDN |
3567 | ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) && |
3568 | !keep_bootcon) { | |
d9a4af56 TG |
3569 | struct hlist_node *tmp; |
3570 | ||
3571 | hlist_for_each_entry_safe(con, tmp, &console_list, node) { | |
5e8ba485 | 3572 | if (con->flags & CON_BOOT) |
4dc64682 | 3573 | unregister_console_locked(con); |
d9a4af56 | 3574 | } |
8259cf43 | 3575 | } |
4dc64682 JO |
3576 | unlock: |
3577 | console_list_unlock(); | |
1da177e4 LT |
3578 | } |
3579 | EXPORT_SYMBOL(register_console); | |
3580 | ||
4dc64682 JO |
3581 | /* Must be called under console_list_lock(). */ |
3582 | static int unregister_console_locked(struct console *console) | |
1da177e4 | 3583 | { |
bbeddf52 | 3584 | int res; |
1da177e4 | 3585 | |
4dc64682 JO |
3586 | lockdep_assert_console_list_lock_held(); |
3587 | ||
1fc0ca9e | 3588 | con_printk(KERN_INFO, console, "disabled\n"); |
6b802394 | 3589 | |
bbeddf52 | 3590 | res = _braille_unregister_console(console); |
bb72e398 | 3591 | if (res < 0) |
bbeddf52 | 3592 | return res; |
bb72e398 AS |
3593 | if (res > 0) |
3594 | return 0; | |
f7511d5f | 3595 | |
d9a4af56 | 3596 | /* Disable it unconditionally */ |
100bdef2 | 3597 | console_srcu_write_flags(console, console->flags & ~CON_ENABLED); |
d9a4af56 | 3598 | |
848a9c10 | 3599 | if (!console_is_registered_locked(console)) |
d9a4af56 | 3600 | return -ENODEV; |
40dc5651 | 3601 | |
6c4afa79 | 3602 | hlist_del_init_rcu(&console->node); |
e78bedbd | 3603 | |
69331af7 | 3604 | /* |
d9a4af56 | 3605 | * <HISTORICAL> |
ab4af03a GE |
3606 | * If this isn't the last console and it has CON_CONSDEV set, we |
3607 | * need to set it on the next preferred console. | |
d9a4af56 TG |
3608 | * </HISTORICAL> |
3609 | * | |
3610 | * The above makes no sense as there is no guarantee that the next | |
3611 | * console has any device attached. Oh well.... | |
1da177e4 | 3612 | */ |
d9a4af56 | 3613 | if (!hlist_empty(&console_list) && console->flags & CON_CONSDEV) |
100bdef2 | 3614 | console_srcu_write_flags(console_first(), console_first()->flags | CON_CONSDEV); |
1da177e4 | 3615 | |
6c4afa79 JO |
3616 | /* |
3617 | * Ensure that all SRCU list walks have completed. All contexts | |
3618 | * must not be able to see this console in the list so that any | |
3619 | * exit/cleanup routines can be performed safely. | |
3620 | */ | |
3621 | synchronize_srcu(&console_srcu); | |
1da177e4 | 3622 | |
6b93bb41 | 3623 | if (console->flags & CON_NBCON) |
5634c90f | 3624 | nbcon_free(console); |
6b93bb41 | 3625 | |
fbc92a34 | 3626 | console_sysfs_notify(); |
e78bedbd | 3627 | |
ed31685c AS |
3628 | if (console->exit) |
3629 | res = console->exit(console); | |
3630 | ||
e78bedbd | 3631 | return res; |
1da177e4 | 3632 | } |
e78bedbd | 3633 | |
4dc64682 JO |
3634 | int unregister_console(struct console *console) |
3635 | { | |
3636 | int res; | |
e78bedbd | 3637 | |
4dc64682 JO |
3638 | console_list_lock(); |
3639 | res = unregister_console_locked(console); | |
3640 | console_list_unlock(); | |
1da177e4 LT |
3641 | return res; |
3642 | } | |
3643 | EXPORT_SYMBOL(unregister_console); | |
d59745ce | 3644 | |
6f883675 JO |
3645 | /** |
3646 | * console_force_preferred_locked - force a registered console preferred | |
3647 | * @con: The registered console to force preferred. | |
3648 | * | |
3649 | * Must be called under console_list_lock(). | |
3650 | */ | |
3651 | void console_force_preferred_locked(struct console *con) | |
3652 | { | |
3653 | struct console *cur_pref_con; | |
3654 | ||
3655 | if (!console_is_registered_locked(con)) | |
3656 | return; | |
3657 | ||
3658 | cur_pref_con = console_first(); | |
3659 | ||
3660 | /* Already preferred? */ | |
3661 | if (cur_pref_con == con) | |
3662 | return; | |
3663 | ||
3664 | /* | |
3665 | * Delete, but do not re-initialize the entry. This allows the console | |
3666 | * to continue to appear registered (via any hlist_unhashed_lockless() | |
3667 | * checks), even though it was briefly removed from the console list. | |
3668 | */ | |
3669 | hlist_del_rcu(&con->node); | |
3670 | ||
3671 | /* | |
3672 | * Ensure that all SRCU list walks have completed so that the console | |
3673 | * can be added to the beginning of the console list and its forward | |
3674 | * list pointer can be re-initialized. | |
3675 | */ | |
3676 | synchronize_srcu(&console_srcu); | |
3677 | ||
3678 | con->flags |= CON_CONSDEV; | |
3679 | WARN_ON(!con->device); | |
3680 | ||
3681 | /* Only the new head can have CON_CONSDEV set. */ | |
3682 | console_srcu_write_flags(cur_pref_con, cur_pref_con->flags & ~CON_CONSDEV); | |
3683 | hlist_add_head_rcu(&con->node, &console_list); | |
3684 | } | |
3685 | EXPORT_SYMBOL(console_force_preferred_locked); | |
3686 | ||
0c688614 NP |
3687 | /* |
3688 | * Initialize the console device. This is called *early*, so | |
3689 | * we can't necessarily depend on lots of kernel help here. | |
3690 | * Just do some early initializations, and do the complex setup | |
3691 | * later. | |
3692 | */ | |
3693 | void __init console_init(void) | |
3694 | { | |
58eacfff | 3695 | int ret; |
1b1eeca7 AB |
3696 | initcall_t call; |
3697 | initcall_entry_t *ce; | |
0c688614 NP |
3698 | |
3699 | /* Setup the default TTY line discipline. */ | |
3700 | n_tty_init(); | |
3701 | ||
3702 | /* | |
3703 | * set up the console device so that later boot sequences can | |
3704 | * inform about problems etc.. | |
3705 | */ | |
1b1eeca7 | 3706 | ce = __con_initcall_start; |
58eacfff | 3707 | trace_initcall_level("console"); |
1b1eeca7 AB |
3708 | while (ce < __con_initcall_end) { |
3709 | call = initcall_from_entry(ce); | |
3710 | trace_initcall_start(call); | |
3711 | ret = call(); | |
3712 | trace_initcall_finish(call, ret); | |
3713 | ce++; | |
0c688614 NP |
3714 | } |
3715 | } | |
3716 | ||
81cc26f2 TR |
3717 | /* |
3718 | * Some boot consoles access data that is in the init section and which will | |
3719 | * be discarded after the initcalls have been run. To make sure that no code | |
3720 | * will access this data, unregister the boot consoles in a late initcall. | |
3721 | * | |
3722 | * If for some reason, such as deferred probe or the driver being a loadable | |
3723 | * module, the real console hasn't registered yet at this point, there will | |
3724 | * be a brief interval in which no messages are logged to the console, which | |
3725 | * makes it difficult to diagnose problems that occur during this time. | |
3726 | * | |
3727 | * To mitigate this problem somewhat, only unregister consoles whose memory | |
2b1be689 | 3728 | * intersects with the init section. Note that all other boot consoles will |
acebb559 | 3729 | * get unregistered when the real preferred console is registered. |
81cc26f2 | 3730 | */ |
034260d6 | 3731 | static int __init printk_late_init(void) |
0c5564bd | 3732 | { |
d9a4af56 | 3733 | struct hlist_node *tmp; |
4d091611 | 3734 | struct console *con; |
90b14889 | 3735 | int ret; |
4d091611 | 3736 | |
4dc64682 | 3737 | console_list_lock(); |
d9a4af56 | 3738 | hlist_for_each_entry_safe(con, tmp, &console_list, node) { |
5a814231 PM |
3739 | if (!(con->flags & CON_BOOT)) |
3740 | continue; | |
3741 | ||
3742 | /* Check addresses that might be used for enabled consoles. */ | |
3743 | if (init_section_intersects(con, sizeof(*con)) || | |
3744 | init_section_contains(con->write, 0) || | |
3745 | init_section_contains(con->read, 0) || | |
3746 | init_section_contains(con->device, 0) || | |
3747 | init_section_contains(con->unblank, 0) || | |
3748 | init_section_contains(con->data, 0)) { | |
81cc26f2 | 3749 | /* |
2b1be689 MR |
3750 | * Please, consider moving the reported consoles out |
3751 | * of the init section. | |
81cc26f2 | 3752 | */ |
2b1be689 MR |
3753 | pr_warn("bootconsole [%s%d] uses init memory and must be disabled even before the real one is ready\n", |
3754 | con->name, con->index); | |
4dc64682 | 3755 | unregister_console_locked(con); |
cb00e99c | 3756 | } |
0c5564bd | 3757 | } |
4dc64682 | 3758 | console_list_unlock(); |
d9a4af56 | 3759 | |
90b14889 SAS |
3760 | ret = cpuhp_setup_state_nocalls(CPUHP_PRINTK_DEAD, "printk:dead", NULL, |
3761 | console_cpu_notify); | |
3762 | WARN_ON(ret < 0); | |
3763 | ret = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "printk:online", | |
3764 | console_cpu_notify, NULL); | |
3765 | WARN_ON(ret < 0); | |
faaa357a | 3766 | printk_sysctl_init(); |
0c5564bd RG |
3767 | return 0; |
3768 | } | |
034260d6 | 3769 | late_initcall(printk_late_init); |
0c5564bd | 3770 | |
7ef3d2fd | 3771 | #if defined CONFIG_PRINTK |
3b604ca8 JO |
3772 | /* If @con is specified, only wait for that console. Otherwise wait for all. */ |
3773 | static bool __pr_flush(struct console *con, int timeout_ms, bool reset_on_progress) | |
3774 | { | |
29fda1ad PM |
3775 | unsigned long timeout_jiffies = msecs_to_jiffies(timeout_ms); |
3776 | unsigned long remaining_jiffies = timeout_jiffies; | |
3b604ca8 JO |
3777 | struct console *c; |
3778 | u64 last_diff = 0; | |
3779 | u64 printk_seq; | |
ad56ebd1 | 3780 | short flags; |
eb7f1ed2 | 3781 | int cookie; |
3b604ca8 JO |
3782 | u64 diff; |
3783 | u64 seq; | |
3784 | ||
3785 | might_sleep(); | |
3786 | ||
ac7d7844 | 3787 | seq = prb_next_reserve_seq(prb); |
3b604ca8 | 3788 | |
054c22bd JO |
3789 | /* Flush the consoles so that records up to @seq are printed. */ |
3790 | console_lock(); | |
3791 | console_unlock(); | |
3792 | ||
3b604ca8 | 3793 | for (;;) { |
29fda1ad PM |
3794 | unsigned long begin_jiffies; |
3795 | unsigned long slept_jiffies; | |
3796 | ||
3b604ca8 JO |
3797 | diff = 0; |
3798 | ||
eb7f1ed2 JO |
3799 | /* |
3800 | * Hold the console_lock to guarantee safe access to | |
054c22bd JO |
3801 | * console->seq. Releasing console_lock flushes more |
3802 | * records in case @seq is still not printed on all | |
3803 | * usable consoles. | |
eb7f1ed2 | 3804 | */ |
3b604ca8 | 3805 | console_lock(); |
9023ca08 | 3806 | |
eb7f1ed2 JO |
3807 | cookie = console_srcu_read_lock(); |
3808 | for_each_console_srcu(c) { | |
3b604ca8 JO |
3809 | if (con && con != c) |
3810 | continue; | |
ad56ebd1 TG |
3811 | |
3812 | flags = console_srcu_read_flags(c); | |
3813 | ||
9e70a5e1 JO |
3814 | /* |
3815 | * If consoles are not usable, it cannot be expected | |
3816 | * that they make forward progress, so only increment | |
3817 | * @diff for usable consoles. | |
3818 | */ | |
3b604ca8 JO |
3819 | if (!console_is_usable(c)) |
3820 | continue; | |
ad56ebd1 TG |
3821 | |
3822 | if (flags & CON_NBCON) { | |
3823 | printk_seq = nbcon_seq_read(c); | |
3824 | } else { | |
3825 | printk_seq = c->seq; | |
3826 | } | |
3827 | ||
3b604ca8 JO |
3828 | if (printk_seq < seq) |
3829 | diff += seq - printk_seq; | |
3830 | } | |
eb7f1ed2 | 3831 | console_srcu_read_unlock(cookie); |
3b604ca8 | 3832 | |
9e70a5e1 | 3833 | if (diff != last_diff && reset_on_progress) |
29fda1ad | 3834 | remaining_jiffies = timeout_jiffies; |
3b604ca8 | 3835 | |
9023ca08 JO |
3836 | console_unlock(); |
3837 | ||
9e70a5e1 | 3838 | /* Note: @diff is 0 if there are no usable consoles. */ |
29fda1ad | 3839 | if (diff == 0 || remaining_jiffies == 0) |
3b604ca8 JO |
3840 | break; |
3841 | ||
29fda1ad PM |
3842 | /* msleep(1) might sleep much longer. Check time by jiffies. */ |
3843 | begin_jiffies = jiffies; | |
3844 | msleep(1); | |
3845 | slept_jiffies = jiffies - begin_jiffies; | |
3846 | ||
3847 | remaining_jiffies -= min(slept_jiffies, remaining_jiffies); | |
3b604ca8 JO |
3848 | |
3849 | last_diff = diff; | |
3850 | } | |
3851 | ||
3852 | return (diff == 0); | |
3853 | } | |
3854 | ||
3855 | /** | |
3856 | * pr_flush() - Wait for printing threads to catch up. | |
3857 | * | |
3858 | * @timeout_ms: The maximum time (in ms) to wait. | |
3859 | * @reset_on_progress: Reset the timeout if forward progress is seen. | |
3860 | * | |
3861 | * A value of 0 for @timeout_ms means no waiting will occur. A value of -1 | |
3862 | * represents infinite waiting. | |
3863 | * | |
3864 | * If @reset_on_progress is true, the timeout will be reset whenever any | |
3865 | * printer has been seen to make some forward progress. | |
3866 | * | |
3867 | * Context: Process context. May sleep while acquiring console lock. | |
9e70a5e1 | 3868 | * Return: true if all usable printers are caught up. |
3b604ca8 | 3869 | */ |
c60ba2d3 | 3870 | static bool pr_flush(int timeout_ms, bool reset_on_progress) |
3b604ca8 JO |
3871 | { |
3872 | return __pr_flush(NULL, timeout_ms, reset_on_progress); | |
3873 | } | |
3b604ca8 | 3874 | |
dc72c32e FW |
3875 | /* |
3876 | * Delayed printk version, for scheduler-internal messages: | |
3877 | */ | |
5831788a PM |
3878 | #define PRINTK_PENDING_WAKEUP 0x01 |
3879 | #define PRINTK_PENDING_OUTPUT 0x02 | |
dc72c32e FW |
3880 | |
3881 | static DEFINE_PER_CPU(int, printk_pending); | |
dc72c32e FW |
3882 | |
3883 | static void wake_up_klogd_work_func(struct irq_work *irq_work) | |
3884 | { | |
2ba3673d | 3885 | int pending = this_cpu_xchg(printk_pending, 0); |
dc72c32e | 3886 | |
5831788a | 3887 | if (pending & PRINTK_PENDING_OUTPUT) { |
458df9fd SR |
3888 | /* If trylock fails, someone else is doing the printing */ |
3889 | if (console_trylock()) | |
3890 | console_unlock(); | |
dc72c32e FW |
3891 | } |
3892 | ||
3893 | if (pending & PRINTK_PENDING_WAKEUP) | |
809631e2 | 3894 | wake_up_interruptible(&log_wait); |
dc72c32e FW |
3895 | } |
3896 | ||
7a9f50a0 PZ |
3897 | static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = |
3898 | IRQ_WORK_INIT_LAZY(wake_up_klogd_work_func); | |
dc72c32e | 3899 | |
5341b93d | 3900 | static void __wake_up_klogd(int val) |
dc72c32e | 3901 | { |
ab6f762f SS |
3902 | if (!printk_percpu_data_ready()) |
3903 | return; | |
3904 | ||
dc72c32e | 3905 | preempt_disable(); |
1f5d7830 JO |
3906 | /* |
3907 | * Guarantee any new records can be seen by tasks preparing to wait | |
3908 | * before this context checks if the wait queue is empty. | |
3909 | * | |
3910 | * The full memory barrier within wq_has_sleeper() pairs with the full | |
3911 | * memory barrier within set_current_state() of | |
3912 | * prepare_to_wait_event(), which is called after ___wait_event() adds | |
3913 | * the waiter but before it has checked the wait condition. | |
3914 | * | |
5831788a | 3915 | * This pairs with devkmsg_read:A and syslog_print:A. |
1f5d7830 | 3916 | */ |
5341b93d | 3917 | if (wq_has_sleeper(&log_wait) || /* LMM(__wake_up_klogd:A) */ |
5831788a | 3918 | (val & PRINTK_PENDING_OUTPUT)) { |
5341b93d | 3919 | this_cpu_or(printk_pending, val); |
bb964a92 | 3920 | irq_work_queue(this_cpu_ptr(&wake_up_klogd_work)); |
dc72c32e FW |
3921 | } |
3922 | preempt_enable(); | |
3923 | } | |
717115e1 | 3924 | |
696ffaf5 JO |
3925 | /** |
3926 | * wake_up_klogd - Wake kernel logging daemon | |
3927 | * | |
3928 | * Use this function when new records have been added to the ringbuffer | |
3929 | * and the console printing of those records has already occurred or is | |
3930 | * known to be handled by some other context. This function will only | |
3931 | * wake the logging daemon. | |
3932 | * | |
3933 | * Context: Any context. | |
3934 | */ | |
5341b93d | 3935 | void wake_up_klogd(void) |
600e1458 | 3936 | { |
5341b93d JO |
3937 | __wake_up_klogd(PRINTK_PENDING_WAKEUP); |
3938 | } | |
ab6f762f | 3939 | |
696ffaf5 JO |
3940 | /** |
3941 | * defer_console_output - Wake kernel logging daemon and trigger | |
3942 | * console printing in a deferred context | |
3943 | * | |
3944 | * Use this function when new records have been added to the ringbuffer, | |
3945 | * this context is responsible for console printing those records, but | |
3946 | * the current context is not allowed to perform the console printing. | |
3947 | * Trigger an irq_work context to perform the console printing. This | |
3948 | * function also wakes the logging daemon. | |
3949 | * | |
3950 | * Context: Any context. | |
3951 | */ | |
5341b93d JO |
3952 | void defer_console_output(void) |
3953 | { | |
3954 | /* | |
3955 | * New messages may have been added directly to the ringbuffer | |
3956 | * using vprintk_store(), so wake any waiters as well. | |
3957 | */ | |
5831788a | 3958 | __wake_up_klogd(PRINTK_PENDING_WAKEUP | PRINTK_PENDING_OUTPUT); |
a338f84d PM |
3959 | } |
3960 | ||
5d5e4522 NP |
3961 | void printk_trigger_flush(void) |
3962 | { | |
3963 | defer_console_output(); | |
3964 | } | |
3965 | ||
a338f84d PM |
3966 | int vprintk_deferred(const char *fmt, va_list args) |
3967 | { | |
696ffaf5 | 3968 | return vprintk_emit(0, LOGLEVEL_SCHED, NULL, fmt, args); |
600e1458 PZ |
3969 | } |
3970 | ||
33701557 | 3971 | int _printk_deferred(const char *fmt, ...) |
719f6a70 PM |
3972 | { |
3973 | va_list args; | |
3974 | int r; | |
3975 | ||
3976 | va_start(args, fmt); | |
3977 | r = vprintk_deferred(fmt, args); | |
3978 | va_end(args); | |
3979 | ||
3980 | return r; | |
3981 | } | |
3982 | ||
1da177e4 LT |
3983 | /* |
3984 | * printk rate limiting, lifted from the networking subsystem. | |
3985 | * | |
641de9d8 UKK |
3986 | * This enforces a rate limit: not more than 10 kernel messages |
3987 | * every 5s to make a denial-of-service attack impossible. | |
1da177e4 | 3988 | */ |
641de9d8 UKK |
3989 | DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10); |
3990 | ||
5c828713 | 3991 | int __printk_ratelimit(const char *func) |
1da177e4 | 3992 | { |
5c828713 | 3993 | return ___ratelimit(&printk_ratelimit_state, func); |
1da177e4 | 3994 | } |
5c828713 | 3995 | EXPORT_SYMBOL(__printk_ratelimit); |
f46c4833 AM |
3996 | |
3997 | /** | |
3998 | * printk_timed_ratelimit - caller-controlled printk ratelimiting | |
3999 | * @caller_jiffies: pointer to caller's state | |
4000 | * @interval_msecs: minimum interval between prints | |
4001 | * | |
4002 | * printk_timed_ratelimit() returns true if more than @interval_msecs | |
4003 | * milliseconds have elapsed since the last time printk_timed_ratelimit() | |
4004 | * returned true. | |
4005 | */ | |
4006 | bool printk_timed_ratelimit(unsigned long *caller_jiffies, | |
4007 | unsigned int interval_msecs) | |
4008 | { | |
249771b8 AE |
4009 | unsigned long elapsed = jiffies - *caller_jiffies; |
4010 | ||
4011 | if (*caller_jiffies && elapsed <= msecs_to_jiffies(interval_msecs)) | |
4012 | return false; | |
4013 | ||
4014 | *caller_jiffies = jiffies; | |
4015 | return true; | |
f46c4833 AM |
4016 | } |
4017 | EXPORT_SYMBOL(printk_timed_ratelimit); | |
456b565c SK |
4018 | |
4019 | static DEFINE_SPINLOCK(dump_list_lock); | |
4020 | static LIST_HEAD(dump_list); | |
4021 | ||
4022 | /** | |
4023 | * kmsg_dump_register - register a kernel log dumper. | |
6485536b | 4024 | * @dumper: pointer to the kmsg_dumper structure |
456b565c SK |
4025 | * |
4026 | * Adds a kernel log dumper to the system. The dump callback in the | |
4027 | * structure will be called when the kernel oopses or panics and must be | |
4028 | * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise. | |
4029 | */ | |
4030 | int kmsg_dump_register(struct kmsg_dumper *dumper) | |
4031 | { | |
4032 | unsigned long flags; | |
4033 | int err = -EBUSY; | |
4034 | ||
4035 | /* The dump callback needs to be set */ | |
4036 | if (!dumper->dump) | |
4037 | return -EINVAL; | |
4038 | ||
4039 | spin_lock_irqsave(&dump_list_lock, flags); | |
4040 | /* Don't allow registering multiple times */ | |
4041 | if (!dumper->registered) { | |
4042 | dumper->registered = 1; | |
fb842b00 | 4043 | list_add_tail_rcu(&dumper->list, &dump_list); |
456b565c SK |
4044 | err = 0; |
4045 | } | |
4046 | spin_unlock_irqrestore(&dump_list_lock, flags); | |
4047 | ||
4048 | return err; | |
4049 | } | |
4050 | EXPORT_SYMBOL_GPL(kmsg_dump_register); | |
4051 | ||
4052 | /** | |
4053 | * kmsg_dump_unregister - unregister a kmsg dumper. | |
6485536b | 4054 | * @dumper: pointer to the kmsg_dumper structure |
456b565c SK |
4055 | * |
4056 | * Removes a dump device from the system. Returns zero on success and | |
4057 | * %-EINVAL otherwise. | |
4058 | */ | |
4059 | int kmsg_dump_unregister(struct kmsg_dumper *dumper) | |
4060 | { | |
4061 | unsigned long flags; | |
4062 | int err = -EINVAL; | |
4063 | ||
4064 | spin_lock_irqsave(&dump_list_lock, flags); | |
4065 | if (dumper->registered) { | |
4066 | dumper->registered = 0; | |
fb842b00 | 4067 | list_del_rcu(&dumper->list); |
456b565c SK |
4068 | err = 0; |
4069 | } | |
4070 | spin_unlock_irqrestore(&dump_list_lock, flags); | |
fb842b00 | 4071 | synchronize_rcu(); |
456b565c SK |
4072 | |
4073 | return err; | |
4074 | } | |
4075 | EXPORT_SYMBOL_GPL(kmsg_dump_unregister); | |
4076 | ||
7ff9554b KS |
4077 | static bool always_kmsg_dump; |
4078 | module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR); | |
4079 | ||
fb13cb8a KC |
4080 | const char *kmsg_dump_reason_str(enum kmsg_dump_reason reason) |
4081 | { | |
4082 | switch (reason) { | |
4083 | case KMSG_DUMP_PANIC: | |
4084 | return "Panic"; | |
4085 | case KMSG_DUMP_OOPS: | |
4086 | return "Oops"; | |
4087 | case KMSG_DUMP_EMERG: | |
4088 | return "Emergency"; | |
4089 | case KMSG_DUMP_SHUTDOWN: | |
4090 | return "Shutdown"; | |
4091 | default: | |
4092 | return "Unknown"; | |
4093 | } | |
4094 | } | |
4095 | EXPORT_SYMBOL_GPL(kmsg_dump_reason_str); | |
4096 | ||
456b565c SK |
4097 | /** |
4098 | * kmsg_dump - dump kernel log to kernel message dumpers. | |
4099 | * @reason: the reason (oops, panic etc) for dumping | |
4100 | * | |
e2ae715d KS |
4101 | * Call each of the registered dumper's dump() callback, which can |
4102 | * retrieve the kmsg records with kmsg_dump_get_line() or | |
4103 | * kmsg_dump_get_buffer(). | |
456b565c SK |
4104 | */ |
4105 | void kmsg_dump(enum kmsg_dump_reason reason) | |
4106 | { | |
456b565c | 4107 | struct kmsg_dumper *dumper; |
456b565c | 4108 | |
e2ae715d KS |
4109 | rcu_read_lock(); |
4110 | list_for_each_entry_rcu(dumper, &dump_list, list) { | |
b1f6f161 PT |
4111 | enum kmsg_dump_reason max_reason = dumper->max_reason; |
4112 | ||
4113 | /* | |
4114 | * If client has not provided a specific max_reason, default | |
4115 | * to KMSG_DUMP_OOPS, unless always_kmsg_dump was set. | |
4116 | */ | |
4117 | if (max_reason == KMSG_DUMP_UNDEF) { | |
4118 | max_reason = always_kmsg_dump ? KMSG_DUMP_MAX : | |
4119 | KMSG_DUMP_OOPS; | |
4120 | } | |
4121 | if (reason > max_reason) | |
e2ae715d KS |
4122 | continue; |
4123 | ||
e2ae715d KS |
4124 | /* invoke dumper which will iterate over records */ |
4125 | dumper->dump(dumper, reason); | |
e2ae715d KS |
4126 | } |
4127 | rcu_read_unlock(); | |
4128 | } | |
4129 | ||
4130 | /** | |
a4f98765 | 4131 | * kmsg_dump_get_line - retrieve one kmsg log line |
f9f3f02d | 4132 | * @iter: kmsg dump iterator |
e2ae715d KS |
4133 | * @syslog: include the "<4>" prefixes |
4134 | * @line: buffer to copy the line to | |
4135 | * @size: maximum size of the buffer | |
4136 | * @len: length of line placed into buffer | |
4137 | * | |
4138 | * Start at the beginning of the kmsg buffer, with the oldest kmsg | |
4139 | * record, and copy one record into the provided buffer. | |
4140 | * | |
4141 | * Consecutive calls will return the next available record moving | |
4142 | * towards the end of the buffer with the youngest messages. | |
4143 | * | |
4144 | * A return value of FALSE indicates that there are no more records to | |
4145 | * read. | |
4146 | */ | |
a4f98765 JO |
4147 | bool kmsg_dump_get_line(struct kmsg_dump_iter *iter, bool syslog, |
4148 | char *line, size_t size, size_t *len) | |
e2ae715d | 4149 | { |
f9f3f02d | 4150 | u64 min_seq = latched_seq_read_nolock(&clear_seq); |
896fbe20 JO |
4151 | struct printk_info info; |
4152 | unsigned int line_count; | |
4153 | struct printk_record r; | |
e2ae715d KS |
4154 | size_t l = 0; |
4155 | bool ret = false; | |
4156 | ||
f9f3f02d JO |
4157 | if (iter->cur_seq < min_seq) |
4158 | iter->cur_seq = min_seq; | |
4159 | ||
f35efc78 | 4160 | prb_rec_init_rd(&r, &info, line, size); |
896fbe20 | 4161 | |
896fbe20 JO |
4162 | /* Read text or count text lines? */ |
4163 | if (line) { | |
f9f3f02d | 4164 | if (!prb_read_valid(prb, iter->cur_seq, &r)) |
896fbe20 JO |
4165 | goto out; |
4166 | l = record_print_text(&r, syslog, printk_time); | |
4167 | } else { | |
f9f3f02d | 4168 | if (!prb_read_valid_info(prb, iter->cur_seq, |
896fbe20 JO |
4169 | &info, &line_count)) { |
4170 | goto out; | |
4171 | } | |
4172 | l = get_record_print_text_size(&info, line_count, syslog, | |
4173 | printk_time); | |
456b565c | 4174 | |
896fbe20 | 4175 | } |
e2ae715d | 4176 | |
f9f3f02d | 4177 | iter->cur_seq = r.info->seq + 1; |
e2ae715d | 4178 | ret = true; |
e2ae715d KS |
4179 | out: |
4180 | if (len) | |
4181 | *len = l; | |
4182 | return ret; | |
4183 | } | |
4184 | EXPORT_SYMBOL_GPL(kmsg_dump_get_line); | |
4185 | ||
4186 | /** | |
4187 | * kmsg_dump_get_buffer - copy kmsg log lines | |
f9f3f02d | 4188 | * @iter: kmsg dump iterator |
e2ae715d | 4189 | * @syslog: include the "<4>" prefixes |
4f0f4af5 | 4190 | * @buf: buffer to copy the line to |
e2ae715d | 4191 | * @size: maximum size of the buffer |
726b5097 | 4192 | * @len_out: length of line placed into buffer |
e2ae715d KS |
4193 | * |
4194 | * Start at the end of the kmsg buffer and fill the provided buffer | |
547bbf7d | 4195 | * with as many of the *youngest* kmsg records that fit into it. |
e2ae715d KS |
4196 | * If the buffer is large enough, all available kmsg records will be |
4197 | * copied with a single call. | |
4198 | * | |
4199 | * Consecutive calls will fill the buffer with the next block of | |
4200 | * available older records, not including the earlier retrieved ones. | |
4201 | * | |
4202 | * A return value of FALSE indicates that there are no more records to | |
4203 | * read. | |
4204 | */ | |
f9f3f02d | 4205 | bool kmsg_dump_get_buffer(struct kmsg_dump_iter *iter, bool syslog, |
726b5097 | 4206 | char *buf, size_t size, size_t *len_out) |
e2ae715d | 4207 | { |
f9f3f02d | 4208 | u64 min_seq = latched_seq_read_nolock(&clear_seq); |
896fbe20 | 4209 | struct printk_info info; |
896fbe20 | 4210 | struct printk_record r; |
e2ae715d | 4211 | u64 seq; |
e2ae715d | 4212 | u64 next_seq; |
726b5097 | 4213 | size_t len = 0; |
e2ae715d | 4214 | bool ret = false; |
e80c1a9d | 4215 | bool time = printk_time; |
e2ae715d | 4216 | |
5f6c7648 | 4217 | if (!buf || !size) |
e2ae715d KS |
4218 | goto out; |
4219 | ||
f9f3f02d JO |
4220 | if (iter->cur_seq < min_seq) |
4221 | iter->cur_seq = min_seq; | |
4222 | ||
f9f3f02d JO |
4223 | if (prb_read_valid_info(prb, iter->cur_seq, &info, NULL)) { |
4224 | if (info.seq != iter->cur_seq) { | |
13791c80 | 4225 | /* messages are gone, move to first available one */ |
f9f3f02d | 4226 | iter->cur_seq = info.seq; |
13791c80 | 4227 | } |
e2ae715d KS |
4228 | } |
4229 | ||
4230 | /* last entry */ | |
93d102f0 | 4231 | if (iter->cur_seq >= iter->next_seq) |
e2ae715d | 4232 | goto out; |
e2ae715d | 4233 | |
726b5097 JO |
4234 | /* |
4235 | * Find first record that fits, including all following records, | |
4260e0e5 JO |
4236 | * into the user-provided buffer for this dump. Pass in size-1 |
4237 | * because this function (by way of record_print_text()) will | |
4238 | * not write more than size-1 bytes of text into @buf. | |
726b5097 | 4239 | */ |
f9f3f02d | 4240 | seq = find_first_fitting_seq(iter->cur_seq, iter->next_seq, |
4260e0e5 | 4241 | size - 1, syslog, time); |
e2ae715d | 4242 | |
726b5097 JO |
4243 | /* |
4244 | * Next kmsg_dump_get_buffer() invocation will dump block of | |
4245 | * older records stored right before this one. | |
4246 | */ | |
e2ae715d | 4247 | next_seq = seq; |
e2ae715d | 4248 | |
726b5097 JO |
4249 | prb_rec_init_rd(&r, &info, buf, size); |
4250 | ||
726b5097 | 4251 | prb_for_each_record(seq, prb, seq, &r) { |
f9f3f02d | 4252 | if (r.info->seq >= iter->next_seq) |
896fbe20 JO |
4253 | break; |
4254 | ||
726b5097 | 4255 | len += record_print_text(&r, syslog, time); |
896fbe20 | 4256 | |
726b5097 JO |
4257 | /* Adjust record to store to remaining buffer space. */ |
4258 | prb_rec_init_rd(&r, &info, buf + len, size - len); | |
e2ae715d KS |
4259 | } |
4260 | ||
f9f3f02d | 4261 | iter->next_seq = next_seq; |
e2ae715d | 4262 | ret = true; |
e2ae715d | 4263 | out: |
726b5097 JO |
4264 | if (len_out) |
4265 | *len_out = len; | |
e2ae715d KS |
4266 | return ret; |
4267 | } | |
4268 | EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer); | |
456b565c | 4269 | |
e2ae715d | 4270 | /** |
325606af | 4271 | * kmsg_dump_rewind - reset the iterator |
f9f3f02d | 4272 | * @iter: kmsg dump iterator |
e2ae715d KS |
4273 | * |
4274 | * Reset the dumper's iterator so that kmsg_dump_get_line() and | |
4275 | * kmsg_dump_get_buffer() can be called again and used multiple | |
4276 | * times within the same dumper.dump() callback. | |
4277 | */ | |
f9f3f02d | 4278 | void kmsg_dump_rewind(struct kmsg_dump_iter *iter) |
e2ae715d | 4279 | { |
a4f98765 JO |
4280 | iter->cur_seq = latched_seq_read_nolock(&clear_seq); |
4281 | iter->next_seq = prb_next_seq(prb); | |
456b565c | 4282 | } |
e2ae715d | 4283 | EXPORT_SYMBOL_GPL(kmsg_dump_rewind); |
196779b9 | 4284 | |
7ef3d2fd | 4285 | #endif |
766c268b JO |
4286 | |
4287 | #ifdef CONFIG_SMP | |
faebd693 JO |
4288 | static atomic_t printk_cpu_sync_owner = ATOMIC_INIT(-1); |
4289 | static atomic_t printk_cpu_sync_nested = ATOMIC_INIT(0); | |
766c268b JO |
4290 | |
4291 | /** | |
faebd693 JO |
4292 | * __printk_cpu_sync_wait() - Busy wait until the printk cpu-reentrant |
4293 | * spinning lock is not owned by any CPU. | |
766c268b JO |
4294 | * |
4295 | * Context: Any context. | |
4296 | */ | |
faebd693 | 4297 | void __printk_cpu_sync_wait(void) |
766c268b JO |
4298 | { |
4299 | do { | |
4300 | cpu_relax(); | |
faebd693 | 4301 | } while (atomic_read(&printk_cpu_sync_owner) != -1); |
766c268b | 4302 | } |
faebd693 | 4303 | EXPORT_SYMBOL(__printk_cpu_sync_wait); |
766c268b JO |
4304 | |
4305 | /** | |
faebd693 JO |
4306 | * __printk_cpu_sync_try_get() - Try to acquire the printk cpu-reentrant |
4307 | * spinning lock. | |
766c268b JO |
4308 | * |
4309 | * If no processor has the lock, the calling processor takes the lock and | |
4310 | * becomes the owner. If the calling processor is already the owner of the | |
4311 | * lock, this function succeeds immediately. | |
4312 | * | |
4313 | * Context: Any context. Expects interrupts to be disabled. | |
4314 | * Return: 1 on success, otherwise 0. | |
4315 | */ | |
faebd693 | 4316 | int __printk_cpu_sync_try_get(void) |
766c268b JO |
4317 | { |
4318 | int cpu; | |
4319 | int old; | |
4320 | ||
4321 | cpu = smp_processor_id(); | |
4322 | ||
3342aa8e JO |
4323 | /* |
4324 | * Guarantee loads and stores from this CPU when it is the lock owner | |
4325 | * are _not_ visible to the previous lock owner. This pairs with | |
faebd693 | 4326 | * __printk_cpu_sync_put:B. |
3342aa8e JO |
4327 | * |
4328 | * Memory barrier involvement: | |
4329 | * | |
faebd693 JO |
4330 | * If __printk_cpu_sync_try_get:A reads from __printk_cpu_sync_put:B, |
4331 | * then __printk_cpu_sync_put:A can never read from | |
4332 | * __printk_cpu_sync_try_get:B. | |
3342aa8e JO |
4333 | * |
4334 | * Relies on: | |
4335 | * | |
faebd693 | 4336 | * RELEASE from __printk_cpu_sync_put:A to __printk_cpu_sync_put:B |
3342aa8e JO |
4337 | * of the previous CPU |
4338 | * matching | |
faebd693 JO |
4339 | * ACQUIRE from __printk_cpu_sync_try_get:A to |
4340 | * __printk_cpu_sync_try_get:B of this CPU | |
3342aa8e | 4341 | */ |
faebd693 JO |
4342 | old = atomic_cmpxchg_acquire(&printk_cpu_sync_owner, -1, |
4343 | cpu); /* LMM(__printk_cpu_sync_try_get:A) */ | |
766c268b | 4344 | if (old == -1) { |
3342aa8e JO |
4345 | /* |
4346 | * This CPU is now the owner and begins loading/storing | |
faebd693 | 4347 | * data: LMM(__printk_cpu_sync_try_get:B) |
3342aa8e | 4348 | */ |
766c268b | 4349 | return 1; |
3342aa8e | 4350 | |
766c268b JO |
4351 | } else if (old == cpu) { |
4352 | /* This CPU is already the owner. */ | |
faebd693 | 4353 | atomic_inc(&printk_cpu_sync_nested); |
766c268b JO |
4354 | return 1; |
4355 | } | |
4356 | ||
4357 | return 0; | |
4358 | } | |
faebd693 | 4359 | EXPORT_SYMBOL(__printk_cpu_sync_try_get); |
766c268b JO |
4360 | |
4361 | /** | |
faebd693 | 4362 | * __printk_cpu_sync_put() - Release the printk cpu-reentrant spinning lock. |
766c268b JO |
4363 | * |
4364 | * The calling processor must be the owner of the lock. | |
4365 | * | |
4366 | * Context: Any context. Expects interrupts to be disabled. | |
4367 | */ | |
faebd693 | 4368 | void __printk_cpu_sync_put(void) |
766c268b | 4369 | { |
faebd693 JO |
4370 | if (atomic_read(&printk_cpu_sync_nested)) { |
4371 | atomic_dec(&printk_cpu_sync_nested); | |
766c268b JO |
4372 | return; |
4373 | } | |
4374 | ||
3342aa8e JO |
4375 | /* |
4376 | * This CPU is finished loading/storing data: | |
faebd693 | 4377 | * LMM(__printk_cpu_sync_put:A) |
3342aa8e JO |
4378 | */ |
4379 | ||
4380 | /* | |
4381 | * Guarantee loads and stores from this CPU when it was the | |
4382 | * lock owner are visible to the next lock owner. This pairs | |
faebd693 | 4383 | * with __printk_cpu_sync_try_get:A. |
3342aa8e JO |
4384 | * |
4385 | * Memory barrier involvement: | |
4386 | * | |
faebd693 JO |
4387 | * If __printk_cpu_sync_try_get:A reads from __printk_cpu_sync_put:B, |
4388 | * then __printk_cpu_sync_try_get:B reads from __printk_cpu_sync_put:A. | |
3342aa8e JO |
4389 | * |
4390 | * Relies on: | |
4391 | * | |
faebd693 | 4392 | * RELEASE from __printk_cpu_sync_put:A to __printk_cpu_sync_put:B |
3342aa8e JO |
4393 | * of this CPU |
4394 | * matching | |
faebd693 JO |
4395 | * ACQUIRE from __printk_cpu_sync_try_get:A to |
4396 | * __printk_cpu_sync_try_get:B of the next CPU | |
3342aa8e | 4397 | */ |
faebd693 JO |
4398 | atomic_set_release(&printk_cpu_sync_owner, |
4399 | -1); /* LMM(__printk_cpu_sync_put:B) */ | |
766c268b | 4400 | } |
faebd693 | 4401 | EXPORT_SYMBOL(__printk_cpu_sync_put); |
766c268b | 4402 | #endif /* CONFIG_SMP */ |