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
 744static bool __read_mostly ignore_loglevel;
 745
 746static int __init ignore_loglevel_setup(char *str)
 747{
 748        ignore_loglevel = 1;
 749        printk(KERN_INFO "debug: ignoring loglevel setting.\n");
 750
 751        return 0;
 752}
 753
 754early_param("ignore_loglevel", ignore_loglevel_setup);
 755module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
 756MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
 757        "print all kernel messages to the console.");
 758
 759#ifdef CONFIG_BOOT_PRINTK_DELAY
 760
 761static int boot_delay; /* msecs delay after each printk during bootup */
 762static unsigned long long loops_per_msec;       /* based on boot_delay */
 763
 764static int __init boot_delay_setup(char *str)
 765{
 766        unsigned long lpj;
 767
 768        lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
 769        loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
 770
 771        get_option(&str, &boot_delay);
 772        if (boot_delay > 10 * 1000)
 773                boot_delay = 0;
 774
 775        pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
 776                "HZ: %d, loops_per_msec: %llu\n",
 777                boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
 778        return 1;
 779}
 780__setup("boot_delay=", boot_delay_setup);
 781
 782static void boot_delay_msec(int level)
 783{
 784        unsigned long long k;
 785        unsigned long timeout;
 786
 787        if ((boot_delay == 0 || system_state != SYSTEM_BOOTING)
 788                || (level >= console_loglevel && !ignore_loglevel)) {
 789                return;
 790        }
 791
 792        k = (unsigned long long)loops_per_msec * boot_delay;
 793
 794        timeout = jiffies + msecs_to_jiffies(boot_delay);
 795        while (k) {
 796                k--;
 797                cpu_relax();
 798                /*
 799                 * use (volatile) jiffies to prevent
 800                 * compiler reduction; loop termination via jiffies
 801                 * is secondary and may or may not happen.
 802                 */
 803                if (time_after(jiffies, timeout))
 804                        break;
 805                touch_nmi_watchdog();
 806        }
 807}
 808#else
 809static inline void boot_delay_msec(int level)
 810{
 811}
 812#endif
 813
 814#ifdef CONFIG_SECURITY_DMESG_RESTRICT
 815int dmesg_restrict = 1;
 816#else
 817int dmesg_restrict;
 818#endif
 819
 820static int syslog_action_restricted(int type)
 821{
 822        if (dmesg_restrict)
 823                return 1;
 824        /* Unless restricted, we allow "read all" and "get buffer size" for everybody */
 825        return type != SYSLOG_ACTION_READ_ALL && type != SYSLOG_ACTION_SIZE_BUFFER;
 826}
 827
 828static int check_syslog_permissions(int type, bool from_file)
 829{
 830        /*
 831         * If this is from /proc/kmsg and we've already opened it, then we've
 832         * already done the capabilities checks at open time.
 833         */
 834        if (from_file && type != SYSLOG_ACTION_OPEN)
 835                return 0;
 836
 837        if (syslog_action_restricted(type)) {
 838                if (capable(CAP_SYSLOG))
 839                        return 0;
 840                /* For historical reasons, accept CAP_SYS_ADMIN too, with a warning */
 841                if (capable(CAP_SYS_ADMIN)) {
 842                        printk_once(KERN_WARNING "%s (%d): "
 843                                 "Attempt to access syslog with CAP_SYS_ADMIN "
 844                                 "but no CAP_SYSLOG (deprecated).\n",
 845                                 current->comm, task_pid_nr(current));
 846                        return 0;
 847                }
 848                return -EPERM;
 849        }
 850        return 0;
 851}
 852
 853#if defined(CONFIG_PRINTK_TIME)
 854static bool printk_time = 1;
 855#else
 856static bool printk_time;
 857#endif
 858module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
 859
 860static size_t print_time(u64 ts, char *buf)
 861{
 862        unsigned long rem_nsec;
 863
 864        if (!printk_time)
 865                return 0;
 866
 867        rem_nsec = do_div(ts, 1000000000);
 868
 869        if (!buf)
 870                return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
 871
 872        return sprintf(buf, "[%5lu.%06lu] ",
 873                       (unsigned long)ts, rem_nsec / 1000);
 874}
 875
 876static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
 877{
 878        size_t len = 0;
 879        unsigned int prefix = (msg->facility << 3) | msg->level;
 880
 881        if (syslog) {
 882                if (buf) {
 883                        len += sprintf(buf, "<%u>", prefix);
 884                } else {
 885                        len += 3;
 886                        if (prefix > 999)
 887                                len += 3;
 888                        else if (prefix > 99)
 889                                len += 2;
 890                        else if (prefix > 9)
 891                                len++;
 892                }
 893        }
 894
 895        len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
 896        return len;
 897}
 898
 899static size_t msg_print_text(const struct log *msg, enum log_flags prev,
 900                             bool syslog, char *buf, size_t size)
 901{
 902        const char *text = log_text(msg);
 903        size_t text_size = msg->text_len;
 904        bool prefix = true;
 905        bool newline = true;
 906        size_t len = 0;
 907
 908        if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
 909                prefix = false;
 910
 911        if (msg->flags & LOG_CONT) {
 912                if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
 913                        prefix = false;
 914
 915                if (!(msg->flags & LOG_NEWLINE))
 916                        newline = false;
 917        }
 918
 919        do {
 920                const char *next = memchr(text, '\n', text_size);
 921                size_t text_len;
 922
 923                if (next) {
 924                        text_len = next - text;
 925                        next++;
 926                        text_size -= next - text;
 927                } else {
 928                        text_len = text_size;
 929                }
 930
 931                if (buf) {
 932                        if (print_prefix(msg, syslog, NULL) +
 933                            text_len + 1 >= size - len)
 934                                break;
 935
 936                        if (prefix)
 937                                len += print_prefix(msg, syslog, buf + len);
 938                        memcpy(buf + len, text, text_len);
 939                        len += text_len;
 940                        if (next || newline)
 941                                buf[len++] = '\n';
 942                } else {
 943                        /* SYSLOG_ACTION_* buffer size only calculation */
 944                        if (prefix)
 945                                len += print_prefix(msg, syslog, NULL);
 946                        len += text_len;
 947                        if (next || newline)
 948                                len++;
 949                }
 950
 951                prefix = true;
 952                text = next;
 953        } while (text);
 954
 955        return len;
 956}
 957
 958static int syslog_print(char __user *buf, int size)
 959{
 960        char *text;
 961        struct log *msg;
 962        int len = 0;
 963
 964        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
 965        if (!text)
 966                return -ENOMEM;
 967
 968        while (size > 0) {
 969                size_t n;
 970                size_t skip;
 971
 972                raw_spin_lock_irq(&logbuf_lock);
 973                if (syslog_seq < log_first_seq) {
 974                        /* messages are gone, move to first one */
 975                        syslog_seq = log_first_seq;
 976                        syslog_idx = log_first_idx;
 977                        syslog_prev = 0;
 978                        syslog_partial = 0;
 979                }
 980                if (syslog_seq == log_next_seq) {
 981                        raw_spin_unlock_irq(&logbuf_lock);
 982                        break;
 983                }
 984
 985                skip = syslog_partial;
 986                msg = log_from_idx(syslog_idx);
 987                n = msg_print_text(msg, syslog_prev, true, text,
 988                                   LOG_LINE_MAX + PREFIX_MAX);
 989                if (n - syslog_partial <= size) {
 990                        /* message fits into buffer, move forward */
 991                        syslog_idx = log_next(syslog_idx);
 992                        syslog_seq++;
 993                        syslog_prev = msg->flags;
 994                        n -= syslog_partial;
 995                        syslog_partial = 0;
 996                } else if (!len){
 997                        /* partial read(), remember position */
 998                        n = size;
 999                        syslog_partial += n;
1000                } else
1001                        n = 0;
1002                raw_spin_unlock_irq(&logbuf_lock);
1003
1004                if (!n)
1005                        break;
1006
1007                if (copy_to_user(buf, text + skip, n)) {
1008                        if (!len)
1009                                len = -EFAULT;
1010                        break;
1011                }
1012
1013                len += n;
1014                size -= n;
1015                buf += n;
1016        }
1017
1018        kfree(text);
1019        return len;
1020}
1021
1022static int syslog_print_all(char __user *buf, int size, bool clear)
1023{
1024        char *text;
1025        int len = 0;
1026
1027        text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
1028        if (!text)
1029                return -ENOMEM;
1030
1031        raw_spin_lock_irq(&logbuf_lock);
1032        if (buf) {
1033                u64 next_seq;
1034                u64 seq;
1035                u32 idx;
1036                enum log_flags prev;
1037
1038                if (clear_seq < log_first_seq) {
1039                        /* messages are gone, move to first available one */
1040                        clear_seq = log_first_seq;
1041                        clear_idx = log_first_idx;
1042                }
1043
1044                /*
1045                 * Find first record that fits, including all following records,
1046                 * into the user-provided buffer for this dump.
1047                 */
1048                seq = clear_seq;
1049                idx = clear_idx;
1050                prev = 0;
1051                while (seq < log_next_seq) {
1052                        struct log *msg = log_from_idx(idx);
1053
1054                        len += msg_print_text(msg, prev, true, NULL, 0);
1055                        prev = msg->flags;
1056                        idx = log_next(idx);
1057                        seq++;
1058                }
1059
1060                /* move first record forward until length fits into the buffer */
1061                seq = clear_seq;
1062                idx = clear_idx;
1063                prev = 0;
1064                while (len > size && seq < log_next_seq) {
1065                        struct log *msg = log_from_idx(idx);
1066
1067                        len -= msg_print_text(msg, prev, true, NULL, 0);
1068                        prev = msg->flags;
1069                        idx = log_next(idx);
1070                        seq++;
1071                }
1072
1073                /* last message fitting into this dump */
1074                next_seq = log_next_seq;
1075
1076                len = 0;
1077                prev = 0;
1078                while (len >= 0 && seq < next_seq) {
1079                        struct log *msg = log_from_idx(idx);
1080                        int textlen;
1081
1082                        textlen = msg_print_text(msg, prev, true, text,
1083                                                 LOG_LINE_MAX + PREFIX_MAX);
1084                        if (textlen < 0) {
1085                                len = textlen;
1086                                break;
1087                        }
1088                        idx = log_next(idx);
1089                        seq++;
1090                        prev = msg->flags;
1091
1092                        raw_spin_unlock_irq(&logbuf_lock);
1093                        if (copy_to_user(buf + len, text, textlen))
1094                                len = -EFAULT;
1095                        else
1096                                len += textlen;
1097                        raw_spin_lock_irq(&logbuf_lock);
1098
1099                        if (seq < log_first_seq) {
1100                                /* messages are gone, move to next one */
1101                                seq = log_first_seq;
1102                                idx = log_first_idx;
1103                                prev = 0;
1104                        }
1105                }
1106        }
1107
1108        if (clear) {
1109                clear_seq = log_next_seq;
1110                clear_idx = log_next_idx;
1111        }
1112        raw_spin_unlock_irq(&logbuf_lock);
1113
1114        kfree(text);
1115        return len;
1116}
1117
1118int do_syslog(int type, char __user *buf, int len, bool from_file)
1119{
1120        bool clear = false;
1121        static int saved_console_loglevel = -1;
1122        int error;
1123
1124        error = check_syslog_permissions(type, from_file);
1125        if (error)
1126                goto out;
1127
1128        error = security_syslog(type);
1129        if (error)
1130                return error;
1131
1132        switch (type) {
1133        case SYSLOG_ACTION_CLOSE:       /* Close log */
1134                break;
1135        case SYSLOG_ACTION_OPEN:        /* Open log */
1136                break;
1137        case SYSLOG_ACTION_READ:        /* Read from log */
1138                error = -EINVAL;
1139                if (!buf || len < 0)
1140                        goto out;
1141                error = 0;
1142                if (!len)
1143                        goto out;
1144                if (!access_ok(VERIFY_WRITE, buf, len)) {
1145                        error = -EFAULT;
1146                        goto out;
1147                }
1148                error = wait_event_interruptible(log_wait,
1149                                                 syslog_seq != log_next_seq);
1150                if (error)
1151                        goto out;
1152                error = syslog_print(buf, len);
1153                break;
1154        /* Read/clear last kernel messages */
1155        case SYSLOG_ACTION_READ_CLEAR:
1156                clear = true;
1157                /* FALL THRU */
1158        /* Read last kernel messages */
1159        case SYSLOG_ACTION_READ_ALL:
1160                error = -EINVAL;
1161                if (!buf || len < 0)
1162                        goto out;
1163                error = 0;
1164                if (!len)
1165                        goto out;
1166                if (!access_ok(VERIFY_WRITE, buf, len)) {
1167                        error = -EFAULT;
1168                        goto out;
1169                }
1170                error = syslog_print_all(buf, len, clear);
1171                break;
1172        /* Clear ring buffer */
1173        case SYSLOG_ACTION_CLEAR:
1174                syslog_print_all(NULL, 0, true);
1175                break;
1176        /* Disable logging to console */
1177        case SYSLOG_ACTION_CONSOLE_OFF:
1178                if (saved_console_loglevel == -1)
1179                        saved_console_loglevel = console_loglevel;
1180                console_loglevel = minimum_console_loglevel;
1181                break;
1182        /* Enable logging to console */
1183        case SYSLOG_ACTION_CONSOLE_ON:
1184                if (saved_console_loglevel != -1) {
1185                        console_loglevel = saved_console_loglevel;
1186                        saved_console_loglevel = -1;
1187                }
1188                break;
1189        /* Set level of messages printed to console */
1190        case SYSLOG_ACTION_CONSOLE_LEVEL:
1191                error = -EINVAL;
1192                if (len < 1 || len > 8)
1193                        goto out;
1194                if (len < minimum_console_loglevel)
1195                        len = minimum_console_loglevel;
1196                console_loglevel = len;
1197                /* Implicitly re-enable logging to console */
1198                saved_console_loglevel = -1;
1199                error = 0;
1200                break;
1201        /* Number of chars in the log buffer */
1202        case SYSLOG_ACTION_SIZE_UNREAD:
1203                raw_spin_lock_irq(&logbuf_lock);
1204                if (syslog_seq < log_first_seq) {
1205                        /* messages are gone, move to first one */
1206                        syslog_seq = log_first_seq;
1207                        syslog_idx = log_first_idx;
1208                        syslog_prev = 0;
1209                        syslog_partial = 0;
1210                }
1211                if (from_file) {
1212                        /*
1213                         * Short-cut for poll(/"proc/kmsg") which simply checks
1214                         * for pending data, not the size; return the count of
1215                         * records, not the length.
1216                         */
1217                        error = log_next_idx - syslog_idx;
1218                } else {
1219                        u64 seq = syslog_seq;
1220                        u32 idx = syslog_idx;
1221                        enum log_flags prev = syslog_prev;
1222
1223                        error = 0;
1224                        while (seq < log_next_seq) {
1225                                struct log *msg = log_from_idx(idx);
1226
1227                                error += msg_print_text(msg, prev, true, NULL, 0);
1228                                idx = log_next(idx);
1229                                seq++;
1230                                prev = msg->flags;
1231                        }
1232                        error -= syslog_partial;
1233                }
1234                raw_spin_unlock_irq(&logbuf_lock);
1235                break;
1236        /* Size of the log buffer */
1237        case SYSLOG_ACTION_SIZE_BUFFER:
1238                error = log_buf_len;
1239                break;
1240        default:
1241                error = -EINVAL;
1242                break;
1243        }
1244out:
1245        return error;
1246}
1247
1248SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
1249{
1250        return do_syslog(type, buf, len, SYSLOG_FROM_CALL);
1251}
1252
1253/*
1254 * Call the console drivers, asking them to write out
1255 * log_buf[start] to log_buf[end - 1].
1256 * The console_lock must be held.
1257 */
1258static void call_console_drivers(int level, const char *text, size_t len)
1259{
1260        struct console *con;
1261
1262        trace_console(text, 0, len, len);
1263
1264        if (level >= console_loglevel && !ignore_loglevel)
1265                return;
1266        if (!console_drivers)
1267                return;
1268
1269        for_each_console(con) {
1270                if (exclusive_console && con != exclusive_console)
1271                        continue;
1272                if (!(con->flags & CON_ENABLED))
1273                        continue;
1274                if (!con->write)
1275                        continue;
1276                if (!cpu_online(smp_processor_id()) &&
1277                    !(con->flags & CON_ANYTIME))
1278                        continue;
1279                con->write(con, text, len);
1280        }
1281}
1282
1283/*
1284 * Zap console related locks when oopsing. Only zap at most once
1285 * every 10 seconds, to leave time for slow consoles to print a
1286 * full oops.
1287 */
1288static void zap_locks(void)
1289{
1290        static unsigned long oops_timestamp;
1291
1292        if (time_after_eq(jiffies, oops_timestamp) &&
1293                        !time_after(jiffies, oops_timestamp + 30 * HZ))
1294                return;
1295
1296        oops_timestamp = jiffies;
1297
1298        debug_locks_off();
1299        /* If a crash is occurring, make sure we can't deadlock */
1300        raw_spin_lock_init(&logbuf_lock);
1301        /* And make sure that we print immediately */
1302        sema_init(&console_sem, 1);
1303}
1304
1305/* Check if we have any console registered that can be called early in boot. */
1306static int have_callable_console(void)
1307{
1308        struct console *con;
1309
1310        for_each_console(con)
1311                if (con->flags & CON_ANYTIME)
1312                        return 1;
1313
1314        return 0;
1315}
1316
1317/*
1318 * Can we actually use the console at this time on this cpu?
1319 *
1320 * Console drivers may assume that per-cpu resources have
1321 * been allocated. So unless they're explicitly marked as
1322 * being able to cope (CON_ANYTIME) don't call them until
1323 * this CPU is officially up.
1324 */
1325static inline int can_use_console(unsigned int cpu)
1326{
1327        return cpu_online(cpu) || have_callable_console();
1328}
1329
1330/*
1331 * Try to get console ownership to actually show the kernel
1332 * messages from a 'printk'. Return true (and with the
1333 * console_lock held, and 'console_locked' set) if it
1334 * is successful, false otherwise.
1335 *
1336 * This gets called with the 'logbuf_lock' spinlock held and
1337 * interrupts disabled. It should return with 'lockbuf_lock'
1338 * released but interrupts still disabled.
1339 */
1340static int console_trylock_for_printk(unsigned int cpu)
1341        __releases(&logbuf_lock)
1342{
1343        int retval = 0, wake = 0;
1344
1345        if (console_trylock()) {
1346                retval = 1;
1347
1348                /*
1349                 * If we can't use the console, we need to release
1350                 * the console semaphore by hand to avoid flushing
1351                 * the buffer. We need to hold the console semaphore
1352                 * in order to do this test safely.
1353                 */
1354                if (!can_use_console(cpu)) {
1355                        console_locked = 0;
1356                        wake = 1;
1357                        retval = 0;
1358                }
1359        }
1360        logbuf_cpu = UINT_MAX;
1361        if (wake)
1362                up(&console_sem);
1363        raw_spin_unlock(&logbuf_lock);
1364        return retval;
1365}
1366
1367int printk_delay_msec __read_mostly;
1368
1369static inline void printk_delay(void)
1370{
1371        if (unlikely(printk_delay_msec)) {
1372                int m = printk_delay_msec;
1373
1374                while (m--) {
1375                        mdelay(1);
1376                        touch_nmi_watchdog();
1377                }
1378        }
1379}
1380
1381/*
1382 * Continuation lines are buffered, and not committed to the record buffer
1383 * until the line is complete, or a race forces it. The line fragments
1384 * though, are printed immediately to the consoles to ensure everything has
1385 * reached the console in case of a kernel crash.
1386 */
1387static struct cont {
1388        char buf[LOG_LINE_MAX];
1389        size_t len;                     /* length == 0 means unused buffer */
1390        size_t cons;                    /* bytes written to console */
1391        struct task_struct *owner;      /* task of first print*/
1392        u64 ts_nsec;                    /* time of first print */
1393        u8 level;                       /* log level of first message */
1394        u8 facility;                    /* log level of first message */
1395        enum log_flags flags;           /* prefix, newline flags */
1396        bool flushed:1;                 /* buffer sealed and committed */
1397} cont;
1398
1399static void cont_flush(enum log_flags flags)
1400{
1401        if (cont.flushed)
1402                return;
1403        if (cont.len == 0)
1404                return;
1405
1406        if (cont.cons) {
1407                /*
1408                 * If a fragment of this line was directly flushed to the
1409                 * console; wait for the console to pick up the rest of the
1410                 * line. LOG_NOCONS suppresses a duplicated output.
1411                 */
1412                log_store(cont.facility, cont.level, flags | LOG_NOCONS,
1413                          cont.ts_nsec, NULL, 0, cont.buf, cont.len);
1414                cont.flags = flags;
1415                cont.flushed = true;
1416        } else {
1417                /*
1418                 * If no fragment of this line ever reached the console,
1419                 * just submit it to the store and free the buffer.
1420                 */
1421                log_store(cont.facility, cont.level, flags, 0,
1422                          NULL, 0, cont.buf, cont.len);
1423                cont.len = 0;
1424        }
1425}
1426
1427static bool cont_add(int facility, int level, const char *text, size_t len)
1428{
1429        if (cont.len && cont.flushed)
1430                return false;
1431
1432        if (cont.len + len > sizeof(cont.buf)) {
1433                /* the line gets too long, split it up in separate records */
1434                cont_flush(LOG_CONT);
1435                return false;
1436        }
1437
1438        if (!cont.len) {
1439                cont.facility = facility;
1440                cont.level = level;
1441                cont.owner = current;
1442                cont.ts_nsec = local_clock();
1443                cont.flags = 0;
1444                cont.cons = 0;
1445                cont.flushed = false;
1446        }
1447
1448        memcpy(cont.buf + cont.len, text, len);
1449        cont.len += len;
1450
1451        if (cont.len > (sizeof(cont.buf) * 80) / 100)
1452                cont_flush(LOG_CONT);
1453
1454        return true;
1455}
1456
1457static size_t cont_print_text(char *text, size_t size)
1458{
1459        size_t textlen = 0;
1460        size_t len;
1461
1462        if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
1463                textlen += print_time(cont.ts_nsec, text);
1464                size -= textlen;
1465        }
1466
1467        len = cont.len - cont.cons;
1468        if (len > 0) {
1469                if (len+1 > size)
1470                        len = size-1;
1471                memcpy(text + textlen, cont.buf + cont.cons, len);
1472                textlen += len;
1473                cont.cons = cont.len;
1474        }
1475
1476        if (cont.flushed) {
1477                if (cont.flags & LOG_NEWLINE)
1478                        text[textlen++] = '\n';
1479                /* got everything, release buffer */
1480                cont.len = 0;
1481        }
1482        return textlen;
1483}
1484
1485asmlinkage int vprintk_emit(int facility, int level,
1486                            const char *dict, size_t dictlen,
1487                            const char *fmt, va_list args)
1488{
1489        static int recursion_bug;
1490        static char textbuf[LOG_LINE_MAX];
1491        char *text = textbuf;
1492        size_t text_len;
1493        enum log_flags lflags = 0;
1494        unsigned long flags;
1495        int this_cpu;
1496        int printed_len = 0;
1497
1498        boot_delay_msec(level);
1499        printk_delay();
1500
1501        /* This stops the holder of console_sem just where we want him */
1502        local_irq_save(flags);
1503        this_cpu = smp_processor_id();
1504
1505        /*
1506         * Ouch, printk recursed into itself!
1507         */
1508        if (unlikely(logbuf_cpu == this_cpu)) {
1509                /*
1510                 * If a crash is occurring during printk() on this CPU,
1511                 * then try to get the crash message out but make sure
1512                 * we can't deadlock. Otherwise just return to avoid the
1513                 * recursion and return - but flag the recursion so that
1514                 * it can be printed at the next appropriate moment:
1515                 */
1516                if (!oops_in_progress && !lockdep_recursing(current)) {
1517                        recursion_bug = 1;
1518                        goto out_restore_irqs;
1519                }
1520                zap_locks();
1521        }
1522
1523        lockdep_off();
1524        raw_spin_lock(&logbuf_lock);
1525        logbuf_cpu = this_cpu;
1526
1527        if (recursion_bug) {
1528                static const char recursion_msg[] =
1529                        "BUG: recent printk recursion!";
1530
1531                recursion_bug = 0;
1532                printed_len += strlen(recursion_msg);
1533                /* emit KERN_CRIT message */
1534                log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
1535                          NULL, 0, recursion_msg, printed_len);
1536        }
1537
1538        /*
1539         * The printf needs to come first; we need the syslog
1540         * prefix which might be passed-in as a parameter.
1541         */
1542        text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
1543
1544        /* mark and strip a trailing newline */
1545        if (text_len && text[text_len-1] == '\n') {
1546                text_len--;
1547                lflags |= LOG_NEWLINE;
1548        }
1549
1550        /* strip kernel syslog prefix and extract log level or control flags */
1551        if (facility == 0) {
1552                int kern_level = printk_get_level(text);
1553
1554                if (kern_level) {
1555                        const char *end_of_header = printk_skip_level(text);
1556                        switch (kern_level) {
1557                        case '0' ... '7':
1558                                if (level == -1)
1559                                        level = kern_level - '0';
1560                        case 'd':       /* KERN_DEFAULT */
1561                                lflags |= LOG_PREFIX;
1562                        case 'c':       /* KERN_CONT */
1563                                break;
1564                        }
1565                        text_len -= end_of_header - text;
1566                        text = (char *)end_of_header;
1567                }
1568        }
1569
1570        if (level == -1)
1571                level = default_message_loglevel;
1572
1573        if (dict)
1574                lflags |= LOG_PREFIX|LOG_NEWLINE;
1575
1576        if (!(lflags & LOG_NEWLINE)) {
1577                /*
1578                 * Flush the conflicting buffer. An earlier newline was missing,
1579                 * or another task also prints continuation lines.
1580                 */
1581                if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
1582                        cont_flush(LOG_NEWLINE);
1583
1584                /* buffer line if possible, otherwise store it right away */
1585                if (!cont_add(facility, level, text, text_len))
1586                        log_store(facility, level, lflags | LOG_CONT, 0,
1587                                  dict, dictlen, text, text_len);
1588        } else {
1589                bool stored = false;
1590
1591                /*
1592                 * If an earlier newline was missing and it was the same task,
1593                 * either merge it with the current buffer and flush, or if
1594                 * there was a race with interrupts (prefix == true) then just
1595                 * flush it out and store this line separately.
1596                 */
1597                if (cont.len && cont.owner == current) {
1598                        if (!(lflags & LOG_PREFIX))
1599                                stored = cont_add(facility, level, text, text_len);
1600                        cont_flush(LOG_NEWLINE);
1601                }
1602
1603                if (!stored)
1604                        log_store(facility, level, lflags, 0,
1605                                  dict, dictlen, text, text_len);
1606        }
1607        printed_len += text_len;
1608
1609        /*
1610         * Try to acquire and then immediately release the console semaphore.
1611         * The release will print out buffers and wake up /dev/kmsg and syslog()
1612         * users.
1613         *
1614         * The console_trylock_for_printk() function will release 'logbuf_lock'
1615         * regardless of whether it actually gets the console semaphore or not.
1616         */
1617        if (console_trylock_for_printk(this_cpu))
1618                console_unlock();
1619
1620        lockdep_on();
1621out_restore_irqs:
1622        local_irq_restore(flags);
1623
1624        return printed_len;
1625}
1626EXPORT_SYMBOL(vprintk_emit);
1627
1628asmlinkage int vprintk(const char *fmt, va_list args)
1629{
1630        return vprintk_emit(0, -1, NULL, 0, fmt, args);
1631}
1632EXPORT_SYMBOL(vprintk);
1633
1634asmlinkage int printk_emit(int facility, int level,
1635                           const char *dict, size_t dictlen,
1636                           const char *fmt, ...)
1637{
1638        va_list args;
1639        int r;
1640
1641        va_start(args, fmt);
1642        r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
1643        va_end(args);
1644
1645        return r;
1646}
1647EXPORT_SYMBOL(printk_emit);
1648
1649/**
1650 * printk - print a kernel message
1651 * @fmt: format string
1652 *
1653 * This is printk(). It can be called from any context. We want it to work.
1654 *
1655 * We try to grab the console_lock. If we succeed, it's easy - we log the
1656 * output and call the console drivers.  If we fail to get the semaphore, we
1657 * place the output into the log buffer and return. The current holder of
1658 * the console_sem will notice the new output in console_unlock(); and will
1659 * send it to the consoles before releasing the lock.
1660 *
1661 * One effect of this deferred printing is that code which calls printk() and
1662 * then changes console_loglevel may break. This is because console_loglevel
1663 * is inspected when the actual printing occurs.
1664 *
1665 * See also:
1666 * printf(3)
1667 *
1668 * See the vsnprintf() documentation for format string extensions over C99.
1669 */
1670asmlinkage int printk(const char *fmt, ...)
1671{
1672        va_list args;
1673        int r;
1674
1675#ifdef CONFIG_KGDB_KDB
1676        if (unlikely(kdb_trap_printk)) {
1677                va_start(args, fmt);
1678                r = vkdb_printf(fmt, args);
1679                va_end(args);
1680                return r;
1681        }
1682#endif
1683        va_start(args, fmt);
1684        r = vprintk_emit(0, -1, NULL, 0, fmt, args);
1685        va_end(args);
1686
1687        return r;
1688}
1689EXPORT_SYMBOL(printk);
1690
1691#else /* CONFIG_PRINTK */
1692
1693#define LOG_LINE_MAX            0
1694#define PREFIX_MAX              0
1695#define LOG_LINE_MAX 0
1696static u64 syslog_seq;
1697static u32 syslog_idx;
1698static u64 console_seq;
1699static u32 console_idx;
1700static enum log_flags syslog_prev;
1701static u64 log_first_seq;
1702static u32 log_first_idx;
1703static u64 log_next_seq;
1704static enum log_flags console_prev;
1705static struct cont {
1706        size_t len;
1707        size_t cons;
1708        u8 level;
1709        bool flushed:1;
1710} cont;
1711static struct log *log_from_idx(u32 idx) { return NULL; }
1712static u32 log_next(u32 idx) { return 0; }
1713static void call_console_drivers(int level, const char *text, size_t len) {}
1714static size_t msg_print_text(const struct log *msg, enum log_flags prev,
1715                             bool syslog, char *buf, size_t size) { return 0; }
1716static size_t cont_print_text(char *text, size_t size) { return 0; }
1717
1718#endif /* CONFIG_PRINTK */
1719
1720static int __add_preferred_console(char *name, int idx, char *options,
1721                                   char *brl_options)
1722{
1723        struct console_cmdline *c;
1724        int i;
1725
1726        /*
1727         *      See if this tty is not yet registered, and
1728         *      if we have a slot free.
1729         */
1730        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1731                if (strcmp(console_cmdline[i].name, name) == 0 &&
1732                          console_cmdline[i].index == idx) {
1733                                if (!brl_options)
1734                                        selected_console = i;
1735                                return 0;
1736                }
1737        if (i == MAX_CMDLINECONSOLES)
1738                return -E2BIG;
1739        if (!brl_options)
1740                selected_console = i;
1741        c = &console_cmdline[i];
1742        strlcpy(c->name, name, sizeof(c->name));
1743        c->options = options;
1744#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1745        c->brl_options = brl_options;
1746#endif
1747        c->index = idx;
1748        return 0;
1749}
1750/*
1751 * Set up a list of consoles.  Called from init/main.c
1752 */
1753static int __init console_setup(char *str)
1754{
1755        char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
1756        char *s, *options, *brl_options = NULL;
1757        int idx;
1758
1759#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
1760        if (!memcmp(str, "brl,", 4)) {
1761                brl_options = "";
1762                str += 4;
1763        } else if (!memcmp(str, "brl=", 4)) {
1764                brl_options = str + 4;
1765                str = strchr(brl_options, ',');
1766                if (!str) {
1767                        printk(KERN_ERR "need port name after brl=\n");
1768                        return 1;
1769                }
1770                *(str++) = 0;
1771        }
1772#endif
1773
1774        /*
1775         * Decode str into name, index, options.
1776         */
1777        if (str[0] >= '0' && str[0] <= '9') {
1778                strcpy(buf, "ttyS");
1779                strncpy(buf + 4, str, sizeof(buf) - 5);
1780        } else {
1781                strncpy(buf, str, sizeof(buf) - 1);
1782        }
1783        buf[sizeof(buf) - 1] = 0;
1784        if ((options = strchr(str, ',')) != NULL)
1785                *(options++) = 0;
1786#ifdef __sparc__
1787        if (!strcmp(str, "ttya"))
1788                strcpy(buf, "ttyS0");
1789        if (!strcmp(str, "ttyb"))
1790                strcpy(buf, "ttyS1");
1791#endif
1792        for (s = buf; *s; s++)
1793                if ((*s >= '0' && *s <= '9') || *s == ',')
1794                        break;
1795        idx = simple_strtoul(s, NULL, 10);
1796        *s = 0;
1797
1798        __add_preferred_console(buf, idx, options, brl_options);
1799        console_set_on_cmdline = 1;
1800        return 1;
1801}
1802__setup("console=", console_setup);
1803
1804/**
1805 * add_preferred_console - add a device to the list of preferred consoles.
1806 * @name: device name
1807 * @idx: device index
1808 * @options: options for this console
1809 *
1810 * The last preferred console added will be used for kernel messages
1811 * and stdin/out/err for init.  Normally this is used by console_setup
1812 * above to handle user-supplied console arguments; however it can also
1813 * be used by arch-specific code either to override the user or more
1814 * commonly to provide a default console (ie from PROM variables) when
1815 * the user has not supplied one.
1816 */
1817int add_preferred_console(char *name, int idx, char *options)
1818{
1819        return __add_preferred_console(name, idx, options, NULL);
1820}
1821
1822int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
1823{
1824        struct console_cmdline *c;
1825        int i;
1826
1827        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
1828                if (strcmp(console_cmdline[i].name, name) == 0 &&
1829                          console_cmdline[i].index == idx) {
1830                                c = &console_cmdline[i];
1831                                strlcpy(c->name, name_new, sizeof(c->name));
1832                                c->name[sizeof(c->name) - 1] = 0;
1833                                c->options = options;
1834                                c->index = idx_new;
1835                                return i;
1836                }
1837        /* not found */
1838        return -1;
1839}
1840
1841bool console_suspend_enabled = 1;
1842EXPORT_SYMBOL(console_suspend_enabled);
1843
1844static int __init console_suspend_disable(char *str)
1845{
1846        console_suspend_enabled = 0;
1847        return 1;
1848}
1849__setup("no_console_suspend", console_suspend_disable);
1850module_param_named(console_suspend, console_suspend_enabled,
1851                bool, S_IRUGO | S_IWUSR);
1852MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
1853        " and hibernate operations");
1854
1855/**
1856 * suspend_console - suspend the console subsystem
1857 *
1858 * This disables printk() while we go into suspend states
1859 */
1860void suspend_console(void)
1861{
1862        if (!console_suspend_enabled)
1863                return;
1864        printk("Suspending console(s) (use no_console_suspend to debug)\n");
1865        console_lock();
1866        console_suspended = 1;
1867        up(&console_sem);
1868}
1869
1870void resume_console(void)
1871{
1872        if (!console_suspend_enabled)
1873                return;
1874        down(&console_sem);
1875        console_suspended = 0;
1876        console_unlock();
1877}
1878
1879/**
1880 * console_cpu_notify - print deferred console messages after CPU hotplug
1881 * @self: notifier struct
1882 * @action: CPU hotplug event
1883 * @hcpu: unused
1884 *
1885 * If printk() is called from a CPU that is not online yet, the messages
1886 * will be spooled but will not show up on the console.  This function is
1887 * called when a new CPU comes online (or fails to come up), and ensures
1888 * that any such output gets printed.
1889 */
1890static int __cpuinit console_cpu_notify(struct notifier_block *self,
1891        unsigned long action, void *hcpu)
1892{
1893        switch (action) {
1894        case CPU_ONLINE:
1895        case CPU_DEAD:
1896        case CPU_DOWN_FAILED:
1897        case CPU_UP_CANCELED:
1898                console_lock();
1899                console_unlock();
1900        }
1901        return NOTIFY_OK;
1902}
1903
1904/**
1905 * console_lock - lock the console system for exclusive use.
1906 *
1907 * Acquires a lock which guarantees that the caller has
1908 * exclusive access to the console system and the console_drivers list.
1909 *
1910 * Can sleep, returns nothing.
1911 */
1912void console_lock(void)
1913{
1914        might_sleep();
1915
1916        down(&console_sem);
1917        if (console_suspended)
1918                return;
1919        console_locked = 1;
1920        console_may_schedule = 1;
1921}
1922EXPORT_SYMBOL(console_lock);
1923
1924/**
1925 * console_trylock - try to lock the console system for exclusive use.
1926 *
1927 * Tried to acquire a lock which guarantees that the caller has
1928 * exclusive access to the console system and the console_drivers list.
1929 *
1930 * returns 1 on success, and 0 on failure to acquire the lock.
1931 */
1932int console_trylock(void)
1933{
1934        if (down_trylock(&console_sem))
1935                return 0;
1936        if (console_suspended) {
1937                up(&console_sem);
1938                return 0;
1939        }
1940        console_locked = 1;
1941        console_may_schedule = 0;
1942        return 1;
1943}
1944EXPORT_SYMBOL(console_trylock);
1945
1946int is_console_locked(void)
1947{
1948        return console_locked;
1949}
1950
1951/*
1952 * Delayed printk version, for scheduler-internal messages:
1953 */
1954#define PRINTK_BUF_SIZE         512
1955
1956#define PRINTK_PENDING_WAKEUP   0x01
1957#define PRINTK_PENDING_SCHED    0x02
1958
1959static DEFINE_PER_CPU(int, printk_pending);
1960static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
1961
1962void printk_tick(void)
1963{
1964        if (__this_cpu_read(printk_pending)) {
1965                int pending = __this_cpu_xchg(printk_pending, 0);
1966                if (pending & PRINTK_PENDING_SCHED) {
1967                        char *buf = __get_cpu_var(printk_sched_buf);
1968                        printk(KERN_WARNING "[sched_delayed] %s", buf);
1969                }
1970                if (pending & PRINTK_PENDING_WAKEUP)
1971                        wake_up_interruptible(&log_wait);
1972        }
1973}
1974
1975int printk_needs_cpu(int cpu)
1976{
1977        if (cpu_is_offline(cpu))
1978                printk_tick();
1979        return __this_cpu_read(printk_pending);
1980}
1981
1982void wake_up_klogd(void)
1983{
1984        if (waitqueue_active(&log_wait))
1985                this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
1986}
1987
1988static void console_cont_flush(char *text, size_t size)
1989{
1990        unsigned long flags;
1991        size_t len;
1992
1993        raw_spin_lock_irqsave(&logbuf_lock, flags);
1994
1995        if (!cont.len)
1996                goto out;
1997
1998        /*
1999         * We still queue earlier records, likely because the console was
2000         * busy. The earlier ones need to be printed before this one, we
2001         * did not flush any fragment so far, so just let it queue up.
2002         */
2003        if (console_seq < log_next_seq && !cont.cons)
2004                goto out;
2005
2006        len = cont_print_text(text, size);
2007        raw_spin_unlock(&logbuf_lock);
2008        stop_critical_timings();
2009        call_console_drivers(cont.level, text, len);
2010        start_critical_timings();
2011        local_irq_restore(flags);
2012        return;
2013out:
2014        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2015}
2016
2017/**
2018 * console_unlock - unlock the console system
2019 *
2020 * Releases the console_lock which the caller holds on the console system
2021 * and the console driver list.
2022 *
2023 * While the console_lock was held, console output may have been buffered
2024 * by printk().  If this is the case, console_unlock(); emits
2025 * the output prior to releasing the lock.
2026 *
2027 * If there is output waiting, we wake /dev/kmsg and syslog() users.
2028 *
2029 * console_unlock(); may be called from any context.
2030 */
2031void console_unlock(void)
2032{
2033        static char text[LOG_LINE_MAX + PREFIX_MAX];
2034        static u64 seen_seq;
2035        unsigned long flags;
2036        bool wake_klogd = false;
2037        bool retry;
2038
2039        if (console_suspended) {
2040                up(&console_sem);
2041                return;
2042        }
2043
2044        console_may_schedule = 0;
2045
2046        /* flush buffered message fragment immediately to console */
2047        console_cont_flush(text, sizeof(text));
2048again:
2049        for (;;) {
2050                struct log *msg;
2051                size_t len;
2052                int level;
2053
2054                raw_spin_lock_irqsave(&logbuf_lock, flags);
2055                if (seen_seq != log_next_seq) {
2056                        wake_klogd = true;
2057                        seen_seq = log_next_seq;
2058                }
2059
2060                if (console_seq < log_first_seq) {
2061                        /* messages are gone, move to first one */
2062                        console_seq = log_first_seq;
2063                        console_idx = log_first_idx;
2064                        console_prev = 0;
2065                }
2066skip:
2067                if (console_seq == log_next_seq)
2068                        break;
2069
2070                msg = log_from_idx(console_idx);
2071                if (msg->flags & LOG_NOCONS) {
2072                        /*
2073                         * Skip record we have buffered and already printed
2074                         * directly to the console when we received it.
2075                         */
2076                        console_idx = log_next(console_idx);
2077                        console_seq++;
2078                        /*
2079                         * We will get here again when we register a new
2080                         * CON_PRINTBUFFER console. Clear the flag so we
2081                         * will properly dump everything later.
2082                         */
2083                        msg->flags &= ~LOG_NOCONS;
2084                        console_prev = msg->flags;
2085                        goto skip;
2086                }
2087
2088                level = msg->level;
2089                len = msg_print_text(msg, console_prev, false,
2090                                     text, sizeof(text));
2091                console_idx = log_next(console_idx);
2092                console_seq++;
2093                console_prev = msg->flags;
2094                raw_spin_unlock(&logbuf_lock);
2095
2096                stop_critical_timings();        /* don't trace print latency */
2097                call_console_drivers(level, text, len);
2098                start_critical_timings();
2099                local_irq_restore(flags);
2100        }
2101        console_locked = 0;
2102
2103        /* Release the exclusive_console once it is used */
2104        if (unlikely(exclusive_console))
2105                exclusive_console = NULL;
2106
2107        raw_spin_unlock(&logbuf_lock);
2108
2109        up(&console_sem);
2110
2111        /*
2112         * Someone could have filled up the buffer again, so re-check if there's
2113         * something to flush. In case we cannot trylock the console_sem again,
2114         * there's a new owner and the console_unlock() from them will do the
2115         * flush, no worries.
2116         */
2117        raw_spin_lock(&logbuf_lock);
2118        retry = console_seq != log_next_seq;
2119        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2120
2121        if (retry && console_trylock())
2122                goto again;
2123
2124        if (wake_klogd)
2125                wake_up_klogd();
2126}
2127EXPORT_SYMBOL(console_unlock);
2128
2129/**
2130 * console_conditional_schedule - yield the CPU if required
2131 *
2132 * If the console code is currently allowed to sleep, and
2133 * if this CPU should yield the CPU to another task, do
2134 * so here.
2135 *
2136 * Must be called within console_lock();.
2137 */
2138void __sched console_conditional_schedule(void)
2139{
2140        if (console_may_schedule)
2141                cond_resched();
2142}
2143EXPORT_SYMBOL(console_conditional_schedule);
2144
2145void console_unblank(void)
2146{
2147        struct console *c;
2148
2149        /*
2150         * console_unblank can no longer be called in interrupt context unless
2151         * oops_in_progress is set to 1..
2152         */
2153        if (oops_in_progress) {
2154                if (down_trylock(&console_sem) != 0)
2155                        return;
2156        } else
2157                console_lock();
2158
2159        console_locked = 1;
2160        console_may_schedule = 0;
2161        for_each_console(c)
2162                if ((c->flags & CON_ENABLED) && c->unblank)
2163                        c->unblank();
2164        console_unlock();
2165}
2166
2167/*
2168 * Return the console tty driver structure and its associated index
2169 */
2170struct tty_driver *console_device(int *index)
2171{
2172        struct console *c;
2173        struct tty_driver *driver = NULL;
2174
2175        console_lock();
2176        for_each_console(c) {
2177                if (!c->device)
2178                        continue;
2179                driver = c->device(c, index);
2180                if (driver)
2181                        break;
2182        }
2183        console_unlock();
2184        return driver;
2185}
2186
2187/*
2188 * Prevent further output on the passed console device so that (for example)
2189 * serial drivers can disable console output before suspending a port, and can
2190 * re-enable output afterwards.
2191 */
2192void console_stop(struct console *console)
2193{
2194        console_lock();
2195        console->flags &= ~CON_ENABLED;
2196        console_unlock();
2197}
2198EXPORT_SYMBOL(console_stop);
2199
2200void console_start(struct console *console)
2201{
2202        console_lock();
2203        console->flags |= CON_ENABLED;
2204        console_unlock();
2205}
2206EXPORT_SYMBOL(console_start);
2207
2208static int __read_mostly keep_bootcon;
2209
2210static int __init keep_bootcon_setup(char *str)
2211{
2212        keep_bootcon = 1;
2213        printk(KERN_INFO "debug: skip boot console de-registration.\n");
2214
2215        return 0;
2216}
2217
2218early_param("keep_bootcon", keep_bootcon_setup);
2219
2220/*
2221 * The console driver calls this routine during kernel initialization
2222 * to register the console printing procedure with printk() and to
2223 * print any messages that were printed by the kernel before the
2224 * console driver was initialized.
2225 *
2226 * This can happen pretty early during the boot process (because of
2227 * early_printk) - sometimes before setup_arch() completes - be careful
2228 * of what kernel features are used - they may not be initialised yet.
2229 *
2230 * There are two types of consoles - bootconsoles (early_printk) and
2231 * "real" consoles (everything which is not a bootconsole) which are
2232 * handled differently.
2233 *  - Any number of bootconsoles can be registered at any time.
2234 *  - As soon as a "real" console is registered, all bootconsoles
2235 *    will be unregistered automatically.
2236 *  - Once a "real" console is registered, any attempt to register a
2237 *    bootconsoles will be rejected
2238 */
2239void register_console(struct console *newcon)
2240{
2241        int i;
2242        unsigned long flags;
2243        struct console *bcon = NULL;
2244
2245        /*
2246         * before we register a new CON_BOOT console, make sure we don't
2247         * already have a valid console
2248         */
2249        if (console_drivers && newcon->flags & CON_BOOT) {
2250                /* find the last or real console */
2251                for_each_console(bcon) {
2252                        if (!(bcon->flags & CON_BOOT)) {
2253                                printk(KERN_INFO "Too late to register bootconsole %s%d\n",
2254                                        newcon->name, newcon->index);
2255                                return;
2256                        }
2257                }
2258        }
2259
2260        if (console_drivers && console_drivers->flags & CON_BOOT)
2261                bcon = console_drivers;
2262
2263        if (preferred_console < 0 || bcon || !console_drivers)
2264                preferred_console = selected_console;
2265
2266        if (newcon->early_setup)
2267                newcon->early_setup();
2268
2269        /*
2270         *      See if we want to use this console driver. If we
2271         *      didn't select a console we take the first one
2272         *      that registers here.
2273         */
2274        if (preferred_console < 0) {
2275                if (newcon->index < 0)
2276                        newcon->index = 0;
2277                if (newcon->setup == NULL ||
2278                    newcon->setup(newcon, NULL) == 0) {
2279                        newcon->flags |= CON_ENABLED;
2280                        if (newcon->device) {
2281                                newcon->flags |= CON_CONSDEV;
2282                                preferred_console = 0;
2283                        }
2284                }
2285        }
2286
2287        /*
2288         *      See if this console matches one we selected on
2289         *      the command line.
2290         */
2291        for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
2292                        i++) {
2293                if (strcmp(console_cmdline[i].name, newcon->name) != 0)
2294                        continue;
2295                if (newcon->index >= 0 &&
2296                    newcon->index != console_cmdline[i].index)
2297                        continue;
2298                if (newcon->index < 0)
2299                        newcon->index = console_cmdline[i].index;
2300#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2301                if (console_cmdline[i].brl_options) {
2302                        newcon->flags |= CON_BRL;
2303                        braille_register_console(newcon,
2304                                        console_cmdline[i].index,
2305                                        console_cmdline[i].options,
2306                                        console_cmdline[i].brl_options);
2307                        return;
2308                }
2309#endif
2310                if (newcon->setup &&
2311                    newcon->setup(newcon, console_cmdline[i].options) != 0)
2312                        break;
2313                newcon->flags |= CON_ENABLED;
2314                newcon->index = console_cmdline[i].index;
2315                if (i == selected_console) {
2316                        newcon->flags |= CON_CONSDEV;
2317                        preferred_console = selected_console;
2318                }
2319                break;
2320        }
2321
2322        if (!(newcon->flags & CON_ENABLED))
2323                return;
2324
2325        /*
2326         * If we have a bootconsole, and are switching to a real console,
2327         * don't print everything out again, since when the boot console, and
2328         * the real console are the same physical device, it's annoying to
2329         * see the beginning boot messages twice
2330         */
2331        if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
2332                newcon->flags &= ~CON_PRINTBUFFER;
2333
2334        /*
2335         *      Put this console in the list - keep the
2336         *      preferred driver at the head of the list.
2337         */
2338        console_lock();
2339        if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
2340                newcon->next = console_drivers;
2341                console_drivers = newcon;
2342                if (newcon->next)
2343                        newcon->next->flags &= ~CON_CONSDEV;
2344        } else {
2345                newcon->next = console_drivers->next;
2346                console_drivers->next = newcon;
2347        }
2348        if (newcon->flags & CON_PRINTBUFFER) {
2349                /*
2350                 * console_unlock(); will print out the buffered messages
2351                 * for us.
2352                 */
2353                raw_spin_lock_irqsave(&logbuf_lock, flags);
2354                console_seq = syslog_seq;
2355                console_idx = syslog_idx;
2356                console_prev = syslog_prev;
2357                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2358                /*
2359                 * We're about to replay the log buffer.  Only do this to the
2360                 * just-registered console to avoid excessive message spam to
2361                 * the already-registered consoles.
2362                 */
2363                exclusive_console = newcon;
2364        }
2365        console_unlock();
2366        console_sysfs_notify();
2367
2368        /*
2369         * By unregistering the bootconsoles after we enable the real console
2370         * we get the "console xxx enabled" message on all the consoles -
2371         * boot consoles, real consoles, etc - this is to ensure that end
2372         * users know there might be something in the kernel's log buffer that
2373         * went to the bootconsole (that they do not see on the real console)
2374         */
2375        if (bcon &&
2376            ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
2377            !keep_bootcon) {
2378                /* we need to iterate through twice, to make sure we print
2379                 * everything out, before we unregister the console(s)
2380                 */
2381                printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
2382                        newcon->name, newcon->index);
2383                for_each_console(bcon)
2384                        if (bcon->flags & CON_BOOT)
2385                                unregister_console(bcon);
2386        } else {
2387                printk(KERN_INFO "%sconsole [%s%d] enabled\n",
2388                        (newcon->flags & CON_BOOT) ? "boot" : "" ,
2389                        newcon->name, newcon->index);
2390        }
2391}
2392EXPORT_SYMBOL(register_console);
2393
2394int unregister_console(struct console *console)
2395{
2396        struct console *a, *b;
2397        int res = 1;
2398
2399#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
2400        if (console->flags & CON_BRL)
2401                return braille_unregister_console(console);
2402#endif
2403
2404        console_lock();
2405        if (console_drivers == console) {
2406                console_drivers=console->next;
2407                res = 0;
2408        } else if (console_drivers) {
2409                for (a=console_drivers->next, b=console_drivers ;
2410                     a; b=a, a=b->next) {
2411                        if (a == console) {
2412                                b->next = a->next;
2413                                res = 0;
2414                                break;
2415                        }
2416                }
2417        }
2418
2419        /*
2420         * If this isn't the last console and it has CON_CONSDEV set, we
2421         * need to set it on the next preferred console.
2422         */
2423        if (console_drivers != NULL && console->flags & CON_CONSDEV)
2424                console_drivers->flags |= CON_CONSDEV;
2425
2426        console_unlock();
2427        console_sysfs_notify();
2428        return res;
2429}
2430EXPORT_SYMBOL(unregister_console);
2431
2432static int __init printk_late_init(void)
2433{
2434        struct console *con;
2435
2436        for_each_console(con) {
2437                if (!keep_bootcon && con->flags & CON_BOOT) {
2438                        printk(KERN_INFO "turn off boot console %s%d\n",
2439                                con->name, con->index);
2440                        unregister_console(con);
2441                }
2442        }
2443        hotcpu_notifier(console_cpu_notify, 0);
2444        return 0;
2445}
2446late_initcall(printk_late_init);
2447
2448#if defined CONFIG_PRINTK
2449
2450int printk_sched(const char *fmt, ...)
2451{
2452        unsigned long flags;
2453        va_list args;
2454        char *buf;
2455        int r;
2456
2457        local_irq_save(flags);
2458        buf = __get_cpu_var(printk_sched_buf);
2459
2460        va_start(args, fmt);
2461        r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
2462        va_end(args);
2463
2464        __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
2465        local_irq_restore(flags);
2466
2467        return r;
2468}
2469
2470/*
2471 * printk rate limiting, lifted from the networking subsystem.
2472 *
2473 * This enforces a rate limit: not more than 10 kernel messages
2474 * every 5s to make a denial-of-service attack impossible.
2475 */
2476DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
2477
2478int __printk_ratelimit(const char *func)
2479{
2480        return ___ratelimit(&printk_ratelimit_state, func);
2481}
2482EXPORT_SYMBOL(__printk_ratelimit);
2483
2484/**
2485 * printk_timed_ratelimit - caller-controlled printk ratelimiting
2486 * @caller_jiffies: pointer to caller's state
2487 * @interval_msecs: minimum interval between prints
2488 *
2489 * printk_timed_ratelimit() returns true if more than @interval_msecs
2490 * milliseconds have elapsed since the last time printk_timed_ratelimit()
2491 * returned true.
2492 */
2493bool printk_timed_ratelimit(unsigned long *caller_jiffies,
2494                        unsigned int interval_msecs)
2495{
2496        if (*caller_jiffies == 0
2497                        || !time_in_range(jiffies, *caller_jiffies,
2498                                        *caller_jiffies
2499                                        + msecs_to_jiffies(interval_msecs))) {
2500                *caller_jiffies = jiffies;
2501                return true;
2502        }
2503        return false;
2504}
2505EXPORT_SYMBOL(printk_timed_ratelimit);
2506
2507static DEFINE_SPINLOCK(dump_list_lock);
2508static LIST_HEAD(dump_list);
2509
2510/**
2511 * kmsg_dump_register - register a kernel log dumper.
2512 * @dumper: pointer to the kmsg_dumper structure
2513 *
2514 * Adds a kernel log dumper to the system. The dump callback in the
2515 * structure will be called when the kernel oopses or panics and must be
2516 * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
2517 */
2518int kmsg_dump_register(struct kmsg_dumper *dumper)
2519{
2520        unsigned long flags;
2521        int err = -EBUSY;
2522
2523        /* The dump callback needs to be set */
2524        if (!dumper->dump)
2525                return -EINVAL;
2526
2527        spin_lock_irqsave(&dump_list_lock, flags);
2528        /* Don't allow registering multiple times */
2529        if (!dumper->registered) {
2530                dumper->registered = 1;
2531                list_add_tail_rcu(&dumper->list, &dump_list);
2532                err = 0;
2533        }
2534        spin_unlock_irqrestore(&dump_list_lock, flags);
2535
2536        return err;
2537}
2538EXPORT_SYMBOL_GPL(kmsg_dump_register);
2539
2540/**
2541 * kmsg_dump_unregister - unregister a kmsg dumper.
2542 * @dumper: pointer to the kmsg_dumper structure
2543 *
2544 * Removes a dump device from the system. Returns zero on success and
2545 * %-EINVAL otherwise.
2546 */
2547int kmsg_dump_unregister(struct kmsg_dumper *dumper)
2548{
2549        unsigned long flags;
2550        int err = -EINVAL;
2551
2552        spin_lock_irqsave(&dump_list_lock, flags);
2553        if (dumper->registered) {
2554                dumper->registered = 0;
2555                list_del_rcu(&dumper->list);
2556                err = 0;
2557        }
2558        spin_unlock_irqrestore(&dump_list_lock, flags);
2559        synchronize_rcu();
2560
2561        return err;
2562}
2563EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
2564
2565static bool always_kmsg_dump;
2566module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
2567
2568/**
2569 * kmsg_dump - dump kernel log to kernel message dumpers.
2570 * @reason: the reason (oops, panic etc) for dumping
2571 *
2572 * Call each of the registered dumper's dump() callback, which can
2573 * retrieve the kmsg records with kmsg_dump_get_line() or
2574 * kmsg_dump_get_buffer().
2575 */
2576void kmsg_dump(enum kmsg_dump_reason reason)
2577{
2578        struct kmsg_dumper *dumper;
2579        unsigned long flags;
2580
2581        if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
2582                return;
2583
2584        rcu_read_lock();
2585        list_for_each_entry_rcu(dumper, &dump_list, list) {
2586                if (dumper->max_reason && reason > dumper->max_reason)
2587                        continue;
2588
2589                /* initialize iterator with data about the stored records */
2590                dumper->active = true;
2591
2592                raw_spin_lock_irqsave(&logbuf_lock, flags);
2593                dumper->cur_seq = clear_seq;
2594                dumper->cur_idx = clear_idx;
2595                dumper->next_seq = log_next_seq;
2596                dumper->next_idx = log_next_idx;
2597                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2598
2599                /* invoke dumper which will iterate over records */
2600                dumper->dump(dumper, reason);
2601
2602                /* reset iterator */
2603                dumper->active = false;
2604        }
2605        rcu_read_unlock();
2606}
2607
2608/**
2609 * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
2610 * @dumper: registered kmsg dumper
2611 * @syslog: include the "<4>" prefixes
2612 * @line: buffer to copy the line to
2613 * @size: maximum size of the buffer
2614 * @len: length of line placed into buffer
2615 *
2616 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2617 * record, and copy one record into the provided buffer.
2618 *
2619 * Consecutive calls will return the next available record moving
2620 * towards the end of the buffer with the youngest messages.
2621 *
2622 * A return value of FALSE indicates that there are no more records to
2623 * read.
2624 *
2625 * The function is similar to kmsg_dump_get_line(), but grabs no locks.
2626 */
2627bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
2628                               char *line, size_t size, size_t *len)
2629{
2630        struct log *msg;
2631        size_t l = 0;
2632        bool ret = false;
2633
2634        if (!dumper->active)
2635                goto out;
2636
2637        if (dumper->cur_seq < log_first_seq) {
2638                /* messages are gone, move to first available one */
2639                dumper->cur_seq = log_first_seq;
2640                dumper->cur_idx = log_first_idx;
2641        }
2642
2643        /* last entry */
2644        if (dumper->cur_seq >= log_next_seq)
2645                goto out;
2646
2647        msg = log_from_idx(dumper->cur_idx);
2648        l = msg_print_text(msg, 0, syslog, line, size);
2649
2650        dumper->cur_idx = log_next(dumper->cur_idx);
2651        dumper->cur_seq++;
2652        ret = true;
2653out:
2654        if (len)
2655                *len = l;
2656        return ret;
2657}
2658
2659/**
2660 * kmsg_dump_get_line - retrieve one kmsg log line
2661 * @dumper: registered kmsg dumper
2662 * @syslog: include the "<4>" prefixes
2663 * @line: buffer to copy the line to
2664 * @size: maximum size of the buffer
2665 * @len: length of line placed into buffer
2666 *
2667 * Start at the beginning of the kmsg buffer, with the oldest kmsg
2668 * record, and copy one record into the provided buffer.
2669 *
2670 * Consecutive calls will return the next available record moving
2671 * towards the end of the buffer with the youngest messages.
2672 *
2673 * A return value of FALSE indicates that there are no more records to
2674 * read.
2675 */
2676bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
2677                        char *line, size_t size, size_t *len)
2678{
2679        unsigned long flags;
2680        bool ret;
2681
2682        raw_spin_lock_irqsave(&logbuf_lock, flags);
2683        ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
2684        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2685
2686        return ret;
2687}
2688EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
2689
2690/**
2691 * kmsg_dump_get_buffer - copy kmsg log lines
2692 * @dumper: registered kmsg dumper
2693 * @syslog: include the "<4>" prefixes
2694 * @buf: buffer to copy the line to
2695 * @size: maximum size of the buffer
2696 * @len: length of line placed into buffer
2697 *
2698 * Start at the end of the kmsg buffer and fill the provided buffer
2699 * with as many of the the *youngest* kmsg records that fit into it.
2700 * If the buffer is large enough, all available kmsg records will be
2701 * copied with a single call.
2702 *
2703 * Consecutive calls will fill the buffer with the next block of
2704 * available older records, not including the earlier retrieved ones.
2705 *
2706 * A return value of FALSE indicates that there are no more records to
2707 * read.
2708 */
2709bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
2710                          char *buf, size_t size, size_t *len)
2711{
2712        unsigned long flags;
2713        u64 seq;
2714        u32 idx;
2715        u64 next_seq;
2716        u32 next_idx;
2717        enum log_flags prev;
2718        size_t l = 0;
2719        bool ret = false;
2720
2721        if (!dumper->active)
2722                goto out;
2723
2724        raw_spin_lock_irqsave(&logbuf_lock, flags);
2725        if (dumper->cur_seq < log_first_seq) {
2726                /* messages are gone, move to first available one */
2727                dumper->cur_seq = log_first_seq;
2728                dumper->cur_idx = log_first_idx;
2729        }
2730
2731        /* last entry */
2732        if (dumper->cur_seq >= dumper->next_seq) {
2733                raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2734                goto out;
2735        }
2736
2737        /* calculate length of entire buffer */
2738        seq = dumper->cur_seq;
2739        idx = dumper->cur_idx;
2740        prev = 0;
2741        while (seq < dumper->next_seq) {
2742                struct log *msg = log_from_idx(idx);
2743
2744                l += msg_print_text(msg, prev, true, NULL, 0);
2745                idx = log_next(idx);
2746                seq++;
2747                prev = msg->flags;
2748        }
2749
2750        /* move first record forward until length fits into the buffer */
2751        seq = dumper->cur_seq;
2752        idx = dumper->cur_idx;
2753        prev = 0;
2754        while (l > size && seq < dumper->next_seq) {
2755                struct log *msg = log_from_idx(idx);
2756
2757                l -= msg_print_text(msg, prev, true, NULL, 0);
2758                idx = log_next(idx);
2759                seq++;
2760                prev = msg->flags;
2761        }
2762
2763        /* last message in next interation */
2764        next_seq = seq;
2765        next_idx = idx;
2766
2767        l = 0;
2768        prev = 0;
2769        while (seq < dumper->next_seq) {
2770                struct log *msg = log_from_idx(idx);
2771
2772                l += msg_print_text(msg, prev, syslog, buf + l, size - l);
2773                idx = log_next(idx);
2774                seq++;
2775                prev = msg->flags;
2776        }
2777
2778        dumper->next_seq = next_seq;
2779        dumper->next_idx = next_idx;
2780        ret = true;
2781        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2782out:
2783        if (len)
2784                *len = l;
2785        return ret;
2786}
2787EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
2788
2789/**
2790 * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
2791 * @dumper: registered kmsg dumper
2792 *
2793 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2794 * kmsg_dump_get_buffer() can be called again and used multiple
2795 * times within the same dumper.dump() callback.
2796 *
2797 * The function is similar to kmsg_dump_rewind(), but grabs no locks.
2798 */
2799void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
2800{
2801        dumper->cur_seq = clear_seq;
2802        dumper->cur_idx = clear_idx;
2803        dumper->next_seq = log_next_seq;
2804        dumper->next_idx = log_next_idx;
2805}
2806
2807/**
2808 * kmsg_dump_rewind - reset the interator
2809 * @dumper: registered kmsg dumper
2810 *
2811 * Reset the dumper's iterator so that kmsg_dump_get_line() and
2812 * kmsg_dump_get_buffer() can be called again and used multiple
2813 * times within the same dumper.dump() callback.
2814 */
2815void kmsg_dump_rewind(struct kmsg_dumper *dumper)
2816{
2817        unsigned long flags;
2818
2819        raw_spin_lock_irqsave(&logbuf_lock, flags);
2820        kmsg_dump_rewind_nolock(dumper);
2821        raw_spin_unlock_irqrestore(&logbuf_lock, flags);
2822}
2823EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
2824#endif
2825
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.