linux/kernel/printk.c
<<
>>
Prefs
   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#ifdef CONFIG_PRINTK
 220/* the next printk record to read by syslog(READ) or /proc/kmsg */
 221static u64 syslog_seq;
 222static u32 syslog_idx;
 223static enum log_flags syslog_prev;
 224static size_t syslog_partial;
 225
 226/* index and sequence number of the first record stored in the buffer */
 227static u64 log_first_seq;
 228static u32 log_first_idx;
 229
 230/* index and sequence number of the next record to store in the buffer */
 231static u64 log_next_seq;
 232static u32 log_next_idx;
 233
 234/* the next printk record to write to the console */
 235static u64 console_seq;
 236static u32 console_idx;
 237static enum log_flags console_prev;
 238
 239/* the next printk record to read after the last 'clear' command */
 240static u64 clear_seq;
 241static u32 clear_idx;
 242
 243#define PREFIX_MAX              32
 244#define LOG_LINE_MAX            1024 - PREFIX_MAX
 245
 246/* record buffer */
 247#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
 248#define LOG_ALIGN 4
 249#else
 250#define LOG_ALIGN __alignof__(struct log)
 251#endif
 252#define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
 253static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
 254static char *log_buf = __log_buf;
 255static u32 log_buf_len = __LOG_BUF_LEN;
 256
 257/* cpu currently holding logbuf_lock */
 258static volatile unsigned int logbuf_cpu = UINT_MAX;
 259
 260/* human readable text of the record */
 261static char *log_text(const struct log *msg)
 262{
 263        return (char *)msg + sizeof(struct log);
 264}
 265
 266/* optional key/value pair dictionary attached to the record */
 267static char *log_dict(const struct log *msg)
 268{
 269        return (char *)msg + sizeof(struct log) + msg->text_len;
 270}
 271
 272/* get record by index; idx must point to valid msg */
 273static struct log *log_from_idx(u32 idx)
 274{
 275        struct log *msg = (struct log *)(log_buf + idx);
 276
 277        /*
 278         * A length == 0 record is the end of buffer marker. Wrap around and
 279         * read the message at the start of the buffer.
 280         */
 281        if (!msg->len)
 282                return (struct log *)log_buf;
 283        return msg;
 284}
 285
 286/* get next record; idx must point to valid msg */
 287static u32 log_next(u32 idx)
 288{
 289        struct log *msg = (struct log *)(log_buf + idx);
 290
 291        /* length == 0 indicates the end of the buffer; wrap */
 292        /*
 293         * A length == 0 record is the end of buffer marker. Wrap around and
 294         * read the message at the start of the buffer as *this* one, and
 295         * return the one after that.
 296         */
 297        if (!msg->len) {
 298                msg = (struct log *)log_buf;
 299                return msg->len;
 300        }
 301        return idx + msg->len;
 302}
 303
 304/* insert record into the buffer, discard old ones, update heads */
 305static void log_store(int facility, int level,
 306                      enum log_flags flags, u64 ts_nsec,
 307                      const char *dict, u16 dict_len,
 308                      const char *text, u16 text_len)
 309{
 310        struct log *msg;
 311        u32 size, pad_len;
 312
 313        /* number of '\0' padding bytes to next message */
 314        size = sizeof(struct log) + text_len + dict_len;
 315        pad_len = (-size) & (LOG_ALIGN - 1);
 316        size += pad_len;
 317
 318        while (log_first_seq < log_next_seq) {
 319                u32 free;
 320
 321                if (log_next_idx > log_first_idx)
 322                        free = max(log_buf_len - log_next_idx, log_first_idx);
 323                else
 324                        free = log_first_idx - log_next_idx;
 325
 326                if (free > size + sizeof(struct log))
 327                        break;
 328
 329                /* drop old messages until we have enough contiuous space */
 330                log_first_idx = log_next(log_first_idx);
 331                log_first_seq++;
 332        }
 333
 334        if (log_next_idx + size + sizeof(struct log) >= log_buf_len) {
 335                /*
 336                 * This message + an additional empty header does not fit
 337                 * at the end of the buffer. Add an empty header with len == 0
 338                 * to signify a wrap around.
 339                 */
 340                memset(log_buf + log_next_idx, 0, sizeof(struct log));
 341                log_next_idx = 0;
 342        }
 343
 344        /* fill message */
 345        msg = (struct log *)(log_buf + log_next_idx);
 346        memcpy(log_text(msg), text, text_len);
 347        msg->text_len = text_len;
 348        memcpy(log_dict(msg), dict, dict_len);
 349        msg->dict_len = dict_len;
 350        msg->facility = facility;
 351        msg->level = level & 7;
 352        msg->flags = flags & 0x1f;
 353        if (ts_nsec > 0)
 354                msg->ts_nsec = ts_nsec;
 355        else
 356                msg->ts_nsec = local_clock();
 357        memset(log_dict(msg) + dict_len, 0, pad_len);
 358        msg->len = sizeof(struct log) + text_len + dict_len + pad_len;
 359
 360        /* insert message */
 361        log_next_idx += msg->len;
 362        log_next_seq++;
 363}
 364
 365/* /dev/kmsg - userspace message inject/listen interface */
 366struct devkmsg_user {
 367        u64 seq;
 368        u32 idx;
 369        enum log_flags prev;
 370        struct mutex lock;
 371        char buf[8192];
 372};
 373
 374static ssize_t devkmsg_writev(struct kiocb *iocb, const struct iovec *iv,
 375                              unsigned long count, loff_t pos)
 376{
 377        char *buf, *line;
 378        int i;
 379        int level = default_message_loglevel;
 380        int facility = 1;       /* LOG_USER */
 381        size_t len = iov_length(iv, count);
 382        ssize_t ret = len;
 383
 384        if (len > LOG_LINE_MAX)
 385                return -EINVAL;
 386        buf = kmalloc(len+1, GFP_KERNEL);
 387        if (buf == NULL)
 388                return -ENOMEM;
 389
 390        line = buf;
 391        for (i = 0; i < count; i++) {
 392                if (copy_from_user(line, iv[i].iov_base, iv[i].iov_len)) {
 393                        ret = -EFAULT;
 394                        goto out;
 395                }
 396                line += iv[i].iov_len;
 397        }
 398
 399        /*
 400         * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
 401         * the decimal value represents 32bit, the lower 3 bit are the log
 402         * level, the rest are the log facility.
 403         *
 404         * If no prefix or no userspace facility is specified, we
 405         * enforce LOG_USER, to be able to reliably distinguish
 406         * kernel-generated messages from userspace-injected ones.
 407         */
 408        line = buf;
 409        if (line[0] == '<') {
 410                char *endp = NULL;
 411
 412                i = simple_strtoul(line+1, &endp, 10);
 413                if (endp && endp[0] == '>') {
 414                        level = i & 7;
 415                        if (i >> 3)
 416                                facility = i >> 3;
 417                        endp++;
 418                        len -= endp - line;
 419                        line = endp;
 420                }
 421        }
 422        line[len] = '\0';
 423
 424        printk_emit(facility, level, NULL, 0, "%s", line);
 425out:
 426        kfree(buf);
 427        return ret;
 428}
 429
 430static ssize_t devkmsg_read(struct file *file, char __user *buf,
 431                            size_t count, loff_t *ppos)
 432{
 433        struct devkmsg_user *user = file->private_data;
 434        struct log *msg;
 435        u64 ts_usec;
 436        size_t i;
 437        char cont = '-';
 438        size_t len;
 439        ssize_t ret;
 440
 441        if (!user)
 442                return -EBADF;
 443
 444        ret = mutex_lock_interruptible(&user->lock);
 445        if (ret)
 446                return ret;
 447        raw_spin_lock_irq(&logbuf_lock);
 448        while (user->seq == log_next_seq) {
 449                if (file->f_flags & O_NONBLOCK) {
 450                        ret = -EAGAIN;
 451                        raw_spin_unlock_irq(&logbuf_lock);
 452                        goto out;
 453                }
 454
 455                raw_spin_unlock_irq(&logbuf_lock);
 456                ret = wait_event_interruptible(log_wait,
 457                                               user->seq != log_next_seq);
 458                if (ret)
 459                        goto out;
 460                raw_spin_lock_irq(&logbuf_lock);
 461        }
 462
 463        if (user->seq < log_first_seq) {
 464                /* our last seen message is gone, return error and reset */
 465                user->idx = log_first_idx;
 466                user->seq = log_first_seq;
 467                ret = -EPIPE;
 468                raw_spin_unlock_irq(&logbuf_lock);
 469                goto out;
 470        }
 471
 472        msg = log_from_idx(user->idx);
 473        ts_usec = msg->ts_nsec;
 474        do_div(ts_usec, 1000);
 475
 476        /*
 477         * If we couldn't merge continuation line fragments during the print,
 478         * export the stored flags to allow an optional external merge of the
 479         * records. Merging the records isn't always neccessarily correct, like
 480         * when we hit a race during printing. In most cases though, it produces
 481         * better readable output. 'c' in the record flags mark the first
 482         * fragment of a line, '+' the following.
 483         */
 484        if (msg->flags & LOG_CONT && !(user->prev & LOG_CONT))
 485                cont = 'c';
 486        else if ((msg->flags & LOG_CONT) ||
 487                 ((user->prev & LOG_CONT) && !(msg->flags & LOG_PREFIX)))
 488                cont = '+';
 489
 490        len = sprintf(user->buf, "%u,%llu,%llu,%c;",
 491                      (msg->facility << 3) | msg->level,
 492                      user->seq, ts_usec, cont);
 493        user->prev = msg->flags;
 494
 495        /* escape non-printable characters */
 496        for (i = 0; i < msg->text_len; i++) {
 497                unsigned char c = log_text(msg)[i];
 498
 499                if (c < ' ' || c >= 127 || c == '\\')
 500                        len += sprintf(user->buf + len, "\\x%02x", c);
 501                else
 502                        user->buf[len++] = c;
 503        }
 504        user->buf[len++] = '\n';
 505
 506        if (msg->dict_len) {
 507                bool line = true;
 508
 509                for (i = 0; i < msg->dict_len; i++) {
 510                        unsigned char c = log_dict(msg)[i];
 511
 512                        if (line) {
 513                                user->buf[len++] = ' ';
 514                                line = false;
 515                        }
 516
 517                        if (c == '\0') {
 518                                user->buf[len++] = '\n';
 519                                line = true;
 520                                continue;
 521                        }
 522
 523                        if (c < ' ' || c >= 127 || c == '\\') {
 524                                len += sprintf(user->buf + len, "\\x%02x", c);
 525                                continue;
 526                        }
 527
 528                        user->buf[len++] = c;
 529                }
 530                user->buf[len++] = '\n';
 531        }
 532
 533        user->idx = log_next(user->idx);
 534        user->seq++;
 535        raw_spin_unlock_irq(&logbuf_lock);
 536
 537        if (len > count) {
 538                ret = -EINVAL;
 539                goto out;
 540        }
 541
 542        if (copy_to_user(buf, user->buf, len)) {
 543                ret = -EFAULT;
 544                goto out;
 545        }
 546        ret = len;
 547out:
 548        mutex_unlock(&user->lock);
 549        return ret;
 550}
 551
 552static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
 553{
 554        struct devkmsg_user *user = file->private_data;
 555        loff_t ret = 0;
 556
 557        if (!user)
 558                return -EBADF;
 559        if (offset)
 560                return -ESPIPE;
 561
 562        raw_spin_lock_irq(&logbuf_lock);
 563        switch (whence) {
 564        case SEEK_SET:
 565                /* the first record */
 566                user->idx = log_first_idx;
 567                user->seq = log_first_seq;
 568                break;
 569        case SEEK_DATA:
 570                /*
 571                 * The first record after the last SYSLOG_ACTION_CLEAR,
 572                 * like issued by 'dmesg -c'. Reading /dev/kmsg itself
 573                 * changes no global state, and does not clear anything.
 574                 */
 575                user->idx = clear_idx;
 576                user->seq = clear_seq;
 577                break;
 578        case SEEK_END:
 579                /* after the last record */
 580                user->idx = log_next_idx;
 581                user->seq = log_next_seq;
 582                break;
 583        default:
 584                ret = -EINVAL;
 585        }
 586        raw_spin_unlock_irq(&logbuf_lock);
 587        return ret;
 588}
 589
 590static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
 591{
 592        struct devkmsg_user *user = file->private_data;
 593        int ret = 0;
 594
 595        if (!user)
 596                return POLLERR|POLLNVAL;
 597
 598        poll_wait(file, &log_wait, wait);
 599
 600        raw_spin_lock_irq(&logbuf_lock);
 601        if (user->seq < log_next_seq) {
 602                /* return error when data has vanished underneath us */
 603                if (user->seq < log_first_seq)
 604                        ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
 605                ret = POLLIN|POLLRDNORM;
 606        }
 607        raw_spin_unlock_irq(&logbuf_lock);
 608
 609        return ret;
 610}
 611
 612static int devkmsg_open(struct inode *inode, struct file *file)
 613{
 614        struct devkmsg_user *user;
 615        int err;
 616
 617        /* write-only does not need any file context */
 618        if ((file->f_flags & O_ACCMODE) == O_WRONLY)
 619                return 0;
 620
 621        err = security_syslog(SYSLOG_ACTION_READ_ALL);
 622        if (err)
 623                return err;
 624
 625        user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
 626        if (!user)
 627                return -ENOMEM;
 628
 629        mutex_init(&user->lock);
 630
 631        raw_spin_lock_irq(&logbuf_lock);
 632        user->idx = log_first_idx;
 633        user->seq = log_first_seq;
 634        raw_spin_unlock_irq(&logbuf_lock);
 635
 636        file->private_data = user;
 637        return 0;
 638}
 639
 640static int devkmsg_release(struct inode *inode, struct file *file)
 641{
 642        struct devkmsg_user *user = file->private_data;
 643
 644        if (!user)
 645                return 0;
 646
 647        mutex_destroy(&user->lock);
 648        kfree(user);
 649        return 0;
 650}
 651
 652const struct file_operations kmsg_fops = {
 653        .open = devkmsg_open,
 654        .read = devkmsg_read,
 655        .aio_write = devkmsg_writev,
 656        .llseek = devkmsg_llseek,
 657        .poll = devkmsg_poll,
 658        .release = devkmsg_release,
 659};
 660
 661#ifdef CONFIG_KEXEC
 662/*
 663 * This appends the listed symbols to /proc/vmcoreinfo
 664 *
 665 * /proc/vmcoreinfo is used by various utiilties, like crash and makedumpfile to
 666 * obtain access to symbols that are otherwise very difficult to locate.  These
 667 * symbols are specifically used so that utilities can access and extract the
 668 * dmesg log from a vmcore file after a crash.
 669 */
 670void log_buf_kexec_setup(void)
 671{
 672        VMCOREINFO_SYMBOL(log_buf);
 673        VMCOREINFO_SYMBOL(log_buf_len);
 674        VMCOREINFO_SYMBOL(log_first_idx);
 675        VMCOREINFO_SYMBOL(log_next_idx);
 676        /*
 677         * Export struct log size and field offsets. User space tools can
 678         * parse it and detect any changes to structure down the line.
 679         */
 680        VMCOREINFO_STRUCT_SIZE(log);
 681        VMCOREINFO_OFFSET(log, ts_nsec);
 682        VMCOREINFO_OFFSET(log, len);
 683        VMCOREINFO_OFFSET(log, text_len);
 684        VMCOREINFO_OFFSET(log, dict_len);
 685}
 686#endif
 687
 688/* requested log_buf_len from kernel cmdline */
 689static unsigned long __initdata new_log_buf_len;
 690
 691/* save requested log_buf_len since it's too early to process it */
 692static int __init log_buf_len_setup(char *str)
 693{
 694        unsigned size = memparse(str, &str);
 695
 696        if (size)
 697                size = roundup_pow_of_two(size);
 698        if (size > log_buf_len)
 699                new_log_buf_len = size;
 700
 701        return 0;
 702}
 703early_param("log_buf_len", log_buf_len_setup);
 704
 705void __init setup_log_buf(int early)
 706{
 707        unsigned long flags;
 708        char *new_log_buf;
 709        int free;
 710
 711        if (!new_log_buf_len)
 712                return;
 713
 714        if (early) {
 715                unsigned long mem;
 716
 717                mem = memblock_alloc(new_log_buf_len, PAGE_SIZE);
 718                if (!mem)
 719                        return;
 720                new_log_buf = __va(mem);
 721        } else {
 722                new_log_buf = alloc_bootmem_nopanic(new_log_buf_len);
 723        }
 724
 725        if (unlikely(!new_log_buf)) {
 726                pr_err("log_buf_len: %ld bytes not available\n",
 727                        new_log_buf_len);
 728                return;
 729        }
 730
 731        raw_spin_lock_irqsave(&logbuf_lock, flags);
 732        log_buf_len = new_log_buf_len;
 733        log_buf = new_log_buf;
 734        new_log_buf_len = 0;
 735        free = __LOG_BUF_LEN - log_next_idx;
 736        memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
 737        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
 738
 739        pr_info("log_buf_len: %d\n", log_buf_len);
 740        pr_info("early log buf free: %d(%d%%)\n",
 741                free, (free * 100) / __LOG_BUF_LEN);
 742}
 743
 744#ifdef CONFIG_BOOT_PRINTK_DELAY
 745
 746static int boot_delay; /* msecs delay after each printk during bootup */
 747static unsigned long long loops_per_msec;       /* based on boot_delay */
 748
 749static int __init boot_delay_setup(char *str)
 750{
 751        unsigned long lpj;
 752
 753        lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
 754        loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
 755
 756        get_option(&str, &boot_delay);
 757        if (boot_delay > 10 * 1000)
 758                boot_delay = 0;
 759
 760        pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
 761                "HZ: %d, loops_per_msec: %llu\n",
 762                boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
 763        return 1;
 764}
 765__setup("boot_delay=", boot_delay_setup);
 766
 767static void boot_delay_msec(void)
 768{
 769        unsigned long long k;
 770        unsigned long timeout;
 771
 772        if (boot_delay == 0 || system_state != SYSTEM_BOOTING)
 773                return;
 774
 775        k = (unsigned long long)loops_per_msec * boot_delay;
 776
 777        timeout = jiffies + msecs_to_jiffies(boot_delay);
 778        while (k) {
 779                k--;
 780                cpu_relax();
 781                /*
 782                 * use (volatile) jiffies to prevent
 783                 * compiler reduction; loop termination via jiffies
 784                 * is secondary and may or may not happen.
 785                 */
 786                if (time_after(jiffies, timeout))
 787                        break;
 788                touch_nmi_watchdog();
 789        }
 790}
 791#else
 792static inline void boot_delay_msec(void)
 793{
 794}
 795#endif
 796
 797#ifdef CONFIG_SECURITY_DMESG_RESTRICT
 798int dmesg_restrict = 1;
 799#else
 800int dmesg_restrict;
 801#endif
 802
 803static int syslog_action_restricted(int type)
 804{
 805        if (dmesg_restrict)
 806                return 1;
 807        /* Unless restricted, we allow "read all" and "get buffer size" for everybody */
 808        return type != SYSLOG_ACTION_READ_ALL && type != SYSLOG_ACTION_SIZE_BUFFER;
 809}
 810
 811static int check_syslog_permissions(int type, bool from_file)
 812{
 813        /*
 814         * If this is from /proc/kmsg and we've already opened it, then we've
 815         * already done the capabilities checks at open time.
 816         */
 817        if (from_file && type != SYSLOG_ACTION_OPEN)
 818                return 0;
 819
 820        if (syslog_action_restricted(type)) {
 821                if (capable(CAP_SYSLOG))
 822                        return 0;
 823                /* For historical reasons, accept CAP_SYS_ADMIN too, with a warning */
 824                if (capable(CAP_SYS_ADMIN)) {
 825                        printk_once(KERN_WARNING "%s (%d): "
 826                                 "Attempt to access syslog with CAP_SYS_ADMIN "
 827                                 "but no CAP_SYSLOG (deprecated).\n",
 828                                 current->comm, task_pid_nr(current));
 829                        return 0;
 830                }
 831                return -EPERM;
 832        }
 833        return 0;
 834}
 835
 836#if defined(CONFIG_PRINTK_TIME)
 837static bool printk_time = 1;
 838#else
 839static bool printk_time;
 840#endif
 841module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
 842
 843static size_t print_time(u64 ts, char *buf)
 844{
 845        unsigned long rem_nsec;
 846
 847        if (!printk_time)
 848                return 0;
 849
 850        rem_nsec = do_div(ts, 1000000000);
 851
 852        if (!buf)
 853                return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
 854
 855        return sprintf(buf, "[%5lu.%06lu] ",
 856                       (unsigned long)ts, rem_nsec / 1000);
 857}
 858
 859static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
 860{
 861        size_t len = 0;
 862        unsigned int prefix = (msg->facility << 3) | msg->level;
 863
 864        if (syslog) {
 865                if (buf) {
 866                        len += sprintf(buf, "<%u>", prefix);
 867                } else {
 868                        len += 3;
 869                        if (prefix > 999)
 870                                len += 3;
 871                        else if (prefix > 99)
 872                                len += 2;
 873                        else if (prefix > 9)
 874                                len++;
 875                }
 876        }
 877
 878        len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
 879        return len;
 880}
 881
 882static size_t msg_print_text(const struct log *msg, enum log_flags prev,
 883                             bool syslog, char *buf, size_t size)
 884{
 885        const char *text = log_text(msg);
 886        size_t text_size = msg->text_len;
 887        bool prefix = true;
 888        bool newline = true;
 889        size_t len = 0;
 890
 891        if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
 892                prefix = false;
 893
 894        if (msg->flags & LOG_CONT) {
 895                if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
 896                        prefix = false;
 897
 898                if (!(msg->flags & LOG_NEWLINE))
 899                        newline = false;
 900        }
 901
 902        do {
 903                const char *next = memchr(text, '\n', text_size);
 904                size_t text_len;
 905
 906                if (next) {
 907                        text_len = next - text;
 908                        next++;
 909                        text_size -= next - text;
 910                } else {
 911                        text_len = text_size;
 912                }
 913
 914                if (buf) {
 915                        if (print_prefix(msg, syslog, NULL) +
 916                            text_len + 1 >= size - len)
 917                                break;
 918
 919                        if (prefix)
 920                                len += print_prefix(msg, syslog, buf + len);
 921                        memcpy(buf + len, text, text_len);
 922                        len += text_len;
 923                        if (next || newline)
 924                                buf[len++] = '\n';
 925                } else {
 926                        /* SYSLOG_ACTION_* buffer size only calculation */
 927                        if (prefix)
 928                                len += print_prefix(msg, syslog, NULL);
 929                        len += text_len;
 930                        if (next || newline)
 931                                len++;
 932                }
 933
 934                prefix = true;
 935                text = next;
 936        } while (text);
 937
 938        return len;
 939}
 940
 941static int syslog_print(char __user *buf, int size)
 942{
 943        char *text;
 944        struct log *msg;
 945        int len = 0;
 946
 947        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
 948        if (!text)
 949                return -ENOMEM;
 950
 951        while (size > 0) {
 952                size_t n;
 953                size_t skip;
 954
 955                raw_spin_lock_irq(&logbuf_lock);
 956                if (syslog_seq < log_first_seq) {
 957                        /* messages are gone, move to first one */
 958                        syslog_seq = log_first_seq;
 959                        syslog_idx = log_first_idx;
 960                        syslog_prev = 0;
 961                        syslog_partial = 0;
 962                }
 963                if (syslog_seq == log_next_seq) {
 964                        raw_spin_unlock_irq(&logbuf_lock);
 965                        break;
 966                }
 967
 968                skip = syslog_partial;
 969                msg = log_from_idx(syslog_idx);
 970                n = msg_print_text(msg, syslog_prev, true, text,
 971                                   LOG_LINE_MAX + PREFIX_MAX);
 972                if (n - syslog_partial <= size) {
 973                        /* message fits into buffer, move forward */
 974                        syslog_idx = log_next(syslog_idx);
 975                        syslog_seq++;
 976                        syslog_prev = msg->flags;
 977                        n -= syslog_partial;
 978                        syslog_partial = 0;
 979                } else if (!len){
 980                        /* partial read(), remember position */
 981                        n = size;
 982                        syslog_partial += n;
 983                } else
 984                        n = 0;
 985                raw_spin_unlock_irq(&logbuf_lock);
 986
 987                if (!n)
 988                        break;
 989
 990                if (copy_to_user(buf, text + skip, n)) {
 991                        if (!len)
 992                                len = -EFAULT;
 993                        break;
 994                }
 995
 996                len += n;
 997                size -= n;
 998                buf += n;
 999        }
1000
1001        kfree(text);
1002        return len;
1003}
1004
1005static int syslog_print_all(char __user *buf, int size, bool clear)
1006{
1007        char *text;
1008        int len = 0;
1009
1010        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1011        if (!text)
1012                return -ENOMEM;
1013
1014        raw_spin_lock_irq(&logbuf_lock);
1015        if (buf) {
1016                u64 next_seq;
1017                u64 seq;
1018                u32 idx;
1019                enum log_flags prev;
1020
1021                if (clear_seq < log_first_seq) {
1022                        /* messages are gone, move to first available one */
1023                        clear_seq = log_first_seq;
1024                        clear_idx = log_first_idx;
1025                }
1026
1027                /*
1028                 * Find first record that fits, including all following records,
1029                 * into the user-provided buffer for this dump.
1030                 */
1031                seq = clear_seq;
1032                idx = clear_idx;
1033                prev = 0;
1034                while (seq < log_next_seq) {
1035                        struct log *msg = log_from_idx(idx);
1036
1037                        len += msg_print_text(msg, prev, true, NULL, 0);
1038                        prev = msg->flags;
1039                        idx = log_next(idx);
1040                        seq++;
1041                }
1042
1043                /* move first record forward until length fits into the buffer */
1044                seq = clear_seq;
1045                idx = clear_idx;
1046                prev = 0;
1047                while (len > size && seq < log_next_seq) {
1048                        struct log *msg = log_from_idx(idx);
1049
1050                        len -= msg_print_text(msg, prev, true, NULL, 0);
1051                        prev = msg->flags;
1052                        idx = log_next(idx);
1053                        seq++;
1054                }
1055
1056                /* last message fitting into this dump */
1057                next_seq = log_next_seq;
1058
1059                len = 0;
1060                prev = 0;
1061                while (len >= 0 && seq < next_seq) {
1062                        struct log *msg = log_from_idx(idx);
1063                        int textlen;
1064
1065                        textlen = msg_print_text(msg, prev, true, text,
1066                                                 LOG_LINE_MAX + PREFIX_MAX);
1067                        if (textlen < 0) {
1068                                len = textlen;
1069                                break;
1070                        }
1071                        idx = log_next(idx);
1072                        seq++;
1073                        prev = msg->flags;
1074
1075                        raw_spin_unlock_irq(&logbuf_lock);
1076                        if (copy_to_user(buf + len, text, textlen))
1077                                len = -EFAULT;
1078                        else
1079                                len += textlen;
1080                        raw_spin_lock_irq(&logbuf_lock);
1081
1082                        if (seq < log_first_seq) {
1083                                /* messages are gone, move to next one */
1084                                seq = log_first_seq;
1085                                idx = log_first_idx;
1086                                prev = 0;
1087                        }
1088                }
1089        }
1090
1091        if (clear) {
1092                clear_seq = log_next_seq;
1093                clear_idx = log_next_idx;
1094        }
1095        raw_spin_unlock_irq(&logbuf_lock);
1096
1097        kfree(text);
1098        return len;
1099}
1100
1101int do_syslog(int type, char __user *buf, int len, bool from_file)
1102{
1103        bool clear = false;
1104        static int saved_console_loglevel = -1;
1105        int error;
1106
1107        error = check_syslog_permissions(type, from_file);
1108        if (error)
1109                goto out;
1110
1111        error = security_syslog(type);
1112        if (error)
1113                return error;
1114
1115        switch (type) {
1116        case SYSLOG_ACTION_CLOSE:       /* Close log */
1117                break;
1118        case SYSLOG_ACTION_OPEN:        /* Open log */
1119                break;
1120        case SYSLOG_ACTION_READ:        /* Read from log */
1121                error = -EINVAL;
1122                if (!buf || len < 0)
1123                        goto out;
1124                error = 0;
1125                if (!len)
1126                        goto out;
1127                if (!access_ok(VERIFY_WRITE, buf, len)) {
1128                        error = -EFAULT;
1129                        goto out;
1130                }
1131                error = wait_event_interruptible(log_wait,
1132                                                 syslog_seq != log_next_seq);
1133                if (error)
1134                        goto out;
1135                error = syslog_print(buf, len);
1136                break;
1137        /* Read/clear last kernel messages */
1138        case SYSLOG_ACTION_READ_CLEAR:
1139                clear = true;
1140                /* FALL THRU */
1141        /* Read last kernel messages */
1142        case SYSLOG_ACTION_READ_ALL:
1143                error = -EINVAL;
1144                if (!buf || len < 0)
1145                        goto out;
1146                error = 0;
1147                if (!len)
1148                        goto out;
1149                if (!access_ok(VERIFY_WRITE, buf, len)) {
1150                        error = -EFAULT;
1151                        goto out;
1152                }
1153                error = syslog_print_all(buf, len, clear);
1154                break;
1155        /* Clear ring buffer */
1156        case SYSLOG_ACTION_CLEAR:
1157                syslog_print_all(NULL, 0, true);
1158                break;
1159        /* Disable logging to console */
1160        case SYSLOG_ACTION_CONSOLE_OFF:
1161                if (saved_console_loglevel == -1)
1162                        saved_console_loglevel = console_loglevel;
1163                console_loglevel = minimum_console_loglevel;
1164                break;
1165        /* Enable logging to console */
1166        case SYSLOG_ACTION_CONSOLE_ON:
1167                if (saved_console_loglevel != -1) {
1168                        console_loglevel = saved_console_loglevel;
1169                        saved_console_loglevel = -1;
1170                }
1171                break;
1172        /* Set level of messages printed to console */
1173        case SYSLOG_ACTION_CONSOLE_LEVEL:
1174                error = -EINVAL;
1175                if (len < 1 || len > 8)
1176                        goto out;
1177                if (len < minimum_console_loglevel)
1178                        len = minimum_console_loglevel;
1179                console_loglevel = len;
1180                /* Implicitly re-enable logging to console */
1181                saved_console_loglevel = -1;
1182                error = 0;
1183                break;
1184        /* Number of chars in the log buffer */
1185        case SYSLOG_ACTION_SIZE_UNREAD:
1186                raw_spin_lock_irq(&logbuf_lock);
1187                if (syslog_seq < log_first_seq) {
1188                        /* messages are gone, move to first one */
1189                        syslog_seq = log_first_seq;
1190                        syslog_idx = log_first_idx;
1191                        syslog_prev = 0;
1192                        syslog_partial = 0;
1193                }
1194                if (from_file) {
1195                        /*
1196                         * Short-cut for poll(/"proc/kmsg") which simply checks
1197                         * for pending data, not the size; return the count of
1198                         * records, not the length.
1199                         */
1200                        error = log_next_idx - syslog_idx;
1201                } else {
1202                        u64 seq = syslog_seq;
1203                        u32 idx = syslog_idx;
1204                        enum log_flags prev = syslog_prev;
1205
1206                        error = 0;
1207                        while (seq < log_next_seq) {
1208                                struct log *msg = log_from_idx(idx);
1209
1210                                error += msg_print_text(msg, prev, true, NULL, 0);
1211                                idx = log_next(idx);
1212                                seq++;
1213                                prev = msg->flags;
1214                        }
1215                        error -= syslog_partial;
1216                }
1217                raw_spin_unlock_irq(&logbuf_lock);
1218                break;
1219        /* Size of the log buffer */
1220        case SYSLOG_ACTION_SIZE_BUFFER:
1221                error = log_buf_len;
1222                break;
1223        default:
1224                error = -EINVAL;
1225                break;
1226        }
1227out:
1228        return error;
1229}
1230
1231SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1232{
1233        return do_syslog(type, buf, len, SYSLOG_FROM_CALL);
1234}
1235
1236static bool __read_mostly ignore_loglevel;
1237
1238static int __init ignore_loglevel_setup(char *str)
1239{
1240        ignore_loglevel = 1;
1241        printk(KERN_INFO "debug: ignoring loglevel setting.\n");
1242
1243        return 0;
1244}
1245
1246early_param("ignore_loglevel", ignore_loglevel_setup);
1247module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1248MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
1249        "print all kernel messages to the console.");
1250
1251/*
1252 * Call the console drivers, asking them to write out
1253 * log_buf[start] to log_buf[end - 1].
1254 * The console_lock must be held.
1255 */
1256static void call_console_drivers(int level, const char *text, size_t len)
1257{
1258        struct console *con;
1259
1260        trace_console(text, 0, len, len);
1261
1262        if (level >= console_loglevel && !ignore_loglevel)
1263                return;
1264        if (!console_drivers)
1265                return;
1266
1267        for_each_console(con) {
1268                if (exclusive_console && con != exclusive_console)
1269                        continue;
1270                if (!(con->flags & CON_ENABLED))
1271                        continue;
1272                if (!con->write)
1273                        continue;
1274                if (!cpu_online(smp_processor_id()) &&
1275                    !(con->flags & CON_ANYTIME))
1276                        continue;
1277                con->write(con, text, len);
1278        }
1279}
1280
1281/*
1282 * Zap console related locks when oopsing. Only zap at most once
1283 * every 10 seconds, to leave time for slow consoles to print a
1284 * full oops.
1285 */
1286static void zap_locks(void)
1287{
1288        static unsigned long oops_timestamp;
1289
1290        if (time_after_eq(jiffies, oops_timestamp) &&
1291                        !time_after(jiffies, oops_timestamp + 30 * HZ))
1292                return;
1293
1294        oops_timestamp = jiffies;
1295
1296        debug_locks_off();
1297        /* If a crash is occurring, make sure we can't deadlock */
1298        raw_spin_lock_init(&logbuf_lock);
1299        /* And make sure that we print immediately */
1300        sema_init(&console_sem, 1);
1301}
1302
1303/* Check if we have any console registered that can be called early in boot. */
1304static int have_callable_console(void)
1305{
1306        struct console *con;
1307
1308        for_each_console(con)
1309                if (con->flags & CON_ANYTIME)
1310                        return 1;
1311
1312        return 0;
1313}
1314
1315/*
1316 * Can we actually use the console at this time on this cpu?
1317 *
1318 * Console drivers may assume that per-cpu resources have
1319 * been allocated. So unless they're explicitly marked as
1320 * being able to cope (CON_ANYTIME) don't call them until
1321 * this CPU is officially up.
1322 */
1323static inline int can_use_console(unsigned int cpu)
1324{
1325        return cpu_online(cpu) || have_callable_console();
1326}
1327
1328/*
1329 * Try to get console ownership to actually show the kernel
1330 * messages from a 'printk'. Return true (and with the
1331 * console_lock held, and 'console_locked' set) if it
1332 * is successful, false otherwise.
1333 *
1334 * This gets called with the 'logbuf_lock' spinlock held and
1335 * interrupts disabled. It should return with 'lockbuf_lock'
1336 * released but interrupts still disabled.
1337 */
1338static int console_trylock_for_printk(unsigned int cpu)
1339        __releases(&logbuf_lock)
1340{
1341        int retval = 0, wake = 0;
1342
1343        if (console_trylock()) {
1344                retval = 1;
1345
1346                /*
1347                 * If we can't use the console, we need to release
1348                 * the console semaphore by hand to avoid flushing
1349                 * the buffer. We need to hold the console semaphore
1350                 * in order to do this test safely.
1351                 */
1352                if (!can_use_console(cpu)) {
1353                        console_locked = 0;
1354                        wake = 1;
1355                        retval = 0;
1356                }
1357        }
1358        logbuf_cpu = UINT_MAX;
1359        if (wake)
1360                up(&console_sem);
1361        raw_spin_unlock(&logbuf_lock);
1362        return retval;
1363}
1364
1365int printk_delay_msec __read_mostly;
1366
1367static inline void printk_delay(void)
1368{
1369        if (unlikely(printk_delay_msec)) {
1370                int m = printk_delay_msec;
1371
1372                while (m--) {
1373                        mdelay(1);
1374                        touch_nmi_watchdog();
1375                }
1376        }
1377}
1378
1379/*
1380 * Continuation lines are buffered, and not committed to the record buffer
1381 * until the line is complete, or a race forces it. The line fragments
1382 * though, are printed immediately to the consoles to ensure everything has
1383 * reached the console in case of a kernel crash.
1384 */
1385static struct cont {
1386        char buf[LOG_LINE_MAX];
1387        size_t len;                     /* length == 0 means unused buffer */
1388        size_t cons;                    /* bytes written to console */
1389        struct task_struct *owner;      /* task of first print*/
1390        u64 ts_nsec;                    /* time of first print */
1391        u8 level;                       /* log level of first message */
1392        u8 facility;                    /* log level of first message */
1393        enum log_flags flags;           /* prefix, newline flags */
1394        bool flushed:1;                 /* buffer sealed and committed */
1395} cont;
1396
1397static void cont_flush(enum log_flags flags)
1398{
1399        if (cont.flushed)
1400                return;
1401        if (cont.len == 0)
1402                return;
1403
1404        if (cont.cons) {
1405                /*
1406                 * If a fragment of this line was directly flushed to the
1407                 * console; wait for the console to pick up the rest of the
1408                 * line. LOG_NOCONS suppresses a duplicated output.
1409                 */
1410                log_store(cont.facility, cont.level, flags | LOG_NOCONS,
1411                          cont.ts_nsec, NULL, 0, cont.buf, cont.len);
1412                cont.flags = flags;
1413                cont.flushed = true;
1414        } else {
1415                /*
1416                 * If no fragment of this line ever reached the console,
1417                 * just submit it to the store and free the buffer.
1418                 */
1419                log_store(cont.facility, cont.level, flags, 0,
1420                          NULL, 0, cont.buf, cont.len);
1421                cont.len = 0;
1422        }
1423}
1424
1425static bool cont_add(int facility, int level, const char *text, size_t len)
1426{
1427        if (cont.len && cont.flushed)
1428                return false;
1429
1430        if (cont.len + len > sizeof(cont.buf)) {
1431                /* the line gets too long, split it up in separate records */
1432                cont_flush(LOG_CONT);
1433                return false;
1434        }
1435
1436        if (!cont.len) {
1437                cont.facility = facility;
1438                cont.level = level;
1439                cont.owner = current;
1440                cont.ts_nsec = local_clock();
1441                cont.flags = 0;
1442                cont.cons = 0;
1443                cont.flushed = false;
1444        }
1445
1446        memcpy(cont.buf + cont.len, text, len);
1447        cont.len += len;
1448
1449        if (cont.len > (sizeof(cont.buf) * 80) / 100)
1450                cont_flush(LOG_CONT);
1451
1452        return true;
1453}
1454
1455static size_t cont_print_text(char *text, size_t size)
1456{
1457        size_t textlen = 0;
1458        size_t len;
1459
1460        if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
1461                textlen += print_time(cont.ts_nsec, text);
1462                size -= textlen;
1463        }
1464
1465        len = cont.len - cont.cons;
1466        if (len > 0) {
1467                if (len+1 > size)
1468                        len = size-1;
1469                memcpy(text + textlen, cont.buf + cont.cons, len);
1470                textlen += len;
1471                cont.cons = cont.len;
1472        }
1473
1474        if (cont.flushed) {
1475                if (cont.flags & LOG_NEWLINE)
1476                        text[textlen++] = '\n';
1477                /* got everything, release buffer */
1478                cont.len = 0;
1479        }
1480        return textlen;
1481}
1482
1483asmlinkage int vprintk_emit(int facility, int level,
1484                            const char *dict, size_t dictlen,
1485                            const char *fmt, va_list args)
1486{
1487        static int recursion_bug;
1488        static char textbuf[LOG_LINE_MAX];
1489        char *text = textbuf;
1490        size_t text_len;
1491        enum log_flags lflags = 0;
1492        unsigned long flags;
1493        int this_cpu;
1494        int printed_len = 0;
1495
1496        boot_delay_msec();
1497        printk_delay();
1498
1499        /* This stops the holder of console_sem just where we want him */
1500        local_irq_save(flags);
1501        this_cpu = smp_processor_id();
1502
1503        /*
1504         * Ouch, printk recursed into itself!
1505         */
1506        if (unlikely(logbuf_cpu == this_cpu)) {
1507                /*
1508                 * If a crash is occurring during printk() on this CPU,
1509                 * then try to get the crash message out but make sure
1510                 * we can't deadlock. Otherwise just return to avoid the
1511                 * recursion and return - but flag the recursion so that
1512                 * it can be printed at the next appropriate moment:
1513                 */
1514                if (!oops_in_progress && !lockdep_recursing(current)) {
1515                        recursion_bug = 1;
1516                        goto out_restore_irqs;
1517                }
1518                zap_locks();
1519        }
1520
1521        lockdep_off();
1522        raw_spin_lock(&logbuf_lock);
1523        logbuf_cpu = this_cpu;
1524
1525        if (recursion_bug) {
1526                static const char recursion_msg[] =
1527                        "BUG: recent printk recursion!";
1528
1529                recursion_bug = 0;
1530                printed_len += strlen(recursion_msg);
1531                /* emit KERN_CRIT message */
1532                log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1533                          NULL, 0, recursion_msg, printed_len);
1534        }
1535
1536        /*
1537         * The printf needs to come first; we need the syslog
1538         * prefix which might be passed-in as a parameter.
1539         */
1540        text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1541
1542        /* mark and strip a trailing newline */
1543        if (text_len && text[text_len-1] == '\n') {
1544                text_len--;
1545                lflags |= LOG_NEWLINE;
1546        }
1547
1548        /* strip kernel syslog prefix and extract log level or control flags */
1549        if (facility == 0) {
1550                int kern_level = printk_get_level(text);
1551
1552                if (kern_level) {
1553                        const char *end_of_header = printk_skip_level(text);
1554                        switch (kern_level) {
1555                        case '0' ... '7':
1556                                if (level == -1)
1557                                        level = kern_level - '0';
1558                        case 'd':       /* KERN_DEFAULT */
1559                                lflags |= LOG_PREFIX;
1560                        case 'c':       /* KERN_CONT */
1561                                break;
1562                        }
1563                        text_len -= end_of_header - text;
1564                        text = (char *)end_of_header;
1565                }
1566        }
1567
1568        if (level == -1)
1569                level = default_message_loglevel;
1570
1571        if (dict)
1572                lflags |= LOG_PREFIX|LOG_NEWLINE;
1573
1574        if (!(lflags & LOG_NEWLINE)) {
1575                /*
1576                 * Flush the conflicting buffer. An earlier newline was missing,
1577                 * or another task also prints continuation lines.
1578                 */
1579                if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
1580                        cont_flush(LOG_NEWLINE);
1581
1582                /* buffer line if possible, otherwise store it right away */
1583                if (!cont_add(facility, level, text, text_len))
1584                        log_store(facility, level, lflags | LOG_CONT, 0,
1585                                  dict, dictlen, text, text_len);
1586        } else {
1587                bool stored = false;
1588
1589                /*
1590                 * If an earlier newline was missing and it was the same task,
1591                 * either merge it with the current buffer and flush, or if
1592                 * there was a race with interrupts (prefix == true) then just
1593                 * flush it out and store this line separately.
1594                 */
1595                if (cont.len && cont.owner == current) {
1596                        if (!(lflags & LOG_PREFIX))
1597                                stored = cont_add(facility, level, text, text_len);
1598                        cont_flush(LOG_NEWLINE);
1599                }
1600
1601                if (!stored)
1602                        log_store(facility, level, lflags, 0,
1603                                  dict, dictlen, text, text_len);
1604        }
1605        printed_len += text_len;
1606
1607        /*
1608         * Try to acquire and then immediately release the console semaphore.
1609         * The release will print out buffers and wake up /dev/kmsg and syslog()
1610         * users.
1611         *
1612         * The console_trylock_for_printk() function will release 'logbuf_lock'
1613         * regardless of whether it actually gets the console semaphore or not.
1614         */
1615        if (console_trylock_for_printk(this_cpu))
1616                console_unlock();
1617
1618        lockdep_on();
1619out_restore_irqs:
1620        local_irq_restore(flags);
1621
1622        return printed_len;
1623}
1624EXPORT_SYMBOL(vprintk_emit);
1625
1626asmlinkage int vprintk(const char *fmt, va_list args)
1627{
1628        return vprintk_emit(0, -1, NULL, 0, fmt, args);
1629}
1630EXPORT_SYMBOL(vprintk);
1631
1632asmlinkage int printk_emit(int facility, int level,
1633                           const char *dict, size_t dictlen,
1634                           const char *fmt, ...)
1635{
1636        va_list args;
1637        int r;
1638
1639        va_start(args, fmt);
1640        r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1641        va_end(args);
1642
1643        return r;
1644}
1645EXPORT_SYMBOL(printk_emit);
1646
1647/**
1648 * printk - print a kernel message
1649 * @fmt: format string
1650 *
1651 * This is printk(). It can be called from any context. We want it to work.
1652 *
1653 * We try to grab the console_lock. If we succeed, it's easy - we log the
1654 * output and call the console drivers.  If we fail to get the semaphore, we
1655 * place the output into the log buffer and return. The current holder of
1656 * the console_sem will notice the new output in console_unlock(); and will
1657 * send it to the consoles before releasing the lock.
1658 *
1659 * One effect of this deferred printing is that code which calls printk() and
1660 * then changes console_loglevel may break. This is because console_loglevel
1661 * is inspected when the actual printing occurs.
1662 *
1663 * See also:
1664 * printf(3)
1665 *
1666 * See the vsnprintf() documentation for format string extensions over C99.
1667 */
1668asmlinkage int printk(const char *fmt, ...)
1669{
1670        va_list args;
1671        int r;
1672
1673#ifdef CONFIG_KGDB_KDB
1674        if (unlikely(kdb_trap_printk)) {
1675                va_start(args, fmt);
1676                r = vkdb_printf(fmt, args);
1677                va_end(args);
1678                return r;
1679        }
1680#endif
1681        va_start(args, fmt);
1682        r = vprintk_emit(0, -1, NULL, 0, fmt, args);
1683        va_end(args);
1684
1685        return r;
1686}
1687EXPORT_SYMBOL(printk);
1688
1689#else /* CONFIG_PRINTK */
1690
1691#define LOG_LINE_MAX            0
1692#define PREFIX_MAX              0
1693#define LOG_LINE_MAX 0
1694static u64 syslog_seq;
1695static u32 syslog_idx;
1696static u64 console_seq;
1697static u32 console_idx;
1698static enum log_flags syslog_prev;
1699static u64 log_first_seq;
1700static u32 log_first_idx;
1701static u64 log_next_seq;
1702static enum log_flags console_prev;
1703static struct cont {
1704        size_t len;
1705        size_t cons;
1706        u8 level;
1707        bool flushed:1;
1708} cont;
1709static struct log *log_from_idx(u32 idx) { return NULL; }
1710static u32 log_next(u32 idx) { return 0; }
1711static void call_console_drivers(int level, const char *text, size_t len) {}
1712static size_t msg_print_text(const struct log *msg, enum log_flags prev,
1713                             bool syslog, char *buf, size_t size) { return 0; }
1714static size_t cont_print_text(char *text, size_t size) { return 0; }
1715
1716#endif /* CONFIG_PRINTK */
1717
1718static int __add_preferred_console(char *name, int idx, char *options,
1719                                   char *brl_options)
1720{
1721        struct console_cmdline *c;
1722        int i;
1723
1724        /*
1725         *      See if this tty is not yet registered, and
1726         *      if we have a slot free.
1727         */
1728        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1729                if (strcmp(console_cmdline[i].name, name) == 0 &&
1730                          console_cmdline[i].index == idx) {
1731                                if (!brl_options)
1732                                        selected_console = i;
1733                                return 0;
1734                }
1735        if (i == MAX_CMDLINECONSOLES)
1736                return -E2BIG;
1737        if (!brl_options)
1738                selected_console = i;
1739        c = &console_cmdline[i];
1740        strlcpy(c->name, name, sizeof(c->name));
1741        c->options = options;
1742#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1743        c->brl_options = brl_options;
1744#endif
1745        c->index = idx;
1746        return 0;
1747}
1748/*
1749 * Set up a list of consoles.  Called from init/main.c
1750 */
1751static int __init console_setup(char *str)
1752{
1753        char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
1754        char *s, *options, *brl_options = NULL;
1755        int idx;
1756
1757#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1758        if (!memcmp(str, "brl,", 4)) {
1759                brl_options = "";
1760                str += 4;
1761        } else if (!memcmp(str, "brl=", 4)) {
1762                brl_options = str + 4;
1763                str = strchr(brl_options, ',');
1764                if (!str) {
1765                        printk(KERN_ERR "need port name after brl=\n");
1766                        return 1;
1767                }
1768                *(str++) = 0;
1769        }
1770#endif
1771
1772        /*
1773         * Decode str into name, index, options.
1774         */
1775        if (str[0] >= '0' && str[0] <= '9') {
1776                strcpy(buf, "ttyS");
1777                strncpy(buf + 4, str, sizeof(buf) - 5);
1778        } else {
1779                strncpy(buf, str, sizeof(buf) - 1);
1780        }
1781        buf[sizeof(buf) - 1] = 0;
1782        if ((options = strchr(str, ',')) != NULL)
1783                *(options++) = 0;
1784#ifdef __sparc__
1785        if (!strcmp(str, "ttya"))
1786                strcpy(buf, "ttyS0");
1787        if (!strcmp(str, "ttyb"))
1788                strcpy(buf, "ttyS1");
1789#endif
1790        for (s = buf; *s; s++)
1791                if ((*s >= '0' && *s <= '9') || *s == ',')
1792                        break;
1793        idx = simple_strtoul(s, NULL, 10);
1794        *s = 0;
1795
1796        __add_preferred_console(buf, idx, options, brl_options);
1797        console_set_on_cmdline = 1;
1798        return 1;
1799}
1800__setup("console=", console_setup);
1801
1802/**
1803 * add_preferred_console - add a device to the list of preferred consoles.
1804 * @name: device name
1805 * @idx: device index
1806 * @options: options for this console
1807 *
1808 * The last preferred console added will be used for kernel messages
1809 * and stdin/out/err for init.  Normally this is used by console_setup
1810 * above to handle user-supplied console arguments; however it can also
1811 * be used by arch-specific code either to override the user or more
1812 * commonly to provide a default console (ie from PROM variables) when
1813 * the user has not supplied one.
1814 */
1815int add_preferred_console(char *name, int idx, char *options)
1816{
1817        return __add_preferred_console(name, idx, options, NULL);
1818}
1819
1820int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
1821{
1822        struct console_cmdline *c;
1823        int i;
1824
1825        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1826                if (strcmp(console_cmdline[i].name, name) == 0 &&
1827                          console_cmdline[i].index == idx) {
1828                                c = &console_cmdline[i];
1829                                strlcpy(c->name, name_new, sizeof(c->name));
1830                                c->name[sizeof(c->name) - 1] = 0;
1831                                c->options = options;
1832                                c->index = idx_new;
1833                                return i;
1834                }
1835        /* not found */
1836        return -1;
1837}
1838
1839bool console_suspend_enabled = 1;
1840EXPORT_SYMBOL(console_suspend_enabled);
1841
1842static int __init console_suspend_disable(char *str)
1843{
1844        console_suspend_enabled = 0;
1845        return 1;
1846}
1847__setup("no_console_suspend", console_suspend_disable);
1848module_param_named(console_suspend, console_suspend_enabled,
1849                bool, S_IRUGO | S_IWUSR);
1850MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
1851        " and hibernate operations");
1852
1853/**
1854 * suspend_console - suspend the console subsystem
1855 *
1856 * This disables printk() while we go into suspend states
1857 */
1858void suspend_console(void)
1859{
1860        if (!console_suspend_enabled)
1861                return;
1862        printk("Suspending console(s) (use no_console_suspend to debug)\n");
1863        console_lock();
1864        console_suspended = 1;
1865        up(&console_sem);
1866}
1867
1868void resume_console(void)
1869{
1870        if (!console_suspend_enabled)
1871                return;
1872        down(&console_sem);
1873        console_suspended = 0;
1874        console_unlock();
1875}
1876
1877/**
1878 * console_cpu_notify - print deferred console messages after CPU hotplug
1879 * @self: notifier struct
1880 * @action: CPU hotplug event
1881 * @hcpu: unused
1882 *
1883 * If printk() is called from a CPU that is not online yet, the messages
1884 * will be spooled but will not show up on the console.  This function is
1885 * called when a new CPU comes online (or fails to come up), and ensures
1886 * that any such output gets printed.
1887 */
1888static int __cpuinit console_cpu_notify(struct notifier_block *self,
1889        unsigned long action, void *hcpu)
1890{
1891        switch (action) {
1892        case CPU_ONLINE:
1893        case CPU_DEAD:
1894        case CPU_DOWN_FAILED:
1895        case CPU_UP_CANCELED:
1896                console_lock();
1897                console_unlock();
1898        }
1899        return NOTIFY_OK;
1900}
1901
1902/**
1903 * console_lock - lock the console system for exclusive use.
1904 *
1905 * Acquires a lock which guarantees that the caller has
1906 * exclusive access to the console system and the console_drivers list.
1907 *
1908 * Can sleep, returns nothing.
1909 */
1910void console_lock(void)
1911{
1912        BUG_ON(in_interrupt());
1913        down(&console_sem);
1914        if (console_suspended)
1915                return;
1916        console_locked = 1;
1917        console_may_schedule = 1;
1918}
1919EXPORT_SYMBOL(console_lock);
1920
1921/**
1922 * console_trylock - try to lock the console system for exclusive use.
1923 *
1924 * Tried to acquire a lock which guarantees that the caller has
1925 * exclusive access to the console system and the console_drivers list.
1926 *
1927 * returns 1 on success, and 0 on failure to acquire the lock.
1928 */
1929int console_trylock(void)
1930{
1931        if (down_trylock(&console_sem))
1932                return 0;
1933        if (console_suspended) {
1934                up(&console_sem);
1935                return 0;
1936        }
1937        console_locked = 1;
1938        console_may_schedule = 0;
1939        return 1;
1940}
1941EXPORT_SYMBOL(console_trylock);
1942
1943int is_console_locked(void)
1944{
1945        return console_locked;
1946}
1947
1948/*
1949 * Delayed printk version, for scheduler-internal messages:
1950 */
1951#define PRINTK_BUF_SIZE         512
1952
1953#define PRINTK_PENDING_WAKEUP   0x01
1954#define PRINTK_PENDING_SCHED    0x02
1955
1956static DEFINE_PER_CPU(int, printk_pending);
1957static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
1958
1959void printk_tick(void)
1960{
1961        if (__this_cpu_read(printk_pending)) {
1962                int pending = __this_cpu_xchg(printk_pending, 0);
1963                if (pending & PRINTK_PENDING_SCHED) {
1964                        char *buf = __get_cpu_var(printk_sched_buf);
1965                        printk(KERN_WARNING "[sched_delayed] %s", buf);
1966                }
1967                if (pending & PRINTK_PENDING_WAKEUP)
1968                        wake_up_interruptible(&log_wait);
1969        }
1970}
1971
1972int printk_needs_cpu(int cpu)
1973{
1974        if (cpu_is_offline(cpu))
1975                printk_tick();
1976        return __this_cpu_read(printk_pending);
1977}
1978
1979void wake_up_klogd(void)
1980{
1981        if (waitqueue_active(&log_wait))
1982                this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
1983}
1984
1985static void console_cont_flush(char *text, size_t size)
1986{
1987        unsigned long flags;
1988        size_t len;
1989
1990        raw_spin_lock_irqsave(&logbuf_lock, flags);
1991
1992        if (!cont.len)
1993                goto out;
1994
1995        /*
1996         * We still queue earlier records, likely because the console was
1997         * busy. The earlier ones need to be printed before this one, we
1998         * did not flush any fragment so far, so just let it queue up.
1999         */
2000        if (console_seq < log_next_seq && !cont.cons)
2001                goto out;
2002
2003        len = cont_print_text(text, size);
2004        raw_spin_unlock(&logbuf_lock);
2005        stop_critical_timings();
2006        call_console_drivers(cont.level, text, len);
2007        start_critical_timings();
2008        local_irq_restore(flags);
2009        return;
2010out:
2011        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2012}
2013
2014/**
2015 * console_unlock - unlock the console system
2016 *
2017 * Releases the console_lock which the caller holds on the console system
2018 * and the console driver list.
2019 *
2020 * While the console_lock was held, console output may have been buffered
2021 * by printk().  If this is the case, console_unlock(); emits
2022 * the output prior to releasing the lock.
2023 *
2024 * If there is output waiting, we wake /dev/kmsg and syslog() users.
2025 *
2026 * console_unlock(); may be called from any context.
2027 */
2028void console_unlock(void)
2029{
2030        static char text[LOG_LINE_MAX + PREFIX_MAX];
2031        static u64 seen_seq;
2032        unsigned long flags;
2033        bool wake_klogd = false;
2034        bool retry;
2035
2036        if (console_suspended) {
2037                up(&console_sem);
2038                return;
2039        }
2040
2041        console_may_schedule = 0;
2042
2043        /* flush buffered message fragment immediately to console */
2044        console_cont_flush(text, sizeof(text));
2045again:
2046        for (;;) {
2047                struct log *msg;
2048                size_t len;
2049                int level;
2050
2051                raw_spin_lock_irqsave(&logbuf_lock, flags);
2052                if (seen_seq != log_next_seq) {
2053                        wake_klogd = true;
2054                        seen_seq = log_next_seq;
2055                }
2056
2057                if (console_seq < log_first_seq) {
2058                        /* messages are gone, move to first one */
2059                        console_seq = log_first_seq;
2060                        console_idx = log_first_idx;
2061                        console_prev = 0;
2062                }
2063skip:
2064                if (console_seq == log_next_seq)
2065                        break;
2066
2067                msg = log_from_idx(console_idx);
2068                if (msg->flags & LOG_NOCONS) {
2069                        /*
2070                         * Skip record we have buffered and already printed
2071                         * directly to the console when we received it.
2072                         */
2073                        console_idx = log_next(console_idx);
2074                        console_seq++;
2075                        /*
2076                         * We will get here again when we register a new
2077                         * CON_PRINTBUFFER console. Clear the flag so we
2078                         * will properly dump everything later.
2079                         */
2080                        msg->flags &= ~LOG_NOCONS;
2081                        console_prev = msg->flags;
2082                        goto skip;
2083                }
2084
2085                level = msg->level;
2086                len = msg_print_text(msg, console_prev, false,
2087                                     text, sizeof(text));
2088                console_idx = log_next(console_idx);
2089                console_seq++;
2090                console_prev = msg->flags;
2091                raw_spin_unlock(&logbuf_lock);
2092
2093                stop_critical_timings();        /* don't trace print latency */
2094                call_console_drivers(level, text, len);
2095                start_critical_timings();
2096                local_irq_restore(flags);
2097        }
2098        console_locked = 0;
2099
2100        /* Release the exclusive_console once it is used */
2101        if (unlikely(exclusive_console))
2102                exclusive_console = NULL;
2103
2104        raw_spin_unlock(&logbuf_lock);
2105
2106        up(&console_sem);
2107
2108        /*
2109         * Someone could have filled up the buffer again, so re-check if there's
2110         * something to flush. In case we cannot trylock the console_sem again,
2111         * there's a new owner and the console_unlock() from them will do the
2112         * flush, no worries.
2113         */
2114        raw_spin_lock(&logbuf_lock);
2115        retry = console_seq != log_next_seq;
2116        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2117
2118        if (retry && console_trylock())
2119                goto again;
2120
2121        if (wake_klogd)
2122                wake_up_klogd();
2123}
2124EXPORT_SYMBOL(console_unlock);
2125
2126/**
2127 * console_conditional_schedule - yield the CPU if required
2128 *
2129 * If the console code is currently allowed to sleep, and
2130 * if this CPU should yield the CPU to another task, do
2131 * so here.
2132 *
2133 * Must be called within console_lock();.
2134 */
2135void __sched console_conditional_schedule(void)
2136{
2137        if (console_may_schedule)
2138                cond_resched();
2139}
2140EXPORT_SYMBOL(console_conditional_schedule);
2141
2142void console_unblank(void)
2143{
2144        struct console *c;
2145
2146        /*
2147         * console_unblank can no longer be called in interrupt context unless
2148         * oops_in_progress is set to 1..
2149         */
2150        if (oops_in_progress) {
2151                if (down_trylock(&console_sem) != 0)
2152                        return;
2153        } else
2154                console_lock();
2155
2156        console_locked = 1;
2157        console_may_schedule = 0;
2158        for_each_console(c)
2159                if ((c->flags & CON_ENABLED) && c->unblank)
2160                        c->unblank();
2161        console_unlock();
2162}
2163
2164/*
2165 * Return the console tty driver structure and its associated index
2166 */
2167struct tty_driver *console_device(int *index)
2168{
2169        struct console *c;
2170        struct tty_driver *driver = NULL;
2171
2172        console_lock();
2173        for_each_console(c) {
2174                if (!c->device)
2175                        continue;
2176                driver = c->device(c, index);
2177                if (driver)
2178                        break;
2179        }
2180        console_unlock();
2181        return driver;
2182}
2183
2184/*
2185 * Prevent further output on the passed console device so that (for example)
2186 * serial drivers can disable console output before suspending a port, and can
2187 * re-enable output afterwards.
2188 */
2189void console_stop(struct console *console)
2190{
2191        console_lock();
2192        console->flags &= ~CON_ENABLED;
2193        console_unlock();
2194}
2195EXPORT_SYMBOL(console_stop);
2196
2197void console_start(struct console *console)
2198{
2199        console_lock();
2200        console->flags |= CON_ENABLED;
2201        console_unlock();
2202}
2203EXPORT_SYMBOL(console_start);
2204
2205static int __read_mostly keep_bootcon;
2206
2207static int __init keep_bootcon_setup(char *str)
2208{
2209        keep_bootcon = 1;
2210        printk(KERN_INFO "debug: skip boot console de-registration.\n");
2211
2212        return 0;
2213}
2214
2215early_param("keep_bootcon", keep_bootcon_setup);
2216
2217/*
2218 * The console driver calls this routine during kernel initialization
2219 * to register the console printing procedure with printk() and to
2220 * print any messages that were printed by the kernel before the
2221 * console driver was initialized.
2222 *
2223 * This can happen pretty early during the boot process (because of
2224 * early_printk) - sometimes before setup_arch() completes - be careful
2225 * of what kernel features are used - they may not be initialised yet.
2226 *
2227 * There are two types of consoles - bootconsoles (early_printk) and
2228 * "real" consoles (everything which is not a bootconsole) which are
2229 * handled differently.
2230 *  - Any number of bootconsoles can be registered at any time.
2231 *  - As soon as a "real" console is registered, all bootconsoles
2232 *    will be unregistered automatically.
2233 *  - Once a "real" console is registered, any attempt to register a
2234 *    bootconsoles will be rejected
2235 */
2236void register_console(struct console *newcon)
2237{
2238        int i;
2239        unsigned long flags;
2240        struct console *bcon = NULL;
2241
2242        /*
2243         * before we register a new CON_BOOT console, make sure we don't
2244         * already have a valid console
2245         */
2246        if (console_drivers && newcon->flags & CON_BOOT) {
2247                /* find the last or real console */
2248                for_each_console(bcon) {
2249                        if (!(bcon->flags & CON_BOOT)) {
2250                                printk(KERN_INFO "Too late to register bootconsole %s%d\n",
2251                                        newcon->name, newcon->index);
2252                                return;
2253                        }
2254                }
2255        }
2256
2257        if (console_drivers && console_drivers->flags & CON_BOOT)
2258                bcon = console_drivers;
2259
2260        if (preferred_console < 0 || bcon || !console_drivers)
2261                preferred_console = selected_console;
2262
2263        if (newcon->early_setup)
2264                newcon->early_setup();
2265
2266        /*
2267         *      See if we want to use this console driver. If we
2268         *      didn't select a console we take the first one
2269         *      that registers here.
2270         */
2271        if (preferred_console < 0) {
2272                if (newcon->index < 0)
2273                        newcon->index = 0;
2274                if (newcon->setup == NULL ||
2275                    newcon->setup(newcon, NULL) == 0) {
2276                        newcon->flags |= CON_ENABLED;
2277                        if (newcon->device) {
2278                                newcon->flags |= CON_CONSDEV;
2279                                preferred_console = 0;
2280                        }
2281                }
2282        }
2283
2284        /*
2285         *      See if this console matches one we selected on
2286         *      the command line.
2287         */
2288        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
2289                        i++) {
2290                if (strcmp(console_cmdline[i].name, newcon->name) != 0)
2291                        continue;
2292                if (newcon->index >= 0 &&
2293                    newcon->index != console_cmdline[i].index)
2294                        continue;
2295                if (newcon->index < 0)
2296                        newcon->index = console_cmdline[i].index;
2297#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2298                if (console_cmdline[i].brl_options) {
2299                        newcon->flags |= CON_BRL;
2300                        braille_register_console(newcon,
2301                                        console_cmdline[i].index,
2302                                        console_cmdline[i].options,
2303                                        console_cmdline[i].brl_options);
2304                        return;
2305                }
2306#endif
2307                if (newcon->setup &&
2308                    newcon->setup(newcon, console_cmdline[i].options) != 0)
2309                        break;
2310                newcon->flags |= CON_ENABLED;
2311                newcon->index = console_cmdline[i].index;
2312                if (i == selected_console) {
2313                        newcon->flags |= CON_CONSDEV;
2314                        preferred_console = selected_console;
2315                }
2316                break;
2317        }
2318
2319        if (!(newcon->flags & CON_ENABLED))
2320                return;
2321
2322        /*
2323         * If we have a bootconsole, and are switching to a real console,
2324         * don't print everything out again, since when the boot console, and
2325         * the real console are the same physical device, it's annoying to
2326         * see the beginning boot messages twice
2327         */
2328        if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2329                newcon->flags &= ~CON_PRINTBUFFER;
2330
2331        /*
2332         *      Put this console in the list - keep the
2333         *      preferred driver at the head of the list.
2334         */
2335        console_lock();
2336        if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2337                newcon->next = console_drivers;
2338                console_drivers = newcon;
2339                if (newcon->next)
2340                        newcon->next->flags &= ~CON_CONSDEV;
2341        } else {
2342                newcon->next = console_drivers->next;
2343                console_drivers->next = newcon;
2344        }
2345        if (newcon->flags & CON_PRINTBUFFER) {
2346                /*
2347                 * console_unlock(); will print out the buffered messages
2348                 * for us.
2349                 */
2350                raw_spin_lock_irqsave(&logbuf_lock, flags);
2351                console_seq = syslog_seq;
2352                console_idx = syslog_idx;
2353                console_prev = syslog_prev;
2354                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2355                /*
2356                 * We're about to replay the log buffer.  Only do this to the
2357                 * just-registered console to avoid excessive message spam to
2358                 * the already-registered consoles.
2359                 */
2360                exclusive_console = newcon;
2361        }
2362        console_unlock();
2363        console_sysfs_notify();
2364
2365        /*
2366         * By unregistering the bootconsoles after we enable the real console
2367         * we get the "console xxx enabled" message on all the consoles -
2368         * boot consoles, real consoles, etc - this is to ensure that end
2369         * users know there might be something in the kernel's log buffer that
2370         * went to the bootconsole (that they do not see on the real console)
2371         */
2372        if (bcon &&
2373            ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2374            !keep_bootcon) {
2375                /* we need to iterate through twice, to make sure we print
2376                 * everything out, before we unregister the console(s)
2377                 */
2378                printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
2379                        newcon->name, newcon->index);
2380                for_each_console(bcon)
2381                        if (bcon->flags & CON_BOOT)
2382                                unregister_console(bcon);
2383        } else {
2384                printk(KERN_INFO "%sconsole [%s%d] enabled\n",
2385                        (newcon->flags & CON_BOOT) ? "boot" : "" ,
2386                        newcon->name, newcon->index);
2387        }
2388}
2389EXPORT_SYMBOL(register_console);
2390
2391int unregister_console(struct console *console)
2392{
2393        struct console *a, *b;
2394        int res = 1;
2395
2396#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2397        if (console->flags & CON_BRL)
2398                return braille_unregister_console(console);
2399#endif
2400
2401        console_lock();
2402        if (console_drivers == console) {
2403                console_drivers=console->next;
2404                res = 0;
2405        } else if (console_drivers) {
2406                for (a=console_drivers->next, b=console_drivers ;
2407                     a; b=a, a=b->next) {
2408                        if (a == console) {
2409                                b->next = a->next;
2410                                res = 0;
2411                                break;
2412                        }
2413                }
2414        }
2415
2416        /*
2417         * If this isn't the last console and it has CON_CONSDEV set, we
2418         * need to set it on the next preferred console.
2419         */
2420        if (console_drivers != NULL && console->flags & CON_CONSDEV)
2421                console_drivers->flags |= CON_CONSDEV;
2422
2423        console_unlock();
2424        console_sysfs_notify();
2425        return res;
2426}
2427EXPORT_SYMBOL(unregister_console);
2428
2429static int __init printk_late_init(void)
2430{
2431        struct console *con;
2432
2433        for_each_console(con) {
2434                if (!keep_bootcon && con->flags & CON_BOOT) {
2435                        printk(KERN_INFO "turn off boot console %s%d\n",
2436                                con->name, con->index);
2437                        unregister_console(con);
2438                }
2439        }
2440        hotcpu_notifier(console_cpu_notify, 0);
2441        return 0;
2442}
2443late_initcall(printk_late_init);
2444
2445#if defined CONFIG_PRINTK
2446
2447int printk_sched(const char *fmt, ...)
2448{
2449        unsigned long flags;
2450        va_list args;
2451        char *buf;
2452        int r;
2453
2454        local_irq_save(flags);
2455        buf = __get_cpu_var(printk_sched_buf);
2456
2457        va_start(args, fmt);
2458        r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
2459        va_end(args);
2460
2461        __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
2462        local_irq_restore(flags);
2463
2464        return r;
2465}
2466
2467/*
2468 * printk rate limiting, lifted from the networking subsystem.
2469 *
2470 * This enforces a rate limit: not more than 10 kernel messages
2471 * every 5s to make a denial-of-service attack impossible.
2472 */
2473DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2474
2475int __printk_ratelimit(const char *func)
2476{
2477        return ___ratelimit(&printk_ratelimit_state, func);
2478}
2479EXPORT_SYMBOL(__printk_ratelimit);
2480
2481/**
2482 * printk_timed_ratelimit - caller-controlled printk ratelimiting
2483 * @caller_jiffies: pointer to caller's state
2484 * @interval_msecs: minimum interval between prints
2485 *
2486 * printk_timed_ratelimit() returns true if more than @interval_msecs
2487 * milliseconds have elapsed since the last time printk_timed_ratelimit()
2488 * returned true.
2489 */
2490bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2491                        unsigned int interval_msecs)
2492{
2493        if (*caller_jiffies == 0
2494                        || !time_in_range(jiffies, *caller_jiffies,
2495                                        *caller_jiffies
2496                                        + msecs_to_jiffies(interval_msecs))) {
2497                *caller_jiffies = jiffies;
2498                return true;
2499        }
2500        return false;
2501}
2502EXPORT_SYMBOL(printk_timed_ratelimit);
2503
2504static DEFINE_SPINLOCK(dump_list_lock);
2505static LIST_HEAD(dump_list);
2506
2507/**
2508 * kmsg_dump_register - register a kernel log dumper.
2509 * @dumper: pointer to the kmsg_dumper structure
2510 *
2511 * Adds a kernel log dumper to the system. The dump callback in the
2512 * structure will be called when the kernel oopses or panics and must be
2513 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2514 */
2515int kmsg_dump_register(struct kmsg_dumper *dumper)
2516{
2517        unsigned long flags;
2518        int err = -EBUSY;
2519
2520        /* The dump callback needs to be set */
2521        if (!dumper->dump)
2522                return -EINVAL;
2523
2524        spin_lock_irqsave(&dump_list_lock, flags);
2525        /* Don't allow registering multiple times */
2526        if (!dumper->registered) {
2527                dumper->registered = 1;
2528                list_add_tail_rcu(&dumper->list, &dump_list);
2529                err = 0;
2530        }
2531        spin_unlock_irqrestore(&dump_list_lock, flags);
2532
2533        return err;
2534}
2535EXPORT_SYMBOL_GPL(kmsg_dump_register);
2536
2537/**
2538 * kmsg_dump_unregister - unregister a kmsg dumper.
2539 * @dumper: pointer to the kmsg_dumper structure
2540 *
2541 * Removes a dump device from the system. Returns zero on success and
2542 * %-EINVAL otherwise.
2543 */
2544int kmsg_dump_unregister(struct kmsg_dumper *dumper)
2545{
2546        unsigned long flags;
2547        int err = -EINVAL;
2548
2549        spin_lock_irqsave(&dump_list_lock, flags);
2550        if (dumper->registered) {
2551                dumper->registered = 0;
2552                list_del_rcu(&dumper->list);
2553                err = 0;
2554        }
2555        spin_unlock_irqrestore(&dump_list_lock, flags);
2556        synchronize_rcu();
2557
2558        return err;
2559}
2560EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
2561
2562static bool always_kmsg_dump;
2563module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
2564
2565/**
2566 * kmsg_dump - dump kernel log to kernel message dumpers.
2567 * @reason: the reason (oops, panic etc) for dumping
2568 *
2569 * Call each of the registered dumper's dump() callback, which can
2570 * retrieve the kmsg records with kmsg_dump_get_line() or
2571 * kmsg_dump_get_buffer().
2572 */
2573void kmsg_dump(enum kmsg_dump_reason reason)
2574{
2575        struct kmsg_dumper *dumper;
2576        unsigned long flags;
2577
2578        if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
2579                return;
2580
2581        rcu_read_lock();
2582        list_for_each_entry_rcu(dumper, &dump_list, list) {
2583                if (dumper->max_reason && reason > dumper->max_reason)
2584                        continue;
2585
2586                /* initialize iterator with data about the stored records */
2587                dumper->active = true;
2588
2589                raw_spin_lock_irqsave(&logbuf_lock, flags);
2590                dumper->cur_seq = clear_seq;
2591                dumper->cur_idx = clear_idx;
2592                dumper->next_seq = log_next_seq;
2593                dumper->next_idx = log_next_idx;
2594                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2595
2596                /* invoke dumper which will iterate over records */
2597                dumper->dump(dumper, reason);
2598
2599                /* reset iterator */
2600                dumper->active = false;
2601        }
2602        rcu_read_unlock();
2603}
2604
2605/**
2606 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
2607 * @dumper: registered kmsg dumper
2608 * @syslog: include the "<4>" prefixes
2609 * @line: buffer to copy the line to
2610 * @size: maximum size of the buffer
2611 * @len: length of line placed into buffer
2612 *
2613 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2614 * record, and copy one record into the provided buffer.
2615 *
2616 * Consecutive calls will return the next available record moving
2617 * towards the end of the buffer with the youngest messages.
2618 *
2619 * A return value of FALSE indicates that there are no more records to
2620 * read.
2621 *
2622 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
2623 */
2624bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
2625                               char *line, size_t size, size_t *len)
2626{
2627        struct log *msg;
2628        size_t l = 0;
2629        bool ret = false;
2630
2631        if (!dumper->active)
2632                goto out;
2633
2634        if (dumper->cur_seq < log_first_seq) {
2635                /* messages are gone, move to first available one */
2636                dumper->cur_seq = log_first_seq;
2637                dumper->cur_idx = log_first_idx;
2638        }
2639
2640        /* last entry */
2641        if (dumper->cur_seq >= log_next_seq)
2642                goto out;
2643
2644        msg = log_from_idx(dumper->cur_idx);
2645        l = msg_print_text(msg, 0, syslog, line, size);
2646
2647        dumper->cur_idx = log_next(dumper->cur_idx);
2648        dumper->cur_seq++;
2649        ret = true;
2650out:
2651        if (len)
2652                *len = l;
2653        return ret;
2654}
2655
2656/**
2657 * kmsg_dump_get_line - retrieve one kmsg log line
2658 * @dumper: registered kmsg dumper
2659 * @syslog: include the "<4>" prefixes
2660 * @line: buffer to copy the line to
2661 * @size: maximum size of the buffer
2662 * @len: length of line placed into buffer
2663 *
2664 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2665 * record, and copy one record into the provided buffer.
2666 *
2667 * Consecutive calls will return the next available record moving
2668 * towards the end of the buffer with the youngest messages.
2669 *
2670 * A return value of FALSE indicates that there are no more records to
2671 * read.
2672 */
2673bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
2674                        char *line, size_t size, size_t *len)
2675{
2676        unsigned long flags;
2677        bool ret;
2678
2679        raw_spin_lock_irqsave(&logbuf_lock, flags);
2680        ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
2681        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2682
2683        return ret;
2684}
2685EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
2686
2687/**
2688 * kmsg_dump_get_buffer - copy kmsg log lines
2689 * @dumper: registered kmsg dumper
2690 * @syslog: include the "<4>" prefixes
2691 * @buf: buffer to copy the line to
2692 * @size: maximum size of the buffer
2693 * @len: length of line placed into buffer
2694 *
2695 * Start at the end of the kmsg buffer and fill the provided buffer
2696 * with as many of the the *youngest* kmsg records that fit into it.
2697 * If the buffer is large enough, all available kmsg records will be
2698 * copied with a single call.
2699 *
2700 * Consecutive calls will fill the buffer with the next block of
2701 * available older records, not including the earlier retrieved ones.
2702 *
2703 * A return value of FALSE indicates that there are no more records to
2704 * read.
2705 */
2706bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
2707                          char *buf, size_t size, size_t *len)
2708{
2709        unsigned long flags;
2710        u64 seq;
2711        u32 idx;
2712        u64 next_seq;
2713        u32 next_idx;
2714        enum log_flags prev;
2715        size_t l = 0;
2716        bool ret = false;
2717
2718        if (!dumper->active)
2719                goto out;
2720
2721        raw_spin_lock_irqsave(&logbuf_lock, flags);
2722        if (dumper->cur_seq < log_first_seq) {
2723                /* messages are gone, move to first available one */
2724                dumper->cur_seq = log_first_seq;
2725                dumper->cur_idx = log_first_idx;
2726        }
2727
2728        /* last entry */
2729        if (dumper->cur_seq >= dumper->next_seq) {
2730                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2731                goto out;
2732        }
2733
2734        /* calculate length of entire buffer */
2735        seq = dumper->cur_seq;
2736        idx = dumper->cur_idx;
2737        prev = 0;
2738        while (seq < dumper->next_seq) {
2739                struct log *msg = log_from_idx(idx);
2740
2741                l += msg_print_text(msg, prev, true, NULL, 0);
2742                idx = log_next(idx);
2743                seq++;
2744                prev = msg->flags;
2745        }
2746
2747        /* move first record forward until length fits into the buffer */
2748        seq = dumper->cur_seq;
2749        idx = dumper->cur_idx;
2750        prev = 0;
2751        while (l > size && seq < dumper->next_seq) {
2752                struct log *msg = log_from_idx(idx);
2753
2754                l -= msg_print_text(msg, prev, true, NULL, 0);
2755                idx = log_next(idx);
2756                seq++;
2757                prev = msg->flags;
2758        }
2759
2760        /* last message in next interation */
2761        next_seq = seq;
2762        next_idx = idx;
2763
2764        l = 0;
2765        prev = 0;
2766        while (seq < dumper->next_seq) {
2767                struct log *msg = log_from_idx(idx);
2768
2769                l += msg_print_text(msg, prev, syslog, buf + l, size - l);
2770                idx = log_next(idx);
2771                seq++;
2772                prev = msg->flags;
2773        }
2774
2775        dumper->next_seq = next_seq;
2776        dumper->next_idx = next_idx;
2777        ret = true;
2778        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2779out:
2780        if (len)
2781                *len = l;
2782        return ret;
2783}
2784EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
2785
2786/**
2787 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
2788 * @dumper: registered kmsg dumper
2789 *
2790 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2791 * kmsg_dump_get_buffer() can be called again and used multiple
2792 * times within the same dumper.dump() callback.
2793 *
2794 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
2795 */
2796void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
2797{
2798        dumper->cur_seq = clear_seq;
2799        dumper->cur_idx = clear_idx;
2800        dumper->next_seq = log_next_seq;
2801        dumper->next_idx = log_next_idx;
2802}
2803
2804/**
2805 * kmsg_dump_rewind - reset the interator
2806 * @dumper: registered kmsg dumper
2807 *
2808 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2809 * kmsg_dump_get_buffer() can be called again and used multiple
2810 * times within the same dumper.dump() callback.
2811 */
2812void kmsg_dump_rewind(struct kmsg_dumper *dumper)
2813{
2814        unsigned long flags;
2815
2816        raw_spin_lock_irqsave(&logbuf_lock, flags);
2817        kmsg_dump_rewind_nolock(dumper);
2818        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2819}
2820EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
2821#endif
2822
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.