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        if (!buf)
 851                return 15;
 852
 853        rem_nsec = do_div(ts, 1000000000);
 854        return sprintf(buf, "[%5lu.%06lu] ",
 855                       (unsigned long)ts, rem_nsec / 1000);
 856}
 857
 858static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
 859{
 860        size_t len = 0;
 861        unsigned int prefix = (msg->facility << 3) | msg->level;
 862
 863        if (syslog) {
 864                if (buf) {
 865                        len += sprintf(buf, "<%u>", prefix);
 866                } else {
 867                        len += 3;
 868                        if (prefix > 999)
 869                                len += 3;
 870                        else if (prefix > 99)
 871                                len += 2;
 872                        else if (prefix > 9)
 873                                len++;
 874                }
 875        }
 876
 877        len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
 878        return len;
 879}
 880
 881static size_t msg_print_text(const struct log *msg, enum log_flags prev,
 882                             bool syslog, char *buf, size_t size)
 883{
 884        const char *text = log_text(msg);
 885        size_t text_size = msg->text_len;
 886        bool prefix = true;
 887        bool newline = true;
 888        size_t len = 0;
 889
 890        if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
 891                prefix = false;
 892
 893        if (msg->flags & LOG_CONT) {
 894                if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
 895                        prefix = false;
 896
 897                if (!(msg->flags & LOG_NEWLINE))
 898                        newline = false;
 899        }
 900
 901        do {
 902                const char *next = memchr(text, '\n', text_size);
 903                size_t text_len;
 904
 905                if (next) {
 906                        text_len = next - text;
 907                        next++;
 908                        text_size -= next - text;
 909                } else {
 910                        text_len = text_size;
 911                }
 912
 913                if (buf) {
 914                        if (print_prefix(msg, syslog, NULL) +
 915                            text_len + 1 >= size - len)
 916                                break;
 917
 918                        if (prefix)
 919                                len += print_prefix(msg, syslog, buf + len);
 920                        memcpy(buf + len, text, text_len);
 921                        len += text_len;
 922                        if (next || newline)
 923                                buf[len++] = '\n';
 924                } else {
 925                        /* SYSLOG_ACTION_* buffer size only calculation */
 926                        if (prefix)
 927                                len += print_prefix(msg, syslog, NULL);
 928                        len += text_len;
 929                        if (next || newline)
 930                                len++;
 931                }
 932
 933                prefix = true;
 934                text = next;
 935        } while (text);
 936
 937        return len;
 938}
 939
 940static int syslog_print(char __user *buf, int size)
 941{
 942        char *text;
 943        struct log *msg;
 944        int len = 0;
 945
 946        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
 947        if (!text)
 948                return -ENOMEM;
 949
 950        while (size > 0) {
 951                size_t n;
 952                size_t skip;
 953
 954                raw_spin_lock_irq(&logbuf_lock);
 955                if (syslog_seq < log_first_seq) {
 956                        /* messages are gone, move to first one */
 957                        syslog_seq = log_first_seq;
 958                        syslog_idx = log_first_idx;
 959                        syslog_prev = 0;
 960                        syslog_partial = 0;
 961                }
 962                if (syslog_seq == log_next_seq) {
 963                        raw_spin_unlock_irq(&logbuf_lock);
 964                        break;
 965                }
 966
 967                skip = syslog_partial;
 968                msg = log_from_idx(syslog_idx);
 969                n = msg_print_text(msg, syslog_prev, true, text,
 970                                   LOG_LINE_MAX + PREFIX_MAX);
 971                if (n - syslog_partial <= size) {
 972                        /* message fits into buffer, move forward */
 973                        syslog_idx = log_next(syslog_idx);
 974                        syslog_seq++;
 975                        syslog_prev = msg->flags;
 976                        n -= syslog_partial;
 977                        syslog_partial = 0;
 978                } else if (!len){
 979                        /* partial read(), remember position */
 980                        n = size;
 981                        syslog_partial += n;
 982                } else
 983                        n = 0;
 984                raw_spin_unlock_irq(&logbuf_lock);
 985
 986                if (!n)
 987                        break;
 988
 989                if (copy_to_user(buf, text + skip, n)) {
 990                        if (!len)
 991                                len = -EFAULT;
 992                        break;
 993                }
 994
 995                len += n;
 996                size -= n;
 997                buf += n;
 998        }
 999
1000        kfree(text);
1001        return len;
1002}
1003
1004static int syslog_print_all(char __user *buf, int size, bool clear)
1005{
1006        char *text;
1007        int len = 0;
1008
1009        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1010        if (!text)
1011                return -ENOMEM;
1012
1013        raw_spin_lock_irq(&logbuf_lock);
1014        if (buf) {
1015                u64 next_seq;
1016                u64 seq;
1017                u32 idx;
1018                enum log_flags prev;
1019
1020                if (clear_seq < log_first_seq) {
1021                        /* messages are gone, move to first available one */
1022                        clear_seq = log_first_seq;
1023                        clear_idx = log_first_idx;
1024                }
1025
1026                /*
1027                 * Find first record that fits, including all following records,
1028                 * into the user-provided buffer for this dump.
1029                 */
1030                seq = clear_seq;
1031                idx = clear_idx;
1032                prev = 0;
1033                while (seq < log_next_seq) {
1034                        struct log *msg = log_from_idx(idx);
1035
1036                        len += msg_print_text(msg, prev, true, NULL, 0);
1037                        prev = msg->flags;
1038                        idx = log_next(idx);
1039                        seq++;
1040                }
1041
1042                /* move first record forward until length fits into the buffer */
1043                seq = clear_seq;
1044                idx = clear_idx;
1045                prev = 0;
1046                while (len > size && seq < log_next_seq) {
1047                        struct log *msg = log_from_idx(idx);
1048
1049                        len -= msg_print_text(msg, prev, true, NULL, 0);
1050                        prev = msg->flags;
1051                        idx = log_next(idx);
1052                        seq++;
1053                }
1054
1055                /* last message fitting into this dump */
1056                next_seq = log_next_seq;
1057
1058                len = 0;
1059                prev = 0;
1060                while (len >= 0 && seq < next_seq) {
1061                        struct log *msg = log_from_idx(idx);
1062                        int textlen;
1063
1064                        textlen = msg_print_text(msg, prev, true, text,
1065                                                 LOG_LINE_MAX + PREFIX_MAX);
1066                        if (textlen < 0) {
1067                                len = textlen;
1068                                break;
1069                        }
1070                        idx = log_next(idx);
1071                        seq++;
1072                        prev = msg->flags;
1073
1074                        raw_spin_unlock_irq(&logbuf_lock);
1075                        if (copy_to_user(buf + len, text, textlen))
1076                                len = -EFAULT;
1077                        else
1078                                len += textlen;
1079                        raw_spin_lock_irq(&logbuf_lock);
1080
1081                        if (seq < log_first_seq) {
1082                                /* messages are gone, move to next one */
1083                                seq = log_first_seq;
1084                                idx = log_first_idx;
1085                                prev = 0;
1086                        }
1087                }
1088        }
1089
1090        if (clear) {
1091                clear_seq = log_next_seq;
1092                clear_idx = log_next_idx;
1093        }
1094        raw_spin_unlock_irq(&logbuf_lock);
1095
1096        kfree(text);
1097        return len;
1098}
1099
1100int do_syslog(int type, char __user *buf, int len, bool from_file)
1101{
1102        bool clear = false;
1103        static int saved_console_loglevel = -1;
1104        int error;
1105
1106        error = check_syslog_permissions(type, from_file);
1107        if (error)
1108                goto out;
1109
1110        error = security_syslog(type);
1111        if (error)
1112                return error;
1113
1114        switch (type) {
1115        case SYSLOG_ACTION_CLOSE:       /* Close log */
1116                break;
1117        case SYSLOG_ACTION_OPEN:        /* Open log */
1118                break;
1119        case SYSLOG_ACTION_READ:        /* Read from log */
1120                error = -EINVAL;
1121                if (!buf || len < 0)
1122                        goto out;
1123                error = 0;
1124                if (!len)
1125                        goto out;
1126                if (!access_ok(VERIFY_WRITE, buf, len)) {
1127                        error = -EFAULT;
1128                        goto out;
1129                }
1130                error = wait_event_interruptible(log_wait,
1131                                                 syslog_seq != log_next_seq);
1132                if (error)
1133                        goto out;
1134                error = syslog_print(buf, len);
1135                break;
1136        /* Read/clear last kernel messages */
1137        case SYSLOG_ACTION_READ_CLEAR:
1138                clear = true;
1139                /* FALL THRU */
1140        /* Read last kernel messages */
1141        case SYSLOG_ACTION_READ_ALL:
1142                error = -EINVAL;
1143                if (!buf || len < 0)
1144                        goto out;
1145                error = 0;
1146                if (!len)
1147                        goto out;
1148                if (!access_ok(VERIFY_WRITE, buf, len)) {
1149                        error = -EFAULT;
1150                        goto out;
1151                }
1152                error = syslog_print_all(buf, len, clear);
1153                break;
1154        /* Clear ring buffer */
1155        case SYSLOG_ACTION_CLEAR:
1156                syslog_print_all(NULL, 0, true);
1157                break;
1158        /* Disable logging to console */
1159        case SYSLOG_ACTION_CONSOLE_OFF:
1160                if (saved_console_loglevel == -1)
1161                        saved_console_loglevel = console_loglevel;
1162                console_loglevel = minimum_console_loglevel;
1163                break;
1164        /* Enable logging to console */
1165        case SYSLOG_ACTION_CONSOLE_ON:
1166                if (saved_console_loglevel != -1) {
1167                        console_loglevel = saved_console_loglevel;
1168                        saved_console_loglevel = -1;
1169                }
1170                break;
1171        /* Set level of messages printed to console */
1172        case SYSLOG_ACTION_CONSOLE_LEVEL:
1173                error = -EINVAL;
1174                if (len < 1 || len > 8)
1175                        goto out;
1176                if (len < minimum_console_loglevel)
1177                        len = minimum_console_loglevel;
1178                console_loglevel = len;
1179                /* Implicitly re-enable logging to console */
1180                saved_console_loglevel = -1;
1181                error = 0;
1182                break;
1183        /* Number of chars in the log buffer */
1184        case SYSLOG_ACTION_SIZE_UNREAD:
1185                raw_spin_lock_irq(&logbuf_lock);
1186                if (syslog_seq < log_first_seq) {
1187                        /* messages are gone, move to first one */
1188                        syslog_seq = log_first_seq;
1189                        syslog_idx = log_first_idx;
1190                        syslog_prev = 0;
1191                        syslog_partial = 0;
1192                }
1193                if (from_file) {
1194                        /*
1195                         * Short-cut for poll(/"proc/kmsg") which simply checks
1196                         * for pending data, not the size; return the count of
1197                         * records, not the length.
1198                         */
1199                        error = log_next_idx - syslog_idx;
1200                } else {
1201                        u64 seq = syslog_seq;
1202                        u32 idx = syslog_idx;
1203                        enum log_flags prev = syslog_prev;
1204
1205                        error = 0;
1206                        while (seq < log_next_seq) {
1207                                struct log *msg = log_from_idx(idx);
1208
1209                                error += msg_print_text(msg, prev, true, NULL, 0);
1210                                idx = log_next(idx);
1211                                seq++;
1212                                prev = msg->flags;
1213                        }
1214                        error -= syslog_partial;
1215                }
1216                raw_spin_unlock_irq(&logbuf_lock);
1217                break;
1218        /* Size of the log buffer */
1219        case SYSLOG_ACTION_SIZE_BUFFER:
1220                error = log_buf_len;
1221                break;
1222        default:
1223                error = -EINVAL;
1224                break;
1225        }
1226out:
1227        return error;
1228}
1229
1230SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1231{
1232        return do_syslog(type, buf, len, SYSLOG_FROM_CALL);
1233}
1234
1235static bool __read_mostly ignore_loglevel;
1236
1237static int __init ignore_loglevel_setup(char *str)
1238{
1239        ignore_loglevel = 1;
1240        printk(KERN_INFO "debug: ignoring loglevel setting.\n");
1241
1242        return 0;
1243}
1244
1245early_param("ignore_loglevel", ignore_loglevel_setup);
1246module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
1247MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
1248        "print all kernel messages to the console.");
1249
1250/*
1251 * Call the console drivers, asking them to write out
1252 * log_buf[start] to log_buf[end - 1].
1253 * The console_lock must be held.
1254 */
1255static void call_console_drivers(int level, const char *text, size_t len)
1256{
1257        struct console *con;
1258
1259        trace_console(text, 0, len, len);
1260
1261        if (level >= console_loglevel && !ignore_loglevel)
1262                return;
1263        if (!console_drivers)
1264                return;
1265
1266        for_each_console(con) {
1267                if (exclusive_console && con != exclusive_console)
1268                        continue;
1269                if (!(con->flags & CON_ENABLED))
1270                        continue;
1271                if (!con->write)
1272                        continue;
1273                if (!cpu_online(smp_processor_id()) &&
1274                    !(con->flags & CON_ANYTIME))
1275                        continue;
1276                con->write(con, text, len);
1277        }
1278}
1279
1280/*
1281 * Zap console related locks when oopsing. Only zap at most once
1282 * every 10 seconds, to leave time for slow consoles to print a
1283 * full oops.
1284 */
1285static void zap_locks(void)
1286{
1287        static unsigned long oops_timestamp;
1288
1289        if (time_after_eq(jiffies, oops_timestamp) &&
1290                        !time_after(jiffies, oops_timestamp + 30 * HZ))
1291                return;
1292
1293        oops_timestamp = jiffies;
1294
1295        debug_locks_off();
1296        /* If a crash is occurring, make sure we can't deadlock */
1297        raw_spin_lock_init(&logbuf_lock);
1298        /* And make sure that we print immediately */
1299        sema_init(&console_sem, 1);
1300}
1301
1302/* Check if we have any console registered that can be called early in boot. */
1303static int have_callable_console(void)
1304{
1305        struct console *con;
1306
1307        for_each_console(con)
1308                if (con->flags & CON_ANYTIME)
1309                        return 1;
1310
1311        return 0;
1312}
1313
1314/*
1315 * Can we actually use the console at this time on this cpu?
1316 *
1317 * Console drivers may assume that per-cpu resources have
1318 * been allocated. So unless they're explicitly marked as
1319 * being able to cope (CON_ANYTIME) don't call them until
1320 * this CPU is officially up.
1321 */
1322static inline int can_use_console(unsigned int cpu)
1323{
1324        return cpu_online(cpu) || have_callable_console();
1325}
1326
1327/*
1328 * Try to get console ownership to actually show the kernel
1329 * messages from a 'printk'. Return true (and with the
1330 * console_lock held, and 'console_locked' set) if it
1331 * is successful, false otherwise.
1332 *
1333 * This gets called with the 'logbuf_lock' spinlock held and
1334 * interrupts disabled. It should return with 'lockbuf_lock'
1335 * released but interrupts still disabled.
1336 */
1337static int console_trylock_for_printk(unsigned int cpu)
1338        __releases(&logbuf_lock)
1339{
1340        int retval = 0, wake = 0;
1341
1342        if (console_trylock()) {
1343                retval = 1;
1344
1345                /*
1346                 * If we can't use the console, we need to release
1347                 * the console semaphore by hand to avoid flushing
1348                 * the buffer. We need to hold the console semaphore
1349                 * in order to do this test safely.
1350                 */
1351                if (!can_use_console(cpu)) {
1352                        console_locked = 0;
1353                        wake = 1;
1354                        retval = 0;
1355                }
1356        }
1357        logbuf_cpu = UINT_MAX;
1358        if (wake)
1359                up(&console_sem);
1360        raw_spin_unlock(&logbuf_lock);
1361        return retval;
1362}
1363
1364int printk_delay_msec __read_mostly;
1365
1366static inline void printk_delay(void)
1367{
1368        if (unlikely(printk_delay_msec)) {
1369                int m = printk_delay_msec;
1370
1371                while (m--) {
1372                        mdelay(1);
1373                        touch_nmi_watchdog();
1374                }
1375        }
1376}
1377
1378/*
1379 * Continuation lines are buffered, and not committed to the record buffer
1380 * until the line is complete, or a race forces it. The line fragments
1381 * though, are printed immediately to the consoles to ensure everything has
1382 * reached the console in case of a kernel crash.
1383 */
1384static struct cont {
1385        char buf[LOG_LINE_MAX];
1386        size_t len;                     /* length == 0 means unused buffer */
1387        size_t cons;                    /* bytes written to console */
1388        struct task_struct *owner;      /* task of first print*/
1389        u64 ts_nsec;                    /* time of first print */
1390        u8 level;                       /* log level of first message */
1391        u8 facility;                    /* log level of first message */
1392        enum log_flags flags;           /* prefix, newline flags */
1393        bool flushed:1;                 /* buffer sealed and committed */
1394} cont;
1395
1396static void cont_flush(enum log_flags flags)
1397{
1398        if (cont.flushed)
1399                return;
1400        if (cont.len == 0)
1401                return;
1402
1403        if (cont.cons) {
1404                /*
1405                 * If a fragment of this line was directly flushed to the
1406                 * console; wait for the console to pick up the rest of the
1407                 * line. LOG_NOCONS suppresses a duplicated output.
1408                 */
1409                log_store(cont.facility, cont.level, flags | LOG_NOCONS,
1410                          cont.ts_nsec, NULL, 0, cont.buf, cont.len);
1411                cont.flags = flags;
1412                cont.flushed = true;
1413        } else {
1414                /*
1415                 * If no fragment of this line ever reached the console,
1416                 * just submit it to the store and free the buffer.
1417                 */
1418                log_store(cont.facility, cont.level, flags, 0,
1419                          NULL, 0, cont.buf, cont.len);
1420                cont.len = 0;
1421        }
1422}
1423
1424static bool cont_add(int facility, int level, const char *text, size_t len)
1425{
1426        if (cont.len && cont.flushed)
1427                return false;
1428
1429        if (cont.len + len > sizeof(cont.buf)) {
1430                /* the line gets too long, split it up in separate records */
1431                cont_flush(LOG_CONT);
1432                return false;
1433        }
1434
1435        if (!cont.len) {
1436                cont.facility = facility;
1437                cont.level = level;
1438                cont.owner = current;
1439                cont.ts_nsec = local_clock();
1440                cont.flags = 0;
1441                cont.cons = 0;
1442                cont.flushed = false;
1443        }
1444
1445        memcpy(cont.buf + cont.len, text, len);
1446        cont.len += len;
1447
1448        if (cont.len > (sizeof(cont.buf) * 80) / 100)
1449                cont_flush(LOG_CONT);
1450
1451        return true;
1452}
1453
1454static size_t cont_print_text(char *text, size_t size)
1455{
1456        size_t textlen = 0;
1457        size_t len;
1458
1459        if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
1460                textlen += print_time(cont.ts_nsec, text);
1461                size -= textlen;
1462        }
1463
1464        len = cont.len - cont.cons;
1465        if (len > 0) {
1466                if (len+1 > size)
1467                        len = size-1;
1468                memcpy(text + textlen, cont.buf + cont.cons, len);
1469                textlen += len;
1470                cont.cons = cont.len;
1471        }
1472
1473        if (cont.flushed) {
1474                if (cont.flags & LOG_NEWLINE)
1475                        text[textlen++] = '\n';
1476                /* got everything, release buffer */
1477                cont.len = 0;
1478        }
1479        return textlen;
1480}
1481
1482asmlinkage int vprintk_emit(int facility, int level,
1483                            const char *dict, size_t dictlen,
1484                            const char *fmt, va_list args)
1485{
1486        static int recursion_bug;
1487        static char textbuf[LOG_LINE_MAX];
1488        char *text = textbuf;
1489        size_t text_len;
1490        enum log_flags lflags = 0;
1491        unsigned long flags;
1492        int this_cpu;
1493        int printed_len = 0;
1494
1495        boot_delay_msec();
1496        printk_delay();
1497
1498        /* This stops the holder of console_sem just where we want him */
1499        local_irq_save(flags);
1500        this_cpu = smp_processor_id();
1501
1502        /*
1503         * Ouch, printk recursed into itself!
1504         */
1505        if (unlikely(logbuf_cpu == this_cpu)) {
1506                /*
1507                 * If a crash is occurring during printk() on this CPU,
1508                 * then try to get the crash message out but make sure
1509                 * we can't deadlock. Otherwise just return to avoid the
1510                 * recursion and return - but flag the recursion so that
1511                 * it can be printed at the next appropriate moment:
1512                 */
1513                if (!oops_in_progress && !lockdep_recursing(current)) {
1514                        recursion_bug = 1;
1515                        goto out_restore_irqs;
1516                }
1517                zap_locks();
1518        }
1519
1520        lockdep_off();
1521        raw_spin_lock(&logbuf_lock);
1522        logbuf_cpu = this_cpu;
1523
1524        if (recursion_bug) {
1525                static const char recursion_msg[] =
1526                        "BUG: recent printk recursion!";
1527
1528                recursion_bug = 0;
1529                printed_len += strlen(recursion_msg);
1530                /* emit KERN_CRIT message */
1531                log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1532                          NULL, 0, recursion_msg, printed_len);
1533        }
1534
1535        /*
1536         * The printf needs to come first; we need the syslog
1537         * prefix which might be passed-in as a parameter.
1538         */
1539        text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1540
1541        /* mark and strip a trailing newline */
1542        if (text_len && text[text_len-1] == '\n') {
1543                text_len--;
1544                lflags |= LOG_NEWLINE;
1545        }
1546
1547        /* strip kernel syslog prefix and extract log level or control flags */
1548        if (facility == 0) {
1549                int kern_level = printk_get_level(text);
1550
1551                if (kern_level) {
1552                        const char *end_of_header = printk_skip_level(text);
1553                        switch (kern_level) {
1554                        case '0' ... '7':
1555                                if (level == -1)
1556                                        level = kern_level - '0';
1557                        case 'd':       /* KERN_DEFAULT */
1558                                lflags |= LOG_PREFIX;
1559                        case 'c':       /* KERN_CONT */
1560                                break;
1561                        }
1562                        text_len -= end_of_header - text;
1563                        text = (char *)end_of_header;
1564                }
1565        }
1566
1567        if (level == -1)
1568                level = default_message_loglevel;
1569
1570        if (dict)
1571                lflags |= LOG_PREFIX|LOG_NEWLINE;
1572
1573        if (!(lflags & LOG_NEWLINE)) {
1574                /*
1575                 * Flush the conflicting buffer. An earlier newline was missing,
1576                 * or another task also prints continuation lines.
1577                 */
1578                if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
1579                        cont_flush(LOG_NEWLINE);
1580
1581                /* buffer line if possible, otherwise store it right away */
1582                if (!cont_add(facility, level, text, text_len))
1583                        log_store(facility, level, lflags | LOG_CONT, 0,
1584                                  dict, dictlen, text, text_len);
1585        } else {
1586                bool stored = false;
1587
1588                /*
1589                 * If an earlier newline was missing and it was the same task,
1590                 * either merge it with the current buffer and flush, or if
1591                 * there was a race with interrupts (prefix == true) then just
1592                 * flush it out and store this line separately.
1593                 */
1594                if (cont.len && cont.owner == current) {
1595                        if (!(lflags & LOG_PREFIX))
1596                                stored = cont_add(facility, level, text, text_len);
1597                        cont_flush(LOG_NEWLINE);
1598                }
1599
1600                if (!stored)
1601                        log_store(facility, level, lflags, 0,
1602                                  dict, dictlen, text, text_len);
1603        }
1604        printed_len += text_len;
1605
1606        /*
1607         * Try to acquire and then immediately release the console semaphore.
1608         * The release will print out buffers and wake up /dev/kmsg and syslog()
1609         * users.
1610         *
1611         * The console_trylock_for_printk() function will release 'logbuf_lock'
1612         * regardless of whether it actually gets the console semaphore or not.
1613         */
1614        if (console_trylock_for_printk(this_cpu))
1615                console_unlock();
1616
1617        lockdep_on();
1618out_restore_irqs:
1619        local_irq_restore(flags);
1620
1621        return printed_len;
1622}
1623EXPORT_SYMBOL(vprintk_emit);
1624
1625asmlinkage int vprintk(const char *fmt, va_list args)
1626{
1627        return vprintk_emit(0, -1, NULL, 0, fmt, args);
1628}
1629EXPORT_SYMBOL(vprintk);
1630
1631asmlinkage int printk_emit(int facility, int level,
1632                           const char *dict, size_t dictlen,
1633                           const char *fmt, ...)
1634{
1635        va_list args;
1636        int r;
1637
1638        va_start(args, fmt);
1639        r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1640        va_end(args);
1641
1642        return r;
1643}
1644EXPORT_SYMBOL(printk_emit);
1645
1646/**
1647 * printk - print a kernel message
1648 * @fmt: format string
1649 *
1650 * This is printk(). It can be called from any context. We want it to work.
1651 *
1652 * We try to grab the console_lock. If we succeed, it's easy - we log the
1653 * output and call the console drivers.  If we fail to get the semaphore, we
1654 * place the output into the log buffer and return. The current holder of
1655 * the console_sem will notice the new output in console_unlock(); and will
1656 * send it to the consoles before releasing the lock.
1657 *
1658 * One effect of this deferred printing is that code which calls printk() and
1659 * then changes console_loglevel may break. This is because console_loglevel
1660 * is inspected when the actual printing occurs.
1661 *
1662 * See also:
1663 * printf(3)
1664 *
1665 * See the vsnprintf() documentation for format string extensions over C99.
1666 */
1667asmlinkage int printk(const char *fmt, ...)
1668{
1669        va_list args;
1670        int r;
1671
1672#ifdef CONFIG_KGDB_KDB
1673        if (unlikely(kdb_trap_printk)) {
1674                va_start(args, fmt);
1675                r = vkdb_printf(fmt, args);
1676                va_end(args);
1677                return r;
1678        }
1679#endif
1680        va_start(args, fmt);
1681        r = vprintk_emit(0, -1, NULL, 0, fmt, args);
1682        va_end(args);
1683
1684        return r;
1685}
1686EXPORT_SYMBOL(printk);
1687
1688#else /* CONFIG_PRINTK */
1689
1690#define LOG_LINE_MAX            0
1691#define PREFIX_MAX              0
1692#define LOG_LINE_MAX 0
1693static u64 syslog_seq;
1694static u32 syslog_idx;
1695static u64 console_seq;
1696static u32 console_idx;
1697static enum log_flags syslog_prev;
1698static u64 log_first_seq;
1699static u32 log_first_idx;
1700static u64 log_next_seq;
1701static enum log_flags console_prev;
1702static struct cont {
1703        size_t len;
1704        size_t cons;
1705        u8 level;
1706        bool flushed:1;
1707} cont;
1708static struct log *log_from_idx(u32 idx) { return NULL; }
1709static u32 log_next(u32 idx) { return 0; }
1710static void call_console_drivers(int level, const char *text, size_t len) {}
1711static size_t msg_print_text(const struct log *msg, enum log_flags prev,
1712                             bool syslog, char *buf, size_t size) { return 0; }
1713static size_t cont_print_text(char *text, size_t size) { return 0; }
1714
1715#endif /* CONFIG_PRINTK */
1716
1717static int __add_preferred_console(char *name, int idx, char *options,
1718                                   char *brl_options)
1719{
1720        struct console_cmdline *c;
1721        int i;
1722
1723        /*
1724         *      See if this tty is not yet registered, and
1725         *      if we have a slot free.
1726         */
1727        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1728                if (strcmp(console_cmdline[i].name, name) == 0 &&
1729                          console_cmdline[i].index == idx) {
1730                                if (!brl_options)
1731                                        selected_console = i;
1732                                return 0;
1733                }
1734        if (i == MAX_CMDLINECONSOLES)
1735                return -E2BIG;
1736        if (!brl_options)
1737                selected_console = i;
1738        c = &console_cmdline[i];
1739        strlcpy(c->name, name, sizeof(c->name));
1740        c->options = options;
1741#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1742        c->brl_options = brl_options;
1743#endif
1744        c->index = idx;
1745        return 0;
1746}
1747/*
1748 * Set up a list of consoles.  Called from init/main.c
1749 */
1750static int __init console_setup(char *str)
1751{
1752        char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
1753        char *s, *options, *brl_options = NULL;
1754        int idx;
1755
1756#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1757        if (!memcmp(str, "brl,", 4)) {
1758                brl_options = "";
1759                str += 4;
1760        } else if (!memcmp(str, "brl=", 4)) {
1761                brl_options = str + 4;
1762                str = strchr(brl_options, ',');
1763                if (!str) {
1764                        printk(KERN_ERR "need port name after brl=\n");
1765                        return 1;
1766                }
1767                *(str++) = 0;
1768        }
1769#endif
1770
1771        /*
1772         * Decode str into name, index, options.
1773         */
1774        if (str[0] >= '0' && str[0] <= '9') {
1775                strcpy(buf, "ttyS");
1776                strncpy(buf + 4, str, sizeof(buf) - 5);
1777        } else {
1778                strncpy(buf, str, sizeof(buf) - 1);
1779        }
1780        buf[sizeof(buf) - 1] = 0;
1781        if ((options = strchr(str, ',')) != NULL)
1782                *(options++) = 0;
1783#ifdef __sparc__
1784        if (!strcmp(str, "ttya"))
1785                strcpy(buf, "ttyS0");
1786        if (!strcmp(str, "ttyb"))
1787                strcpy(buf, "ttyS1");
1788#endif
1789        for (s = buf; *s; s++)
1790                if ((*s >= '0' && *s <= '9') || *s == ',')
1791                        break;
1792        idx = simple_strtoul(s, NULL, 10);
1793        *s = 0;
1794
1795        __add_preferred_console(buf, idx, options, brl_options);
1796        console_set_on_cmdline = 1;
1797        return 1;
1798}
1799__setup("console=", console_setup);
1800
1801/**
1802 * add_preferred_console - add a device to the list of preferred consoles.
1803 * @name: device name
1804 * @idx: device index
1805 * @options: options for this console
1806 *
1807 * The last preferred console added will be used for kernel messages
1808 * and stdin/out/err for init.  Normally this is used by console_setup
1809 * above to handle user-supplied console arguments; however it can also
1810 * be used by arch-specific code either to override the user or more
1811 * commonly to provide a default console (ie from PROM variables) when
1812 * the user has not supplied one.
1813 */
1814int add_preferred_console(char *name, int idx, char *options)
1815{
1816        return __add_preferred_console(name, idx, options, NULL);
1817}
1818
1819int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
1820{
1821        struct console_cmdline *c;
1822        int i;
1823
1824        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1825                if (strcmp(console_cmdline[i].name, name) == 0 &&
1826                          console_cmdline[i].index == idx) {
1827                                c = &console_cmdline[i];
1828                                strlcpy(c->name, name_new, sizeof(c->name));
1829                                c->name[sizeof(c->name) - 1] = 0;
1830                                c->options = options;
1831                                c->index = idx_new;
1832                                return i;
1833                }
1834        /* not found */
1835        return -1;
1836}
1837
1838bool console_suspend_enabled = 1;
1839EXPORT_SYMBOL(console_suspend_enabled);
1840
1841static int __init console_suspend_disable(char *str)
1842{
1843        console_suspend_enabled = 0;
1844        return 1;
1845}
1846__setup("no_console_suspend", console_suspend_disable);
1847module_param_named(console_suspend, console_suspend_enabled,
1848                bool, S_IRUGO | S_IWUSR);
1849MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
1850        " and hibernate operations");
1851
1852/**
1853 * suspend_console - suspend the console subsystem
1854 *
1855 * This disables printk() while we go into suspend states
1856 */
1857void suspend_console(void)
1858{
1859        if (!console_suspend_enabled)
1860                return;
1861        printk("Suspending console(s) (use no_console_suspend to debug)\n");
1862        console_lock();
1863        console_suspended = 1;
1864        up(&console_sem);
1865}
1866
1867void resume_console(void)
1868{
1869        if (!console_suspend_enabled)
1870                return;
1871        down(&console_sem);
1872        console_suspended = 0;
1873        console_unlock();
1874}
1875
1876/**
1877 * console_cpu_notify - print deferred console messages after CPU hotplug
1878 * @self: notifier struct
1879 * @action: CPU hotplug event
1880 * @hcpu: unused
1881 *
1882 * If printk() is called from a CPU that is not online yet, the messages
1883 * will be spooled but will not show up on the console.  This function is
1884 * called when a new CPU comes online (or fails to come up), and ensures
1885 * that any such output gets printed.
1886 */
1887static int __cpuinit console_cpu_notify(struct notifier_block *self,
1888        unsigned long action, void *hcpu)
1889{
1890        switch (action) {
1891        case CPU_ONLINE:
1892        case CPU_DEAD:
1893        case CPU_DOWN_FAILED:
1894        case CPU_UP_CANCELED:
1895                console_lock();
1896                console_unlock();
1897        }
1898        return NOTIFY_OK;
1899}
1900
1901/**
1902 * console_lock - lock the console system for exclusive use.
1903 *
1904 * Acquires a lock which guarantees that the caller has
1905 * exclusive access to the console system and the console_drivers list.
1906 *
1907 * Can sleep, returns nothing.
1908 */
1909void console_lock(void)
1910{
1911        BUG_ON(in_interrupt());
1912        down(&console_sem);
1913        if (console_suspended)
1914                return;
1915        console_locked = 1;
1916        console_may_schedule = 1;
1917}
1918EXPORT_SYMBOL(console_lock);
1919
1920/**
1921 * console_trylock - try to lock the console system for exclusive use.
1922 *
1923 * Tried to acquire a lock which guarantees that the caller has
1924 * exclusive access to the console system and the console_drivers list.
1925 *
1926 * returns 1 on success, and 0 on failure to acquire the lock.
1927 */
1928int console_trylock(void)
1929{
1930        if (down_trylock(&console_sem))
1931                return 0;
1932        if (console_suspended) {
1933                up(&console_sem);
1934                return 0;
1935        }
1936        console_locked = 1;
1937        console_may_schedule = 0;
1938        return 1;
1939}
1940EXPORT_SYMBOL(console_trylock);
1941
1942int is_console_locked(void)
1943{
1944        return console_locked;
1945}
1946
1947/*
1948 * Delayed printk version, for scheduler-internal messages:
1949 */
1950#define PRINTK_BUF_SIZE         512
1951
1952#define PRINTK_PENDING_WAKEUP   0x01
1953#define PRINTK_PENDING_SCHED    0x02
1954
1955static DEFINE_PER_CPU(int, printk_pending);
1956static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
1957
1958void printk_tick(void)
1959{
1960        if (__this_cpu_read(printk_pending)) {
1961                int pending = __this_cpu_xchg(printk_pending, 0);
1962                if (pending & PRINTK_PENDING_SCHED) {
1963                        char *buf = __get_cpu_var(printk_sched_buf);
1964                        printk(KERN_WARNING "[sched_delayed] %s", buf);
1965                }
1966                if (pending & PRINTK_PENDING_WAKEUP)
1967                        wake_up_interruptible(&log_wait);
1968        }
1969}
1970
1971int printk_needs_cpu(int cpu)
1972{
1973        if (cpu_is_offline(cpu))
1974                printk_tick();
1975        return __this_cpu_read(printk_pending);
1976}
1977
1978void wake_up_klogd(void)
1979{
1980        if (waitqueue_active(&log_wait))
1981                this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
1982}
1983
1984static void console_cont_flush(char *text, size_t size)
1985{
1986        unsigned long flags;
1987        size_t len;
1988
1989        raw_spin_lock_irqsave(&logbuf_lock, flags);
1990
1991        if (!cont.len)
1992                goto out;
1993
1994        /*
1995         * We still queue earlier records, likely because the console was
1996         * busy. The earlier ones need to be printed before this one, we
1997         * did not flush any fragment so far, so just let it queue up.
1998         */
1999        if (console_seq < log_next_seq && !cont.cons)
2000                goto out;
2001
2002        len = cont_print_text(text, size);
2003        raw_spin_unlock(&logbuf_lock);
2004        stop_critical_timings();
2005        call_console_drivers(cont.level, text, len);
2006        start_critical_timings();
2007        local_irq_restore(flags);
2008        return;
2009out:
2010        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2011}
2012
2013/**
2014 * console_unlock - unlock the console system
2015 *
2016 * Releases the console_lock which the caller holds on the console system
2017 * and the console driver list.
2018 *
2019 * While the console_lock was held, console output may have been buffered
2020 * by printk().  If this is the case, console_unlock(); emits
2021 * the output prior to releasing the lock.
2022 *
2023 * If there is output waiting, we wake /dev/kmsg and syslog() users.
2024 *
2025 * console_unlock(); may be called from any context.
2026 */
2027void console_unlock(void)
2028{
2029        static char text[LOG_LINE_MAX + PREFIX_MAX];
2030        static u64 seen_seq;
2031        unsigned long flags;
2032        bool wake_klogd = false;
2033        bool retry;
2034
2035        if (console_suspended) {
2036                up(&console_sem);
2037                return;
2038        }
2039
2040        console_may_schedule = 0;
2041
2042        /* flush buffered message fragment immediately to console */
2043        console_cont_flush(text, sizeof(text));
2044again:
2045        for (;;) {
2046                struct log *msg;
2047                size_t len;
2048                int level;
2049
2050                raw_spin_lock_irqsave(&logbuf_lock, flags);
2051                if (seen_seq != log_next_seq) {
2052                        wake_klogd = true;
2053                        seen_seq = log_next_seq;
2054                }
2055
2056                if (console_seq < log_first_seq) {
2057                        /* messages are gone, move to first one */
2058                        console_seq = log_first_seq;
2059                        console_idx = log_first_idx;
2060                        console_prev = 0;
2061                }
2062skip:
2063                if (console_seq == log_next_seq)
2064                        break;
2065
2066                msg = log_from_idx(console_idx);
2067                if (msg->flags & LOG_NOCONS) {
2068                        /*
2069                         * Skip record we have buffered and already printed
2070                         * directly to the console when we received it.
2071                         */
2072                        console_idx = log_next(console_idx);
2073                        console_seq++;
2074                        /*
2075                         * We will get here again when we register a new
2076                         * CON_PRINTBUFFER console. Clear the flag so we
2077                         * will properly dump everything later.
2078                         */
2079                        msg->flags &= ~LOG_NOCONS;
2080                        console_prev = msg->flags;
2081                        goto skip;
2082                }
2083
2084                level = msg->level;
2085                len = msg_print_text(msg, console_prev, false,
2086                                     text, sizeof(text));
2087                console_idx = log_next(console_idx);
2088                console_seq++;
2089                console_prev = msg->flags;
2090                raw_spin_unlock(&logbuf_lock);
2091
2092                stop_critical_timings();        /* don't trace print latency */
2093                call_console_drivers(level, text, len);
2094                start_critical_timings();
2095                local_irq_restore(flags);
2096        }
2097        console_locked = 0;
2098
2099        /* Release the exclusive_console once it is used */
2100        if (unlikely(exclusive_console))
2101                exclusive_console = NULL;
2102
2103        raw_spin_unlock(&logbuf_lock);
2104
2105        up(&console_sem);
2106
2107        /*
2108         * Someone could have filled up the buffer again, so re-check if there's
2109         * something to flush. In case we cannot trylock the console_sem again,
2110         * there's a new owner and the console_unlock() from them will do the
2111         * flush, no worries.
2112         */
2113        raw_spin_lock(&logbuf_lock);
2114        retry = console_seq != log_next_seq;
2115        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2116
2117        if (retry && console_trylock())
2118                goto again;
2119
2120        if (wake_klogd)
2121                wake_up_klogd();
2122}
2123EXPORT_SYMBOL(console_unlock);
2124
2125/**
2126 * console_conditional_schedule - yield the CPU if required
2127 *
2128 * If the console code is currently allowed to sleep, and
2129 * if this CPU should yield the CPU to another task, do
2130 * so here.
2131 *
2132 * Must be called within console_lock();.
2133 */
2134void __sched console_conditional_schedule(void)
2135{
2136        if (console_may_schedule)
2137                cond_resched();
2138}
2139EXPORT_SYMBOL(console_conditional_schedule);
2140
2141void console_unblank(void)
2142{
2143        struct console *c;
2144
2145        /*
2146         * console_unblank can no longer be called in interrupt context unless
2147         * oops_in_progress is set to 1..
2148         */
2149        if (oops_in_progress) {
2150                if (down_trylock(&console_sem) != 0)
2151                        return;
2152        } else
2153                console_lock();
2154
2155        console_locked = 1;
2156        console_may_schedule = 0;
2157        for_each_console(c)
2158                if ((c->flags & CON_ENABLED) && c->unblank)
2159                        c->unblank();
2160        console_unlock();
2161}
2162
2163/*
2164 * Return the console tty driver structure and its associated index
2165 */
2166struct tty_driver *console_device(int *index)
2167{
2168        struct console *c;
2169        struct tty_driver *driver = NULL;
2170
2171        console_lock();
2172        for_each_console(c) {
2173                if (!c->device)
2174                        continue;
2175                driver = c->device(c, index);
2176                if (driver)
2177                        break;
2178        }
2179        console_unlock();
2180        return driver;
2181}
2182
2183/*
2184 * Prevent further output on the passed console device so that (for example)
2185 * serial drivers can disable console output before suspending a port, and can
2186 * re-enable output afterwards.
2187 */
2188void console_stop(struct console *console)
2189{
2190        console_lock();
2191        console->flags &= ~CON_ENABLED;
2192        console_unlock();
2193}
2194EXPORT_SYMBOL(console_stop);
2195
2196void console_start(struct console *console)
2197{
2198        console_lock();
2199        console->flags |= CON_ENABLED;
2200        console_unlock();
2201}
2202EXPORT_SYMBOL(console_start);
2203
2204static int __read_mostly keep_bootcon;
2205
2206static int __init keep_bootcon_setup(char *str)
2207{
2208        keep_bootcon = 1;
2209        printk(KERN_INFO "debug: skip boot console de-registration.\n");
2210
2211        return 0;
2212}
2213
2214early_param("keep_bootcon", keep_bootcon_setup);
2215
2216/*
2217 * The console driver calls this routine during kernel initialization
2218 * to register the console printing procedure with printk() and to
2219 * print any messages that were printed by the kernel before the
2220 * console driver was initialized.
2221 *
2222 * This can happen pretty early during the boot process (because of
2223 * early_printk) - sometimes before setup_arch() completes - be careful
2224 * of what kernel features are used - they may not be initialised yet.
2225 *
2226 * There are two types of consoles - bootconsoles (early_printk) and
2227 * "real" consoles (everything which is not a bootconsole) which are
2228 * handled differently.
2229 *  - Any number of bootconsoles can be registered at any time.
2230 *  - As soon as a "real" console is registered, all bootconsoles
2231 *    will be unregistered automatically.
2232 *  - Once a "real" console is registered, any attempt to register a
2233 *    bootconsoles will be rejected
2234 */
2235void register_console(struct console *newcon)
2236{
2237        int i;
2238        unsigned long flags;
2239        struct console *bcon = NULL;
2240
2241        /*
2242         * before we register a new CON_BOOT console, make sure we don't
2243         * already have a valid console
2244         */
2245        if (console_drivers && newcon->flags & CON_BOOT) {
2246                /* find the last or real console */
2247                for_each_console(bcon) {
2248                        if (!(bcon->flags & CON_BOOT)) {
2249                                printk(KERN_INFO "Too late to register bootconsole %s%d\n",
2250                                        newcon->name, newcon->index);
2251                                return;
2252                        }
2253                }
2254        }
2255
2256        if (console_drivers && console_drivers->flags & CON_BOOT)
2257                bcon = console_drivers;
2258
2259        if (preferred_console < 0 || bcon || !console_drivers)
2260                preferred_console = selected_console;
2261
2262        if (newcon->early_setup)
2263                newcon->early_setup();
2264
2265        /*
2266         *      See if we want to use this console driver. If we
2267         *      didn't select a console we take the first one
2268         *      that registers here.
2269         */
2270        if (preferred_console < 0) {
2271                if (newcon->index < 0)
2272                        newcon->index = 0;
2273                if (newcon->setup == NULL ||
2274                    newcon->setup(newcon, NULL) == 0) {
2275                        newcon->flags |= CON_ENABLED;
2276                        if (newcon->device) {
2277                                newcon->flags |= CON_CONSDEV;
2278                                preferred_console = 0;
2279                        }
2280                }
2281        }
2282
2283        /*
2284         *      See if this console matches one we selected on
2285         *      the command line.
2286         */
2287        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
2288                        i++) {
2289                if (strcmp(console_cmdline[i].name, newcon->name) != 0)
2290                        continue;
2291                if (newcon->index >= 0 &&
2292                    newcon->index != console_cmdline[i].index)
2293                        continue;
2294                if (newcon->index < 0)
2295                        newcon->index = console_cmdline[i].index;
2296#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2297                if (console_cmdline[i].brl_options) {
2298                        newcon->flags |= CON_BRL;
2299                        braille_register_console(newcon,
2300                                        console_cmdline[i].index,
2301                                        console_cmdline[i].options,
2302                                        console_cmdline[i].brl_options);
2303                        return;
2304                }
2305#endif
2306                if (newcon->setup &&
2307                    newcon->setup(newcon, console_cmdline[i].options) != 0)
2308                        break;
2309                newcon->flags |= CON_ENABLED;
2310                newcon->index = console_cmdline[i].index;
2311                if (i == selected_console) {
2312                        newcon->flags |= CON_CONSDEV;
2313                        preferred_console = selected_console;
2314                }
2315                break;
2316        }
2317
2318        if (!(newcon->flags & CON_ENABLED))
2319                return;
2320
2321        /*
2322         * If we have a bootconsole, and are switching to a real console,
2323         * don't print everything out again, since when the boot console, and
2324         * the real console are the same physical device, it's annoying to
2325         * see the beginning boot messages twice
2326         */
2327        if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2328                newcon->flags &= ~CON_PRINTBUFFER;
2329
2330        /*
2331         *      Put this console in the list - keep the
2332         *      preferred driver at the head of the list.
2333         */
2334        console_lock();
2335        if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2336                newcon->next = console_drivers;
2337                console_drivers = newcon;
2338                if (newcon->next)
2339                        newcon->next->flags &= ~CON_CONSDEV;
2340        } else {
2341                newcon->next = console_drivers->next;
2342                console_drivers->next = newcon;
2343        }
2344        if (newcon->flags & CON_PRINTBUFFER) {
2345                /*
2346                 * console_unlock(); will print out the buffered messages
2347                 * for us.
2348                 */
2349                raw_spin_lock_irqsave(&logbuf_lock, flags);
2350                console_seq = syslog_seq;
2351                console_idx = syslog_idx;
2352                console_prev = syslog_prev;
2353                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2354                /*
2355                 * We're about to replay the log buffer.  Only do this to the
2356                 * just-registered console to avoid excessive message spam to
2357                 * the already-registered consoles.
2358                 */
2359                exclusive_console = newcon;
2360        }
2361        console_unlock();
2362        console_sysfs_notify();
2363
2364        /*
2365         * By unregistering the bootconsoles after we enable the real console
2366         * we get the "console xxx enabled" message on all the consoles -
2367         * boot consoles, real consoles, etc - this is to ensure that end
2368         * users know there might be something in the kernel's log buffer that
2369         * went to the bootconsole (that they do not see on the real console)
2370         */
2371        if (bcon &&
2372            ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2373            !keep_bootcon) {
2374                /* we need to iterate through twice, to make sure we print
2375                 * everything out, before we unregister the console(s)
2376                 */
2377                printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
2378                        newcon->name, newcon->index);
2379                for_each_console(bcon)
2380                        if (bcon->flags & CON_BOOT)
2381                                unregister_console(bcon);
2382        } else {
2383                printk(KERN_INFO "%sconsole [%s%d] enabled\n",
2384                        (newcon->flags & CON_BOOT) ? "boot" : "" ,
2385                        newcon->name, newcon->index);
2386        }
2387}
2388EXPORT_SYMBOL(register_console);
2389
2390int unregister_console(struct console *console)
2391{
2392        struct console *a, *b;
2393        int res = 1;
2394
2395#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2396        if (console->flags & CON_BRL)
2397                return braille_unregister_console(console);
2398#endif
2399
2400        console_lock();
2401        if (console_drivers == console) {
2402                console_drivers=console->next;
2403                res = 0;
2404        } else if (console_drivers) {
2405                for (a=console_drivers->next, b=console_drivers ;
2406                     a; b=a, a=b->next) {
2407                        if (a == console) {
2408                                b->next = a->next;
2409                                res = 0;
2410                                break;
2411                        }
2412                }
2413        }
2414
2415        /*
2416         * If this isn't the last console and it has CON_CONSDEV set, we
2417         * need to set it on the next preferred console.
2418         */
2419        if (console_drivers != NULL && console->flags & CON_CONSDEV)
2420                console_drivers->flags |= CON_CONSDEV;
2421
2422        console_unlock();
2423        console_sysfs_notify();
2424        return res;
2425}
2426EXPORT_SYMBOL(unregister_console);
2427
2428static int __init printk_late_init(void)
2429{
2430        struct console *con;
2431
2432        for_each_console(con) {
2433                if (!keep_bootcon && con->flags & CON_BOOT) {
2434                        printk(KERN_INFO "turn off boot console %s%d\n",
2435                                con->name, con->index);
2436                        unregister_console(con);
2437                }
2438        }
2439        hotcpu_notifier(console_cpu_notify, 0);
2440        return 0;
2441}
2442late_initcall(printk_late_init);
2443
2444#if defined CONFIG_PRINTK
2445
2446int printk_sched(const char *fmt, ...)
2447{
2448        unsigned long flags;
2449        va_list args;
2450        char *buf;
2451        int r;
2452
2453        local_irq_save(flags);
2454        buf = __get_cpu_var(printk_sched_buf);
2455
2456        va_start(args, fmt);
2457        r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
2458        va_end(args);
2459
2460        __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
2461        local_irq_restore(flags);
2462
2463        return r;
2464}
2465
2466/*
2467 * printk rate limiting, lifted from the networking subsystem.
2468 *
2469 * This enforces a rate limit: not more than 10 kernel messages
2470 * every 5s to make a denial-of-service attack impossible.
2471 */
2472DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2473
2474int __printk_ratelimit(const char *func)
2475{
2476        return ___ratelimit(&printk_ratelimit_state, func);
2477}
2478EXPORT_SYMBOL(__printk_ratelimit);
2479
2480/**
2481 * printk_timed_ratelimit - caller-controlled printk ratelimiting
2482 * @caller_jiffies: pointer to caller's state
2483 * @interval_msecs: minimum interval between prints
2484 *
2485 * printk_timed_ratelimit() returns true if more than @interval_msecs
2486 * milliseconds have elapsed since the last time printk_timed_ratelimit()
2487 * returned true.
2488 */
2489bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2490                        unsigned int interval_msecs)
2491{
2492        if (*caller_jiffies == 0
2493                        || !time_in_range(jiffies, *caller_jiffies,
2494                                        *caller_jiffies
2495                                        + msecs_to_jiffies(interval_msecs))) {
2496                *caller_jiffies = jiffies;
2497                return true;
2498        }
2499        return false;
2500}
2501EXPORT_SYMBOL(printk_timed_ratelimit);
2502
2503static DEFINE_SPINLOCK(dump_list_lock);
2504static LIST_HEAD(dump_list);
2505
2506/**
2507 * kmsg_dump_register - register a kernel log dumper.
2508 * @dumper: pointer to the kmsg_dumper structure
2509 *
2510 * Adds a kernel log dumper to the system. The dump callback in the
2511 * structure will be called when the kernel oopses or panics and must be
2512 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2513 */
2514int kmsg_dump_register(struct kmsg_dumper *dumper)
2515{
2516        unsigned long flags;
2517        int err = -EBUSY;
2518
2519        /* The dump callback needs to be set */
2520        if (!dumper->dump)
2521                return -EINVAL;
2522
2523        spin_lock_irqsave(&dump_list_lock, flags);
2524        /* Don't allow registering multiple times */
2525        if (!dumper->registered) {
2526                dumper->registered = 1;
2527                list_add_tail_rcu(&dumper->list, &dump_list);
2528                err = 0;
2529        }
2530        spin_unlock_irqrestore(&dump_list_lock, flags);
2531
2532        return err;
2533}
2534EXPORT_SYMBOL_GPL(kmsg_dump_register);
2535
2536/**
2537 * kmsg_dump_unregister - unregister a kmsg dumper.
2538 * @dumper: pointer to the kmsg_dumper structure
2539 *
2540 * Removes a dump device from the system. Returns zero on success and
2541 * %-EINVAL otherwise.
2542 */
2543int kmsg_dump_unregister(struct kmsg_dumper *dumper)
2544{
2545        unsigned long flags;
2546        int err = -EINVAL;
2547
2548        spin_lock_irqsave(&dump_list_lock, flags);
2549        if (dumper->registered) {
2550                dumper->registered = 0;
2551                list_del_rcu(&dumper->list);
2552                err = 0;
2553        }
2554        spin_unlock_irqrestore(&dump_list_lock, flags);
2555        synchronize_rcu();
2556
2557        return err;
2558}
2559EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
2560
2561static bool always_kmsg_dump;
2562module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
2563
2564/**
2565 * kmsg_dump - dump kernel log to kernel message dumpers.
2566 * @reason: the reason (oops, panic etc) for dumping
2567 *
2568 * Call each of the registered dumper's dump() callback, which can
2569 * retrieve the kmsg records with kmsg_dump_get_line() or
2570 * kmsg_dump_get_buffer().
2571 */
2572void kmsg_dump(enum kmsg_dump_reason reason)
2573{
2574        struct kmsg_dumper *dumper;
2575        unsigned long flags;
2576
2577        if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
2578                return;
2579
2580        rcu_read_lock();
2581        list_for_each_entry_rcu(dumper, &dump_list, list) {
2582                if (dumper->max_reason && reason > dumper->max_reason)
2583                        continue;
2584
2585                /* initialize iterator with data about the stored records */
2586                dumper->active = true;
2587
2588                raw_spin_lock_irqsave(&logbuf_lock, flags);
2589                dumper->cur_seq = clear_seq;
2590                dumper->cur_idx = clear_idx;
2591                dumper->next_seq = log_next_seq;
2592                dumper->next_idx = log_next_idx;
2593                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2594
2595                /* invoke dumper which will iterate over records */
2596                dumper->dump(dumper, reason);
2597
2598                /* reset iterator */
2599                dumper->active = false;
2600        }
2601        rcu_read_unlock();
2602}
2603
2604/**
2605 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
2606 * @dumper: registered kmsg dumper
2607 * @syslog: include the "<4>" prefixes
2608 * @line: buffer to copy the line to
2609 * @size: maximum size of the buffer
2610 * @len: length of line placed into buffer
2611 *
2612 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2613 * record, and copy one record into the provided buffer.
2614 *
2615 * Consecutive calls will return the next available record moving
2616 * towards the end of the buffer with the youngest messages.
2617 *
2618 * A return value of FALSE indicates that there are no more records to
2619 * read.
2620 *
2621 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
2622 */
2623bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
2624                               char *line, size_t size, size_t *len)
2625{
2626        struct log *msg;
2627        size_t l = 0;
2628        bool ret = false;
2629
2630        if (!dumper->active)
2631                goto out;
2632
2633        if (dumper->cur_seq < log_first_seq) {
2634                /* messages are gone, move to first available one */
2635                dumper->cur_seq = log_first_seq;
2636                dumper->cur_idx = log_first_idx;
2637        }
2638
2639        /* last entry */
2640        if (dumper->cur_seq >= log_next_seq)
2641                goto out;
2642
2643        msg = log_from_idx(dumper->cur_idx);
2644        l = msg_print_text(msg, 0, syslog, line, size);
2645
2646        dumper->cur_idx = log_next(dumper->cur_idx);
2647        dumper->cur_seq++;
2648        ret = true;
2649out:
2650        if (len)
2651                *len = l;
2652        return ret;
2653}
2654
2655/**
2656 * kmsg_dump_get_line - retrieve one kmsg log line
2657 * @dumper: registered kmsg dumper
2658 * @syslog: include the "<4>" prefixes
2659 * @line: buffer to copy the line to
2660 * @size: maximum size of the buffer
2661 * @len: length of line placed into buffer
2662 *
2663 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2664 * record, and copy one record into the provided buffer.
2665 *
2666 * Consecutive calls will return the next available record moving
2667 * towards the end of the buffer with the youngest messages.
2668 *
2669 * A return value of FALSE indicates that there are no more records to
2670 * read.
2671 */
2672bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
2673                        char *line, size_t size, size_t *len)
2674{
2675        unsigned long flags;
2676        bool ret;
2677
2678        raw_spin_lock_irqsave(&logbuf_lock, flags);
2679        ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
2680        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2681
2682        return ret;
2683}
2684EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
2685
2686/**
2687 * kmsg_dump_get_buffer - copy kmsg log lines
2688 * @dumper: registered kmsg dumper
2689 * @syslog: include the "<4>" prefixes
2690 * @buf: buffer to copy the line to
2691 * @size: maximum size of the buffer
2692 * @len: length of line placed into buffer
2693 *
2694 * Start at the end of the kmsg buffer and fill the provided buffer
2695 * with as many of the the *youngest* kmsg records that fit into it.
2696 * If the buffer is large enough, all available kmsg records will be
2697 * copied with a single call.
2698 *
2699 * Consecutive calls will fill the buffer with the next block of
2700 * available older records, not including the earlier retrieved ones.
2701 *
2702 * A return value of FALSE indicates that there are no more records to
2703 * read.
2704 */
2705bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
2706                          char *buf, size_t size, size_t *len)
2707{
2708        unsigned long flags;
2709        u64 seq;
2710        u32 idx;
2711        u64 next_seq;
2712        u32 next_idx;
2713        enum log_flags prev;
2714        size_t l = 0;
2715        bool ret = false;
2716
2717        if (!dumper->active)
2718                goto out;
2719
2720        raw_spin_lock_irqsave(&logbuf_lock, flags);
2721        if (dumper->cur_seq < log_first_seq) {
2722                /* messages are gone, move to first available one */
2723                dumper->cur_seq = log_first_seq;
2724                dumper->cur_idx = log_first_idx;
2725        }
2726
2727        /* last entry */
2728        if (dumper->cur_seq >= dumper->next_seq) {
2729                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2730                goto out;
2731        }
2732
2733        /* calculate length of entire buffer */
2734        seq = dumper->cur_seq;
2735        idx = dumper->cur_idx;
2736        prev = 0;
2737        while (seq < dumper->next_seq) {
2738                struct log *msg = log_from_idx(idx);
2739
2740                l += msg_print_text(msg, prev, true, NULL, 0);
2741                idx = log_next(idx);
2742                seq++;
2743                prev = msg->flags;
2744        }
2745
2746        /* move first record forward until length fits into the buffer */
2747        seq = dumper->cur_seq;
2748        idx = dumper->cur_idx;
2749        prev = 0;
2750        while (l > size && seq < dumper->next_seq) {
2751                struct log *msg = log_from_idx(idx);
2752
2753                l -= msg_print_text(msg, prev, true, NULL, 0);
2754                idx = log_next(idx);
2755                seq++;
2756                prev = msg->flags;
2757        }
2758
2759        /* last message in next interation */
2760        next_seq = seq;
2761        next_idx = idx;
2762
2763        l = 0;
2764        prev = 0;
2765        while (seq < dumper->next_seq) {
2766                struct log *msg = log_from_idx(idx);
2767
2768                l += msg_print_text(msg, prev, syslog, buf + l, size - l);
2769                idx = log_next(idx);
2770                seq++;
2771                prev = msg->flags;
2772        }
2773
2774        dumper->next_seq = next_seq;
2775        dumper->next_idx = next_idx;
2776        ret = true;
2777        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2778out:
2779        if (len)
2780                *len = l;
2781        return ret;
2782}
2783EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
2784
2785/**
2786 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
2787 * @dumper: registered kmsg dumper
2788 *
2789 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2790 * kmsg_dump_get_buffer() can be called again and used multiple
2791 * times within the same dumper.dump() callback.
2792 *
2793 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
2794 */
2795void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
2796{
2797        dumper->cur_seq = clear_seq;
2798        dumper->cur_idx = clear_idx;
2799        dumper->next_seq = log_next_seq;
2800        dumper->next_idx = log_next_idx;
2801}
2802
2803/**
2804 * kmsg_dump_rewind - reset the interator
2805 * @dumper: registered kmsg dumper
2806 *
2807 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2808 * kmsg_dump_get_buffer() can be called again and used multiple
2809 * times within the same dumper.dump() callback.
2810 */
2811void kmsg_dump_rewind(struct kmsg_dumper *dumper)
2812{
2813        unsigned long flags;
2814
2815        raw_spin_lock_irqsave(&logbuf_lock, flags);
2816        kmsg_dump_rewind_nolock(dumper);
2817        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2818}
2819EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
2820#endif
2821
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.