Root/kernel/printk.c

1/*
2 * linux/kernel/printk.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * Modified to make sys_syslog() more flexible: added commands to
7 * return the last 4k of kernel messages, regardless of whether
8 * they've been read or not. Added option to suppress kernel printk's
9 * to the console. Added hook for sending the console messages
10 * elsewhere, in preparation for a serial line console (someday).
11 * Ted Ts'o, 2/11/93.
12 * Modified for sysctl support, 1/8/97, Chris Horn.
13 * Fixed SMP synchronization, 08/08/99, Manfred Spraul
14 * manfred@colorfullife.com
15 * Rewrote bits to get rid of console_lock
16 * 01Mar01 Andrew Morton
17 */
18
19#include <linux/kernel.h>
20#include <linux/mm.h>
21#include <linux/tty.h>
22#include <linux/tty_driver.h>
23#include <linux/console.h>
24#include <linux/init.h>
25#include <linux/jiffies.h>
26#include <linux/nmi.h>
27#include <linux/module.h>
28#include <linux/moduleparam.h>
29#include <linux/interrupt.h> /* For in_interrupt() */
30#include <linux/delay.h>
31#include <linux/smp.h>
32#include <linux/security.h>
33#include <linux/bootmem.h>
34#include <linux/memblock.h>
35#include <linux/syscalls.h>
36#include <linux/kexec.h>
37#include <linux/kdb.h>
38#include <linux/ratelimit.h>
39#include <linux/kmsg_dump.h>
40#include <linux/syslog.h>
41#include <linux/cpu.h>
42#include <linux/notifier.h>
43#include <linux/rculist.h>
44#include <linux/poll.h>
45
46#include <asm/uaccess.h>
47
48#define CREATE_TRACE_POINTS
49#include <trace/events/printk.h>
50
51/*
52 * Architectures can override it:
53 */
54void asmlinkage __attribute__((weak)) early_printk(const char *fmt, ...)
55{
56}
57
58/* printk's without a loglevel use this.. */
59#define DEFAULT_MESSAGE_LOGLEVEL CONFIG_DEFAULT_MESSAGE_LOGLEVEL
60
61/* We show everything that is MORE important than this.. */
62#define MINIMUM_CONSOLE_LOGLEVEL 1 /* Minimum loglevel we let people use */
63#define DEFAULT_CONSOLE_LOGLEVEL 7 /* anything MORE serious than KERN_DEBUG */
64
65DECLARE_WAIT_QUEUE_HEAD(log_wait);
66
67int console_printk[4] = {
68    DEFAULT_CONSOLE_LOGLEVEL, /* console_loglevel */
69    DEFAULT_MESSAGE_LOGLEVEL, /* default_message_loglevel */
70    MINIMUM_CONSOLE_LOGLEVEL, /* minimum_console_loglevel */
71    DEFAULT_CONSOLE_LOGLEVEL, /* default_console_loglevel */
72};
73
74/*
75 * Low level drivers may need that to know if they can schedule in
76 * their unblank() callback or not. So let's export it.
77 */
78int oops_in_progress;
79EXPORT_SYMBOL(oops_in_progress);
80
81/*
82 * console_sem protects the console_drivers list, and also
83 * provides serialisation for access to the entire console
84 * driver system.
85 */
86static DEFINE_SEMAPHORE(console_sem);
87struct console *console_drivers;
88EXPORT_SYMBOL_GPL(console_drivers);
89
90/*
91 * This is used for debugging the mess that is the VT code by
92 * keeping track if we have the console semaphore held. It's
93 * definitely not the perfect debug tool (we don't know if _WE_
94 * hold it are racing, but it helps tracking those weird code
95 * path in the console code where we end up in places I want
96 * locked without the console sempahore held
97 */
98static int console_locked, console_suspended;
99
100/*
101 * If exclusive_console is non-NULL then only this console is to be printed to.
102 */
103static struct console *exclusive_console;
104
105/*
106 * Array of consoles built from command line options (console=)
107 */
108struct console_cmdline
109{
110    char name[8]; /* Name of the driver */
111    int index; /* Minor dev. to use */
112    char *options; /* Options for the driver */
113#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
114    char *brl_options; /* Options for braille driver */
115#endif
116};
117
118#define MAX_CMDLINECONSOLES 8
119
120static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
121static int selected_console = -1;
122static int preferred_console = -1;
123int console_set_on_cmdline;
124EXPORT_SYMBOL(console_set_on_cmdline);
125
126/* Flag: console code may call schedule() */
127static int console_may_schedule;
128
129/*
130 * The printk log buffer consists of a chain of concatenated variable
131 * length records. Every record starts with a record header, containing
132 * the overall length of the record.
133 *
134 * The heads to the first and last entry in the buffer, as well as the
135 * sequence numbers of these both entries are maintained when messages
136 * are stored..
137 *
138 * If the heads indicate available messages, the length in the header
139 * tells the start next message. A length == 0 for the next message
140 * indicates a wrap-around to the beginning of the buffer.
141 *
142 * Every record carries the monotonic timestamp in microseconds, as well as
143 * the standard userspace syslog level and syslog facility. The usual
144 * kernel messages use LOG_KERN; userspace-injected messages always carry
145 * a matching syslog facility, by default LOG_USER. The origin of every
146 * message can be reliably determined that way.
147 *
148 * The human readable log message directly follows the message header. The
149 * length of the message text is stored in the header, the stored message
150 * is not terminated.
151 *
152 * Optionally, a message can carry a dictionary of properties (key/value pairs),
153 * to provide userspace with a machine-readable message context.
154 *
155 * Examples for well-defined, commonly used property names are:
156 * DEVICE=b12:8 device identifier
157 * b12:8 block dev_t
158 * c127:3 char dev_t
159 * n8 netdev ifindex
160 * +sound:card0 subsystem:devname
161 * SUBSYSTEM=pci driver-core subsystem name
162 *
163 * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
164 * follows directly after a '=' character. Every property is terminated by
165 * a '\0' character. The last property is not terminated.
166 *
167 * Example of a message structure:
168 * 0000 ff 8f 00 00 00 00 00 00 monotonic time in nsec
169 * 0008 34 00 record is 52 bytes long
170 * 000a 0b 00 text is 11 bytes long
171 * 000c 1f 00 dictionary is 23 bytes long
172 * 000e 03 00 LOG_KERN (facility) LOG_ERR (level)
173 * 0010 69 74 27 73 20 61 20 6c "it's a l"
174 * 69 6e 65 "ine"
175 * 001b 44 45 56 49 43 "DEVIC"
176 * 45 3d 62 38 3a 32 00 44 "E=b8:2\0D"
177 * 52 49 56 45 52 3d 62 75 "RIVER=bu"
178 * 67 "g"
179 * 0032 00 00 00 padding to next message header
180 *
181 * The 'struct log' buffer header must never be directly exported to
182 * userspace, it is a kernel-private implementation detail that might
183 * need to be changed in the future, when the requirements change.
184 *
185 * /dev/kmsg exports the structured data in the following line format:
186 * "level,sequnum,timestamp;<message text>\n"
187 *
188 * The optional key/value pairs are attached as continuation lines starting
189 * with a space character and terminated by a newline. All possible
190 * non-prinatable characters are escaped in the "\xff" notation.
191 *
192 * Users of the export format should ignore possible additional values
193 * separated by ',', and find the message after the ';' character.
194 */
195
196enum log_flags {
197    LOG_NOCONS = 1, /* already flushed, do not print to console */
198    LOG_NEWLINE = 2, /* text ended with a newline */
199    LOG_PREFIX = 4, /* text started with a prefix */
200    LOG_CONT = 8, /* text is a fragment of a continuation line */
201};
202
203struct log {
204    u64 ts_nsec; /* timestamp in nanoseconds */
205    u16 len; /* length of entire record */
206    u16 text_len; /* length of text buffer */
207    u16 dict_len; /* length of dictionary buffer */
208    u8 facility; /* syslog facility */
209    u8 flags:5; /* internal record flags */
210    u8 level:3; /* syslog level */
211};
212
213/*
214 * The logbuf_lock protects kmsg buffer, indices, counters. It is also
215 * used in interesting ways to provide interlocking in console_unlock();
216 */
217static DEFINE_RAW_SPINLOCK(logbuf_lock);
218
219/* the next printk record to read by syslog(READ) or /proc/kmsg */
220static u64 syslog_seq;
221static u32 syslog_idx;
222static enum log_flags syslog_prev;
223static size_t syslog_partial;
224
225/* index and sequence number of the first record stored in the buffer */
226static u64 log_first_seq;
227static u32 log_first_idx;
228
229/* index and sequence number of the next record to store in the buffer */
230static u64 log_next_seq;
231#ifdef CONFIG_PRINTK
232static u32 log_next_idx;
233
234/* the next printk record to read after the last 'clear' command */
235static u64 clear_seq;
236static u32 clear_idx;
237
238#define LOG_LINE_MAX 1024
239
240/* record buffer */
241#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
242#define LOG_ALIGN 4
243#else
244#define LOG_ALIGN __alignof__(struct log)
245#endif
246#define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
247static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
248static char *log_buf = __log_buf;
249static u32 log_buf_len = __LOG_BUF_LEN;
250
251/* cpu currently holding logbuf_lock */
252static volatile unsigned int logbuf_cpu = UINT_MAX;
253
254/* human readable text of the record */
255static char *log_text(const struct log *msg)
256{
257    return (char *)msg + sizeof(struct log);
258}
259
260/* optional key/value pair dictionary attached to the record */
261static char *log_dict(const struct log *msg)
262{
263    return (char *)msg + sizeof(struct log) + msg->text_len;
264}
265
266/* get record by index; idx must point to valid msg */
267static struct log *log_from_idx(u32 idx)
268{
269    struct log *msg = (struct log *)(log_buf + idx);
270
271    /*
272     * A length == 0 record is the end of buffer marker. Wrap around and
273     * read the message at the start of the buffer.
274     */
275    if (!msg->len)
276        return (struct log *)log_buf;
277    return msg;
278}
279
280/* get next record; idx must point to valid msg */
281static u32 log_next(u32 idx)
282{
283    struct log *msg = (struct log *)(log_buf + idx);
284
285    /* length == 0 indicates the end of the buffer; wrap */
286    /*
287     * A length == 0 record is the end of buffer marker. Wrap around and
288     * read the message at the start of the buffer as *this* one, and
289     * return the one after that.
290     */
291    if (!msg->len) {
292        msg = (struct log *)log_buf;
293        return msg->len;
294    }
295    return idx + msg->len;
296}
297
298/* insert record into the buffer, discard old ones, update heads */
299static void log_store(int facility, int level,
300              enum log_flags flags, u64 ts_nsec,
301              const char *dict, u16 dict_len,
302              const char *text, u16 text_len)
303{
304    struct log *msg;
305    u32 size, pad_len;
306
307    /* number of '\0' padding bytes to next message */
308    size = sizeof(struct log) + text_len + dict_len;
309    pad_len = (-size) & (LOG_ALIGN - 1);
310    size += pad_len;
311
312    while (log_first_seq < log_next_seq) {
313        u32 free;
314
315        if (log_next_idx > log_first_idx)
316            free = max(log_buf_len - log_next_idx, log_first_idx);
317        else
318            free = log_first_idx - log_next_idx;
319
320        if (free > size + sizeof(struct log))
321            break;
322
323        /* drop old messages until we have enough contiuous space */
324        log_first_idx = log_next(log_first_idx);
325        log_first_seq++;
326    }
327
328    if (log_next_idx + size + sizeof(struct log) >= log_buf_len) {
329        /*
330         * This message + an additional empty header does not fit
331         * at the end of the buffer. Add an empty header with len == 0
332         * to signify a wrap around.
333         */
334        memset(log_buf + log_next_idx, 0, sizeof(struct log));
335        log_next_idx = 0;
336    }
337
338    /* fill message */
339    msg = (struct log *)(log_buf + log_next_idx);
340    memcpy(log_text(msg), text, text_len);
341    msg->text_len = text_len;
342    memcpy(log_dict(msg), dict, dict_len);
343    msg->dict_len = dict_len;
344    msg->facility = facility;
345    msg->level = level & 7;
346    msg->flags = flags & 0x1f;
347    if (ts_nsec > 0)
348        msg->ts_nsec = ts_nsec;
349    else
350        msg->ts_nsec = local_clock();
351    memset(log_dict(msg) + dict_len, 0, pad_len);
352    msg->len = sizeof(struct log) + text_len + dict_len + pad_len;
353
354    /* insert message */
355    log_next_idx += msg->len;
356    log_next_seq++;
357}
358
359/* /dev/kmsg - userspace message inject/listen interface */
360struct devkmsg_user {
361    u64 seq;
362    u32 idx;
363    struct mutex lock;
364    char buf[8192];
365};
366
367static ssize_t devkmsg_writev(struct kiocb *iocb, const struct iovec *iv,
368                  unsigned long count, loff_t pos)
369{
370    char *buf, *line;
371    int i;
372    int level = default_message_loglevel;
373    int facility = 1; /* LOG_USER */
374    size_t len = iov_length(iv, count);
375    ssize_t ret = len;
376
377    if (len > LOG_LINE_MAX)
378        return -EINVAL;
379    buf = kmalloc(len+1, GFP_KERNEL);
380    if (buf == NULL)
381        return -ENOMEM;
382
383    line = buf;
384    for (i = 0; i < count; i++) {
385        if (copy_from_user(line, iv[i].iov_base, iv[i].iov_len))
386            goto out;
387        line += iv[i].iov_len;
388    }
389
390    /*
391     * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
392     * the decimal value represents 32bit, the lower 3 bit are the log
393     * level, the rest are the log facility.
394     *
395     * If no prefix or no userspace facility is specified, we
396     * enforce LOG_USER, to be able to reliably distinguish
397     * kernel-generated messages from userspace-injected ones.
398     */
399    line = buf;
400    if (line[0] == '<') {
401        char *endp = NULL;
402
403        i = simple_strtoul(line+1, &endp, 10);
404        if (endp && endp[0] == '>') {
405            level = i & 7;
406            if (i >> 3)
407                facility = i >> 3;
408            endp++;
409            len -= endp - line;
410            line = endp;
411        }
412    }
413    line[len] = '\0';
414
415    printk_emit(facility, level, NULL, 0, "%s", line);
416out:
417    kfree(buf);
418    return ret;
419}
420
421static ssize_t devkmsg_read(struct file *file, char __user *buf,
422                size_t count, loff_t *ppos)
423{
424    struct devkmsg_user *user = file->private_data;
425    struct log *msg;
426    u64 ts_usec;
427    size_t i;
428    size_t len;
429    ssize_t ret;
430
431    if (!user)
432        return -EBADF;
433
434    ret = mutex_lock_interruptible(&user->lock);
435    if (ret)
436        return ret;
437    raw_spin_lock_irq(&logbuf_lock);
438    while (user->seq == log_next_seq) {
439        if (file->f_flags & O_NONBLOCK) {
440            ret = -EAGAIN;
441            raw_spin_unlock_irq(&logbuf_lock);
442            goto out;
443        }
444
445        raw_spin_unlock_irq(&logbuf_lock);
446        ret = wait_event_interruptible(log_wait,
447                           user->seq != log_next_seq);
448        if (ret)
449            goto out;
450        raw_spin_lock_irq(&logbuf_lock);
451    }
452
453    if (user->seq < log_first_seq) {
454        /* our last seen message is gone, return error and reset */
455        user->idx = log_first_idx;
456        user->seq = log_first_seq;
457        ret = -EPIPE;
458        raw_spin_unlock_irq(&logbuf_lock);
459        goto out;
460    }
461
462    msg = log_from_idx(user->idx);
463    ts_usec = msg->ts_nsec;
464    do_div(ts_usec, 1000);
465    len = sprintf(user->buf, "%u,%llu,%llu;",
466              (msg->facility << 3) | msg->level, user->seq, ts_usec);
467
468    /* escape non-printable characters */
469    for (i = 0; i < msg->text_len; i++) {
470        unsigned char c = log_text(msg)[i];
471
472        if (c < ' ' || c >= 127 || c == '\\')
473            len += sprintf(user->buf + len, "\\x%02x", c);
474        else
475            user->buf[len++] = c;
476    }
477    user->buf[len++] = '\n';
478
479    if (msg->dict_len) {
480        bool line = true;
481
482        for (i = 0; i < msg->dict_len; i++) {
483            unsigned char c = log_dict(msg)[i];
484
485            if (line) {
486                user->buf[len++] = ' ';
487                line = false;
488            }
489
490            if (c == '\0') {
491                user->buf[len++] = '\n';
492                line = true;
493                continue;
494            }
495
496            if (c < ' ' || c >= 127 || c == '\\') {
497                len += sprintf(user->buf + len, "\\x%02x", c);
498                continue;
499            }
500
501            user->buf[len++] = c;
502        }
503        user->buf[len++] = '\n';
504    }
505
506    user->idx = log_next(user->idx);
507    user->seq++;
508    raw_spin_unlock_irq(&logbuf_lock);
509
510    if (len > count) {
511        ret = -EINVAL;
512        goto out;
513    }
514
515    if (copy_to_user(buf, user->buf, len)) {
516        ret = -EFAULT;
517        goto out;
518    }
519    ret = len;
520out:
521    mutex_unlock(&user->lock);
522    return ret;
523}
524
525static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
526{
527    struct devkmsg_user *user = file->private_data;
528    loff_t ret = 0;
529
530    if (!user)
531        return -EBADF;
532    if (offset)
533        return -ESPIPE;
534
535    raw_spin_lock_irq(&logbuf_lock);
536    switch (whence) {
537    case SEEK_SET:
538        /* the first record */
539        user->idx = log_first_idx;
540        user->seq = log_first_seq;
541        break;
542    case SEEK_DATA:
543        /*
544         * The first record after the last SYSLOG_ACTION_CLEAR,
545         * like issued by 'dmesg -c'. Reading /dev/kmsg itself
546         * changes no global state, and does not clear anything.
547         */
548        user->idx = clear_idx;
549        user->seq = clear_seq;
550        break;
551    case SEEK_END:
552        /* after the last record */
553        user->idx = log_next_idx;
554        user->seq = log_next_seq;
555        break;
556    default:
557        ret = -EINVAL;
558    }
559    raw_spin_unlock_irq(&logbuf_lock);
560    return ret;
561}
562
563static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
564{
565    struct devkmsg_user *user = file->private_data;
566    int ret = 0;
567
568    if (!user)
569        return POLLERR|POLLNVAL;
570
571    poll_wait(file, &log_wait, wait);
572
573    raw_spin_lock_irq(&logbuf_lock);
574    if (user->seq < log_next_seq) {
575        /* return error when data has vanished underneath us */
576        if (user->seq < log_first_seq)
577            ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
578        ret = POLLIN|POLLRDNORM;
579    }
580    raw_spin_unlock_irq(&logbuf_lock);
581
582    return ret;
583}
584
585static int devkmsg_open(struct inode *inode, struct file *file)
586{
587    struct devkmsg_user *user;
588    int err;
589
590    /* write-only does not need any file context */
591    if ((file->f_flags & O_ACCMODE) == O_WRONLY)
592        return 0;
593
594    err = security_syslog(SYSLOG_ACTION_READ_ALL);
595    if (err)
596        return err;
597
598    user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
599    if (!user)
600        return -ENOMEM;
601
602    mutex_init(&user->lock);
603
604    raw_spin_lock_irq(&logbuf_lock);
605    user->idx = log_first_idx;
606    user->seq = log_first_seq;
607    raw_spin_unlock_irq(&logbuf_lock);
608
609    file->private_data = user;
610    return 0;
611}
612
613static int devkmsg_release(struct inode *inode, struct file *file)
614{
615    struct devkmsg_user *user = file->private_data;
616
617    if (!user)
618        return 0;
619
620    mutex_destroy(&user->lock);
621    kfree(user);
622    return 0;
623}
624
625const struct file_operations kmsg_fops = {
626    .open = devkmsg_open,
627    .read = devkmsg_read,
628    .aio_write = devkmsg_writev,
629    .llseek = devkmsg_llseek,
630    .poll = devkmsg_poll,
631    .release = devkmsg_release,
632};
633
634#ifdef CONFIG_KEXEC
635/*
636 * This appends the listed symbols to /proc/vmcoreinfo
637 *
638 * /proc/vmcoreinfo is used by various utiilties, like crash and makedumpfile to
639 * obtain access to symbols that are otherwise very difficult to locate. These
640 * symbols are specifically used so that utilities can access and extract the
641 * dmesg log from a vmcore file after a crash.
642 */
643void log_buf_kexec_setup(void)
644{
645    VMCOREINFO_SYMBOL(log_buf);
646    VMCOREINFO_SYMBOL(log_buf_len);
647    VMCOREINFO_SYMBOL(log_first_idx);
648    VMCOREINFO_SYMBOL(log_next_idx);
649}
650#endif
651
652/* requested log_buf_len from kernel cmdline */
653static unsigned long __initdata new_log_buf_len;
654
655/* save requested log_buf_len since it's too early to process it */
656static int __init log_buf_len_setup(char *str)
657{
658    unsigned size = memparse(str, &str);
659
660    if (size)
661        size = roundup_pow_of_two(size);
662    if (size > log_buf_len)
663        new_log_buf_len = size;
664
665    return 0;
666}
667early_param("log_buf_len", log_buf_len_setup);
668
669void __init setup_log_buf(int early)
670{
671    unsigned long flags;
672    char *new_log_buf;
673    int free;
674
675    if (!new_log_buf_len)
676        return;
677
678    if (early) {
679        unsigned long mem;
680
681        mem = memblock_alloc(new_log_buf_len, PAGE_SIZE);
682        if (!mem)
683            return;
684        new_log_buf = __va(mem);
685    } else {
686        new_log_buf = alloc_bootmem_nopanic(new_log_buf_len);
687    }
688
689    if (unlikely(!new_log_buf)) {
690        pr_err("log_buf_len: %ld bytes not available\n",
691            new_log_buf_len);
692        return;
693    }
694
695    raw_spin_lock_irqsave(&logbuf_lock, flags);
696    log_buf_len = new_log_buf_len;
697    log_buf = new_log_buf;
698    new_log_buf_len = 0;
699    free = __LOG_BUF_LEN - log_next_idx;
700    memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
701    raw_spin_unlock_irqrestore(&logbuf_lock, flags);
702
703    pr_info("log_buf_len: %d\n", log_buf_len);
704    pr_info("early log buf free: %d(%d%%)\n",
705        free, (free * 100) / __LOG_BUF_LEN);
706}
707
708#ifdef CONFIG_BOOT_PRINTK_DELAY
709
710static int boot_delay; /* msecs delay after each printk during bootup */
711static unsigned long long loops_per_msec; /* based on boot_delay */
712
713static int __init boot_delay_setup(char *str)
714{
715    unsigned long lpj;
716
717    lpj = preset_lpj ? preset_lpj : 1000000; /* some guess */
718    loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
719
720    get_option(&str, &boot_delay);
721    if (boot_delay > 10 * 1000)
722        boot_delay = 0;
723
724    pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
725        "HZ: %d, loops_per_msec: %llu\n",
726        boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
727    return 1;
728}
729__setup("boot_delay=", boot_delay_setup);
730
731static void boot_delay_msec(void)
732{
733    unsigned long long k;
734    unsigned long timeout;
735
736    if (boot_delay == 0 || system_state != SYSTEM_BOOTING)
737        return;
738
739    k = (unsigned long long)loops_per_msec * boot_delay;
740
741    timeout = jiffies + msecs_to_jiffies(boot_delay);
742    while (k) {
743        k--;
744        cpu_relax();
745        /*
746         * use (volatile) jiffies to prevent
747         * compiler reduction; loop termination via jiffies
748         * is secondary and may or may not happen.
749         */
750        if (time_after(jiffies, timeout))
751            break;
752        touch_nmi_watchdog();
753    }
754}
755#else
756static inline void boot_delay_msec(void)
757{
758}
759#endif
760
761#ifdef CONFIG_SECURITY_DMESG_RESTRICT
762int dmesg_restrict = 1;
763#else
764int dmesg_restrict;
765#endif
766
767static int syslog_action_restricted(int type)
768{
769    if (dmesg_restrict)
770        return 1;
771    /* Unless restricted, we allow "read all" and "get buffer size" for everybody */
772    return type != SYSLOG_ACTION_READ_ALL && type != SYSLOG_ACTION_SIZE_BUFFER;
773}
774
775static int check_syslog_permissions(int type, bool from_file)
776{
777    /*
778     * If this is from /proc/kmsg and we've already opened it, then we've
779     * already done the capabilities checks at open time.
780     */
781    if (from_file && type != SYSLOG_ACTION_OPEN)
782        return 0;
783
784    if (syslog_action_restricted(type)) {
785        if (capable(CAP_SYSLOG))
786            return 0;
787        /* For historical reasons, accept CAP_SYS_ADMIN too, with a warning */
788        if (capable(CAP_SYS_ADMIN)) {
789            printk_once(KERN_WARNING "%s (%d): "
790                 "Attempt to access syslog with CAP_SYS_ADMIN "
791                 "but no CAP_SYSLOG (deprecated).\n",
792                 current->comm, task_pid_nr(current));
793            return 0;
794        }
795        return -EPERM;
796    }
797    return 0;
798}
799
800#if defined(CONFIG_PRINTK_TIME)
801static bool printk_time = 1;
802#else
803static bool printk_time;
804#endif
805module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
806
807static size_t print_time(u64 ts, char *buf)
808{
809    unsigned long rem_nsec;
810
811    if (!printk_time)
812        return 0;
813
814    if (!buf)
815        return 15;
816
817    rem_nsec = do_div(ts, 1000000000);
818    return sprintf(buf, "[%5lu.%06lu] ",
819               (unsigned long)ts, rem_nsec / 1000);
820}
821
822static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
823{
824    size_t len = 0;
825    unsigned int prefix = (msg->facility << 3) | msg->level;
826
827    if (syslog) {
828        if (buf) {
829            len += sprintf(buf, "<%u>", prefix);
830        } else {
831            len += 3;
832            if (prefix > 999)
833                len += 3;
834            else if (prefix > 99)
835                len += 2;
836            else if (prefix > 9)
837                len++;
838        }
839    }
840
841    len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
842    return len;
843}
844
845static size_t msg_print_text(const struct log *msg, enum log_flags prev,
846                 bool syslog, char *buf, size_t size)
847{
848    const char *text = log_text(msg);
849    size_t text_size = msg->text_len;
850    bool prefix = true;
851    bool newline = true;
852    size_t len = 0;
853
854    if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
855        prefix = false;
856
857    if (msg->flags & LOG_CONT) {
858        if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
859            prefix = false;
860
861        if (!(msg->flags & LOG_NEWLINE))
862            newline = false;
863    }
864
865    do {
866        const char *next = memchr(text, '\n', text_size);
867        size_t text_len;
868
869        if (next) {
870            text_len = next - text;
871            next++;
872            text_size -= next - text;
873        } else {
874            text_len = text_size;
875        }
876
877        if (buf) {
878            if (print_prefix(msg, syslog, NULL) +
879                text_len + 1>= size - len)
880                break;
881
882            if (prefix)
883                len += print_prefix(msg, syslog, buf + len);
884            memcpy(buf + len, text, text_len);
885            len += text_len;
886            if (next || newline)
887                buf[len++] = '\n';
888        } else {
889            /* SYSLOG_ACTION_* buffer size only calculation */
890            if (prefix)
891                len += print_prefix(msg, syslog, NULL);
892            len += text_len;
893            if (next || newline)
894                len++;
895        }
896
897        prefix = true;
898        text = next;
899    } while (text);
900
901    return len;
902}
903
904static int syslog_print(char __user *buf, int size)
905{
906    char *text;
907    struct log *msg;
908    int len = 0;
909
910    text = kmalloc(LOG_LINE_MAX, GFP_KERNEL);
911    if (!text)
912        return -ENOMEM;
913
914    while (size > 0) {
915        size_t n;
916        size_t skip;
917
918        raw_spin_lock_irq(&logbuf_lock);
919        if (syslog_seq < log_first_seq) {
920            /* messages are gone, move to first one */
921            syslog_seq = log_first_seq;
922            syslog_idx = log_first_idx;
923            syslog_prev = 0;
924            syslog_partial = 0;
925        }
926        if (syslog_seq == log_next_seq) {
927            raw_spin_unlock_irq(&logbuf_lock);
928            break;
929        }
930
931        skip = syslog_partial;
932        msg = log_from_idx(syslog_idx);
933        n = msg_print_text(msg, syslog_prev, true, text, LOG_LINE_MAX);
934        if (n - syslog_partial <= size) {
935            /* message fits into buffer, move forward */
936            syslog_idx = log_next(syslog_idx);
937            syslog_seq++;
938            syslog_prev = msg->flags;
939            n -= syslog_partial;
940            syslog_partial = 0;
941        } else if (!len){
942            /* partial read(), remember position */
943            n = size;
944            syslog_partial += n;
945        } else
946            n = 0;
947        raw_spin_unlock_irq(&logbuf_lock);
948
949        if (!n)
950            break;
951
952        if (copy_to_user(buf, text + skip, n)) {
953            if (!len)
954                len = -EFAULT;
955            break;
956        }
957
958        len += n;
959        size -= n;
960        buf += n;
961    }
962
963    kfree(text);
964    return len;
965}
966
967static int syslog_print_all(char __user *buf, int size, bool clear)
968{
969    char *text;
970    int len = 0;
971
972    text = kmalloc(LOG_LINE_MAX, GFP_KERNEL);
973    if (!text)
974        return -ENOMEM;
975
976    raw_spin_lock_irq(&logbuf_lock);
977    if (buf) {
978        u64 next_seq;
979        u64 seq;
980        u32 idx;
981        enum log_flags prev;
982
983        if (clear_seq < log_first_seq) {
984            /* messages are gone, move to first available one */
985            clear_seq = log_first_seq;
986            clear_idx = log_first_idx;
987        }
988
989        /*
990         * Find first record that fits, including all following records,
991         * into the user-provided buffer for this dump.
992         */
993        seq = clear_seq;
994        idx = clear_idx;
995        prev = 0;
996        while (seq < log_next_seq) {
997            struct log *msg = log_from_idx(idx);
998
999            len += msg_print_text(msg, prev, true, NULL, 0);
1000            idx = log_next(idx);
1001            seq++;
1002        }
1003
1004        /* move first record forward until length fits into the buffer */
1005        seq = clear_seq;
1006        idx = clear_idx;
1007        prev = 0;
1008        while (len > size && seq < log_next_seq) {
1009            struct log *msg = log_from_idx(idx);
1010
1011            len -= msg_print_text(msg, prev, true, NULL, 0);
1012            idx = log_next(idx);
1013            seq++;
1014        }
1015
1016        /* last message fitting into this dump */
1017        next_seq = log_next_seq;
1018
1019        len = 0;
1020        prev = 0;
1021        while (len >= 0 && seq < next_seq) {
1022            struct log *msg = log_from_idx(idx);
1023            int textlen;
1024
1025            textlen = msg_print_text(msg, prev, true, text, LOG_LINE_MAX);
1026            if (textlen < 0) {
1027                len = textlen;
1028                break;
1029            }
1030            idx = log_next(idx);
1031            seq++;
1032            prev = msg->flags;
1033
1034            raw_spin_unlock_irq(&logbuf_lock);
1035            if (copy_to_user(buf + len, text, textlen))
1036                len = -EFAULT;
1037            else
1038                len += textlen;
1039            raw_spin_lock_irq(&logbuf_lock);
1040
1041            if (seq < log_first_seq) {
1042                /* messages are gone, move to next one */
1043                seq = log_first_seq;
1044                idx = log_first_idx;
1045                prev = 0;
1046            }
1047        }
1048    }
1049
1050    if (clear) {
1051        clear_seq = log_next_seq;
1052        clear_idx = log_next_idx;
1053    }
1054    raw_spin_unlock_irq(&logbuf_lock);
1055
1056    kfree(text);
1057    return len;
1058}
1059
1060int do_syslog(int type, char __user *buf, int len, bool from_file)
1061{
1062    bool clear = false;
1063    static int saved_console_loglevel = -1;
1064    int error;
1065
1066    error = check_syslog_permissions(type, from_file);
1067    if (error)
1068        goto out;
1069
1070    error = security_syslog(type);
1071    if (error)
1072        return error;
1073
1074    switch (type) {
1075    case SYSLOG_ACTION_CLOSE: /* Close log */
1076        break;
1077    case SYSLOG_ACTION_OPEN: /* Open log */
1078        break;
1079    case SYSLOG_ACTION_READ: /* Read from log */
1080        error = -EINVAL;
1081        if (!buf || len < 0)
1082            goto out;
1083        error = 0;
1084        if (!len)
1085            goto out;
1086        if (!access_ok(VERIFY_WRITE, buf, len)) {
1087            error = -EFAULT;
1088            goto out;
1089        }
1090        error = wait_event_interruptible(log_wait,
1091                         syslog_seq != log_next_seq);
1092        if (error)
1093            goto out;
1094        error = syslog_print(buf, len);
1095        break;
1096    /* Read/clear last kernel messages */
1097    case SYSLOG_ACTION_READ_CLEAR:
1098        clear = true;
1099        /* FALL THRU */
1100    /* Read last kernel messages */
1101    case SYSLOG_ACTION_READ_ALL:
1102        error = -EINVAL;
1103        if (!buf || len < 0)
1104            goto out;
1105        error = 0;
1106        if (!len)
1107            goto out;
1108        if (!access_ok(VERIFY_WRITE, buf, len)) {
1109            error = -EFAULT;
1110            goto out;
1111        }
1112        error = syslog_print_all(buf, len, clear);
1113        break;
1114    /* Clear ring buffer */
1115    case SYSLOG_ACTION_CLEAR:
1116        syslog_print_all(NULL, 0, true);
1117        break;
1118    /* Disable logging to console */
1119    case SYSLOG_ACTION_CONSOLE_OFF:
1120        if (saved_console_loglevel == -1)
1121            saved_console_loglevel = console_loglevel;
1122        console_loglevel = minimum_console_loglevel;
1123        break;
1124    /* Enable logging to console */
1125    case SYSLOG_ACTION_CONSOLE_ON:
1126        if (saved_console_loglevel != -1) {
1127            console_loglevel = saved_console_loglevel;
1128            saved_console_loglevel = -1;
1129        }
1130        break;
1131    /* Set level of messages printed to console */
1132    case SYSLOG_ACTION_CONSOLE_LEVEL:
1133        error = -EINVAL;
1134        if (len < 1 || len > 8)
1135            goto out;
1136        if (len < minimum_console_loglevel)
1137            len = minimum_console_loglevel;
1138        console_loglevel = len;
1139        /* Implicitly re-enable logging to console */
1140        saved_console_loglevel = -1;
1141        error = 0;
1142        break;
1143    /* Number of chars in the log buffer */
1144    case SYSLOG_ACTION_SIZE_UNREAD:
1145        raw_spin_lock_irq(&logbuf_lock);
1146        if (syslog_seq < log_first_seq) {
1147            /* messages are gone, move to first one */
1148            syslog_seq = log_first_seq;
1149            syslog_idx = log_first_idx;
1150            syslog_prev = 0;
1151            syslog_partial = 0;
1152        }
1153        if (from_file) {
1154            /*
1155             * Short-cut for poll(/"proc/kmsg") which simply checks
1156             * for pending data, not the size; return the count of
1157             * records, not the length.
1158             */
1159            error = log_next_idx - syslog_idx;
1160        } else {
1161            u64 seq = syslog_seq;
1162            u32 idx = syslog_idx;
1163            enum log_flags prev = syslog_prev;
1164
1165            error = 0;
1166            while (seq < log_next_seq) {
1167                struct log *msg = log_from_idx(idx);
1168
1169                error += msg_print_text(msg, prev, true, NULL, 0);
1170                idx = log_next(idx);
1171                seq++;
1172                prev = msg->flags;
1173            }
1174            error -= syslog_partial;
1175        }
1176        raw_spin_unlock_irq(&logbuf_lock);
1177        break;
1178    /* Size of the log buffer */
1179    case SYSLOG_ACTION_SIZE_BUFFER:
1180        error = log_buf_len;
1181        break;
1182    default:
1183        error = -EINVAL;
1184        break;
1185    }
1186out:
1187    return error;
1188}
1189
1190SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1191{
1192    return do_syslog(type, buf, len, SYSLOG_FROM_CALL);
1193}
1194
1195static bool __read_mostly ignore_loglevel;
1196
1197static int __init ignore_loglevel_setup(char *str)
1198{
1199    ignore_loglevel = 1;
1200    printk(KERN_INFO "debug: ignoring loglevel setting.\n");
1201
1202    return 0;
1203}
1204
1205early_param("ignore_loglevel", ignore_loglevel_setup);
1206module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1207MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
1208    "print all kernel messages to the console.");
1209
1210/*
1211 * Call the console drivers, asking them to write out
1212 * log_buf[start] to log_buf[end - 1].
1213 * The console_lock must be held.
1214 */
1215static void call_console_drivers(int level, const char *text, size_t len)
1216{
1217    struct console *con;
1218
1219    trace_console(text, 0, len, len);
1220
1221    if (level >= console_loglevel && !ignore_loglevel)
1222        return;
1223    if (!console_drivers)
1224        return;
1225
1226    for_each_console(con) {
1227        if (exclusive_console && con != exclusive_console)
1228            continue;
1229        if (!(con->flags & CON_ENABLED))
1230            continue;
1231        if (!con->write)
1232            continue;
1233        if (!cpu_online(smp_processor_id()) &&
1234            !(con->flags & CON_ANYTIME))
1235            continue;
1236        con->write(con, text, len);
1237    }
1238}
1239
1240/*
1241 * Zap console related locks when oopsing. Only zap at most once
1242 * every 10 seconds, to leave time for slow consoles to print a
1243 * full oops.
1244 */
1245static void zap_locks(void)
1246{
1247    static unsigned long oops_timestamp;
1248
1249    if (time_after_eq(jiffies, oops_timestamp) &&
1250            !time_after(jiffies, oops_timestamp + 30 * HZ))
1251        return;
1252
1253    oops_timestamp = jiffies;
1254
1255    debug_locks_off();
1256    /* If a crash is occurring, make sure we can't deadlock */
1257    raw_spin_lock_init(&logbuf_lock);
1258    /* And make sure that we print immediately */
1259    sema_init(&console_sem, 1);
1260}
1261
1262/* Check if we have any console registered that can be called early in boot. */
1263static int have_callable_console(void)
1264{
1265    struct console *con;
1266
1267    for_each_console(con)
1268        if (con->flags & CON_ANYTIME)
1269            return 1;
1270
1271    return 0;
1272}
1273
1274/*
1275 * Can we actually use the console at this time on this cpu?
1276 *
1277 * Console drivers may assume that per-cpu resources have
1278 * been allocated. So unless they're explicitly marked as
1279 * being able to cope (CON_ANYTIME) don't call them until
1280 * this CPU is officially up.
1281 */
1282static inline int can_use_console(unsigned int cpu)
1283{
1284    return cpu_online(cpu) || have_callable_console();
1285}
1286
1287/*
1288 * Try to get console ownership to actually show the kernel
1289 * messages from a 'printk'. Return true (and with the
1290 * console_lock held, and 'console_locked' set) if it
1291 * is successful, false otherwise.
1292 *
1293 * This gets called with the 'logbuf_lock' spinlock held and
1294 * interrupts disabled. It should return with 'lockbuf_lock'
1295 * released but interrupts still disabled.
1296 */
1297static int console_trylock_for_printk(unsigned int cpu)
1298    __releases(&logbuf_lock)
1299{
1300    int retval = 0, wake = 0;
1301
1302    if (console_trylock()) {
1303        retval = 1;
1304
1305        /*
1306         * If we can't use the console, we need to release
1307         * the console semaphore by hand to avoid flushing
1308         * the buffer. We need to hold the console semaphore
1309         * in order to do this test safely.
1310         */
1311        if (!can_use_console(cpu)) {
1312            console_locked = 0;
1313            wake = 1;
1314            retval = 0;
1315        }
1316    }
1317    logbuf_cpu = UINT_MAX;
1318    if (wake)
1319        up(&console_sem);
1320    raw_spin_unlock(&logbuf_lock);
1321    return retval;
1322}
1323
1324int printk_delay_msec __read_mostly;
1325
1326static inline void printk_delay(void)
1327{
1328    if (unlikely(printk_delay_msec)) {
1329        int m = printk_delay_msec;
1330
1331        while (m--) {
1332            mdelay(1);
1333            touch_nmi_watchdog();
1334        }
1335    }
1336}
1337
1338/*
1339 * Continuation lines are buffered, and not committed to the record buffer
1340 * until the line is complete, or a race forces it. The line fragments
1341 * though, are printed immediately to the consoles to ensure everything has
1342 * reached the console in case of a kernel crash.
1343 */
1344static struct cont {
1345    char buf[LOG_LINE_MAX];
1346    size_t len; /* length == 0 means unused buffer */
1347    size_t cons; /* bytes written to console */
1348    struct task_struct *owner; /* task of first print*/
1349    u64 ts_nsec; /* time of first print */
1350    u8 level; /* log level of first message */
1351    u8 facility; /* log level of first message */
1352    bool flushed:1; /* buffer sealed and committed */
1353} cont;
1354
1355static void cont_flush(void)
1356{
1357    if (cont.flushed)
1358        return;
1359    if (cont.len == 0)
1360        return;
1361
1362    log_store(cont.facility, cont.level, LOG_NOCONS, cont.ts_nsec,
1363          NULL, 0, cont.buf, cont.len);
1364
1365    cont.flushed = true;
1366}
1367
1368static bool cont_add(int facility, int level, const char *text, size_t len)
1369{
1370    if (cont.len && cont.flushed)
1371        return false;
1372
1373    if (cont.len + len > sizeof(cont.buf)) {
1374        cont_flush();
1375        return false;
1376    }
1377
1378    if (!cont.len) {
1379        cont.facility = facility;
1380        cont.level = level;
1381        cont.owner = current;
1382        cont.ts_nsec = local_clock();
1383        cont.cons = 0;
1384        cont.flushed = false;
1385    }
1386
1387    memcpy(cont.buf + cont.len, text, len);
1388    cont.len += len;
1389    return true;
1390}
1391
1392static size_t cont_print_text(char *text, size_t size)
1393{
1394    size_t textlen = 0;
1395    size_t len;
1396
1397    if (cont.cons == 0) {
1398        textlen += print_time(cont.ts_nsec, text);
1399        size -= textlen;
1400    }
1401
1402    len = cont.len - cont.cons;
1403    if (len > 0) {
1404        if (len+1 > size)
1405            len = size-1;
1406        memcpy(text + textlen, cont.buf + cont.cons, len);
1407        textlen += len;
1408        cont.cons = cont.len;
1409    }
1410
1411    if (cont.flushed) {
1412        text[textlen++] = '\n';
1413        /* got everything, release buffer */
1414        cont.len = 0;
1415    }
1416    return textlen;
1417}
1418
1419asmlinkage int vprintk_emit(int facility, int level,
1420                const char *dict, size_t dictlen,
1421                const char *fmt, va_list args)
1422{
1423    static int recursion_bug;
1424    static char textbuf[LOG_LINE_MAX];
1425    char *text = textbuf;
1426    size_t text_len;
1427    enum log_flags lflags = 0;
1428    unsigned long flags;
1429    int this_cpu;
1430    int printed_len = 0;
1431
1432    boot_delay_msec();
1433    printk_delay();
1434
1435    /* This stops the holder of console_sem just where we want him */
1436    local_irq_save(flags);
1437    this_cpu = smp_processor_id();
1438
1439    /*
1440     * Ouch, printk recursed into itself!
1441     */
1442    if (unlikely(logbuf_cpu == this_cpu)) {
1443        /*
1444         * If a crash is occurring during printk() on this CPU,
1445         * then try to get the crash message out but make sure
1446         * we can't deadlock. Otherwise just return to avoid the
1447         * recursion and return - but flag the recursion so that
1448         * it can be printed at the next appropriate moment:
1449         */
1450        if (!oops_in_progress && !lockdep_recursing(current)) {
1451            recursion_bug = 1;
1452            goto out_restore_irqs;
1453        }
1454        zap_locks();
1455    }
1456
1457    lockdep_off();
1458    raw_spin_lock(&logbuf_lock);
1459    logbuf_cpu = this_cpu;
1460
1461    if (recursion_bug) {
1462        static const char recursion_msg[] =
1463            "BUG: recent printk recursion!";
1464
1465        recursion_bug = 0;
1466        printed_len += strlen(recursion_msg);
1467        /* emit KERN_CRIT message */
1468        log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1469              NULL, 0, recursion_msg, printed_len);
1470    }
1471
1472    /*
1473     * The printf needs to come first; we need the syslog
1474     * prefix which might be passed-in as a parameter.
1475     */
1476    text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1477
1478    /* mark and strip a trailing newline */
1479    if (text_len && text[text_len-1] == '\n') {
1480        text_len--;
1481        lflags |= LOG_NEWLINE;
1482    }
1483
1484    /* strip syslog prefix and extract log level or control flags */
1485    if (text[0] == '<' && text[1] && text[2] == '>') {
1486        switch (text[1]) {
1487        case '0' ... '7':
1488            if (level == -1)
1489                level = text[1] - '0';
1490        case 'd': /* KERN_DEFAULT */
1491            lflags |= LOG_PREFIX;
1492        case 'c': /* KERN_CONT */
1493            text += 3;
1494            text_len -= 3;
1495        }
1496    }
1497
1498    if (level == -1)
1499        level = default_message_loglevel;
1500
1501    if (dict)
1502        lflags |= LOG_PREFIX|LOG_NEWLINE;
1503
1504    if (!(lflags & LOG_NEWLINE)) {
1505        /*
1506         * Flush the conflicting buffer. An earlier newline was missing,
1507         * or another task also prints continuation lines.
1508         */
1509        if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
1510            cont_flush();
1511
1512        /* buffer line if possible, otherwise store it right away */
1513        if (!cont_add(facility, level, text, text_len))
1514            log_store(facility, level, lflags | LOG_CONT, 0,
1515                  dict, dictlen, text, text_len);
1516    } else {
1517        bool stored = false;
1518
1519        /*
1520         * If an earlier newline was missing and it was the same task,
1521         * either merge it with the current buffer and flush, or if
1522         * there was a race with interrupts (prefix == true) then just
1523         * flush it out and store this line separately.
1524         */
1525        if (cont.len && cont.owner == current) {
1526            if (!(lflags & LOG_PREFIX))
1527                stored = cont_add(facility, level, text, text_len);
1528            cont_flush();
1529        }
1530
1531        if (!stored)
1532            log_store(facility, level, lflags, 0,
1533                  dict, dictlen, text, text_len);
1534    }
1535    printed_len += text_len;
1536
1537    /*
1538     * Try to acquire and then immediately release the console semaphore.
1539     * The release will print out buffers and wake up /dev/kmsg and syslog()
1540     * users.
1541     *
1542     * The console_trylock_for_printk() function will release 'logbuf_lock'
1543     * regardless of whether it actually gets the console semaphore or not.
1544     */
1545    if (console_trylock_for_printk(this_cpu))
1546        console_unlock();
1547
1548    lockdep_on();
1549out_restore_irqs:
1550    local_irq_restore(flags);
1551
1552    return printed_len;
1553}
1554EXPORT_SYMBOL(vprintk_emit);
1555
1556asmlinkage int vprintk(const char *fmt, va_list args)
1557{
1558    return vprintk_emit(0, -1, NULL, 0, fmt, args);
1559}
1560EXPORT_SYMBOL(vprintk);
1561
1562asmlinkage int printk_emit(int facility, int level,
1563               const char *dict, size_t dictlen,
1564               const char *fmt, ...)
1565{
1566    va_list args;
1567    int r;
1568
1569    va_start(args, fmt);
1570    r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1571    va_end(args);
1572
1573    return r;
1574}
1575EXPORT_SYMBOL(printk_emit);
1576
1577/**
1578 * printk - print a kernel message
1579 * @fmt: format string
1580 *
1581 * This is printk(). It can be called from any context. We want it to work.
1582 *
1583 * We try to grab the console_lock. If we succeed, it's easy - we log the
1584 * output and call the console drivers. If we fail to get the semaphore, we
1585 * place the output into the log buffer and return. The current holder of
1586 * the console_sem will notice the new output in console_unlock(); and will
1587 * send it to the consoles before releasing the lock.
1588 *
1589 * One effect of this deferred printing is that code which calls printk() and
1590 * then changes console_loglevel may break. This is because console_loglevel
1591 * is inspected when the actual printing occurs.
1592 *
1593 * See also:
1594 * printf(3)
1595 *
1596 * See the vsnprintf() documentation for format string extensions over C99.
1597 */
1598asmlinkage int printk(const char *fmt, ...)
1599{
1600    va_list args;
1601    int r;
1602
1603#ifdef CONFIG_KGDB_KDB
1604    if (unlikely(kdb_trap_printk)) {
1605        va_start(args, fmt);
1606        r = vkdb_printf(fmt, args);
1607        va_end(args);
1608        return r;
1609    }
1610#endif
1611    va_start(args, fmt);
1612    r = vprintk_emit(0, -1, NULL, 0, fmt, args);
1613    va_end(args);
1614
1615    return r;
1616}
1617EXPORT_SYMBOL(printk);
1618
1619#else
1620
1621#define LOG_LINE_MAX 0
1622static struct cont {
1623    size_t len;
1624    size_t cons;
1625    u8 level;
1626    bool flushed:1;
1627} cont;
1628static struct log *log_from_idx(u32 idx) { return NULL; }
1629static u32 log_next(u32 idx) { return 0; }
1630static void call_console_drivers(int level, const char *text, size_t len) {}
1631static size_t msg_print_text(const struct log *msg, enum log_flags prev,
1632                 bool syslog, char *buf, size_t size) { return 0; }
1633static size_t cont_print_text(char *text, size_t size) { return 0; }
1634
1635#endif /* CONFIG_PRINTK */
1636
1637static int __add_preferred_console(char *name, int idx, char *options,
1638                   char *brl_options)
1639{
1640    struct console_cmdline *c;
1641    int i;
1642
1643    /*
1644     * See if this tty is not yet registered, and
1645     * if we have a slot free.
1646     */
1647    for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1648        if (strcmp(console_cmdline[i].name, name) == 0 &&
1649              console_cmdline[i].index == idx) {
1650                if (!brl_options)
1651                    selected_console = i;
1652                return 0;
1653        }
1654    if (i == MAX_CMDLINECONSOLES)
1655        return -E2BIG;
1656    if (!brl_options)
1657        selected_console = i;
1658    c = &console_cmdline[i];
1659    strlcpy(c->name, name, sizeof(c->name));
1660    c->options = options;
1661#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1662    c->brl_options = brl_options;
1663#endif
1664    c->index = idx;
1665    return 0;
1666}
1667/*
1668 * Set up a list of consoles. Called from init/main.c
1669 */
1670static int __init console_setup(char *str)
1671{
1672    char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
1673    char *s, *options, *brl_options = NULL;
1674    int idx;
1675
1676#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1677    if (!memcmp(str, "brl,", 4)) {
1678        brl_options = "";
1679        str += 4;
1680    } else if (!memcmp(str, "brl=", 4)) {
1681        brl_options = str + 4;
1682        str = strchr(brl_options, ',');
1683        if (!str) {
1684            printk(KERN_ERR "need port name after brl=\n");
1685            return 1;
1686        }
1687        *(str++) = 0;
1688    }
1689#endif
1690
1691    /*
1692     * Decode str into name, index, options.
1693     */
1694    if (str[0] >= '0' && str[0] <= '9') {
1695        strcpy(buf, "ttyS");
1696        strncpy(buf + 4, str, sizeof(buf) - 5);
1697    } else {
1698        strncpy(buf, str, sizeof(buf) - 1);
1699    }
1700    buf[sizeof(buf) - 1] = 0;
1701    if ((options = strchr(str, ',')) != NULL)
1702        *(options++) = 0;
1703#ifdef __sparc__
1704    if (!strcmp(str, "ttya"))
1705        strcpy(buf, "ttyS0");
1706    if (!strcmp(str, "ttyb"))
1707        strcpy(buf, "ttyS1");
1708#endif
1709    for (s = buf; *s; s++)
1710        if ((*s >= '0' && *s <= '9') || *s == ',')
1711            break;
1712    idx = simple_strtoul(s, NULL, 10);
1713    *s = 0;
1714
1715    __add_preferred_console(buf, idx, options, brl_options);
1716    console_set_on_cmdline = 1;
1717    return 1;
1718}
1719__setup("console=", console_setup);
1720
1721/**
1722 * add_preferred_console - add a device to the list of preferred consoles.
1723 * @name: device name
1724 * @idx: device index
1725 * @options: options for this console
1726 *
1727 * The last preferred console added will be used for kernel messages
1728 * and stdin/out/err for init. Normally this is used by console_setup
1729 * above to handle user-supplied console arguments; however it can also
1730 * be used by arch-specific code either to override the user or more
1731 * commonly to provide a default console (ie from PROM variables) when
1732 * the user has not supplied one.
1733 */
1734int add_preferred_console(char *name, int idx, char *options)
1735{
1736    return __add_preferred_console(name, idx, options, NULL);
1737}
1738
1739int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
1740{
1741    struct console_cmdline *c;
1742    int i;
1743
1744    for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1745        if (strcmp(console_cmdline[i].name, name) == 0 &&
1746              console_cmdline[i].index == idx) {
1747                c = &console_cmdline[i];
1748                strlcpy(c->name, name_new, sizeof(c->name));
1749                c->name[sizeof(c->name) - 1] = 0;
1750                c->options = options;
1751                c->index = idx_new;
1752                return i;
1753        }
1754    /* not found */
1755    return -1;
1756}
1757
1758bool console_suspend_enabled = 1;
1759EXPORT_SYMBOL(console_suspend_enabled);
1760
1761static int __init console_suspend_disable(char *str)
1762{
1763    console_suspend_enabled = 0;
1764    return 1;
1765}
1766__setup("no_console_suspend", console_suspend_disable);
1767module_param_named(console_suspend, console_suspend_enabled,
1768        bool, S_IRUGO | S_IWUSR);
1769MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
1770    " and hibernate operations");
1771
1772/**
1773 * suspend_console - suspend the console subsystem
1774 *
1775 * This disables printk() while we go into suspend states
1776 */
1777void suspend_console(void)
1778{
1779    if (!console_suspend_enabled)
1780        return;
1781    printk("Suspending console(s) (use no_console_suspend to debug)\n");
1782    console_lock();
1783    console_suspended = 1;
1784    up(&console_sem);
1785}
1786
1787void resume_console(void)
1788{
1789    if (!console_suspend_enabled)
1790        return;
1791    down(&console_sem);
1792    console_suspended = 0;
1793    console_unlock();
1794}
1795
1796/**
1797 * console_cpu_notify - print deferred console messages after CPU hotplug
1798 * @self: notifier struct
1799 * @action: CPU hotplug event
1800 * @hcpu: unused
1801 *
1802 * If printk() is called from a CPU that is not online yet, the messages
1803 * will be spooled but will not show up on the console. This function is
1804 * called when a new CPU comes online (or fails to come up), and ensures
1805 * that any such output gets printed.
1806 */
1807static int __cpuinit console_cpu_notify(struct notifier_block *self,
1808    unsigned long action, void *hcpu)
1809{
1810    switch (action) {
1811    case CPU_ONLINE:
1812    case CPU_DEAD:
1813    case CPU_DYING:
1814    case CPU_DOWN_FAILED:
1815    case CPU_UP_CANCELED:
1816        console_lock();
1817        console_unlock();
1818    }
1819    return NOTIFY_OK;
1820}
1821
1822/**
1823 * console_lock - lock the console system for exclusive use.
1824 *
1825 * Acquires a lock which guarantees that the caller has
1826 * exclusive access to the console system and the console_drivers list.
1827 *
1828 * Can sleep, returns nothing.
1829 */
1830void console_lock(void)
1831{
1832    BUG_ON(in_interrupt());
1833    down(&console_sem);
1834    if (console_suspended)
1835        return;
1836    console_locked = 1;
1837    console_may_schedule = 1;
1838}
1839EXPORT_SYMBOL(console_lock);
1840
1841/**
1842 * console_trylock - try to lock the console system for exclusive use.
1843 *
1844 * Tried to acquire a lock which guarantees that the caller has
1845 * exclusive access to the console system and the console_drivers list.
1846 *
1847 * returns 1 on success, and 0 on failure to acquire the lock.
1848 */
1849int console_trylock(void)
1850{
1851    if (down_trylock(&console_sem))
1852        return 0;
1853    if (console_suspended) {
1854        up(&console_sem);
1855        return 0;
1856    }
1857    console_locked = 1;
1858    console_may_schedule = 0;
1859    return 1;
1860}
1861EXPORT_SYMBOL(console_trylock);
1862
1863int is_console_locked(void)
1864{
1865    return console_locked;
1866}
1867
1868/*
1869 * Delayed printk version, for scheduler-internal messages:
1870 */
1871#define PRINTK_BUF_SIZE 512
1872
1873#define PRINTK_PENDING_WAKEUP 0x01
1874#define PRINTK_PENDING_SCHED 0x02
1875
1876static DEFINE_PER_CPU(int, printk_pending);
1877static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
1878
1879void printk_tick(void)
1880{
1881    if (__this_cpu_read(printk_pending)) {
1882        int pending = __this_cpu_xchg(printk_pending, 0);
1883        if (pending & PRINTK_PENDING_SCHED) {
1884            char *buf = __get_cpu_var(printk_sched_buf);
1885            printk(KERN_WARNING "[sched_delayed] %s", buf);
1886        }
1887        if (pending & PRINTK_PENDING_WAKEUP)
1888            wake_up_interruptible(&log_wait);
1889    }
1890}
1891
1892int printk_needs_cpu(int cpu)
1893{
1894    if (cpu_is_offline(cpu))
1895        printk_tick();
1896    return __this_cpu_read(printk_pending);
1897}
1898
1899void wake_up_klogd(void)
1900{
1901    if (waitqueue_active(&log_wait))
1902        this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
1903}
1904
1905/* the next printk record to write to the console */
1906static u64 console_seq;
1907static u32 console_idx;
1908static enum log_flags console_prev;
1909
1910/**
1911 * console_unlock - unlock the console system
1912 *
1913 * Releases the console_lock which the caller holds on the console system
1914 * and the console driver list.
1915 *
1916 * While the console_lock was held, console output may have been buffered
1917 * by printk(). If this is the case, console_unlock(); emits
1918 * the output prior to releasing the lock.
1919 *
1920 * If there is output waiting, we wake /dev/kmsg and syslog() users.
1921 *
1922 * console_unlock(); may be called from any context.
1923 */
1924void console_unlock(void)
1925{
1926    static char text[LOG_LINE_MAX];
1927    static u64 seen_seq;
1928    unsigned long flags;
1929    bool wake_klogd = false;
1930    bool retry;
1931
1932    if (console_suspended) {
1933        up(&console_sem);
1934        return;
1935    }
1936
1937    console_may_schedule = 0;
1938
1939    /* flush buffered message fragment immediately to console */
1940    raw_spin_lock_irqsave(&logbuf_lock, flags);
1941    if (cont.len && (cont.cons < cont.len || cont.flushed)) {
1942        size_t len;
1943
1944        len = cont_print_text(text, sizeof(text));
1945        raw_spin_unlock(&logbuf_lock);
1946        stop_critical_timings();
1947        call_console_drivers(cont.level, text, len);
1948        start_critical_timings();
1949        local_irq_restore(flags);
1950    } else
1951        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
1952
1953again:
1954    for (;;) {
1955        struct log *msg;
1956        size_t len;
1957        int level;
1958
1959        raw_spin_lock_irqsave(&logbuf_lock, flags);
1960        if (seen_seq != log_next_seq) {
1961            wake_klogd = true;
1962            seen_seq = log_next_seq;
1963        }
1964
1965        if (console_seq < log_first_seq) {
1966            /* messages are gone, move to first one */
1967            console_seq = log_first_seq;
1968            console_idx = log_first_idx;
1969            console_prev = 0;
1970        }
1971skip:
1972        if (console_seq == log_next_seq)
1973            break;
1974
1975        msg = log_from_idx(console_idx);
1976        if (msg->flags & LOG_NOCONS) {
1977            /*
1978             * Skip record we have buffered and already printed
1979             * directly to the console when we received it.
1980             */
1981            console_idx = log_next(console_idx);
1982            console_seq++;
1983            /*
1984             * We will get here again when we register a new
1985             * CON_PRINTBUFFER console. Clear the flag so we
1986             * will properly dump everything later.
1987             */
1988            msg->flags &= ~LOG_NOCONS;
1989            goto skip;
1990        }
1991
1992        level = msg->level;
1993        len = msg_print_text(msg, console_prev, false,
1994                     text, sizeof(text));
1995        console_idx = log_next(console_idx);
1996        console_seq++;
1997        console_prev = msg->flags;
1998        raw_spin_unlock(&logbuf_lock);
1999
2000        stop_critical_timings(); /* don't trace print latency */
2001        call_console_drivers(level, text, len);
2002        start_critical_timings();
2003        local_irq_restore(flags);
2004    }
2005    console_locked = 0;
2006
2007    /* Release the exclusive_console once it is used */
2008    if (unlikely(exclusive_console))
2009        exclusive_console = NULL;
2010
2011    raw_spin_unlock(&logbuf_lock);
2012
2013    up(&console_sem);
2014
2015    /*
2016     * Someone could have filled up the buffer again, so re-check if there's
2017     * something to flush. In case we cannot trylock the console_sem again,
2018     * there's a new owner and the console_unlock() from them will do the
2019     * flush, no worries.
2020     */
2021    raw_spin_lock(&logbuf_lock);
2022    retry = console_seq != log_next_seq;
2023    raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2024
2025    if (retry && console_trylock())
2026        goto again;
2027
2028    if (wake_klogd)
2029        wake_up_klogd();
2030}
2031EXPORT_SYMBOL(console_unlock);
2032
2033/**
2034 * console_conditional_schedule - yield the CPU if required
2035 *
2036 * If the console code is currently allowed to sleep, and
2037 * if this CPU should yield the CPU to another task, do
2038 * so here.
2039 *
2040 * Must be called within console_lock();.
2041 */
2042void __sched console_conditional_schedule(void)
2043{
2044    if (console_may_schedule)
2045        cond_resched();
2046}
2047EXPORT_SYMBOL(console_conditional_schedule);
2048
2049void console_unblank(void)
2050{
2051    struct console *c;
2052
2053    /*
2054     * console_unblank can no longer be called in interrupt context unless
2055     * oops_in_progress is set to 1..
2056     */
2057    if (oops_in_progress) {
2058        if (down_trylock(&console_sem) != 0)
2059            return;
2060    } else
2061        console_lock();
2062
2063    console_locked = 1;
2064    console_may_schedule = 0;
2065    for_each_console(c)
2066        if ((c->flags & CON_ENABLED) && c->unblank)
2067            c->unblank();
2068    console_unlock();
2069}
2070
2071/*
2072 * Return the console tty driver structure and its associated index
2073 */
2074struct tty_driver *console_device(int *index)
2075{
2076    struct console *c;
2077    struct tty_driver *driver = NULL;
2078
2079    console_lock();
2080    for_each_console(c) {
2081        if (!c->device)
2082            continue;
2083        driver = c->device(c, index);
2084        if (driver)
2085            break;
2086    }
2087    console_unlock();
2088    return driver;
2089}
2090
2091/*
2092 * Prevent further output on the passed console device so that (for example)
2093 * serial drivers can disable console output before suspending a port, and can
2094 * re-enable output afterwards.
2095 */
2096void console_stop(struct console *console)
2097{
2098    console_lock();
2099    console->flags &= ~CON_ENABLED;
2100    console_unlock();
2101}
2102EXPORT_SYMBOL(console_stop);
2103
2104void console_start(struct console *console)
2105{
2106    console_lock();
2107    console->flags |= CON_ENABLED;
2108    console_unlock();
2109}
2110EXPORT_SYMBOL(console_start);
2111
2112static int __read_mostly keep_bootcon;
2113
2114static int __init keep_bootcon_setup(char *str)
2115{
2116    keep_bootcon = 1;
2117    printk(KERN_INFO "debug: skip boot console de-registration.\n");
2118
2119    return 0;
2120}
2121
2122early_param("keep_bootcon", keep_bootcon_setup);
2123
2124/*
2125 * The console driver calls this routine during kernel initialization
2126 * to register the console printing procedure with printk() and to
2127 * print any messages that were printed by the kernel before the
2128 * console driver was initialized.
2129 *
2130 * This can happen pretty early during the boot process (because of
2131 * early_printk) - sometimes before setup_arch() completes - be careful
2132 * of what kernel features are used - they may not be initialised yet.
2133 *
2134 * There are two types of consoles - bootconsoles (early_printk) and
2135 * "real" consoles (everything which is not a bootconsole) which are
2136 * handled differently.
2137 * - Any number of bootconsoles can be registered at any time.
2138 * - As soon as a "real" console is registered, all bootconsoles
2139 * will be unregistered automatically.
2140 * - Once a "real" console is registered, any attempt to register a
2141 * bootconsoles will be rejected
2142 */
2143void register_console(struct console *newcon)
2144{
2145    int i;
2146    unsigned long flags;
2147    struct console *bcon = NULL;
2148
2149    /*
2150     * before we register a new CON_BOOT console, make sure we don't
2151     * already have a valid console
2152     */
2153    if (console_drivers && newcon->flags & CON_BOOT) {
2154        /* find the last or real console */
2155        for_each_console(bcon) {
2156            if (!(bcon->flags & CON_BOOT)) {
2157                printk(KERN_INFO "Too late to register bootconsole %s%d\n",
2158                    newcon->name, newcon->index);
2159                return;
2160            }
2161        }
2162    }
2163
2164    if (console_drivers && console_drivers->flags & CON_BOOT)
2165        bcon = console_drivers;
2166
2167    if (preferred_console < 0 || bcon || !console_drivers)
2168        preferred_console = selected_console;
2169
2170    if (newcon->early_setup)
2171        newcon->early_setup();
2172
2173    /*
2174     * See if we want to use this console driver. If we
2175     * didn't select a console we take the first one
2176     * that registers here.
2177     */
2178    if (preferred_console < 0) {
2179        if (newcon->index < 0)
2180            newcon->index = 0;
2181        if (newcon->setup == NULL ||
2182            newcon->setup(newcon, NULL) == 0) {
2183            newcon->flags |= CON_ENABLED;
2184            if (newcon->device) {
2185                newcon->flags |= CON_CONSDEV;
2186                preferred_console = 0;
2187            }
2188        }
2189    }
2190
2191    /*
2192     * See if this console matches one we selected on
2193     * the command line.
2194     */
2195    for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
2196            i++) {
2197        if (strcmp(console_cmdline[i].name, newcon->name) != 0)
2198            continue;
2199        if (newcon->index >= 0 &&
2200            newcon->index != console_cmdline[i].index)
2201            continue;
2202        if (newcon->index < 0)
2203            newcon->index = console_cmdline[i].index;
2204#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2205        if (console_cmdline[i].brl_options) {
2206            newcon->flags |= CON_BRL;
2207            braille_register_console(newcon,
2208                    console_cmdline[i].index,
2209                    console_cmdline[i].options,
2210                    console_cmdline[i].brl_options);
2211            return;
2212        }
2213#endif
2214        if (newcon->setup &&
2215            newcon->setup(newcon, console_cmdline[i].options) != 0)
2216            break;
2217        newcon->flags |= CON_ENABLED;
2218        newcon->index = console_cmdline[i].index;
2219        if (i == selected_console) {
2220            newcon->flags |= CON_CONSDEV;
2221            preferred_console = selected_console;
2222        }
2223        break;
2224    }
2225
2226    if (!(newcon->flags & CON_ENABLED))
2227        return;
2228
2229    /*
2230     * If we have a bootconsole, and are switching to a real console,
2231     * don't print everything out again, since when the boot console, and
2232     * the real console are the same physical device, it's annoying to
2233     * see the beginning boot messages twice
2234     */
2235    if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2236        newcon->flags &= ~CON_PRINTBUFFER;
2237
2238    /*
2239     * Put this console in the list - keep the
2240     * preferred driver at the head of the list.
2241     */
2242    console_lock();
2243    if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2244        newcon->next = console_drivers;
2245        console_drivers = newcon;
2246        if (newcon->next)
2247            newcon->next->flags &= ~CON_CONSDEV;
2248    } else {
2249        newcon->next = console_drivers->next;
2250        console_drivers->next = newcon;
2251    }
2252    if (newcon->flags & CON_PRINTBUFFER) {
2253        /*
2254         * console_unlock(); will print out the buffered messages
2255         * for us.
2256         */
2257        raw_spin_lock_irqsave(&logbuf_lock, flags);
2258        console_seq = syslog_seq;
2259        console_idx = syslog_idx;
2260        console_prev = syslog_prev;
2261        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2262        /*
2263         * We're about to replay the log buffer. Only do this to the
2264         * just-registered console to avoid excessive message spam to
2265         * the already-registered consoles.
2266         */
2267        exclusive_console = newcon;
2268    }
2269    console_unlock();
2270    console_sysfs_notify();
2271
2272    /*
2273     * By unregistering the bootconsoles after we enable the real console
2274     * we get the "console xxx enabled" message on all the consoles -
2275     * boot consoles, real consoles, etc - this is to ensure that end
2276     * users know there might be something in the kernel's log buffer that
2277     * went to the bootconsole (that they do not see on the real console)
2278     */
2279    if (bcon &&
2280        ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2281        !keep_bootcon) {
2282        /* we need to iterate through twice, to make sure we print
2283         * everything out, before we unregister the console(s)
2284         */
2285        printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
2286            newcon->name, newcon->index);
2287        for_each_console(bcon)
2288            if (bcon->flags & CON_BOOT)
2289                unregister_console(bcon);
2290    } else {
2291        printk(KERN_INFO "%sconsole [%s%d] enabled\n",
2292            (newcon->flags & CON_BOOT) ? "boot" : "" ,
2293            newcon->name, newcon->index);
2294    }
2295}
2296EXPORT_SYMBOL(register_console);
2297
2298int unregister_console(struct console *console)
2299{
2300        struct console *a, *b;
2301    int res = 1;
2302
2303#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2304    if (console->flags & CON_BRL)
2305        return braille_unregister_console(console);
2306#endif
2307
2308    console_lock();
2309    if (console_drivers == console) {
2310        console_drivers=console->next;
2311        res = 0;
2312    } else if (console_drivers) {
2313        for (a=console_drivers->next, b=console_drivers ;
2314             a; b=a, a=b->next) {
2315            if (a == console) {
2316                b->next = a->next;
2317                res = 0;
2318                break;
2319            }
2320        }
2321    }
2322
2323    /*
2324     * If this isn't the last console and it has CON_CONSDEV set, we
2325     * need to set it on the next preferred console.
2326     */
2327    if (console_drivers != NULL && console->flags & CON_CONSDEV)
2328        console_drivers->flags |= CON_CONSDEV;
2329
2330    console_unlock();
2331    console_sysfs_notify();
2332    return res;
2333}
2334EXPORT_SYMBOL(unregister_console);
2335
2336static int __init printk_late_init(void)
2337{
2338    struct console *con;
2339
2340    for_each_console(con) {
2341        if (!keep_bootcon && con->flags & CON_BOOT) {
2342            printk(KERN_INFO "turn off boot console %s%d\n",
2343                con->name, con->index);
2344            unregister_console(con);
2345        }
2346    }
2347    hotcpu_notifier(console_cpu_notify, 0);
2348    return 0;
2349}
2350late_initcall(printk_late_init);
2351
2352#if defined CONFIG_PRINTK
2353
2354int printk_sched(const char *fmt, ...)
2355{
2356    unsigned long flags;
2357    va_list args;
2358    char *buf;
2359    int r;
2360
2361    local_irq_save(flags);
2362    buf = __get_cpu_var(printk_sched_buf);
2363
2364    va_start(args, fmt);
2365    r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
2366    va_end(args);
2367
2368    __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
2369    local_irq_restore(flags);
2370
2371    return r;
2372}
2373
2374/*
2375 * printk rate limiting, lifted from the networking subsystem.
2376 *
2377 * This enforces a rate limit: not more than 10 kernel messages
2378 * every 5s to make a denial-of-service attack impossible.
2379 */
2380DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2381
2382int __printk_ratelimit(const char *func)
2383{
2384    return ___ratelimit(&printk_ratelimit_state, func);
2385}
2386EXPORT_SYMBOL(__printk_ratelimit);
2387
2388/**
2389 * printk_timed_ratelimit - caller-controlled printk ratelimiting
2390 * @caller_jiffies: pointer to caller's state
2391 * @interval_msecs: minimum interval between prints
2392 *
2393 * printk_timed_ratelimit() returns true if more than @interval_msecs
2394 * milliseconds have elapsed since the last time printk_timed_ratelimit()
2395 * returned true.
2396 */
2397bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2398            unsigned int interval_msecs)
2399{
2400    if (*caller_jiffies == 0
2401            || !time_in_range(jiffies, *caller_jiffies,
2402                    *caller_jiffies
2403                    + msecs_to_jiffies(interval_msecs))) {
2404        *caller_jiffies = jiffies;
2405        return true;
2406    }
2407    return false;
2408}
2409EXPORT_SYMBOL(printk_timed_ratelimit);
2410
2411static DEFINE_SPINLOCK(dump_list_lock);
2412static LIST_HEAD(dump_list);
2413
2414/**
2415 * kmsg_dump_register - register a kernel log dumper.
2416 * @dumper: pointer to the kmsg_dumper structure
2417 *
2418 * Adds a kernel log dumper to the system. The dump callback in the
2419 * structure will be called when the kernel oopses or panics and must be
2420 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2421 */
2422int kmsg_dump_register(struct kmsg_dumper *dumper)
2423{
2424    unsigned long flags;
2425    int err = -EBUSY;
2426
2427    /* The dump callback needs to be set */
2428    if (!dumper->dump)
2429        return -EINVAL;
2430
2431    spin_lock_irqsave(&dump_list_lock, flags);
2432    /* Don't allow registering multiple times */
2433    if (!dumper->registered) {
2434        dumper->registered = 1;
2435        list_add_tail_rcu(&dumper->list, &dump_list);
2436        err = 0;
2437    }
2438    spin_unlock_irqrestore(&dump_list_lock, flags);
2439
2440    return err;
2441}
2442EXPORT_SYMBOL_GPL(kmsg_dump_register);
2443
2444/**
2445 * kmsg_dump_unregister - unregister a kmsg dumper.
2446 * @dumper: pointer to the kmsg_dumper structure
2447 *
2448 * Removes a dump device from the system. Returns zero on success and
2449 * %-EINVAL otherwise.
2450 */
2451int kmsg_dump_unregister(struct kmsg_dumper *dumper)
2452{
2453    unsigned long flags;
2454    int err = -EINVAL;
2455
2456    spin_lock_irqsave(&dump_list_lock, flags);
2457    if (dumper->registered) {
2458        dumper->registered = 0;
2459        list_del_rcu(&dumper->list);
2460        err = 0;
2461    }
2462    spin_unlock_irqrestore(&dump_list_lock, flags);
2463    synchronize_rcu();
2464
2465    return err;
2466}
2467EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
2468
2469static bool always_kmsg_dump;
2470module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
2471
2472/**
2473 * kmsg_dump - dump kernel log to kernel message dumpers.
2474 * @reason: the reason (oops, panic etc) for dumping
2475 *
2476 * Call each of the registered dumper's dump() callback, which can
2477 * retrieve the kmsg records with kmsg_dump_get_line() or
2478 * kmsg_dump_get_buffer().
2479 */
2480void kmsg_dump(enum kmsg_dump_reason reason)
2481{
2482    struct kmsg_dumper *dumper;
2483    unsigned long flags;
2484
2485    if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
2486        return;
2487
2488    rcu_read_lock();
2489    list_for_each_entry_rcu(dumper, &dump_list, list) {
2490        if (dumper->max_reason && reason > dumper->max_reason)
2491            continue;
2492
2493        /* initialize iterator with data about the stored records */
2494        dumper->active = true;
2495
2496        raw_spin_lock_irqsave(&logbuf_lock, flags);
2497        dumper->cur_seq = clear_seq;
2498        dumper->cur_idx = clear_idx;
2499        dumper->next_seq = log_next_seq;
2500        dumper->next_idx = log_next_idx;
2501        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2502
2503        /* invoke dumper which will iterate over records */
2504        dumper->dump(dumper, reason);
2505
2506        /* reset iterator */
2507        dumper->active = false;
2508    }
2509    rcu_read_unlock();
2510}
2511
2512/**
2513 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
2514 * @dumper: registered kmsg dumper
2515 * @syslog: include the "<4>" prefixes
2516 * @line: buffer to copy the line to
2517 * @size: maximum size of the buffer
2518 * @len: length of line placed into buffer
2519 *
2520 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2521 * record, and copy one record into the provided buffer.
2522 *
2523 * Consecutive calls will return the next available record moving
2524 * towards the end of the buffer with the youngest messages.
2525 *
2526 * A return value of FALSE indicates that there are no more records to
2527 * read.
2528 *
2529 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
2530 */
2531bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
2532                   char *line, size_t size, size_t *len)
2533{
2534    struct log *msg;
2535    size_t l = 0;
2536    bool ret = false;
2537
2538    if (!dumper->active)
2539        goto out;
2540
2541    if (dumper->cur_seq < log_first_seq) {
2542        /* messages are gone, move to first available one */
2543        dumper->cur_seq = log_first_seq;
2544        dumper->cur_idx = log_first_idx;
2545    }
2546
2547    /* last entry */
2548    if (dumper->cur_seq >= log_next_seq)
2549        goto out;
2550
2551    msg = log_from_idx(dumper->cur_idx);
2552    l = msg_print_text(msg, 0, syslog, line, size);
2553
2554    dumper->cur_idx = log_next(dumper->cur_idx);
2555    dumper->cur_seq++;
2556    ret = true;
2557out:
2558    if (len)
2559        *len = l;
2560    return ret;
2561}
2562
2563/**
2564 * kmsg_dump_get_line - retrieve one kmsg log line
2565 * @dumper: registered kmsg dumper
2566 * @syslog: include the "<4>" prefixes
2567 * @line: buffer to copy the line to
2568 * @size: maximum size of the buffer
2569 * @len: length of line placed into buffer
2570 *
2571 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2572 * record, and copy one record into the provided buffer.
2573 *
2574 * Consecutive calls will return the next available record moving
2575 * towards the end of the buffer with the youngest messages.
2576 *
2577 * A return value of FALSE indicates that there are no more records to
2578 * read.
2579 */
2580bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
2581            char *line, size_t size, size_t *len)
2582{
2583    unsigned long flags;
2584    bool ret;
2585
2586    raw_spin_lock_irqsave(&logbuf_lock, flags);
2587    ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
2588    raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2589
2590    return ret;
2591}
2592EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
2593
2594/**
2595 * kmsg_dump_get_buffer - copy kmsg log lines
2596 * @dumper: registered kmsg dumper
2597 * @syslog: include the "<4>" prefixes
2598 * @buf: buffer to copy the line to
2599 * @size: maximum size of the buffer
2600 * @len: length of line placed into buffer
2601 *
2602 * Start at the end of the kmsg buffer and fill the provided buffer
2603 * with as many of the the *youngest* kmsg records that fit into it.
2604 * If the buffer is large enough, all available kmsg records will be
2605 * copied with a single call.
2606 *
2607 * Consecutive calls will fill the buffer with the next block of
2608 * available older records, not including the earlier retrieved ones.
2609 *
2610 * A return value of FALSE indicates that there are no more records to
2611 * read.
2612 */
2613bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
2614              char *buf, size_t size, size_t *len)
2615{
2616    unsigned long flags;
2617    u64 seq;
2618    u32 idx;
2619    u64 next_seq;
2620    u32 next_idx;
2621    enum log_flags prev;
2622    size_t l = 0;
2623    bool ret = false;
2624
2625    if (!dumper->active)
2626        goto out;
2627
2628    raw_spin_lock_irqsave(&logbuf_lock, flags);
2629    if (dumper->cur_seq < log_first_seq) {
2630        /* messages are gone, move to first available one */
2631        dumper->cur_seq = log_first_seq;
2632        dumper->cur_idx = log_first_idx;
2633    }
2634
2635    /* last entry */
2636    if (dumper->cur_seq >= dumper->next_seq) {
2637        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2638        goto out;
2639    }
2640
2641    /* calculate length of entire buffer */
2642    seq = dumper->cur_seq;
2643    idx = dumper->cur_idx;
2644    prev = 0;
2645    while (seq < dumper->next_seq) {
2646        struct log *msg = log_from_idx(idx);
2647
2648        l += msg_print_text(msg, prev, true, NULL, 0);
2649        idx = log_next(idx);
2650        seq++;
2651        prev = msg->flags;
2652    }
2653
2654    /* move first record forward until length fits into the buffer */
2655    seq = dumper->cur_seq;
2656    idx = dumper->cur_idx;
2657    prev = 0;
2658    while (l > size && seq < dumper->next_seq) {
2659        struct log *msg = log_from_idx(idx);
2660
2661        l -= msg_print_text(msg, prev, true, NULL, 0);
2662        idx = log_next(idx);
2663        seq++;
2664        prev = msg->flags;
2665    }
2666
2667    /* last message in next interation */
2668    next_seq = seq;
2669    next_idx = idx;
2670
2671    l = 0;
2672    prev = 0;
2673    while (seq < dumper->next_seq) {
2674        struct log *msg = log_from_idx(idx);
2675
2676        l += msg_print_text(msg, prev, syslog, buf + l, size - l);
2677        idx = log_next(idx);
2678        seq++;
2679        prev = msg->flags;
2680    }
2681
2682    dumper->next_seq = next_seq;
2683    dumper->next_idx = next_idx;
2684    ret = true;
2685    raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2686out:
2687    if (len)
2688        *len = l;
2689    return ret;
2690}
2691EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
2692
2693/**
2694 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
2695 * @dumper: registered kmsg dumper
2696 *
2697 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2698 * kmsg_dump_get_buffer() can be called again and used multiple
2699 * times within the same dumper.dump() callback.
2700 *
2701 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
2702 */
2703void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
2704{
2705    dumper->cur_seq = clear_seq;
2706    dumper->cur_idx = clear_idx;
2707    dumper->next_seq = log_next_seq;
2708    dumper->next_idx = log_next_idx;
2709}
2710
2711/**
2712 * kmsg_dump_rewind - reset the interator
2713 * @dumper: registered kmsg dumper
2714 *
2715 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2716 * kmsg_dump_get_buffer() can be called again and used multiple
2717 * times within the same dumper.dump() callback.
2718 */
2719void kmsg_dump_rewind(struct kmsg_dumper *dumper)
2720{
2721    unsigned long flags;
2722
2723    raw_spin_lock_irqsave(&logbuf_lock, flags);
2724    kmsg_dump_rewind_nolock(dumper);
2725    raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2726}
2727EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
2728#endif
2729

Archive Download this file



interactive