printk: move to separate directory for easier modification
authorJoe Perches <joe@perches.com>
Wed, 31 Jul 2013 20:53:42 +0000 (13:53 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 31 Jul 2013 21:41:03 +0000 (14:41 -0700)
Make it easier to break up printk into bite-sized chunks.

Remove printk path/filename from comment.

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Samuel Thibault <samuel.thibault@ens-lyon.org>
Cc: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Documentation/DocBook/device-drivers.tmpl
kernel/Makefile
kernel/printk.c [deleted file]
kernel/printk/Makefile [new file with mode: 0644]
kernel/printk/printk.c [new file with mode: 0644]

index cbfdf5486639a144c414cc46334144fccee8a1e5..fe397f90a34f50153f5936cfa13635ed7bc85b7b 100644 (file)
@@ -84,7 +84,7 @@ X!Iinclude/linux/kobject.h
 
      <sect1><title>Kernel utility functions</title>
 !Iinclude/linux/kernel.h
-!Ekernel/printk.c
+!Ekernel/printk/printk.c
 !Ekernel/panic.c
 !Ekernel/sys.c
 !Ekernel/rcupdate.c
index 470839d1a30ed72f68789df711bfa78bdc60e5f4..35ef1185e359457323d5513b578185ad16435bc2 100644 (file)
@@ -2,7 +2,7 @@
 # Makefile for the linux kernel.
 #
 
-obj-y     = fork.o exec_domain.o panic.o printk.o \
+obj-y     = fork.o exec_domain.o panic.o \
            cpu.o exit.o itimer.o time.o softirq.o resource.o \
            sysctl.o sysctl_binary.o capability.o ptrace.o timer.o user.o \
            signal.o sys.o kmod.o workqueue.o pid.o task_work.o \
@@ -24,6 +24,7 @@ endif
 
 obj-y += sched/
 obj-y += power/
+obj-y += printk/
 obj-y += cpu/
 
 obj-$(CONFIG_CHECKPOINT_RESTORE) += kcmp.o
diff --git a/kernel/printk.c b/kernel/printk.c
deleted file mode 100644 (file)
index 69b0890..0000000
+++ /dev/null
@@ -1,2924 +0,0 @@
-/*
- *  linux/kernel/printk.c
- *
- *  Copyright (C) 1991, 1992  Linus Torvalds
- *
- * Modified to make sys_syslog() more flexible: added commands to
- * return the last 4k of kernel messages, regardless of whether
- * they've been read or not.  Added option to suppress kernel printk's
- * to the console.  Added hook for sending the console messages
- * elsewhere, in preparation for a serial line console (someday).
- * Ted Ts'o, 2/11/93.
- * Modified for sysctl support, 1/8/97, Chris Horn.
- * Fixed SMP synchronization, 08/08/99, Manfred Spraul
- *     manfred@colorfullife.com
- * Rewrote bits to get rid of console_lock
- *     01Mar01 Andrew Morton
- */
-
-#include <linux/kernel.h>
-#include <linux/mm.h>
-#include <linux/tty.h>
-#include <linux/tty_driver.h>
-#include <linux/console.h>
-#include <linux/init.h>
-#include <linux/jiffies.h>
-#include <linux/nmi.h>
-#include <linux/module.h>
-#include <linux/moduleparam.h>
-#include <linux/interrupt.h>                   /* For in_interrupt() */
-#include <linux/delay.h>
-#include <linux/smp.h>
-#include <linux/security.h>
-#include <linux/bootmem.h>
-#include <linux/memblock.h>
-#include <linux/aio.h>
-#include <linux/syscalls.h>
-#include <linux/kexec.h>
-#include <linux/kdb.h>
-#include <linux/ratelimit.h>
-#include <linux/kmsg_dump.h>
-#include <linux/syslog.h>
-#include <linux/cpu.h>
-#include <linux/notifier.h>
-#include <linux/rculist.h>
-#include <linux/poll.h>
-#include <linux/irq_work.h>
-#include <linux/utsname.h>
-
-#include <asm/uaccess.h>
-
-#define CREATE_TRACE_POINTS
-#include <trace/events/printk.h>
-
-/* printk's without a loglevel use this.. */
-#define DEFAULT_MESSAGE_LOGLEVEL CONFIG_DEFAULT_MESSAGE_LOGLEVEL
-
-/* We show everything that is MORE important than this.. */
-#define MINIMUM_CONSOLE_LOGLEVEL 1 /* Minimum loglevel we let people use */
-#define DEFAULT_CONSOLE_LOGLEVEL 7 /* anything MORE serious than KERN_DEBUG */
-
-int console_printk[4] = {
-       DEFAULT_CONSOLE_LOGLEVEL,       /* console_loglevel */
-       DEFAULT_MESSAGE_LOGLEVEL,       /* default_message_loglevel */
-       MINIMUM_CONSOLE_LOGLEVEL,       /* minimum_console_loglevel */
-       DEFAULT_CONSOLE_LOGLEVEL,       /* default_console_loglevel */
-};
-
-/*
- * Low level drivers may need that to know if they can schedule in
- * their unblank() callback or not. So let's export it.
- */
-int oops_in_progress;
-EXPORT_SYMBOL(oops_in_progress);
-
-/*
- * console_sem protects the console_drivers list, and also
- * provides serialisation for access to the entire console
- * driver system.
- */
-static DEFINE_SEMAPHORE(console_sem);
-struct console *console_drivers;
-EXPORT_SYMBOL_GPL(console_drivers);
-
-#ifdef CONFIG_LOCKDEP
-static struct lockdep_map console_lock_dep_map = {
-       .name = "console_lock"
-};
-#endif
-
-/*
- * This is used for debugging the mess that is the VT code by
- * keeping track if we have the console semaphore held. It's
- * definitely not the perfect debug tool (we don't know if _WE_
- * hold it are racing, but it helps tracking those weird code
- * path in the console code where we end up in places I want
- * locked without the console sempahore held
- */
-static int console_locked, console_suspended;
-
-/*
- * If exclusive_console is non-NULL then only this console is to be printed to.
- */
-static struct console *exclusive_console;
-
-/*
- *     Array of consoles built from command line options (console=)
- */
-struct console_cmdline
-{
-       char    name[8];                        /* Name of the driver       */
-       int     index;                          /* Minor dev. to use        */
-       char    *options;                       /* Options for the driver   */
-#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
-       char    *brl_options;                   /* Options for braille driver */
-#endif
-};
-
-#define MAX_CMDLINECONSOLES 8
-
-static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
-static int selected_console = -1;
-static int preferred_console = -1;
-int console_set_on_cmdline;
-EXPORT_SYMBOL(console_set_on_cmdline);
-
-/* Flag: console code may call schedule() */
-static int console_may_schedule;
-
-/*
- * The printk log buffer consists of a chain of concatenated variable
- * length records. Every record starts with a record header, containing
- * the overall length of the record.
- *
- * The heads to the first and last entry in the buffer, as well as the
- * sequence numbers of these both entries are maintained when messages
- * are stored..
- *
- * If the heads indicate available messages, the length in the header
- * tells the start next message. A length == 0 for the next message
- * indicates a wrap-around to the beginning of the buffer.
- *
- * Every record carries the monotonic timestamp in microseconds, as well as
- * the standard userspace syslog level and syslog facility. The usual
- * kernel messages use LOG_KERN; userspace-injected messages always carry
- * a matching syslog facility, by default LOG_USER. The origin of every
- * message can be reliably determined that way.
- *
- * The human readable log message directly follows the message header. The
- * length of the message text is stored in the header, the stored message
- * is not terminated.
- *
- * Optionally, a message can carry a dictionary of properties (key/value pairs),
- * to provide userspace with a machine-readable message context.
- *
- * Examples for well-defined, commonly used property names are:
- *   DEVICE=b12:8               device identifier
- *                                b12:8         block dev_t
- *                                c127:3        char dev_t
- *                                n8            netdev ifindex
- *                                +sound:card0  subsystem:devname
- *   SUBSYSTEM=pci              driver-core subsystem name
- *
- * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
- * follows directly after a '=' character. Every property is terminated by
- * a '\0' character. The last property is not terminated.
- *
- * Example of a message structure:
- *   0000  ff 8f 00 00 00 00 00 00      monotonic time in nsec
- *   0008  34 00                        record is 52 bytes long
- *   000a        0b 00                  text is 11 bytes long
- *   000c              1f 00            dictionary is 23 bytes long
- *   000e                    03 00      LOG_KERN (facility) LOG_ERR (level)
- *   0010  69 74 27 73 20 61 20 6c      "it's a l"
- *         69 6e 65                     "ine"
- *   001b           44 45 56 49 43      "DEVIC"
- *         45 3d 62 38 3a 32 00 44      "E=b8:2\0D"
- *         52 49 56 45 52 3d 62 75      "RIVER=bu"
- *         67                           "g"
- *   0032     00 00 00                  padding to next message header
- *
- * The 'struct log' buffer header must never be directly exported to
- * userspace, it is a kernel-private implementation detail that might
- * need to be changed in the future, when the requirements change.
- *
- * /dev/kmsg exports the structured data in the following line format:
- *   "level,sequnum,timestamp;<message text>\n"
- *
- * The optional key/value pairs are attached as continuation lines starting
- * with a space character and terminated by a newline. All possible
- * non-prinatable characters are escaped in the "\xff" notation.
- *
- * Users of the export format should ignore possible additional values
- * separated by ',', and find the message after the ';' character.
- */
-
-enum log_flags {
-       LOG_NOCONS      = 1,    /* already flushed, do not print to console */
-       LOG_NEWLINE     = 2,    /* text ended with a newline */
-       LOG_PREFIX      = 4,    /* text started with a prefix */
-       LOG_CONT        = 8,    /* text is a fragment of a continuation line */
-};
-
-struct log {
-       u64 ts_nsec;            /* timestamp in nanoseconds */
-       u16 len;                /* length of entire record */
-       u16 text_len;           /* length of text buffer */
-       u16 dict_len;           /* length of dictionary buffer */
-       u8 facility;            /* syslog facility */
-       u8 flags:5;             /* internal record flags */
-       u8 level:3;             /* syslog level */
-};
-
-/*
- * The logbuf_lock protects kmsg buffer, indices, counters. It is also
- * used in interesting ways to provide interlocking in console_unlock();
- */
-static DEFINE_RAW_SPINLOCK(logbuf_lock);
-
-#ifdef CONFIG_PRINTK
-DECLARE_WAIT_QUEUE_HEAD(log_wait);
-/* the next printk record to read by syslog(READ) or /proc/kmsg */
-static u64 syslog_seq;
-static u32 syslog_idx;
-static enum log_flags syslog_prev;
-static size_t syslog_partial;
-
-/* index and sequence number of the first record stored in the buffer */
-static u64 log_first_seq;
-static u32 log_first_idx;
-
-/* index and sequence number of the next record to store in the buffer */
-static u64 log_next_seq;
-static u32 log_next_idx;
-
-/* the next printk record to write to the console */
-static u64 console_seq;
-static u32 console_idx;
-static enum log_flags console_prev;
-
-/* the next printk record to read after the last 'clear' command */
-static u64 clear_seq;
-static u32 clear_idx;
-
-#define PREFIX_MAX             32
-#define LOG_LINE_MAX           1024 - PREFIX_MAX
-
-/* record buffer */
-#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
-#define LOG_ALIGN 4
-#else
-#define LOG_ALIGN __alignof__(struct log)
-#endif
-#define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
-static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
-static char *log_buf = __log_buf;
-static u32 log_buf_len = __LOG_BUF_LEN;
-
-/* cpu currently holding logbuf_lock */
-static volatile unsigned int logbuf_cpu = UINT_MAX;
-
-/* human readable text of the record */
-static char *log_text(const struct log *msg)
-{
-       return (char *)msg + sizeof(struct log);
-}
-
-/* optional key/value pair dictionary attached to the record */
-static char *log_dict(const struct log *msg)
-{
-       return (char *)msg + sizeof(struct log) + msg->text_len;
-}
-
-/* get record by index; idx must point to valid msg */
-static struct log *log_from_idx(u32 idx)
-{
-       struct log *msg = (struct log *)(log_buf + idx);
-
-       /*
-        * A length == 0 record is the end of buffer marker. Wrap around and
-        * read the message at the start of the buffer.
-        */
-       if (!msg->len)
-               return (struct log *)log_buf;
-       return msg;
-}
-
-/* get next record; idx must point to valid msg */
-static u32 log_next(u32 idx)
-{
-       struct log *msg = (struct log *)(log_buf + idx);
-
-       /* length == 0 indicates the end of the buffer; wrap */
-       /*
-        * A length == 0 record is the end of buffer marker. Wrap around and
-        * read the message at the start of the buffer as *this* one, and
-        * return the one after that.
-        */
-       if (!msg->len) {
-               msg = (struct log *)log_buf;
-               return msg->len;
-       }
-       return idx + msg->len;
-}
-
-/* insert record into the buffer, discard old ones, update heads */
-static void log_store(int facility, int level,
-                     enum log_flags flags, u64 ts_nsec,
-                     const char *dict, u16 dict_len,
-                     const char *text, u16 text_len)
-{
-       struct log *msg;
-       u32 size, pad_len;
-
-       /* number of '\0' padding bytes to next message */
-       size = sizeof(struct log) + text_len + dict_len;
-       pad_len = (-size) & (LOG_ALIGN - 1);
-       size += pad_len;
-
-       while (log_first_seq < log_next_seq) {
-               u32 free;
-
-               if (log_next_idx > log_first_idx)
-                       free = max(log_buf_len - log_next_idx, log_first_idx);
-               else
-                       free = log_first_idx - log_next_idx;
-
-               if (free > size + sizeof(struct log))
-                       break;
-
-               /* drop old messages until we have enough contiuous space */
-               log_first_idx = log_next(log_first_idx);
-               log_first_seq++;
-       }
-
-       if (log_next_idx + size + sizeof(struct log) >= log_buf_len) {
-               /*
-                * This message + an additional empty header does not fit
-                * at the end of the buffer. Add an empty header with len == 0
-                * to signify a wrap around.
-                */
-               memset(log_buf + log_next_idx, 0, sizeof(struct log));
-               log_next_idx = 0;
-       }
-
-       /* fill message */
-       msg = (struct log *)(log_buf + log_next_idx);
-       memcpy(log_text(msg), text, text_len);
-       msg->text_len = text_len;
-       memcpy(log_dict(msg), dict, dict_len);
-       msg->dict_len = dict_len;
-       msg->facility = facility;
-       msg->level = level & 7;
-       msg->flags = flags & 0x1f;
-       if (ts_nsec > 0)
-               msg->ts_nsec = ts_nsec;
-       else
-               msg->ts_nsec = local_clock();
-       memset(log_dict(msg) + dict_len, 0, pad_len);
-       msg->len = sizeof(struct log) + text_len + dict_len + pad_len;
-
-       /* insert message */
-       log_next_idx += msg->len;
-       log_next_seq++;
-}
-
-#ifdef CONFIG_SECURITY_DMESG_RESTRICT
-int dmesg_restrict = 1;
-#else
-int dmesg_restrict;
-#endif
-
-static int syslog_action_restricted(int type)
-{
-       if (dmesg_restrict)
-               return 1;
-       /*
-        * Unless restricted, we allow "read all" and "get buffer size"
-        * for everybody.
-        */
-       return type != SYSLOG_ACTION_READ_ALL &&
-              type != SYSLOG_ACTION_SIZE_BUFFER;
-}
-
-static int check_syslog_permissions(int type, bool from_file)
-{
-       /*
-        * If this is from /proc/kmsg and we've already opened it, then we've
-        * already done the capabilities checks at open time.
-        */
-       if (from_file && type != SYSLOG_ACTION_OPEN)
-               return 0;
-
-       if (syslog_action_restricted(type)) {
-               if (capable(CAP_SYSLOG))
-                       return 0;
-               /*
-                * For historical reasons, accept CAP_SYS_ADMIN too, with
-                * a warning.
-                */
-               if (capable(CAP_SYS_ADMIN)) {
-                       pr_warn_once("%s (%d): Attempt to access syslog with "
-                                    "CAP_SYS_ADMIN but no CAP_SYSLOG "
-                                    "(deprecated).\n",
-                                current->comm, task_pid_nr(current));
-                       return 0;
-               }
-               return -EPERM;
-       }
-       return security_syslog(type);
-}
-
-
-/* /dev/kmsg - userspace message inject/listen interface */
-struct devkmsg_user {
-       u64 seq;
-       u32 idx;
-       enum log_flags prev;
-       struct mutex lock;
-       char buf[8192];
-};
-
-static ssize_t devkmsg_writev(struct kiocb *iocb, const struct iovec *iv,
-                             unsigned long count, loff_t pos)
-{
-       char *buf, *line;
-       int i;
-       int level = default_message_loglevel;
-       int facility = 1;       /* LOG_USER */
-       size_t len = iov_length(iv, count);
-       ssize_t ret = len;
-
-       if (len > LOG_LINE_MAX)
-               return -EINVAL;
-       buf = kmalloc(len+1, GFP_KERNEL);
-       if (buf == NULL)
-               return -ENOMEM;
-
-       line = buf;
-       for (i = 0; i < count; i++) {
-               if (copy_from_user(line, iv[i].iov_base, iv[i].iov_len)) {
-                       ret = -EFAULT;
-                       goto out;
-               }
-               line += iv[i].iov_len;
-       }
-
-       /*
-        * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
-        * the decimal value represents 32bit, the lower 3 bit are the log
-        * level, the rest are the log facility.
-        *
-        * If no prefix or no userspace facility is specified, we
-        * enforce LOG_USER, to be able to reliably distinguish
-        * kernel-generated messages from userspace-injected ones.
-        */
-       line = buf;
-       if (line[0] == '<') {
-               char *endp = NULL;
-
-               i = simple_strtoul(line+1, &endp, 10);
-               if (endp && endp[0] == '>') {
-                       level = i & 7;
-                       if (i >> 3)
-                               facility = i >> 3;
-                       endp++;
-                       len -= endp - line;
-                       line = endp;
-               }
-       }
-       line[len] = '\0';
-
-       printk_emit(facility, level, NULL, 0, "%s", line);
-out:
-       kfree(buf);
-       return ret;
-}
-
-static ssize_t devkmsg_read(struct file *file, char __user *buf,
-                           size_t count, loff_t *ppos)
-{
-       struct devkmsg_user *user = file->private_data;
-       struct log *msg;
-       u64 ts_usec;
-       size_t i;
-       char cont = '-';
-       size_t len;
-       ssize_t ret;
-
-       if (!user)
-               return -EBADF;
-
-       ret = mutex_lock_interruptible(&user->lock);
-       if (ret)
-               return ret;
-       raw_spin_lock_irq(&logbuf_lock);
-       while (user->seq == log_next_seq) {
-               if (file->f_flags & O_NONBLOCK) {
-                       ret = -EAGAIN;
-                       raw_spin_unlock_irq(&logbuf_lock);
-                       goto out;
-               }
-
-               raw_spin_unlock_irq(&logbuf_lock);
-               ret = wait_event_interruptible(log_wait,
-                                              user->seq != log_next_seq);
-               if (ret)
-                       goto out;
-               raw_spin_lock_irq(&logbuf_lock);
-       }
-
-       if (user->seq < log_first_seq) {
-               /* our last seen message is gone, return error and reset */
-               user->idx = log_first_idx;
-               user->seq = log_first_seq;
-               ret = -EPIPE;
-               raw_spin_unlock_irq(&logbuf_lock);
-               goto out;
-       }
-
-       msg = log_from_idx(user->idx);
-       ts_usec = msg->ts_nsec;
-       do_div(ts_usec, 1000);
-
-       /*
-        * If we couldn't merge continuation line fragments during the print,
-        * export the stored flags to allow an optional external merge of the
-        * records. Merging the records isn't always neccessarily correct, like
-        * when we hit a race during printing. In most cases though, it produces
-        * better readable output. 'c' in the record flags mark the first
-        * fragment of a line, '+' the following.
-        */
-       if (msg->flags & LOG_CONT && !(user->prev & LOG_CONT))
-               cont = 'c';
-       else if ((msg->flags & LOG_CONT) ||
-                ((user->prev & LOG_CONT) && !(msg->flags & LOG_PREFIX)))
-               cont = '+';
-
-       len = sprintf(user->buf, "%u,%llu,%llu,%c;",
-                     (msg->facility << 3) | msg->level,
-                     user->seq, ts_usec, cont);
-       user->prev = msg->flags;
-
-       /* escape non-printable characters */
-       for (i = 0; i < msg->text_len; i++) {
-               unsigned char c = log_text(msg)[i];
-
-               if (c < ' ' || c >= 127 || c == '\\')
-                       len += sprintf(user->buf + len, "\\x%02x", c);
-               else
-                       user->buf[len++] = c;
-       }
-       user->buf[len++] = '\n';
-
-       if (msg->dict_len) {
-               bool line = true;
-
-               for (i = 0; i < msg->dict_len; i++) {
-                       unsigned char c = log_dict(msg)[i];
-
-                       if (line) {
-                               user->buf[len++] = ' ';
-                               line = false;
-                       }
-
-                       if (c == '\0') {
-                               user->buf[len++] = '\n';
-                               line = true;
-                               continue;
-                       }
-
-                       if (c < ' ' || c >= 127 || c == '\\') {
-                               len += sprintf(user->buf + len, "\\x%02x", c);
-                               continue;
-                       }
-
-                       user->buf[len++] = c;
-               }
-               user->buf[len++] = '\n';
-       }
-
-       user->idx = log_next(user->idx);
-       user->seq++;
-       raw_spin_unlock_irq(&logbuf_lock);
-
-       if (len > count) {
-               ret = -EINVAL;
-               goto out;
-       }
-
-       if (copy_to_user(buf, user->buf, len)) {
-               ret = -EFAULT;
-               goto out;
-       }
-       ret = len;
-out:
-       mutex_unlock(&user->lock);
-       return ret;
-}
-
-static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
-{
-       struct devkmsg_user *user = file->private_data;
-       loff_t ret = 0;
-
-       if (!user)
-               return -EBADF;
-       if (offset)
-               return -ESPIPE;
-
-       raw_spin_lock_irq(&logbuf_lock);
-       switch (whence) {
-       case SEEK_SET:
-               /* the first record */
-               user->idx = log_first_idx;
-               user->seq = log_first_seq;
-               break;
-       case SEEK_DATA:
-               /*
-                * The first record after the last SYSLOG_ACTION_CLEAR,
-                * like issued by 'dmesg -c'. Reading /dev/kmsg itself
-                * changes no global state, and does not clear anything.
-                */
-               user->idx = clear_idx;
-               user->seq = clear_seq;
-               break;
-       case SEEK_END:
-               /* after the last record */
-               user->idx = log_next_idx;
-               user->seq = log_next_seq;
-               break;
-       default:
-               ret = -EINVAL;
-       }
-       raw_spin_unlock_irq(&logbuf_lock);
-       return ret;
-}
-
-static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
-{
-       struct devkmsg_user *user = file->private_data;
-       int ret = 0;
-
-       if (!user)
-               return POLLERR|POLLNVAL;
-
-       poll_wait(file, &log_wait, wait);
-
-       raw_spin_lock_irq(&logbuf_lock);
-       if (user->seq < log_next_seq) {
-               /* return error when data has vanished underneath us */
-               if (user->seq < log_first_seq)
-                       ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
-               else
-                       ret = POLLIN|POLLRDNORM;
-       }
-       raw_spin_unlock_irq(&logbuf_lock);
-
-       return ret;
-}
-
-static int devkmsg_open(struct inode *inode, struct file *file)
-{
-       struct devkmsg_user *user;
-       int err;
-
-       /* write-only does not need any file context */
-       if ((file->f_flags & O_ACCMODE) == O_WRONLY)
-               return 0;
-
-       err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
-                                      SYSLOG_FROM_READER);
-       if (err)
-               return err;
-
-       user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
-       if (!user)
-               return -ENOMEM;
-
-       mutex_init(&user->lock);
-
-       raw_spin_lock_irq(&logbuf_lock);
-       user->idx = log_first_idx;
-       user->seq = log_first_seq;
-       raw_spin_unlock_irq(&logbuf_lock);
-
-       file->private_data = user;
-       return 0;
-}
-
-static int devkmsg_release(struct inode *inode, struct file *file)
-{
-       struct devkmsg_user *user = file->private_data;
-
-       if (!user)
-               return 0;
-
-       mutex_destroy(&user->lock);
-       kfree(user);
-       return 0;
-}
-
-const struct file_operations kmsg_fops = {
-       .open = devkmsg_open,
-       .read = devkmsg_read,
-       .aio_write = devkmsg_writev,
-       .llseek = devkmsg_llseek,
-       .poll = devkmsg_poll,
-       .release = devkmsg_release,
-};
-
-#ifdef CONFIG_KEXEC
-/*
- * This appends the listed symbols to /proc/vmcoreinfo
- *
- * /proc/vmcoreinfo is used by various utiilties, like crash and makedumpfile to
- * obtain access to symbols that are otherwise very difficult to locate.  These
- * symbols are specifically used so that utilities can access and extract the
- * dmesg log from a vmcore file after a crash.
- */
-void log_buf_kexec_setup(void)
-{
-       VMCOREINFO_SYMBOL(log_buf);
-       VMCOREINFO_SYMBOL(log_buf_len);
-       VMCOREINFO_SYMBOL(log_first_idx);
-       VMCOREINFO_SYMBOL(log_next_idx);
-       /*
-        * Export struct log size and field offsets. User space tools can
-        * parse it and detect any changes to structure down the line.
-        */
-       VMCOREINFO_STRUCT_SIZE(log);
-       VMCOREINFO_OFFSET(log, ts_nsec);
-       VMCOREINFO_OFFSET(log, len);
-       VMCOREINFO_OFFSET(log, text_len);
-       VMCOREINFO_OFFSET(log, dict_len);
-}
-#endif
-
-/* requested log_buf_len from kernel cmdline */
-static unsigned long __initdata new_log_buf_len;
-
-/* save requested log_buf_len since it's too early to process it */
-static int __init log_buf_len_setup(char *str)
-{
-       unsigned size = memparse(str, &str);
-
-       if (size)
-               size = roundup_pow_of_two(size);
-       if (size > log_buf_len)
-               new_log_buf_len = size;
-
-       return 0;
-}
-early_param("log_buf_len", log_buf_len_setup);
-
-void __init setup_log_buf(int early)
-{
-       unsigned long flags;
-       char *new_log_buf;
-       int free;
-
-       if (!new_log_buf_len)
-               return;
-
-       if (early) {
-               unsigned long mem;
-
-               mem = memblock_alloc(new_log_buf_len, PAGE_SIZE);
-               if (!mem)
-                       return;
-               new_log_buf = __va(mem);
-       } else {
-               new_log_buf = alloc_bootmem_nopanic(new_log_buf_len);
-       }
-
-       if (unlikely(!new_log_buf)) {
-               pr_err("log_buf_len: %ld bytes not available\n",
-                       new_log_buf_len);
-               return;
-       }
-
-       raw_spin_lock_irqsave(&logbuf_lock, flags);
-       log_buf_len = new_log_buf_len;
-       log_buf = new_log_buf;
-       new_log_buf_len = 0;
-       free = __LOG_BUF_LEN - log_next_idx;
-       memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-
-       pr_info("log_buf_len: %d\n", log_buf_len);
-       pr_info("early log buf free: %d(%d%%)\n",
-               free, (free * 100) / __LOG_BUF_LEN);
-}
-
-static bool __read_mostly ignore_loglevel;
-
-static int __init ignore_loglevel_setup(char *str)
-{
-       ignore_loglevel = 1;
-       printk(KERN_INFO "debug: ignoring loglevel setting.\n");
-
-       return 0;
-}
-
-early_param("ignore_loglevel", ignore_loglevel_setup);
-module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
-MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
-       "print all kernel messages to the console.");
-
-#ifdef CONFIG_BOOT_PRINTK_DELAY
-
-static int boot_delay; /* msecs delay after each printk during bootup */
-static unsigned long long loops_per_msec;      /* based on boot_delay */
-
-static int __init boot_delay_setup(char *str)
-{
-       unsigned long lpj;
-
-       lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
-       loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
-
-       get_option(&str, &boot_delay);
-       if (boot_delay > 10 * 1000)
-               boot_delay = 0;
-
-       pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
-               "HZ: %d, loops_per_msec: %llu\n",
-               boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
-       return 1;
-}
-__setup("boot_delay=", boot_delay_setup);
-
-static void boot_delay_msec(int level)
-{
-       unsigned long long k;
-       unsigned long timeout;
-
-       if ((boot_delay == 0 || system_state != SYSTEM_BOOTING)
-               || (level >= console_loglevel && !ignore_loglevel)) {
-               return;
-       }
-
-       k = (unsigned long long)loops_per_msec * boot_delay;
-
-       timeout = jiffies + msecs_to_jiffies(boot_delay);
-       while (k) {
-               k--;
-               cpu_relax();
-               /*
-                * use (volatile) jiffies to prevent
-                * compiler reduction; loop termination via jiffies
-                * is secondary and may or may not happen.
-                */
-               if (time_after(jiffies, timeout))
-                       break;
-               touch_nmi_watchdog();
-       }
-}
-#else
-static inline void boot_delay_msec(int level)
-{
-}
-#endif
-
-#if defined(CONFIG_PRINTK_TIME)
-static bool printk_time = 1;
-#else
-static bool printk_time;
-#endif
-module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
-
-static size_t print_time(u64 ts, char *buf)
-{
-       unsigned long rem_nsec;
-
-       if (!printk_time)
-               return 0;
-
-       rem_nsec = do_div(ts, 1000000000);
-
-       if (!buf)
-               return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
-
-       return sprintf(buf, "[%5lu.%06lu] ",
-                      (unsigned long)ts, rem_nsec / 1000);
-}
-
-static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
-{
-       size_t len = 0;
-       unsigned int prefix = (msg->facility << 3) | msg->level;
-
-       if (syslog) {
-               if (buf) {
-                       len += sprintf(buf, "<%u>", prefix);
-               } else {
-                       len += 3;
-                       if (prefix > 999)
-                               len += 3;
-                       else if (prefix > 99)
-                               len += 2;
-                       else if (prefix > 9)
-                               len++;
-               }
-       }
-
-       len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
-       return len;
-}
-
-static size_t msg_print_text(const struct log *msg, enum log_flags prev,
-                            bool syslog, char *buf, size_t size)
-{
-       const char *text = log_text(msg);
-       size_t text_size = msg->text_len;
-       bool prefix = true;
-       bool newline = true;
-       size_t len = 0;
-
-       if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
-               prefix = false;
-
-       if (msg->flags & LOG_CONT) {
-               if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
-                       prefix = false;
-
-               if (!(msg->flags & LOG_NEWLINE))
-                       newline = false;
-       }
-
-       do {
-               const char *next = memchr(text, '\n', text_size);
-               size_t text_len;
-
-               if (next) {
-                       text_len = next - text;
-                       next++;
-                       text_size -= next - text;
-               } else {
-                       text_len = text_size;
-               }
-
-               if (buf) {
-                       if (print_prefix(msg, syslog, NULL) +
-                           text_len + 1 >= size - len)
-                               break;
-
-                       if (prefix)
-                               len += print_prefix(msg, syslog, buf + len);
-                       memcpy(buf + len, text, text_len);
-                       len += text_len;
-                       if (next || newline)
-                               buf[len++] = '\n';
-               } else {
-                       /* SYSLOG_ACTION_* buffer size only calculation */
-                       if (prefix)
-                               len += print_prefix(msg, syslog, NULL);
-                       len += text_len;
-                       if (next || newline)
-                               len++;
-               }
-
-               prefix = true;
-               text = next;
-       } while (text);
-
-       return len;
-}
-
-static int syslog_print(char __user *buf, int size)
-{
-       char *text;
-       struct log *msg;
-       int len = 0;
-
-       text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
-       if (!text)
-               return -ENOMEM;
-
-       while (size > 0) {
-               size_t n;
-               size_t skip;
-
-               raw_spin_lock_irq(&logbuf_lock);
-               if (syslog_seq < log_first_seq) {
-                       /* messages are gone, move to first one */
-                       syslog_seq = log_first_seq;
-                       syslog_idx = log_first_idx;
-                       syslog_prev = 0;
-                       syslog_partial = 0;
-               }
-               if (syslog_seq == log_next_seq) {
-                       raw_spin_unlock_irq(&logbuf_lock);
-                       break;
-               }
-
-               skip = syslog_partial;
-               msg = log_from_idx(syslog_idx);
-               n = msg_print_text(msg, syslog_prev, true, text,
-                                  LOG_LINE_MAX + PREFIX_MAX);
-               if (n - syslog_partial <= size) {
-                       /* message fits into buffer, move forward */
-                       syslog_idx = log_next(syslog_idx);
-                       syslog_seq++;
-                       syslog_prev = msg->flags;
-                       n -= syslog_partial;
-                       syslog_partial = 0;
-               } else if (!len){
-                       /* partial read(), remember position */
-                       n = size;
-                       syslog_partial += n;
-               } else
-                       n = 0;
-               raw_spin_unlock_irq(&logbuf_lock);
-
-               if (!n)
-                       break;
-
-               if (copy_to_user(buf, text + skip, n)) {
-                       if (!len)
-                               len = -EFAULT;
-                       break;
-               }
-
-               len += n;
-               size -= n;
-               buf += n;
-       }
-
-       kfree(text);
-       return len;
-}
-
-static int syslog_print_all(char __user *buf, int size, bool clear)
-{
-       char *text;
-       int len = 0;
-
-       text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
-       if (!text)
-               return -ENOMEM;
-
-       raw_spin_lock_irq(&logbuf_lock);
-       if (buf) {
-               u64 next_seq;
-               u64 seq;
-               u32 idx;
-               enum log_flags prev;
-
-               if (clear_seq < log_first_seq) {
-                       /* messages are gone, move to first available one */
-                       clear_seq = log_first_seq;
-                       clear_idx = log_first_idx;
-               }
-
-               /*
-                * Find first record that fits, including all following records,
-                * into the user-provided buffer for this dump.
-                */
-               seq = clear_seq;
-               idx = clear_idx;
-               prev = 0;
-               while (seq < log_next_seq) {
-                       struct log *msg = log_from_idx(idx);
-
-                       len += msg_print_text(msg, prev, true, NULL, 0);
-                       prev = msg->flags;
-                       idx = log_next(idx);
-                       seq++;
-               }
-
-               /* move first record forward until length fits into the buffer */
-               seq = clear_seq;
-               idx = clear_idx;
-               prev = 0;
-               while (len > size && seq < log_next_seq) {
-                       struct log *msg = log_from_idx(idx);
-
-                       len -= msg_print_text(msg, prev, true, NULL, 0);
-                       prev = msg->flags;
-                       idx = log_next(idx);
-                       seq++;
-               }
-
-               /* last message fitting into this dump */
-               next_seq = log_next_seq;
-
-               len = 0;
-               prev = 0;
-               while (len >= 0 && seq < next_seq) {
-                       struct log *msg = log_from_idx(idx);
-                       int textlen;
-
-                       textlen = msg_print_text(msg, prev, true, text,
-                                                LOG_LINE_MAX + PREFIX_MAX);
-                       if (textlen < 0) {
-                               len = textlen;
-                               break;
-                       }
-                       idx = log_next(idx);
-                       seq++;
-                       prev = msg->flags;
-
-                       raw_spin_unlock_irq(&logbuf_lock);
-                       if (copy_to_user(buf + len, text, textlen))
-                               len = -EFAULT;
-                       else
-                               len += textlen;
-                       raw_spin_lock_irq(&logbuf_lock);
-
-                       if (seq < log_first_seq) {
-                               /* messages are gone, move to next one */
-                               seq = log_first_seq;
-                               idx = log_first_idx;
-                               prev = 0;
-                       }
-               }
-       }
-
-       if (clear) {
-               clear_seq = log_next_seq;
-               clear_idx = log_next_idx;
-       }
-       raw_spin_unlock_irq(&logbuf_lock);
-
-       kfree(text);
-       return len;
-}
-
-int do_syslog(int type, char __user *buf, int len, bool from_file)
-{
-       bool clear = false;
-       static int saved_console_loglevel = -1;
-       int error;
-
-       error = check_syslog_permissions(type, from_file);
-       if (error)
-               goto out;
-
-       error = security_syslog(type);
-       if (error)
-               return error;
-
-       switch (type) {
-       case SYSLOG_ACTION_CLOSE:       /* Close log */
-               break;
-       case SYSLOG_ACTION_OPEN:        /* Open log */
-               break;
-       case SYSLOG_ACTION_READ:        /* Read from log */
-               error = -EINVAL;
-               if (!buf || len < 0)
-                       goto out;
-               error = 0;
-               if (!len)
-                       goto out;
-               if (!access_ok(VERIFY_WRITE, buf, len)) {
-                       error = -EFAULT;
-                       goto out;
-               }
-               error = wait_event_interruptible(log_wait,
-                                                syslog_seq != log_next_seq);
-               if (error)
-                       goto out;
-               error = syslog_print(buf, len);
-               break;
-       /* Read/clear last kernel messages */
-       case SYSLOG_ACTION_READ_CLEAR:
-               clear = true;
-               /* FALL THRU */
-       /* Read last kernel messages */
-       case SYSLOG_ACTION_READ_ALL:
-               error = -EINVAL;
-               if (!buf || len < 0)
-                       goto out;
-               error = 0;
-               if (!len)
-                       goto out;
-               if (!access_ok(VERIFY_WRITE, buf, len)) {
-                       error = -EFAULT;
-                       goto out;
-               }
-               error = syslog_print_all(buf, len, clear);
-               break;
-       /* Clear ring buffer */
-       case SYSLOG_ACTION_CLEAR:
-               syslog_print_all(NULL, 0, true);
-               break;
-       /* Disable logging to console */
-       case SYSLOG_ACTION_CONSOLE_OFF:
-               if (saved_console_loglevel == -1)
-                       saved_console_loglevel = console_loglevel;
-               console_loglevel = minimum_console_loglevel;
-               break;
-       /* Enable logging to console */
-       case SYSLOG_ACTION_CONSOLE_ON:
-               if (saved_console_loglevel != -1) {
-                       console_loglevel = saved_console_loglevel;
-                       saved_console_loglevel = -1;
-               }
-               break;
-       /* Set level of messages printed to console */
-       case SYSLOG_ACTION_CONSOLE_LEVEL:
-               error = -EINVAL;
-               if (len < 1 || len > 8)
-                       goto out;
-               if (len < minimum_console_loglevel)
-                       len = minimum_console_loglevel;
-               console_loglevel = len;
-               /* Implicitly re-enable logging to console */
-               saved_console_loglevel = -1;
-               error = 0;
-               break;
-       /* Number of chars in the log buffer */
-       case SYSLOG_ACTION_SIZE_UNREAD:
-               raw_spin_lock_irq(&logbuf_lock);
-               if (syslog_seq < log_first_seq) {
-                       /* messages are gone, move to first one */
-                       syslog_seq = log_first_seq;
-                       syslog_idx = log_first_idx;
-                       syslog_prev = 0;
-                       syslog_partial = 0;
-               }
-               if (from_file) {
-                       /*
-                        * Short-cut for poll(/"proc/kmsg") which simply checks
-                        * for pending data, not the size; return the count of
-                        * records, not the length.
-                        */
-                       error = log_next_idx - syslog_idx;
-               } else {
-                       u64 seq = syslog_seq;
-                       u32 idx = syslog_idx;
-                       enum log_flags prev = syslog_prev;
-
-                       error = 0;
-                       while (seq < log_next_seq) {
-                               struct log *msg = log_from_idx(idx);
-
-                               error += msg_print_text(msg, prev, true, NULL, 0);
-                               idx = log_next(idx);
-                               seq++;
-                               prev = msg->flags;
-                       }
-                       error -= syslog_partial;
-               }
-               raw_spin_unlock_irq(&logbuf_lock);
-               break;
-       /* Size of the log buffer */
-       case SYSLOG_ACTION_SIZE_BUFFER:
-               error = log_buf_len;
-               break;
-       default:
-               error = -EINVAL;
-               break;
-       }
-out:
-       return error;
-}
-
-SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
-{
-       return do_syslog(type, buf, len, SYSLOG_FROM_READER);
-}
-
-/*
- * Call the console drivers, asking them to write out
- * log_buf[start] to log_buf[end - 1].
- * The console_lock must be held.
- */
-static void call_console_drivers(int level, const char *text, size_t len)
-{
-       struct console *con;
-
-       trace_console(text, len);
-
-       if (level >= console_loglevel && !ignore_loglevel)
-               return;
-       if (!console_drivers)
-               return;
-
-       for_each_console(con) {
-               if (exclusive_console && con != exclusive_console)
-                       continue;
-               if (!(con->flags & CON_ENABLED))
-                       continue;
-               if (!con->write)
-                       continue;
-               if (!cpu_online(smp_processor_id()) &&
-                   !(con->flags & CON_ANYTIME))
-                       continue;
-               con->write(con, text, len);
-       }
-}
-
-/*
- * Zap console related locks when oopsing. Only zap at most once
- * every 10 seconds, to leave time for slow consoles to print a
- * full oops.
- */
-static void zap_locks(void)
-{
-       static unsigned long oops_timestamp;
-
-       if (time_after_eq(jiffies, oops_timestamp) &&
-                       !time_after(jiffies, oops_timestamp + 30 * HZ))
-               return;
-
-       oops_timestamp = jiffies;
-
-       debug_locks_off();
-       /* If a crash is occurring, make sure we can't deadlock */
-       raw_spin_lock_init(&logbuf_lock);
-       /* And make sure that we print immediately */
-       sema_init(&console_sem, 1);
-}
-
-/* Check if we have any console registered that can be called early in boot. */
-static int have_callable_console(void)
-{
-       struct console *con;
-
-       for_each_console(con)
-               if (con->flags & CON_ANYTIME)
-                       return 1;
-
-       return 0;
-}
-
-/*
- * Can we actually use the console at this time on this cpu?
- *
- * Console drivers may assume that per-cpu resources have
- * been allocated. So unless they're explicitly marked as
- * being able to cope (CON_ANYTIME) don't call them until
- * this CPU is officially up.
- */
-static inline int can_use_console(unsigned int cpu)
-{
-       return cpu_online(cpu) || have_callable_console();
-}
-
-/*
- * Try to get console ownership to actually show the kernel
- * messages from a 'printk'. Return true (and with the
- * console_lock held, and 'console_locked' set) if it
- * is successful, false otherwise.
- *
- * This gets called with the 'logbuf_lock' spinlock held and
- * interrupts disabled. It should return with 'lockbuf_lock'
- * released but interrupts still disabled.
- */
-static int console_trylock_for_printk(unsigned int cpu)
-       __releases(&logbuf_lock)
-{
-       int retval = 0, wake = 0;
-
-       if (console_trylock()) {
-               retval = 1;
-
-               /*
-                * If we can't use the console, we need to release
-                * the console semaphore by hand to avoid flushing
-                * the buffer. We need to hold the console semaphore
-                * in order to do this test safely.
-                */
-               if (!can_use_console(cpu)) {
-                       console_locked = 0;
-                       wake = 1;
-                       retval = 0;
-               }
-       }
-       logbuf_cpu = UINT_MAX;
-       raw_spin_unlock(&logbuf_lock);
-       if (wake)
-               up(&console_sem);
-       return retval;
-}
-
-int printk_delay_msec __read_mostly;
-
-static inline void printk_delay(void)
-{
-       if (unlikely(printk_delay_msec)) {
-               int m = printk_delay_msec;
-
-               while (m--) {
-                       mdelay(1);
-                       touch_nmi_watchdog();
-               }
-       }
-}
-
-/*
- * Continuation lines are buffered, and not committed to the record buffer
- * until the line is complete, or a race forces it. The line fragments
- * though, are printed immediately to the consoles to ensure everything has
- * reached the console in case of a kernel crash.
- */
-static struct cont {
-       char buf[LOG_LINE_MAX];
-       size_t len;                     /* length == 0 means unused buffer */
-       size_t cons;                    /* bytes written to console */
-       struct task_struct *owner;      /* task of first print*/
-       u64 ts_nsec;                    /* time of first print */
-       u8 level;                       /* log level of first message */
-       u8 facility;                    /* log level of first message */
-       enum log_flags flags;           /* prefix, newline flags */
-       bool flushed:1;                 /* buffer sealed and committed */
-} cont;
-
-static void cont_flush(enum log_flags flags)
-{
-       if (cont.flushed)
-               return;
-       if (cont.len == 0)
-               return;
-
-       if (cont.cons) {
-               /*
-                * If a fragment of this line was directly flushed to the
-                * console; wait for the console to pick up the rest of the
-                * line. LOG_NOCONS suppresses a duplicated output.
-                */
-               log_store(cont.facility, cont.level, flags | LOG_NOCONS,
-                         cont.ts_nsec, NULL, 0, cont.buf, cont.len);
-               cont.flags = flags;
-               cont.flushed = true;
-       } else {
-               /*
-                * If no fragment of this line ever reached the console,
-                * just submit it to the store and free the buffer.
-                */
-               log_store(cont.facility, cont.level, flags, 0,
-                         NULL, 0, cont.buf, cont.len);
-               cont.len = 0;
-       }
-}
-
-static bool cont_add(int facility, int level, const char *text, size_t len)
-{
-       if (cont.len && cont.flushed)
-               return false;
-
-       if (cont.len + len > sizeof(cont.buf)) {
-               /* the line gets too long, split it up in separate records */
-               cont_flush(LOG_CONT);
-               return false;
-       }
-
-       if (!cont.len) {
-               cont.facility = facility;
-               cont.level = level;
-               cont.owner = current;
-               cont.ts_nsec = local_clock();
-               cont.flags = 0;
-               cont.cons = 0;
-               cont.flushed = false;
-       }
-
-       memcpy(cont.buf + cont.len, text, len);
-       cont.len += len;
-
-       if (cont.len > (sizeof(cont.buf) * 80) / 100)
-               cont_flush(LOG_CONT);
-
-       return true;
-}
-
-static size_t cont_print_text(char *text, size_t size)
-{
-       size_t textlen = 0;
-       size_t len;
-
-       if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
-               textlen += print_time(cont.ts_nsec, text);
-               size -= textlen;
-       }
-
-       len = cont.len - cont.cons;
-       if (len > 0) {
-               if (len+1 > size)
-                       len = size-1;
-               memcpy(text + textlen, cont.buf + cont.cons, len);
-               textlen += len;
-               cont.cons = cont.len;
-       }
-
-       if (cont.flushed) {
-               if (cont.flags & LOG_NEWLINE)
-                       text[textlen++] = '\n';
-               /* got everything, release buffer */
-               cont.len = 0;
-       }
-       return textlen;
-}
-
-asmlinkage int vprintk_emit(int facility, int level,
-                           const char *dict, size_t dictlen,
-                           const char *fmt, va_list args)
-{
-       static int recursion_bug;
-       static char textbuf[LOG_LINE_MAX];
-       char *text = textbuf;
-       size_t text_len;
-       enum log_flags lflags = 0;
-       unsigned long flags;
-       int this_cpu;
-       int printed_len = 0;
-
-       boot_delay_msec(level);
-       printk_delay();
-
-       /* This stops the holder of console_sem just where we want him */
-       local_irq_save(flags);
-       this_cpu = smp_processor_id();
-
-       /*
-        * Ouch, printk recursed into itself!
-        */
-       if (unlikely(logbuf_cpu == this_cpu)) {
-               /*
-                * If a crash is occurring during printk() on this CPU,
-                * then try to get the crash message out but make sure
-                * we can't deadlock. Otherwise just return to avoid the
-                * recursion and return - but flag the recursion so that
-                * it can be printed at the next appropriate moment:
-                */
-               if (!oops_in_progress && !lockdep_recursing(current)) {
-                       recursion_bug = 1;
-                       goto out_restore_irqs;
-               }
-               zap_locks();
-       }
-
-       lockdep_off();
-       raw_spin_lock(&logbuf_lock);
-       logbuf_cpu = this_cpu;
-
-       if (recursion_bug) {
-               static const char recursion_msg[] =
-                       "BUG: recent printk recursion!";
-
-               recursion_bug = 0;
-               printed_len += strlen(recursion_msg);
-               /* emit KERN_CRIT message */
-               log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
-                         NULL, 0, recursion_msg, printed_len);
-       }
-
-       /*
-        * The printf needs to come first; we need the syslog
-        * prefix which might be passed-in as a parameter.
-        */
-       text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
-
-       /* mark and strip a trailing newline */
-       if (text_len && text[text_len-1] == '\n') {
-               text_len--;
-               lflags |= LOG_NEWLINE;
-       }
-
-       /* strip kernel syslog prefix and extract log level or control flags */
-       if (facility == 0) {
-               int kern_level = printk_get_level(text);
-
-               if (kern_level) {
-                       const char *end_of_header = printk_skip_level(text);
-                       switch (kern_level) {
-                       case '0' ... '7':
-                               if (level == -1)
-                                       level = kern_level - '0';
-                       case 'd':       /* KERN_DEFAULT */
-                               lflags |= LOG_PREFIX;
-                       case 'c':       /* KERN_CONT */
-                               break;
-                       }
-                       text_len -= end_of_header - text;
-                       text = (char *)end_of_header;
-               }
-       }
-
-       if (level == -1)
-               level = default_message_loglevel;
-
-       if (dict)
-               lflags |= LOG_PREFIX|LOG_NEWLINE;
-
-       if (!(lflags & LOG_NEWLINE)) {
-               /*
-                * Flush the conflicting buffer. An earlier newline was missing,
-                * or another task also prints continuation lines.
-                */
-               if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
-                       cont_flush(LOG_NEWLINE);
-
-               /* buffer line if possible, otherwise store it right away */
-               if (!cont_add(facility, level, text, text_len))
-                       log_store(facility, level, lflags | LOG_CONT, 0,
-                                 dict, dictlen, text, text_len);
-       } else {
-               bool stored = false;
-
-               /*
-                * If an earlier newline was missing and it was the same task,
-                * either merge it with the current buffer and flush, or if
-                * there was a race with interrupts (prefix == true) then just
-                * flush it out and store this line separately.
-                */
-               if (cont.len && cont.owner == current) {
-                       if (!(lflags & LOG_PREFIX))
-                               stored = cont_add(facility, level, text, text_len);
-                       cont_flush(LOG_NEWLINE);
-               }
-
-               if (!stored)
-                       log_store(facility, level, lflags, 0,
-                                 dict, dictlen, text, text_len);
-       }
-       printed_len += text_len;
-
-       /*
-        * Try to acquire and then immediately release the console semaphore.
-        * The release will print out buffers and wake up /dev/kmsg and syslog()
-        * users.
-        *
-        * The console_trylock_for_printk() function will release 'logbuf_lock'
-        * regardless of whether it actually gets the console semaphore or not.
-        */
-       if (console_trylock_for_printk(this_cpu))
-               console_unlock();
-
-       lockdep_on();
-out_restore_irqs:
-       local_irq_restore(flags);
-
-       return printed_len;
-}
-EXPORT_SYMBOL(vprintk_emit);
-
-asmlinkage int vprintk(const char *fmt, va_list args)
-{
-       return vprintk_emit(0, -1, NULL, 0, fmt, args);
-}
-EXPORT_SYMBOL(vprintk);
-
-asmlinkage int printk_emit(int facility, int level,
-                          const char *dict, size_t dictlen,
-                          const char *fmt, ...)
-{
-       va_list args;
-       int r;
-
-       va_start(args, fmt);
-       r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
-       va_end(args);
-
-       return r;
-}
-EXPORT_SYMBOL(printk_emit);
-
-/**
- * printk - print a kernel message
- * @fmt: format string
- *
- * This is printk(). It can be called from any context. We want it to work.
- *
- * We try to grab the console_lock. If we succeed, it's easy - we log the
- * output and call the console drivers.  If we fail to get the semaphore, we
- * place the output into the log buffer and return. The current holder of
- * the console_sem will notice the new output in console_unlock(); and will
- * send it to the consoles before releasing the lock.
- *
- * One effect of this deferred printing is that code which calls printk() and
- * then changes console_loglevel may break. This is because console_loglevel
- * is inspected when the actual printing occurs.
- *
- * See also:
- * printf(3)
- *
- * See the vsnprintf() documentation for format string extensions over C99.
- */
-asmlinkage int printk(const char *fmt, ...)
-{
-       va_list args;
-       int r;
-
-#ifdef CONFIG_KGDB_KDB
-       if (unlikely(kdb_trap_printk)) {
-               va_start(args, fmt);
-               r = vkdb_printf(fmt, args);
-               va_end(args);
-               return r;
-       }
-#endif
-       va_start(args, fmt);
-       r = vprintk_emit(0, -1, NULL, 0, fmt, args);
-       va_end(args);
-
-       return r;
-}
-EXPORT_SYMBOL(printk);
-
-#else /* CONFIG_PRINTK */
-
-#define LOG_LINE_MAX           0
-#define PREFIX_MAX             0
-#define LOG_LINE_MAX 0
-static u64 syslog_seq;
-static u32 syslog_idx;
-static u64 console_seq;
-static u32 console_idx;
-static enum log_flags syslog_prev;
-static u64 log_first_seq;
-static u32 log_first_idx;
-static u64 log_next_seq;
-static enum log_flags console_prev;
-static struct cont {
-       size_t len;
-       size_t cons;
-       u8 level;
-       bool flushed:1;
-} cont;
-static struct log *log_from_idx(u32 idx) { return NULL; }
-static u32 log_next(u32 idx) { return 0; }
-static void call_console_drivers(int level, const char *text, size_t len) {}
-static size_t msg_print_text(const struct log *msg, enum log_flags prev,
-                            bool syslog, char *buf, size_t size) { return 0; }
-static size_t cont_print_text(char *text, size_t size) { return 0; }
-
-#endif /* CONFIG_PRINTK */
-
-#ifdef CONFIG_EARLY_PRINTK
-struct console *early_console;
-
-void early_vprintk(const char *fmt, va_list ap)
-{
-       if (early_console) {
-               char buf[512];
-               int n = vscnprintf(buf, sizeof(buf), fmt, ap);
-
-               early_console->write(early_console, buf, n);
-       }
-}
-
-asmlinkage void early_printk(const char *fmt, ...)
-{
-       va_list ap;
-
-       va_start(ap, fmt);
-       early_vprintk(fmt, ap);
-       va_end(ap);
-}
-#endif
-
-static int __add_preferred_console(char *name, int idx, char *options,
-                                  char *brl_options)
-{
-       struct console_cmdline *c;
-       int i;
-
-       /*
-        *      See if this tty is not yet registered, and
-        *      if we have a slot free.
-        */
-       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
-               if (strcmp(console_cmdline[i].name, name) == 0 &&
-                         console_cmdline[i].index == idx) {
-                               if (!brl_options)
-                                       selected_console = i;
-                               return 0;
-               }
-       if (i == MAX_CMDLINECONSOLES)
-               return -E2BIG;
-       if (!brl_options)
-               selected_console = i;
-       c = &console_cmdline[i];
-       strlcpy(c->name, name, sizeof(c->name));
-       c->options = options;
-#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
-       c->brl_options = brl_options;
-#endif
-       c->index = idx;
-       return 0;
-}
-/*
- * Set up a list of consoles.  Called from init/main.c
- */
-static int __init console_setup(char *str)
-{
-       char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
-       char *s, *options, *brl_options = NULL;
-       int idx;
-
-#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
-       if (!memcmp(str, "brl,", 4)) {
-               brl_options = "";
-               str += 4;
-       } else if (!memcmp(str, "brl=", 4)) {
-               brl_options = str + 4;
-               str = strchr(brl_options, ',');
-               if (!str) {
-                       printk(KERN_ERR "need port name after brl=\n");
-                       return 1;
-               }
-               *(str++) = 0;
-       }
-#endif
-
-       /*
-        * Decode str into name, index, options.
-        */
-       if (str[0] >= '0' && str[0] <= '9') {
-               strcpy(buf, "ttyS");
-               strncpy(buf + 4, str, sizeof(buf) - 5);
-       } else {
-               strncpy(buf, str, sizeof(buf) - 1);
-       }
-       buf[sizeof(buf) - 1] = 0;
-       if ((options = strchr(str, ',')) != NULL)
-               *(options++) = 0;
-#ifdef __sparc__
-       if (!strcmp(str, "ttya"))
-               strcpy(buf, "ttyS0");
-       if (!strcmp(str, "ttyb"))
-               strcpy(buf, "ttyS1");
-#endif
-       for (s = buf; *s; s++)
-               if ((*s >= '0' && *s <= '9') || *s == ',')
-                       break;
-       idx = simple_strtoul(s, NULL, 10);
-       *s = 0;
-
-       __add_preferred_console(buf, idx, options, brl_options);
-       console_set_on_cmdline = 1;
-       return 1;
-}
-__setup("console=", console_setup);
-
-/**
- * add_preferred_console - add a device to the list of preferred consoles.
- * @name: device name
- * @idx: device index
- * @options: options for this console
- *
- * The last preferred console added will be used for kernel messages
- * and stdin/out/err for init.  Normally this is used by console_setup
- * above to handle user-supplied console arguments; however it can also
- * be used by arch-specific code either to override the user or more
- * commonly to provide a default console (ie from PROM variables) when
- * the user has not supplied one.
- */
-int add_preferred_console(char *name, int idx, char *options)
-{
-       return __add_preferred_console(name, idx, options, NULL);
-}
-
-int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
-{
-       struct console_cmdline *c;
-       int i;
-
-       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
-               if (strcmp(console_cmdline[i].name, name) == 0 &&
-                         console_cmdline[i].index == idx) {
-                               c = &console_cmdline[i];
-                               strlcpy(c->name, name_new, sizeof(c->name));
-                               c->name[sizeof(c->name) - 1] = 0;
-                               c->options = options;
-                               c->index = idx_new;
-                               return i;
-               }
-       /* not found */
-       return -1;
-}
-
-bool console_suspend_enabled = 1;
-EXPORT_SYMBOL(console_suspend_enabled);
-
-static int __init console_suspend_disable(char *str)
-{
-       console_suspend_enabled = 0;
-       return 1;
-}
-__setup("no_console_suspend", console_suspend_disable);
-module_param_named(console_suspend, console_suspend_enabled,
-               bool, S_IRUGO | S_IWUSR);
-MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
-       " and hibernate operations");
-
-/**
- * suspend_console - suspend the console subsystem
- *
- * This disables printk() while we go into suspend states
- */
-void suspend_console(void)
-{
-       if (!console_suspend_enabled)
-               return;
-       printk("Suspending console(s) (use no_console_suspend to debug)\n");
-       console_lock();
-       console_suspended = 1;
-       up(&console_sem);
-}
-
-void resume_console(void)
-{
-       if (!console_suspend_enabled)
-               return;
-       down(&console_sem);
-       console_suspended = 0;
-       console_unlock();
-}
-
-/**
- * console_cpu_notify - print deferred console messages after CPU hotplug
- * @self: notifier struct
- * @action: CPU hotplug event
- * @hcpu: unused
- *
- * If printk() is called from a CPU that is not online yet, the messages
- * will be spooled but will not show up on the console.  This function is
- * called when a new CPU comes online (or fails to come up), and ensures
- * that any such output gets printed.
- */
-static int console_cpu_notify(struct notifier_block *self,
-       unsigned long action, void *hcpu)
-{
-       switch (action) {
-       case CPU_ONLINE:
-       case CPU_DEAD:
-       case CPU_DOWN_FAILED:
-       case CPU_UP_CANCELED:
-               console_lock();
-               console_unlock();
-       }
-       return NOTIFY_OK;
-}
-
-/**
- * console_lock - lock the console system for exclusive use.
- *
- * Acquires a lock which guarantees that the caller has
- * exclusive access to the console system and the console_drivers list.
- *
- * Can sleep, returns nothing.
- */
-void console_lock(void)
-{
-       might_sleep();
-
-       down(&console_sem);
-       if (console_suspended)
-               return;
-       console_locked = 1;
-       console_may_schedule = 1;
-       mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);
-}
-EXPORT_SYMBOL(console_lock);
-
-/**
- * console_trylock - try to lock the console system for exclusive use.
- *
- * Tried to acquire a lock which guarantees that the caller has
- * exclusive access to the console system and the console_drivers list.
- *
- * returns 1 on success, and 0 on failure to acquire the lock.
- */
-int console_trylock(void)
-{
-       if (down_trylock(&console_sem))
-               return 0;
-       if (console_suspended) {
-               up(&console_sem);
-               return 0;
-       }
-       console_locked = 1;
-       console_may_schedule = 0;
-       mutex_acquire(&console_lock_dep_map, 0, 1, _RET_IP_);
-       return 1;
-}
-EXPORT_SYMBOL(console_trylock);
-
-int is_console_locked(void)
-{
-       return console_locked;
-}
-
-static void console_cont_flush(char *text, size_t size)
-{
-       unsigned long flags;
-       size_t len;
-
-       raw_spin_lock_irqsave(&logbuf_lock, flags);
-
-       if (!cont.len)
-               goto out;
-
-       /*
-        * We still queue earlier records, likely because the console was
-        * busy. The earlier ones need to be printed before this one, we
-        * did not flush any fragment so far, so just let it queue up.
-        */
-       if (console_seq < log_next_seq && !cont.cons)
-               goto out;
-
-       len = cont_print_text(text, size);
-       raw_spin_unlock(&logbuf_lock);
-       stop_critical_timings();
-       call_console_drivers(cont.level, text, len);
-       start_critical_timings();
-       local_irq_restore(flags);
-       return;
-out:
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-}
-
-/**
- * console_unlock - unlock the console system
- *
- * Releases the console_lock which the caller holds on the console system
- * and the console driver list.
- *
- * While the console_lock was held, console output may have been buffered
- * by printk().  If this is the case, console_unlock(); emits
- * the output prior to releasing the lock.
- *
- * If there is output waiting, we wake /dev/kmsg and syslog() users.
- *
- * console_unlock(); may be called from any context.
- */
-void console_unlock(void)
-{
-       static char text[LOG_LINE_MAX + PREFIX_MAX];
-       static u64 seen_seq;
-       unsigned long flags;
-       bool wake_klogd = false;
-       bool retry;
-
-       if (console_suspended) {
-               up(&console_sem);
-               return;
-       }
-
-       console_may_schedule = 0;
-
-       /* flush buffered message fragment immediately to console */
-       console_cont_flush(text, sizeof(text));
-again:
-       for (;;) {
-               struct log *msg;
-               size_t len;
-               int level;
-
-               raw_spin_lock_irqsave(&logbuf_lock, flags);
-               if (seen_seq != log_next_seq) {
-                       wake_klogd = true;
-                       seen_seq = log_next_seq;
-               }
-
-               if (console_seq < log_first_seq) {
-                       /* messages are gone, move to first one */
-                       console_seq = log_first_seq;
-                       console_idx = log_first_idx;
-                       console_prev = 0;
-               }
-skip:
-               if (console_seq == log_next_seq)
-                       break;
-
-               msg = log_from_idx(console_idx);
-               if (msg->flags & LOG_NOCONS) {
-                       /*
-                        * Skip record we have buffered and already printed
-                        * directly to the console when we received it.
-                        */
-                       console_idx = log_next(console_idx);
-                       console_seq++;
-                       /*
-                        * We will get here again when we register a new
-                        * CON_PRINTBUFFER console. Clear the flag so we
-                        * will properly dump everything later.
-                        */
-                       msg->flags &= ~LOG_NOCONS;
-                       console_prev = msg->flags;
-                       goto skip;
-               }
-
-               level = msg->level;
-               len = msg_print_text(msg, console_prev, false,
-                                    text, sizeof(text));
-               console_idx = log_next(console_idx);
-               console_seq++;
-               console_prev = msg->flags;
-               raw_spin_unlock(&logbuf_lock);
-
-               stop_critical_timings();        /* don't trace print latency */
-               call_console_drivers(level, text, len);
-               start_critical_timings();
-               local_irq_restore(flags);
-       }
-       console_locked = 0;
-       mutex_release(&console_lock_dep_map, 1, _RET_IP_);
-
-       /* Release the exclusive_console once it is used */
-       if (unlikely(exclusive_console))
-               exclusive_console = NULL;
-
-       raw_spin_unlock(&logbuf_lock);
-
-       up(&console_sem);
-
-       /*
-        * Someone could have filled up the buffer again, so re-check if there's
-        * something to flush. In case we cannot trylock the console_sem again,
-        * there's a new owner and the console_unlock() from them will do the
-        * flush, no worries.
-        */
-       raw_spin_lock(&logbuf_lock);
-       retry = console_seq != log_next_seq;
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-
-       if (retry && console_trylock())
-               goto again;
-
-       if (wake_klogd)
-               wake_up_klogd();
-}
-EXPORT_SYMBOL(console_unlock);
-
-/**
- * console_conditional_schedule - yield the CPU if required
- *
- * If the console code is currently allowed to sleep, and
- * if this CPU should yield the CPU to another task, do
- * so here.
- *
- * Must be called within console_lock();.
- */
-void __sched console_conditional_schedule(void)
-{
-       if (console_may_schedule)
-               cond_resched();
-}
-EXPORT_SYMBOL(console_conditional_schedule);
-
-void console_unblank(void)
-{
-       struct console *c;
-
-       /*
-        * console_unblank can no longer be called in interrupt context unless
-        * oops_in_progress is set to 1..
-        */
-       if (oops_in_progress) {
-               if (down_trylock(&console_sem) != 0)
-                       return;
-       } else
-               console_lock();
-
-       console_locked = 1;
-       console_may_schedule = 0;
-       for_each_console(c)
-               if ((c->flags & CON_ENABLED) && c->unblank)
-                       c->unblank();
-       console_unlock();
-}
-
-/*
- * Return the console tty driver structure and its associated index
- */
-struct tty_driver *console_device(int *index)
-{
-       struct console *c;
-       struct tty_driver *driver = NULL;
-
-       console_lock();
-       for_each_console(c) {
-               if (!c->device)
-                       continue;
-               driver = c->device(c, index);
-               if (driver)
-                       break;
-       }
-       console_unlock();
-       return driver;
-}
-
-/*
- * Prevent further output on the passed console device so that (for example)
- * serial drivers can disable console output before suspending a port, and can
- * re-enable output afterwards.
- */
-void console_stop(struct console *console)
-{
-       console_lock();
-       console->flags &= ~CON_ENABLED;
-       console_unlock();
-}
-EXPORT_SYMBOL(console_stop);
-
-void console_start(struct console *console)
-{
-       console_lock();
-       console->flags |= CON_ENABLED;
-       console_unlock();
-}
-EXPORT_SYMBOL(console_start);
-
-static int __read_mostly keep_bootcon;
-
-static int __init keep_bootcon_setup(char *str)
-{
-       keep_bootcon = 1;
-       printk(KERN_INFO "debug: skip boot console de-registration.\n");
-
-       return 0;
-}
-
-early_param("keep_bootcon", keep_bootcon_setup);
-
-/*
- * The console driver calls this routine during kernel initialization
- * to register the console printing procedure with printk() and to
- * print any messages that were printed by the kernel before the
- * console driver was initialized.
- *
- * This can happen pretty early during the boot process (because of
- * early_printk) - sometimes before setup_arch() completes - be careful
- * of what kernel features are used - they may not be initialised yet.
- *
- * There are two types of consoles - bootconsoles (early_printk) and
- * "real" consoles (everything which is not a bootconsole) which are
- * handled differently.
- *  - Any number of bootconsoles can be registered at any time.
- *  - As soon as a "real" console is registered, all bootconsoles
- *    will be unregistered automatically.
- *  - Once a "real" console is registered, any attempt to register a
- *    bootconsoles will be rejected
- */
-void register_console(struct console *newcon)
-{
-       int i;
-       unsigned long flags;
-       struct console *bcon = NULL;
-
-       /*
-        * before we register a new CON_BOOT console, make sure we don't
-        * already have a valid console
-        */
-       if (console_drivers && newcon->flags & CON_BOOT) {
-               /* find the last or real console */
-               for_each_console(bcon) {
-                       if (!(bcon->flags & CON_BOOT)) {
-                               printk(KERN_INFO "Too late to register bootconsole %s%d\n",
-                                       newcon->name, newcon->index);
-                               return;
-                       }
-               }
-       }
-
-       if (console_drivers && console_drivers->flags & CON_BOOT)
-               bcon = console_drivers;
-
-       if (preferred_console < 0 || bcon || !console_drivers)
-               preferred_console = selected_console;
-
-       if (newcon->early_setup)
-               newcon->early_setup();
-
-       /*
-        *      See if we want to use this console driver. If we
-        *      didn't select a console we take the first one
-        *      that registers here.
-        */
-       if (preferred_console < 0) {
-               if (newcon->index < 0)
-                       newcon->index = 0;
-               if (newcon->setup == NULL ||
-                   newcon->setup(newcon, NULL) == 0) {
-                       newcon->flags |= CON_ENABLED;
-                       if (newcon->device) {
-                               newcon->flags |= CON_CONSDEV;
-                               preferred_console = 0;
-                       }
-               }
-       }
-
-       /*
-        *      See if this console matches one we selected on
-        *      the command line.
-        */
-       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
-                       i++) {
-               if (strcmp(console_cmdline[i].name, newcon->name) != 0)
-                       continue;
-               if (newcon->index >= 0 &&
-                   newcon->index != console_cmdline[i].index)
-                       continue;
-               if (newcon->index < 0)
-                       newcon->index = console_cmdline[i].index;
-#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
-               if (console_cmdline[i].brl_options) {
-                       newcon->flags |= CON_BRL;
-                       braille_register_console(newcon,
-                                       console_cmdline[i].index,
-                                       console_cmdline[i].options,
-                                       console_cmdline[i].brl_options);
-                       return;
-               }
-#endif
-               if (newcon->setup &&
-                   newcon->setup(newcon, console_cmdline[i].options) != 0)
-                       break;
-               newcon->flags |= CON_ENABLED;
-               newcon->index = console_cmdline[i].index;
-               if (i == selected_console) {
-                       newcon->flags |= CON_CONSDEV;
-                       preferred_console = selected_console;
-               }
-               break;
-       }
-
-       if (!(newcon->flags & CON_ENABLED))
-               return;
-
-       /*
-        * If we have a bootconsole, and are switching to a real console,
-        * don't print everything out again, since when the boot console, and
-        * the real console are the same physical device, it's annoying to
-        * see the beginning boot messages twice
-        */
-       if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
-               newcon->flags &= ~CON_PRINTBUFFER;
-
-       /*
-        *      Put this console in the list - keep the
-        *      preferred driver at the head of the list.
-        */
-       console_lock();
-       if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
-               newcon->next = console_drivers;
-               console_drivers = newcon;
-               if (newcon->next)
-                       newcon->next->flags &= ~CON_CONSDEV;
-       } else {
-               newcon->next = console_drivers->next;
-               console_drivers->next = newcon;
-       }
-       if (newcon->flags & CON_PRINTBUFFER) {
-               /*
-                * console_unlock(); will print out the buffered messages
-                * for us.
-                */
-               raw_spin_lock_irqsave(&logbuf_lock, flags);
-               console_seq = syslog_seq;
-               console_idx = syslog_idx;
-               console_prev = syslog_prev;
-               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-               /*
-                * We're about to replay the log buffer.  Only do this to the
-                * just-registered console to avoid excessive message spam to
-                * the already-registered consoles.
-                */
-               exclusive_console = newcon;
-       }
-       console_unlock();
-       console_sysfs_notify();
-
-       /*
-        * By unregistering the bootconsoles after we enable the real console
-        * we get the "console xxx enabled" message on all the consoles -
-        * boot consoles, real consoles, etc - this is to ensure that end
-        * users know there might be something in the kernel's log buffer that
-        * went to the bootconsole (that they do not see on the real console)
-        */
-       if (bcon &&
-           ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
-           !keep_bootcon) {
-               /* we need to iterate through twice, to make sure we print
-                * everything out, before we unregister the console(s)
-                */
-               printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
-                       newcon->name, newcon->index);
-               for_each_console(bcon)
-                       if (bcon->flags & CON_BOOT)
-                               unregister_console(bcon);
-       } else {
-               printk(KERN_INFO "%sconsole [%s%d] enabled\n",
-                       (newcon->flags & CON_BOOT) ? "boot" : "" ,
-                       newcon->name, newcon->index);
-       }
-}
-EXPORT_SYMBOL(register_console);
-
-int unregister_console(struct console *console)
-{
-        struct console *a, *b;
-       int res = 1;
-
-#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
-       if (console->flags & CON_BRL)
-               return braille_unregister_console(console);
-#endif
-
-       console_lock();
-       if (console_drivers == console) {
-               console_drivers=console->next;
-               res = 0;
-       } else if (console_drivers) {
-               for (a=console_drivers->next, b=console_drivers ;
-                    a; b=a, a=b->next) {
-                       if (a == console) {
-                               b->next = a->next;
-                               res = 0;
-                               break;
-                       }
-               }
-       }
-
-       /*
-        * If this isn't the last console and it has CON_CONSDEV set, we
-        * need to set it on the next preferred console.
-        */
-       if (console_drivers != NULL && console->flags & CON_CONSDEV)
-               console_drivers->flags |= CON_CONSDEV;
-
-       console_unlock();
-       console_sysfs_notify();
-       return res;
-}
-EXPORT_SYMBOL(unregister_console);
-
-static int __init printk_late_init(void)
-{
-       struct console *con;
-
-       for_each_console(con) {
-               if (!keep_bootcon && con->flags & CON_BOOT) {
-                       printk(KERN_INFO "turn off boot console %s%d\n",
-                               con->name, con->index);
-                       unregister_console(con);
-               }
-       }
-       hotcpu_notifier(console_cpu_notify, 0);
-       return 0;
-}
-late_initcall(printk_late_init);
-
-#if defined CONFIG_PRINTK
-/*
- * Delayed printk version, for scheduler-internal messages:
- */
-#define PRINTK_BUF_SIZE                512
-
-#define PRINTK_PENDING_WAKEUP  0x01
-#define PRINTK_PENDING_SCHED   0x02
-
-static DEFINE_PER_CPU(int, printk_pending);
-static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
-
-static void wake_up_klogd_work_func(struct irq_work *irq_work)
-{
-       int pending = __this_cpu_xchg(printk_pending, 0);
-
-       if (pending & PRINTK_PENDING_SCHED) {
-               char *buf = __get_cpu_var(printk_sched_buf);
-               printk(KERN_WARNING "[sched_delayed] %s", buf);
-       }
-
-       if (pending & PRINTK_PENDING_WAKEUP)
-               wake_up_interruptible(&log_wait);
-}
-
-static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
-       .func = wake_up_klogd_work_func,
-       .flags = IRQ_WORK_LAZY,
-};
-
-void wake_up_klogd(void)
-{
-       preempt_disable();
-       if (waitqueue_active(&log_wait)) {
-               this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
-               irq_work_queue(&__get_cpu_var(wake_up_klogd_work));
-       }
-       preempt_enable();
-}
-
-int printk_sched(const char *fmt, ...)
-{
-       unsigned long flags;
-       va_list args;
-       char *buf;
-       int r;
-
-       local_irq_save(flags);
-       buf = __get_cpu_var(printk_sched_buf);
-
-       va_start(args, fmt);
-       r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
-       va_end(args);
-
-       __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
-       irq_work_queue(&__get_cpu_var(wake_up_klogd_work));
-       local_irq_restore(flags);
-
-       return r;
-}
-
-/*
- * printk rate limiting, lifted from the networking subsystem.
- *
- * This enforces a rate limit: not more than 10 kernel messages
- * every 5s to make a denial-of-service attack impossible.
- */
-DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
-
-int __printk_ratelimit(const char *func)
-{
-       return ___ratelimit(&printk_ratelimit_state, func);
-}
-EXPORT_SYMBOL(__printk_ratelimit);
-
-/**
- * printk_timed_ratelimit - caller-controlled printk ratelimiting
- * @caller_jiffies: pointer to caller's state
- * @interval_msecs: minimum interval between prints
- *
- * printk_timed_ratelimit() returns true if more than @interval_msecs
- * milliseconds have elapsed since the last time printk_timed_ratelimit()
- * returned true.
- */
-bool printk_timed_ratelimit(unsigned long *caller_jiffies,
-                       unsigned int interval_msecs)
-{
-       if (*caller_jiffies == 0
-                       || !time_in_range(jiffies, *caller_jiffies,
-                                       *caller_jiffies
-                                       + msecs_to_jiffies(interval_msecs))) {
-               *caller_jiffies = jiffies;
-               return true;
-       }
-       return false;
-}
-EXPORT_SYMBOL(printk_timed_ratelimit);
-
-static DEFINE_SPINLOCK(dump_list_lock);
-static LIST_HEAD(dump_list);
-
-/**
- * kmsg_dump_register - register a kernel log dumper.
- * @dumper: pointer to the kmsg_dumper structure
- *
- * Adds a kernel log dumper to the system. The dump callback in the
- * structure will be called when the kernel oopses or panics and must be
- * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
- */
-int kmsg_dump_register(struct kmsg_dumper *dumper)
-{
-       unsigned long flags;
-       int err = -EBUSY;
-
-       /* The dump callback needs to be set */
-       if (!dumper->dump)
-               return -EINVAL;
-
-       spin_lock_irqsave(&dump_list_lock, flags);
-       /* Don't allow registering multiple times */
-       if (!dumper->registered) {
-               dumper->registered = 1;
-               list_add_tail_rcu(&dumper->list, &dump_list);
-               err = 0;
-       }
-       spin_unlock_irqrestore(&dump_list_lock, flags);
-
-       return err;
-}
-EXPORT_SYMBOL_GPL(kmsg_dump_register);
-
-/**
- * kmsg_dump_unregister - unregister a kmsg dumper.
- * @dumper: pointer to the kmsg_dumper structure
- *
- * Removes a dump device from the system. Returns zero on success and
- * %-EINVAL otherwise.
- */
-int kmsg_dump_unregister(struct kmsg_dumper *dumper)
-{
-       unsigned long flags;
-       int err = -EINVAL;
-
-       spin_lock_irqsave(&dump_list_lock, flags);
-       if (dumper->registered) {
-               dumper->registered = 0;
-               list_del_rcu(&dumper->list);
-               err = 0;
-       }
-       spin_unlock_irqrestore(&dump_list_lock, flags);
-       synchronize_rcu();
-
-       return err;
-}
-EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
-
-static bool always_kmsg_dump;
-module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
-
-/**
- * kmsg_dump - dump kernel log to kernel message dumpers.
- * @reason: the reason (oops, panic etc) for dumping
- *
- * Call each of the registered dumper's dump() callback, which can
- * retrieve the kmsg records with kmsg_dump_get_line() or
- * kmsg_dump_get_buffer().
- */
-void kmsg_dump(enum kmsg_dump_reason reason)
-{
-       struct kmsg_dumper *dumper;
-       unsigned long flags;
-
-       if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
-               return;
-
-       rcu_read_lock();
-       list_for_each_entry_rcu(dumper, &dump_list, list) {
-               if (dumper->max_reason && reason > dumper->max_reason)
-                       continue;
-
-               /* initialize iterator with data about the stored records */
-               dumper->active = true;
-
-               raw_spin_lock_irqsave(&logbuf_lock, flags);
-               dumper->cur_seq = clear_seq;
-               dumper->cur_idx = clear_idx;
-               dumper->next_seq = log_next_seq;
-               dumper->next_idx = log_next_idx;
-               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-
-               /* invoke dumper which will iterate over records */
-               dumper->dump(dumper, reason);
-
-               /* reset iterator */
-               dumper->active = false;
-       }
-       rcu_read_unlock();
-}
-
-/**
- * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
- * @dumper: registered kmsg dumper
- * @syslog: include the "<4>" prefixes
- * @line: buffer to copy the line to
- * @size: maximum size of the buffer
- * @len: length of line placed into buffer
- *
- * Start at the beginning of the kmsg buffer, with the oldest kmsg
- * record, and copy one record into the provided buffer.
- *
- * Consecutive calls will return the next available record moving
- * towards the end of the buffer with the youngest messages.
- *
- * A return value of FALSE indicates that there are no more records to
- * read.
- *
- * The function is similar to kmsg_dump_get_line(), but grabs no locks.
- */
-bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
-                              char *line, size_t size, size_t *len)
-{
-       struct log *msg;
-       size_t l = 0;
-       bool ret = false;
-
-       if (!dumper->active)
-               goto out;
-
-       if (dumper->cur_seq < log_first_seq) {
-               /* messages are gone, move to first available one */
-               dumper->cur_seq = log_first_seq;
-               dumper->cur_idx = log_first_idx;
-       }
-
-       /* last entry */
-       if (dumper->cur_seq >= log_next_seq)
-               goto out;
-
-       msg = log_from_idx(dumper->cur_idx);
-       l = msg_print_text(msg, 0, syslog, line, size);
-
-       dumper->cur_idx = log_next(dumper->cur_idx);
-       dumper->cur_seq++;
-       ret = true;
-out:
-       if (len)
-               *len = l;
-       return ret;
-}
-
-/**
- * kmsg_dump_get_line - retrieve one kmsg log line
- * @dumper: registered kmsg dumper
- * @syslog: include the "<4>" prefixes
- * @line: buffer to copy the line to
- * @size: maximum size of the buffer
- * @len: length of line placed into buffer
- *
- * Start at the beginning of the kmsg buffer, with the oldest kmsg
- * record, and copy one record into the provided buffer.
- *
- * Consecutive calls will return the next available record moving
- * towards the end of the buffer with the youngest messages.
- *
- * A return value of FALSE indicates that there are no more records to
- * read.
- */
-bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
-                       char *line, size_t size, size_t *len)
-{
-       unsigned long flags;
-       bool ret;
-
-       raw_spin_lock_irqsave(&logbuf_lock, flags);
-       ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-
-       return ret;
-}
-EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
-
-/**
- * kmsg_dump_get_buffer - copy kmsg log lines
- * @dumper: registered kmsg dumper
- * @syslog: include the "<4>" prefixes
- * @buf: buffer to copy the line to
- * @size: maximum size of the buffer
- * @len: length of line placed into buffer
- *
- * Start at the end of the kmsg buffer and fill the provided buffer
- * with as many of the the *youngest* kmsg records that fit into it.
- * If the buffer is large enough, all available kmsg records will be
- * copied with a single call.
- *
- * Consecutive calls will fill the buffer with the next block of
- * available older records, not including the earlier retrieved ones.
- *
- * A return value of FALSE indicates that there are no more records to
- * read.
- */
-bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
-                         char *buf, size_t size, size_t *len)
-{
-       unsigned long flags;
-       u64 seq;
-       u32 idx;
-       u64 next_seq;
-       u32 next_idx;
-       enum log_flags prev;
-       size_t l = 0;
-       bool ret = false;
-
-       if (!dumper->active)
-               goto out;
-
-       raw_spin_lock_irqsave(&logbuf_lock, flags);
-       if (dumper->cur_seq < log_first_seq) {
-               /* messages are gone, move to first available one */
-               dumper->cur_seq = log_first_seq;
-               dumper->cur_idx = log_first_idx;
-       }
-
-       /* last entry */
-       if (dumper->cur_seq >= dumper->next_seq) {
-               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-               goto out;
-       }
-
-       /* calculate length of entire buffer */
-       seq = dumper->cur_seq;
-       idx = dumper->cur_idx;
-       prev = 0;
-       while (seq < dumper->next_seq) {
-               struct log *msg = log_from_idx(idx);
-
-               l += msg_print_text(msg, prev, true, NULL, 0);
-               idx = log_next(idx);
-               seq++;
-               prev = msg->flags;
-       }
-
-       /* move first record forward until length fits into the buffer */
-       seq = dumper->cur_seq;
-       idx = dumper->cur_idx;
-       prev = 0;
-       while (l > size && seq < dumper->next_seq) {
-               struct log *msg = log_from_idx(idx);
-
-               l -= msg_print_text(msg, prev, true, NULL, 0);
-               idx = log_next(idx);
-               seq++;
-               prev = msg->flags;
-       }
-
-       /* last message in next interation */
-       next_seq = seq;
-       next_idx = idx;
-
-       l = 0;
-       prev = 0;
-       while (seq < dumper->next_seq) {
-               struct log *msg = log_from_idx(idx);
-
-               l += msg_print_text(msg, prev, syslog, buf + l, size - l);
-               idx = log_next(idx);
-               seq++;
-               prev = msg->flags;
-       }
-
-       dumper->next_seq = next_seq;
-       dumper->next_idx = next_idx;
-       ret = true;
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-out:
-       if (len)
-               *len = l;
-       return ret;
-}
-EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
-
-/**
- * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
- * @dumper: registered kmsg dumper
- *
- * Reset the dumper's iterator so that kmsg_dump_get_line() and
- * kmsg_dump_get_buffer() can be called again and used multiple
- * times within the same dumper.dump() callback.
- *
- * The function is similar to kmsg_dump_rewind(), but grabs no locks.
- */
-void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
-{
-       dumper->cur_seq = clear_seq;
-       dumper->cur_idx = clear_idx;
-       dumper->next_seq = log_next_seq;
-       dumper->next_idx = log_next_idx;
-}
-
-/**
- * kmsg_dump_rewind - reset the interator
- * @dumper: registered kmsg dumper
- *
- * Reset the dumper's iterator so that kmsg_dump_get_line() and
- * kmsg_dump_get_buffer() can be called again and used multiple
- * times within the same dumper.dump() callback.
- */
-void kmsg_dump_rewind(struct kmsg_dumper *dumper)
-{
-       unsigned long flags;
-
-       raw_spin_lock_irqsave(&logbuf_lock, flags);
-       kmsg_dump_rewind_nolock(dumper);
-       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
-}
-EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
-
-static char dump_stack_arch_desc_str[128];
-
-/**
- * dump_stack_set_arch_desc - set arch-specific str to show with task dumps
- * @fmt: printf-style format string
- * @...: arguments for the format string
- *
- * The configured string will be printed right after utsname during task
- * dumps.  Usually used to add arch-specific system identifiers.  If an
- * arch wants to make use of such an ID string, it should initialize this
- * as soon as possible during boot.
- */
-void __init dump_stack_set_arch_desc(const char *fmt, ...)
-{
-       va_list args;
-
-       va_start(args, fmt);
-       vsnprintf(dump_stack_arch_desc_str, sizeof(dump_stack_arch_desc_str),
-                 fmt, args);
-       va_end(args);
-}
-
-/**
- * dump_stack_print_info - print generic debug info for dump_stack()
- * @log_lvl: log level
- *
- * Arch-specific dump_stack() implementations can use this function to
- * print out the same debug information as the generic dump_stack().
- */
-void dump_stack_print_info(const char *log_lvl)
-{
-       printk("%sCPU: %d PID: %d Comm: %.20s %s %s %.*s\n",
-              log_lvl, raw_smp_processor_id(), current->pid, current->comm,
-              print_tainted(), init_utsname()->release,
-              (int)strcspn(init_utsname()->version, " "),
-              init_utsname()->version);
-
-       if (dump_stack_arch_desc_str[0] != '\0')
-               printk("%sHardware name: %s\n",
-                      log_lvl, dump_stack_arch_desc_str);
-
-       print_worker_info(log_lvl, current);
-}
-
-/**
- * show_regs_print_info - print generic debug info for show_regs()
- * @log_lvl: log level
- *
- * show_regs() implementations can use this function to print out generic
- * debug information.
- */
-void show_regs_print_info(const char *log_lvl)
-{
-       dump_stack_print_info(log_lvl);
-
-       printk("%stask: %p ti: %p task.ti: %p\n",
-              log_lvl, current, current_thread_info(),
-              task_thread_info(current));
-}
-
-#endif
diff --git a/kernel/printk/Makefile b/kernel/printk/Makefile
new file mode 100644 (file)
index 0000000..36d306d
--- /dev/null
@@ -0,0 +1 @@
+obj-y  = printk.o
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
new file mode 100644 (file)
index 0000000..69b0890
--- /dev/null
@@ -0,0 +1,2924 @@
+/*
+ *  linux/kernel/printk.c
+ *
+ *  Copyright (C) 1991, 1992  Linus Torvalds
+ *
+ * Modified to make sys_syslog() more flexible: added commands to
+ * return the last 4k of kernel messages, regardless of whether
+ * they've been read or not.  Added option to suppress kernel printk's
+ * to the console.  Added hook for sending the console messages
+ * elsewhere, in preparation for a serial line console (someday).
+ * Ted Ts'o, 2/11/93.
+ * Modified for sysctl support, 1/8/97, Chris Horn.
+ * Fixed SMP synchronization, 08/08/99, Manfred Spraul
+ *     manfred@colorfullife.com
+ * Rewrote bits to get rid of console_lock
+ *     01Mar01 Andrew Morton
+ */
+
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/tty.h>
+#include <linux/tty_driver.h>
+#include <linux/console.h>
+#include <linux/init.h>
+#include <linux/jiffies.h>
+#include <linux/nmi.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/interrupt.h>                   /* For in_interrupt() */
+#include <linux/delay.h>
+#include <linux/smp.h>
+#include <linux/security.h>
+#include <linux/bootmem.h>
+#include <linux/memblock.h>
+#include <linux/aio.h>
+#include <linux/syscalls.h>
+#include <linux/kexec.h>
+#include <linux/kdb.h>
+#include <linux/ratelimit.h>
+#include <linux/kmsg_dump.h>
+#include <linux/syslog.h>
+#include <linux/cpu.h>
+#include <linux/notifier.h>
+#include <linux/rculist.h>
+#include <linux/poll.h>
+#include <linux/irq_work.h>
+#include <linux/utsname.h>
+
+#include <asm/uaccess.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/printk.h>
+
+/* printk's without a loglevel use this.. */
+#define DEFAULT_MESSAGE_LOGLEVEL CONFIG_DEFAULT_MESSAGE_LOGLEVEL
+
+/* We show everything that is MORE important than this.. */
+#define MINIMUM_CONSOLE_LOGLEVEL 1 /* Minimum loglevel we let people use */
+#define DEFAULT_CONSOLE_LOGLEVEL 7 /* anything MORE serious than KERN_DEBUG */
+
+int console_printk[4] = {
+       DEFAULT_CONSOLE_LOGLEVEL,       /* console_loglevel */
+       DEFAULT_MESSAGE_LOGLEVEL,       /* default_message_loglevel */
+       MINIMUM_CONSOLE_LOGLEVEL,       /* minimum_console_loglevel */
+       DEFAULT_CONSOLE_LOGLEVEL,       /* default_console_loglevel */
+};
+
+/*
+ * Low level drivers may need that to know if they can schedule in
+ * their unblank() callback or not. So let's export it.
+ */
+int oops_in_progress;
+EXPORT_SYMBOL(oops_in_progress);
+
+/*
+ * console_sem protects the console_drivers list, and also
+ * provides serialisation for access to the entire console
+ * driver system.
+ */
+static DEFINE_SEMAPHORE(console_sem);
+struct console *console_drivers;
+EXPORT_SYMBOL_GPL(console_drivers);
+
+#ifdef CONFIG_LOCKDEP
+static struct lockdep_map console_lock_dep_map = {
+       .name = "console_lock"
+};
+#endif
+
+/*
+ * This is used for debugging the mess that is the VT code by
+ * keeping track if we have the console semaphore held. It's
+ * definitely not the perfect debug tool (we don't know if _WE_
+ * hold it are racing, but it helps tracking those weird code
+ * path in the console code where we end up in places I want
+ * locked without the console sempahore held
+ */
+static int console_locked, console_suspended;
+
+/*
+ * If exclusive_console is non-NULL then only this console is to be printed to.
+ */
+static struct console *exclusive_console;
+
+/*
+ *     Array of consoles built from command line options (console=)
+ */
+struct console_cmdline
+{
+       char    name[8];                        /* Name of the driver       */
+       int     index;                          /* Minor dev. to use        */
+       char    *options;                       /* Options for the driver   */
+#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
+       char    *brl_options;                   /* Options for braille driver */
+#endif
+};
+
+#define MAX_CMDLINECONSOLES 8
+
+static struct console_cmdline console_cmdline[MAX_CMDLINECONSOLES];
+static int selected_console = -1;
+static int preferred_console = -1;
+int console_set_on_cmdline;
+EXPORT_SYMBOL(console_set_on_cmdline);
+
+/* Flag: console code may call schedule() */
+static int console_may_schedule;
+
+/*
+ * The printk log buffer consists of a chain of concatenated variable
+ * length records. Every record starts with a record header, containing
+ * the overall length of the record.
+ *
+ * The heads to the first and last entry in the buffer, as well as the
+ * sequence numbers of these both entries are maintained when messages
+ * are stored..
+ *
+ * If the heads indicate available messages, the length in the header
+ * tells the start next message. A length == 0 for the next message
+ * indicates a wrap-around to the beginning of the buffer.
+ *
+ * Every record carries the monotonic timestamp in microseconds, as well as
+ * the standard userspace syslog level and syslog facility. The usual
+ * kernel messages use LOG_KERN; userspace-injected messages always carry
+ * a matching syslog facility, by default LOG_USER. The origin of every
+ * message can be reliably determined that way.
+ *
+ * The human readable log message directly follows the message header. The
+ * length of the message text is stored in the header, the stored message
+ * is not terminated.
+ *
+ * Optionally, a message can carry a dictionary of properties (key/value pairs),
+ * to provide userspace with a machine-readable message context.
+ *
+ * Examples for well-defined, commonly used property names are:
+ *   DEVICE=b12:8               device identifier
+ *                                b12:8         block dev_t
+ *                                c127:3        char dev_t
+ *                                n8            netdev ifindex
+ *                                +sound:card0  subsystem:devname
+ *   SUBSYSTEM=pci              driver-core subsystem name
+ *
+ * Valid characters in property names are [a-zA-Z0-9.-_]. The plain text value
+ * follows directly after a '=' character. Every property is terminated by
+ * a '\0' character. The last property is not terminated.
+ *
+ * Example of a message structure:
+ *   0000  ff 8f 00 00 00 00 00 00      monotonic time in nsec
+ *   0008  34 00                        record is 52 bytes long
+ *   000a        0b 00                  text is 11 bytes long
+ *   000c              1f 00            dictionary is 23 bytes long
+ *   000e                    03 00      LOG_KERN (facility) LOG_ERR (level)
+ *   0010  69 74 27 73 20 61 20 6c      "it's a l"
+ *         69 6e 65                     "ine"
+ *   001b           44 45 56 49 43      "DEVIC"
+ *         45 3d 62 38 3a 32 00 44      "E=b8:2\0D"
+ *         52 49 56 45 52 3d 62 75      "RIVER=bu"
+ *         67                           "g"
+ *   0032     00 00 00                  padding to next message header
+ *
+ * The 'struct log' buffer header must never be directly exported to
+ * userspace, it is a kernel-private implementation detail that might
+ * need to be changed in the future, when the requirements change.
+ *
+ * /dev/kmsg exports the structured data in the following line format:
+ *   "level,sequnum,timestamp;<message text>\n"
+ *
+ * The optional key/value pairs are attached as continuation lines starting
+ * with a space character and terminated by a newline. All possible
+ * non-prinatable characters are escaped in the "\xff" notation.
+ *
+ * Users of the export format should ignore possible additional values
+ * separated by ',', and find the message after the ';' character.
+ */
+
+enum log_flags {
+       LOG_NOCONS      = 1,    /* already flushed, do not print to console */
+       LOG_NEWLINE     = 2,    /* text ended with a newline */
+       LOG_PREFIX      = 4,    /* text started with a prefix */
+       LOG_CONT        = 8,    /* text is a fragment of a continuation line */
+};
+
+struct log {
+       u64 ts_nsec;            /* timestamp in nanoseconds */
+       u16 len;                /* length of entire record */
+       u16 text_len;           /* length of text buffer */
+       u16 dict_len;           /* length of dictionary buffer */
+       u8 facility;            /* syslog facility */
+       u8 flags:5;             /* internal record flags */
+       u8 level:3;             /* syslog level */
+};
+
+/*
+ * The logbuf_lock protects kmsg buffer, indices, counters. It is also
+ * used in interesting ways to provide interlocking in console_unlock();
+ */
+static DEFINE_RAW_SPINLOCK(logbuf_lock);
+
+#ifdef CONFIG_PRINTK
+DECLARE_WAIT_QUEUE_HEAD(log_wait);
+/* the next printk record to read by syslog(READ) or /proc/kmsg */
+static u64 syslog_seq;
+static u32 syslog_idx;
+static enum log_flags syslog_prev;
+static size_t syslog_partial;
+
+/* index and sequence number of the first record stored in the buffer */
+static u64 log_first_seq;
+static u32 log_first_idx;
+
+/* index and sequence number of the next record to store in the buffer */
+static u64 log_next_seq;
+static u32 log_next_idx;
+
+/* the next printk record to write to the console */
+static u64 console_seq;
+static u32 console_idx;
+static enum log_flags console_prev;
+
+/* the next printk record to read after the last 'clear' command */
+static u64 clear_seq;
+static u32 clear_idx;
+
+#define PREFIX_MAX             32
+#define LOG_LINE_MAX           1024 - PREFIX_MAX
+
+/* record buffer */
+#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
+#define LOG_ALIGN 4
+#else
+#define LOG_ALIGN __alignof__(struct log)
+#endif
+#define __LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT)
+static char __log_buf[__LOG_BUF_LEN] __aligned(LOG_ALIGN);
+static char *log_buf = __log_buf;
+static u32 log_buf_len = __LOG_BUF_LEN;
+
+/* cpu currently holding logbuf_lock */
+static volatile unsigned int logbuf_cpu = UINT_MAX;
+
+/* human readable text of the record */
+static char *log_text(const struct log *msg)
+{
+       return (char *)msg + sizeof(struct log);
+}
+
+/* optional key/value pair dictionary attached to the record */
+static char *log_dict(const struct log *msg)
+{
+       return (char *)msg + sizeof(struct log) + msg->text_len;
+}
+
+/* get record by index; idx must point to valid msg */
+static struct log *log_from_idx(u32 idx)
+{
+       struct log *msg = (struct log *)(log_buf + idx);
+
+       /*
+        * A length == 0 record is the end of buffer marker. Wrap around and
+        * read the message at the start of the buffer.
+        */
+       if (!msg->len)
+               return (struct log *)log_buf;
+       return msg;
+}
+
+/* get next record; idx must point to valid msg */
+static u32 log_next(u32 idx)
+{
+       struct log *msg = (struct log *)(log_buf + idx);
+
+       /* length == 0 indicates the end of the buffer; wrap */
+       /*
+        * A length == 0 record is the end of buffer marker. Wrap around and
+        * read the message at the start of the buffer as *this* one, and
+        * return the one after that.
+        */
+       if (!msg->len) {
+               msg = (struct log *)log_buf;
+               return msg->len;
+       }
+       return idx + msg->len;
+}
+
+/* insert record into the buffer, discard old ones, update heads */
+static void log_store(int facility, int level,
+                     enum log_flags flags, u64 ts_nsec,
+                     const char *dict, u16 dict_len,
+                     const char *text, u16 text_len)
+{
+       struct log *msg;
+       u32 size, pad_len;
+
+       /* number of '\0' padding bytes to next message */
+       size = sizeof(struct log) + text_len + dict_len;
+       pad_len = (-size) & (LOG_ALIGN - 1);
+       size += pad_len;
+
+       while (log_first_seq < log_next_seq) {
+               u32 free;
+
+               if (log_next_idx > log_first_idx)
+                       free = max(log_buf_len - log_next_idx, log_first_idx);
+               else
+                       free = log_first_idx - log_next_idx;
+
+               if (free > size + sizeof(struct log))
+                       break;
+
+               /* drop old messages until we have enough contiuous space */
+               log_first_idx = log_next(log_first_idx);
+               log_first_seq++;
+       }
+
+       if (log_next_idx + size + sizeof(struct log) >= log_buf_len) {
+               /*
+                * This message + an additional empty header does not fit
+                * at the end of the buffer. Add an empty header with len == 0
+                * to signify a wrap around.
+                */
+               memset(log_buf + log_next_idx, 0, sizeof(struct log));
+               log_next_idx = 0;
+       }
+
+       /* fill message */
+       msg = (struct log *)(log_buf + log_next_idx);
+       memcpy(log_text(msg), text, text_len);
+       msg->text_len = text_len;
+       memcpy(log_dict(msg), dict, dict_len);
+       msg->dict_len = dict_len;
+       msg->facility = facility;
+       msg->level = level & 7;
+       msg->flags = flags & 0x1f;
+       if (ts_nsec > 0)
+               msg->ts_nsec = ts_nsec;
+       else
+               msg->ts_nsec = local_clock();
+       memset(log_dict(msg) + dict_len, 0, pad_len);
+       msg->len = sizeof(struct log) + text_len + dict_len + pad_len;
+
+       /* insert message */
+       log_next_idx += msg->len;
+       log_next_seq++;
+}
+
+#ifdef CONFIG_SECURITY_DMESG_RESTRICT
+int dmesg_restrict = 1;
+#else
+int dmesg_restrict;
+#endif
+
+static int syslog_action_restricted(int type)
+{
+       if (dmesg_restrict)
+               return 1;
+       /*
+        * Unless restricted, we allow "read all" and "get buffer size"
+        * for everybody.
+        */
+       return type != SYSLOG_ACTION_READ_ALL &&
+              type != SYSLOG_ACTION_SIZE_BUFFER;
+}
+
+static int check_syslog_permissions(int type, bool from_file)
+{
+       /*
+        * If this is from /proc/kmsg and we've already opened it, then we've
+        * already done the capabilities checks at open time.
+        */
+       if (from_file && type != SYSLOG_ACTION_OPEN)
+               return 0;
+
+       if (syslog_action_restricted(type)) {
+               if (capable(CAP_SYSLOG))
+                       return 0;
+               /*
+                * For historical reasons, accept CAP_SYS_ADMIN too, with
+                * a warning.
+                */
+               if (capable(CAP_SYS_ADMIN)) {
+                       pr_warn_once("%s (%d): Attempt to access syslog with "
+                                    "CAP_SYS_ADMIN but no CAP_SYSLOG "
+                                    "(deprecated).\n",
+                                current->comm, task_pid_nr(current));
+                       return 0;
+               }
+               return -EPERM;
+       }
+       return security_syslog(type);
+}
+
+
+/* /dev/kmsg - userspace message inject/listen interface */
+struct devkmsg_user {
+       u64 seq;
+       u32 idx;
+       enum log_flags prev;
+       struct mutex lock;
+       char buf[8192];
+};
+
+static ssize_t devkmsg_writev(struct kiocb *iocb, const struct iovec *iv,
+                             unsigned long count, loff_t pos)
+{
+       char *buf, *line;
+       int i;
+       int level = default_message_loglevel;
+       int facility = 1;       /* LOG_USER */
+       size_t len = iov_length(iv, count);
+       ssize_t ret = len;
+
+       if (len > LOG_LINE_MAX)
+               return -EINVAL;
+       buf = kmalloc(len+1, GFP_KERNEL);
+       if (buf == NULL)
+               return -ENOMEM;
+
+       line = buf;
+       for (i = 0; i < count; i++) {
+               if (copy_from_user(line, iv[i].iov_base, iv[i].iov_len)) {
+                       ret = -EFAULT;
+                       goto out;
+               }
+               line += iv[i].iov_len;
+       }
+
+       /*
+        * Extract and skip the syslog prefix <[0-9]*>. Coming from userspace
+        * the decimal value represents 32bit, the lower 3 bit are the log
+        * level, the rest are the log facility.
+        *
+        * If no prefix or no userspace facility is specified, we
+        * enforce LOG_USER, to be able to reliably distinguish
+        * kernel-generated messages from userspace-injected ones.
+        */
+       line = buf;
+       if (line[0] == '<') {
+               char *endp = NULL;
+
+               i = simple_strtoul(line+1, &endp, 10);
+               if (endp && endp[0] == '>') {
+                       level = i & 7;
+                       if (i >> 3)
+                               facility = i >> 3;
+                       endp++;
+                       len -= endp - line;
+                       line = endp;
+               }
+       }
+       line[len] = '\0';
+
+       printk_emit(facility, level, NULL, 0, "%s", line);
+out:
+       kfree(buf);
+       return ret;
+}
+
+static ssize_t devkmsg_read(struct file *file, char __user *buf,
+                           size_t count, loff_t *ppos)
+{
+       struct devkmsg_user *user = file->private_data;
+       struct log *msg;
+       u64 ts_usec;
+       size_t i;
+       char cont = '-';
+       size_t len;
+       ssize_t ret;
+
+       if (!user)
+               return -EBADF;
+
+       ret = mutex_lock_interruptible(&user->lock);
+       if (ret)
+               return ret;
+       raw_spin_lock_irq(&logbuf_lock);
+       while (user->seq == log_next_seq) {
+               if (file->f_flags & O_NONBLOCK) {
+                       ret = -EAGAIN;
+                       raw_spin_unlock_irq(&logbuf_lock);
+                       goto out;
+               }
+
+               raw_spin_unlock_irq(&logbuf_lock);
+               ret = wait_event_interruptible(log_wait,
+                                              user->seq != log_next_seq);
+               if (ret)
+                       goto out;
+               raw_spin_lock_irq(&logbuf_lock);
+       }
+
+       if (user->seq < log_first_seq) {
+               /* our last seen message is gone, return error and reset */
+               user->idx = log_first_idx;
+               user->seq = log_first_seq;
+               ret = -EPIPE;
+               raw_spin_unlock_irq(&logbuf_lock);
+               goto out;
+       }
+
+       msg = log_from_idx(user->idx);
+       ts_usec = msg->ts_nsec;
+       do_div(ts_usec, 1000);
+
+       /*
+        * If we couldn't merge continuation line fragments during the print,
+        * export the stored flags to allow an optional external merge of the
+        * records. Merging the records isn't always neccessarily correct, like
+        * when we hit a race during printing. In most cases though, it produces
+        * better readable output. 'c' in the record flags mark the first
+        * fragment of a line, '+' the following.
+        */
+       if (msg->flags & LOG_CONT && !(user->prev & LOG_CONT))
+               cont = 'c';
+       else if ((msg->flags & LOG_CONT) ||
+                ((user->prev & LOG_CONT) && !(msg->flags & LOG_PREFIX)))
+               cont = '+';
+
+       len = sprintf(user->buf, "%u,%llu,%llu,%c;",
+                     (msg->facility << 3) | msg->level,
+                     user->seq, ts_usec, cont);
+       user->prev = msg->flags;
+
+       /* escape non-printable characters */
+       for (i = 0; i < msg->text_len; i++) {
+               unsigned char c = log_text(msg)[i];
+
+               if (c < ' ' || c >= 127 || c == '\\')
+                       len += sprintf(user->buf + len, "\\x%02x", c);
+               else
+                       user->buf[len++] = c;
+       }
+       user->buf[len++] = '\n';
+
+       if (msg->dict_len) {
+               bool line = true;
+
+               for (i = 0; i < msg->dict_len; i++) {
+                       unsigned char c = log_dict(msg)[i];
+
+                       if (line) {
+                               user->buf[len++] = ' ';
+                               line = false;
+                       }
+
+                       if (c == '\0') {
+                               user->buf[len++] = '\n';
+                               line = true;
+                               continue;
+                       }
+
+                       if (c < ' ' || c >= 127 || c == '\\') {
+                               len += sprintf(user->buf + len, "\\x%02x", c);
+                               continue;
+                       }
+
+                       user->buf[len++] = c;
+               }
+               user->buf[len++] = '\n';
+       }
+
+       user->idx = log_next(user->idx);
+       user->seq++;
+       raw_spin_unlock_irq(&logbuf_lock);
+
+       if (len > count) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       if (copy_to_user(buf, user->buf, len)) {
+               ret = -EFAULT;
+               goto out;
+       }
+       ret = len;
+out:
+       mutex_unlock(&user->lock);
+       return ret;
+}
+
+static loff_t devkmsg_llseek(struct file *file, loff_t offset, int whence)
+{
+       struct devkmsg_user *user = file->private_data;
+       loff_t ret = 0;
+
+       if (!user)
+               return -EBADF;
+       if (offset)
+               return -ESPIPE;
+
+       raw_spin_lock_irq(&logbuf_lock);
+       switch (whence) {
+       case SEEK_SET:
+               /* the first record */
+               user->idx = log_first_idx;
+               user->seq = log_first_seq;
+               break;
+       case SEEK_DATA:
+               /*
+                * The first record after the last SYSLOG_ACTION_CLEAR,
+                * like issued by 'dmesg -c'. Reading /dev/kmsg itself
+                * changes no global state, and does not clear anything.
+                */
+               user->idx = clear_idx;
+               user->seq = clear_seq;
+               break;
+       case SEEK_END:
+               /* after the last record */
+               user->idx = log_next_idx;
+               user->seq = log_next_seq;
+               break;
+       default:
+               ret = -EINVAL;
+       }
+       raw_spin_unlock_irq(&logbuf_lock);
+       return ret;
+}
+
+static unsigned int devkmsg_poll(struct file *file, poll_table *wait)
+{
+       struct devkmsg_user *user = file->private_data;
+       int ret = 0;
+
+       if (!user)
+               return POLLERR|POLLNVAL;
+
+       poll_wait(file, &log_wait, wait);
+
+       raw_spin_lock_irq(&logbuf_lock);
+       if (user->seq < log_next_seq) {
+               /* return error when data has vanished underneath us */
+               if (user->seq < log_first_seq)
+                       ret = POLLIN|POLLRDNORM|POLLERR|POLLPRI;
+               else
+                       ret = POLLIN|POLLRDNORM;
+       }
+       raw_spin_unlock_irq(&logbuf_lock);
+
+       return ret;
+}
+
+static int devkmsg_open(struct inode *inode, struct file *file)
+{
+       struct devkmsg_user *user;
+       int err;
+
+       /* write-only does not need any file context */
+       if ((file->f_flags & O_ACCMODE) == O_WRONLY)
+               return 0;
+
+       err = check_syslog_permissions(SYSLOG_ACTION_READ_ALL,
+                                      SYSLOG_FROM_READER);
+       if (err)
+               return err;
+
+       user = kmalloc(sizeof(struct devkmsg_user), GFP_KERNEL);
+       if (!user)
+               return -ENOMEM;
+
+       mutex_init(&user->lock);
+
+       raw_spin_lock_irq(&logbuf_lock);
+       user->idx = log_first_idx;
+       user->seq = log_first_seq;
+       raw_spin_unlock_irq(&logbuf_lock);
+
+       file->private_data = user;
+       return 0;
+}
+
+static int devkmsg_release(struct inode *inode, struct file *file)
+{
+       struct devkmsg_user *user = file->private_data;
+
+       if (!user)
+               return 0;
+
+       mutex_destroy(&user->lock);
+       kfree(user);
+       return 0;
+}
+
+const struct file_operations kmsg_fops = {
+       .open = devkmsg_open,
+       .read = devkmsg_read,
+       .aio_write = devkmsg_writev,
+       .llseek = devkmsg_llseek,
+       .poll = devkmsg_poll,
+       .release = devkmsg_release,
+};
+
+#ifdef CONFIG_KEXEC
+/*
+ * This appends the listed symbols to /proc/vmcoreinfo
+ *
+ * /proc/vmcoreinfo is used by various utiilties, like crash and makedumpfile to
+ * obtain access to symbols that are otherwise very difficult to locate.  These
+ * symbols are specifically used so that utilities can access and extract the
+ * dmesg log from a vmcore file after a crash.
+ */
+void log_buf_kexec_setup(void)
+{
+       VMCOREINFO_SYMBOL(log_buf);
+       VMCOREINFO_SYMBOL(log_buf_len);
+       VMCOREINFO_SYMBOL(log_first_idx);
+       VMCOREINFO_SYMBOL(log_next_idx);
+       /*
+        * Export struct log size and field offsets. User space tools can
+        * parse it and detect any changes to structure down the line.
+        */
+       VMCOREINFO_STRUCT_SIZE(log);
+       VMCOREINFO_OFFSET(log, ts_nsec);
+       VMCOREINFO_OFFSET(log, len);
+       VMCOREINFO_OFFSET(log, text_len);
+       VMCOREINFO_OFFSET(log, dict_len);
+}
+#endif
+
+/* requested log_buf_len from kernel cmdline */
+static unsigned long __initdata new_log_buf_len;
+
+/* save requested log_buf_len since it's too early to process it */
+static int __init log_buf_len_setup(char *str)
+{
+       unsigned size = memparse(str, &str);
+
+       if (size)
+               size = roundup_pow_of_two(size);
+       if (size > log_buf_len)
+               new_log_buf_len = size;
+
+       return 0;
+}
+early_param("log_buf_len", log_buf_len_setup);
+
+void __init setup_log_buf(int early)
+{
+       unsigned long flags;
+       char *new_log_buf;
+       int free;
+
+       if (!new_log_buf_len)
+               return;
+
+       if (early) {
+               unsigned long mem;
+
+               mem = memblock_alloc(new_log_buf_len, PAGE_SIZE);
+               if (!mem)
+                       return;
+               new_log_buf = __va(mem);
+       } else {
+               new_log_buf = alloc_bootmem_nopanic(new_log_buf_len);
+       }
+
+       if (unlikely(!new_log_buf)) {
+               pr_err("log_buf_len: %ld bytes not available\n",
+                       new_log_buf_len);
+               return;
+       }
+
+       raw_spin_lock_irqsave(&logbuf_lock, flags);
+       log_buf_len = new_log_buf_len;
+       log_buf = new_log_buf;
+       new_log_buf_len = 0;
+       free = __LOG_BUF_LEN - log_next_idx;
+       memcpy(log_buf, __log_buf, __LOG_BUF_LEN);
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+
+       pr_info("log_buf_len: %d\n", log_buf_len);
+       pr_info("early log buf free: %d(%d%%)\n",
+               free, (free * 100) / __LOG_BUF_LEN);
+}
+
+static bool __read_mostly ignore_loglevel;
+
+static int __init ignore_loglevel_setup(char *str)
+{
+       ignore_loglevel = 1;
+       printk(KERN_INFO "debug: ignoring loglevel setting.\n");
+
+       return 0;
+}
+
+early_param("ignore_loglevel", ignore_loglevel_setup);
+module_param(ignore_loglevel, bool, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(ignore_loglevel, "ignore loglevel setting, to"
+       "print all kernel messages to the console.");
+
+#ifdef CONFIG_BOOT_PRINTK_DELAY
+
+static int boot_delay; /* msecs delay after each printk during bootup */
+static unsigned long long loops_per_msec;      /* based on boot_delay */
+
+static int __init boot_delay_setup(char *str)
+{
+       unsigned long lpj;
+
+       lpj = preset_lpj ? preset_lpj : 1000000;        /* some guess */
+       loops_per_msec = (unsigned long long)lpj / 1000 * HZ;
+
+       get_option(&str, &boot_delay);
+       if (boot_delay > 10 * 1000)
+               boot_delay = 0;
+
+       pr_debug("boot_delay: %u, preset_lpj: %ld, lpj: %lu, "
+               "HZ: %d, loops_per_msec: %llu\n",
+               boot_delay, preset_lpj, lpj, HZ, loops_per_msec);
+       return 1;
+}
+__setup("boot_delay=", boot_delay_setup);
+
+static void boot_delay_msec(int level)
+{
+       unsigned long long k;
+       unsigned long timeout;
+
+       if ((boot_delay == 0 || system_state != SYSTEM_BOOTING)
+               || (level >= console_loglevel && !ignore_loglevel)) {
+               return;
+       }
+
+       k = (unsigned long long)loops_per_msec * boot_delay;
+
+       timeout = jiffies + msecs_to_jiffies(boot_delay);
+       while (k) {
+               k--;
+               cpu_relax();
+               /*
+                * use (volatile) jiffies to prevent
+                * compiler reduction; loop termination via jiffies
+                * is secondary and may or may not happen.
+                */
+               if (time_after(jiffies, timeout))
+                       break;
+               touch_nmi_watchdog();
+       }
+}
+#else
+static inline void boot_delay_msec(int level)
+{
+}
+#endif
+
+#if defined(CONFIG_PRINTK_TIME)
+static bool printk_time = 1;
+#else
+static bool printk_time;
+#endif
+module_param_named(time, printk_time, bool, S_IRUGO | S_IWUSR);
+
+static size_t print_time(u64 ts, char *buf)
+{
+       unsigned long rem_nsec;
+
+       if (!printk_time)
+               return 0;
+
+       rem_nsec = do_div(ts, 1000000000);
+
+       if (!buf)
+               return snprintf(NULL, 0, "[%5lu.000000] ", (unsigned long)ts);
+
+       return sprintf(buf, "[%5lu.%06lu] ",
+                      (unsigned long)ts, rem_nsec / 1000);
+}
+
+static size_t print_prefix(const struct log *msg, bool syslog, char *buf)
+{
+       size_t len = 0;
+       unsigned int prefix = (msg->facility << 3) | msg->level;
+
+       if (syslog) {
+               if (buf) {
+                       len += sprintf(buf, "<%u>", prefix);
+               } else {
+                       len += 3;
+                       if (prefix > 999)
+                               len += 3;
+                       else if (prefix > 99)
+                               len += 2;
+                       else if (prefix > 9)
+                               len++;
+               }
+       }
+
+       len += print_time(msg->ts_nsec, buf ? buf + len : NULL);
+       return len;
+}
+
+static size_t msg_print_text(const struct log *msg, enum log_flags prev,
+                            bool syslog, char *buf, size_t size)
+{
+       const char *text = log_text(msg);
+       size_t text_size = msg->text_len;
+       bool prefix = true;
+       bool newline = true;
+       size_t len = 0;
+
+       if ((prev & LOG_CONT) && !(msg->flags & LOG_PREFIX))
+               prefix = false;
+
+       if (msg->flags & LOG_CONT) {
+               if ((prev & LOG_CONT) && !(prev & LOG_NEWLINE))
+                       prefix = false;
+
+               if (!(msg->flags & LOG_NEWLINE))
+                       newline = false;
+       }
+
+       do {
+               const char *next = memchr(text, '\n', text_size);
+               size_t text_len;
+
+               if (next) {
+                       text_len = next - text;
+                       next++;
+                       text_size -= next - text;
+               } else {
+                       text_len = text_size;
+               }
+
+               if (buf) {
+                       if (print_prefix(msg, syslog, NULL) +
+                           text_len + 1 >= size - len)
+                               break;
+
+                       if (prefix)
+                               len += print_prefix(msg, syslog, buf + len);
+                       memcpy(buf + len, text, text_len);
+                       len += text_len;
+                       if (next || newline)
+                               buf[len++] = '\n';
+               } else {
+                       /* SYSLOG_ACTION_* buffer size only calculation */
+                       if (prefix)
+                               len += print_prefix(msg, syslog, NULL);
+                       len += text_len;
+                       if (next || newline)
+                               len++;
+               }
+
+               prefix = true;
+               text = next;
+       } while (text);
+
+       return len;
+}
+
+static int syslog_print(char __user *buf, int size)
+{
+       char *text;
+       struct log *msg;
+       int len = 0;
+
+       text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
+       if (!text)
+               return -ENOMEM;
+
+       while (size > 0) {
+               size_t n;
+               size_t skip;
+
+               raw_spin_lock_irq(&logbuf_lock);
+               if (syslog_seq < log_first_seq) {
+                       /* messages are gone, move to first one */
+                       syslog_seq = log_first_seq;
+                       syslog_idx = log_first_idx;
+                       syslog_prev = 0;
+                       syslog_partial = 0;
+               }
+               if (syslog_seq == log_next_seq) {
+                       raw_spin_unlock_irq(&logbuf_lock);
+                       break;
+               }
+
+               skip = syslog_partial;
+               msg = log_from_idx(syslog_idx);
+               n = msg_print_text(msg, syslog_prev, true, text,
+                                  LOG_LINE_MAX + PREFIX_MAX);
+               if (n - syslog_partial <= size) {
+                       /* message fits into buffer, move forward */
+                       syslog_idx = log_next(syslog_idx);
+                       syslog_seq++;
+                       syslog_prev = msg->flags;
+                       n -= syslog_partial;
+                       syslog_partial = 0;
+               } else if (!len){
+                       /* partial read(), remember position */
+                       n = size;
+                       syslog_partial += n;
+               } else
+                       n = 0;
+               raw_spin_unlock_irq(&logbuf_lock);
+
+               if (!n)
+                       break;
+
+               if (copy_to_user(buf, text + skip, n)) {
+                       if (!len)
+                               len = -EFAULT;
+                       break;
+               }
+
+               len += n;
+               size -= n;
+               buf += n;
+       }
+
+       kfree(text);
+       return len;
+}
+
+static int syslog_print_all(char __user *buf, int size, bool clear)
+{
+       char *text;
+       int len = 0;
+
+       text = kmalloc(LOG_LINE_MAX + PREFIX_MAX, GFP_KERNEL);
+       if (!text)
+               return -ENOMEM;
+
+       raw_spin_lock_irq(&logbuf_lock);
+       if (buf) {
+               u64 next_seq;
+               u64 seq;
+               u32 idx;
+               enum log_flags prev;
+
+               if (clear_seq < log_first_seq) {
+                       /* messages are gone, move to first available one */
+                       clear_seq = log_first_seq;
+                       clear_idx = log_first_idx;
+               }
+
+               /*
+                * Find first record that fits, including all following records,
+                * into the user-provided buffer for this dump.
+                */
+               seq = clear_seq;
+               idx = clear_idx;
+               prev = 0;
+               while (seq < log_next_seq) {
+                       struct log *msg = log_from_idx(idx);
+
+                       len += msg_print_text(msg, prev, true, NULL, 0);
+                       prev = msg->flags;
+                       idx = log_next(idx);
+                       seq++;
+               }
+
+               /* move first record forward until length fits into the buffer */
+               seq = clear_seq;
+               idx = clear_idx;
+               prev = 0;
+               while (len > size && seq < log_next_seq) {
+                       struct log *msg = log_from_idx(idx);
+
+                       len -= msg_print_text(msg, prev, true, NULL, 0);
+                       prev = msg->flags;
+                       idx = log_next(idx);
+                       seq++;
+               }
+
+               /* last message fitting into this dump */
+               next_seq = log_next_seq;
+
+               len = 0;
+               prev = 0;
+               while (len >= 0 && seq < next_seq) {
+                       struct log *msg = log_from_idx(idx);
+                       int textlen;
+
+                       textlen = msg_print_text(msg, prev, true, text,
+                                                LOG_LINE_MAX + PREFIX_MAX);
+                       if (textlen < 0) {
+                               len = textlen;
+                               break;
+                       }
+                       idx = log_next(idx);
+                       seq++;
+                       prev = msg->flags;
+
+                       raw_spin_unlock_irq(&logbuf_lock);
+                       if (copy_to_user(buf + len, text, textlen))
+                               len = -EFAULT;
+                       else
+                               len += textlen;
+                       raw_spin_lock_irq(&logbuf_lock);
+
+                       if (seq < log_first_seq) {
+                               /* messages are gone, move to next one */
+                               seq = log_first_seq;
+                               idx = log_first_idx;
+                               prev = 0;
+                       }
+               }
+       }
+
+       if (clear) {
+               clear_seq = log_next_seq;
+               clear_idx = log_next_idx;
+       }
+       raw_spin_unlock_irq(&logbuf_lock);
+
+       kfree(text);
+       return len;
+}
+
+int do_syslog(int type, char __user *buf, int len, bool from_file)
+{
+       bool clear = false;
+       static int saved_console_loglevel = -1;
+       int error;
+
+       error = check_syslog_permissions(type, from_file);
+       if (error)
+               goto out;
+
+       error = security_syslog(type);
+       if (error)
+               return error;
+
+       switch (type) {
+       case SYSLOG_ACTION_CLOSE:       /* Close log */
+               break;
+       case SYSLOG_ACTION_OPEN:        /* Open log */
+               break;
+       case SYSLOG_ACTION_READ:        /* Read from log */
+               error = -EINVAL;
+               if (!buf || len < 0)
+                       goto out;
+               error = 0;
+               if (!len)
+                       goto out;
+               if (!access_ok(VERIFY_WRITE, buf, len)) {
+                       error = -EFAULT;
+                       goto out;
+               }
+               error = wait_event_interruptible(log_wait,
+                                                syslog_seq != log_next_seq);
+               if (error)
+                       goto out;
+               error = syslog_print(buf, len);
+               break;
+       /* Read/clear last kernel messages */
+       case SYSLOG_ACTION_READ_CLEAR:
+               clear = true;
+               /* FALL THRU */
+       /* Read last kernel messages */
+       case SYSLOG_ACTION_READ_ALL:
+               error = -EINVAL;
+               if (!buf || len < 0)
+                       goto out;
+               error = 0;
+               if (!len)
+                       goto out;
+               if (!access_ok(VERIFY_WRITE, buf, len)) {
+                       error = -EFAULT;
+                       goto out;
+               }
+               error = syslog_print_all(buf, len, clear);
+               break;
+       /* Clear ring buffer */
+       case SYSLOG_ACTION_CLEAR:
+               syslog_print_all(NULL, 0, true);
+               break;
+       /* Disable logging to console */
+       case SYSLOG_ACTION_CONSOLE_OFF:
+               if (saved_console_loglevel == -1)
+                       saved_console_loglevel = console_loglevel;
+               console_loglevel = minimum_console_loglevel;
+               break;
+       /* Enable logging to console */
+       case SYSLOG_ACTION_CONSOLE_ON:
+               if (saved_console_loglevel != -1) {
+                       console_loglevel = saved_console_loglevel;
+                       saved_console_loglevel = -1;
+               }
+               break;
+       /* Set level of messages printed to console */
+       case SYSLOG_ACTION_CONSOLE_LEVEL:
+               error = -EINVAL;
+               if (len < 1 || len > 8)
+                       goto out;
+               if (len < minimum_console_loglevel)
+                       len = minimum_console_loglevel;
+               console_loglevel = len;
+               /* Implicitly re-enable logging to console */
+               saved_console_loglevel = -1;
+               error = 0;
+               break;
+       /* Number of chars in the log buffer */
+       case SYSLOG_ACTION_SIZE_UNREAD:
+               raw_spin_lock_irq(&logbuf_lock);
+               if (syslog_seq < log_first_seq) {
+                       /* messages are gone, move to first one */
+                       syslog_seq = log_first_seq;
+                       syslog_idx = log_first_idx;
+                       syslog_prev = 0;
+                       syslog_partial = 0;
+               }
+               if (from_file) {
+                       /*
+                        * Short-cut for poll(/"proc/kmsg") which simply checks
+                        * for pending data, not the size; return the count of
+                        * records, not the length.
+                        */
+                       error = log_next_idx - syslog_idx;
+               } else {
+                       u64 seq = syslog_seq;
+                       u32 idx = syslog_idx;
+                       enum log_flags prev = syslog_prev;
+
+                       error = 0;
+                       while (seq < log_next_seq) {
+                               struct log *msg = log_from_idx(idx);
+
+                               error += msg_print_text(msg, prev, true, NULL, 0);
+                               idx = log_next(idx);
+                               seq++;
+                               prev = msg->flags;
+                       }
+                       error -= syslog_partial;
+               }
+               raw_spin_unlock_irq(&logbuf_lock);
+               break;
+       /* Size of the log buffer */
+       case SYSLOG_ACTION_SIZE_BUFFER:
+               error = log_buf_len;
+               break;
+       default:
+               error = -EINVAL;
+               break;
+       }
+out:
+       return error;
+}
+
+SYSCALL_DEFINE3(syslog, int, type, char __user *, buf, int, len)
+{
+       return do_syslog(type, buf, len, SYSLOG_FROM_READER);
+}
+
+/*
+ * Call the console drivers, asking them to write out
+ * log_buf[start] to log_buf[end - 1].
+ * The console_lock must be held.
+ */
+static void call_console_drivers(int level, const char *text, size_t len)
+{
+       struct console *con;
+
+       trace_console(text, len);
+
+       if (level >= console_loglevel && !ignore_loglevel)
+               return;
+       if (!console_drivers)
+               return;
+
+       for_each_console(con) {
+               if (exclusive_console && con != exclusive_console)
+                       continue;
+               if (!(con->flags & CON_ENABLED))
+                       continue;
+               if (!con->write)
+                       continue;
+               if (!cpu_online(smp_processor_id()) &&
+                   !(con->flags & CON_ANYTIME))
+                       continue;
+               con->write(con, text, len);
+       }
+}
+
+/*
+ * Zap console related locks when oopsing. Only zap at most once
+ * every 10 seconds, to leave time for slow consoles to print a
+ * full oops.
+ */
+static void zap_locks(void)
+{
+       static unsigned long oops_timestamp;
+
+       if (time_after_eq(jiffies, oops_timestamp) &&
+                       !time_after(jiffies, oops_timestamp + 30 * HZ))
+               return;
+
+       oops_timestamp = jiffies;
+
+       debug_locks_off();
+       /* If a crash is occurring, make sure we can't deadlock */
+       raw_spin_lock_init(&logbuf_lock);
+       /* And make sure that we print immediately */
+       sema_init(&console_sem, 1);
+}
+
+/* Check if we have any console registered that can be called early in boot. */
+static int have_callable_console(void)
+{
+       struct console *con;
+
+       for_each_console(con)
+               if (con->flags & CON_ANYTIME)
+                       return 1;
+
+       return 0;
+}
+
+/*
+ * Can we actually use the console at this time on this cpu?
+ *
+ * Console drivers may assume that per-cpu resources have
+ * been allocated. So unless they're explicitly marked as
+ * being able to cope (CON_ANYTIME) don't call them until
+ * this CPU is officially up.
+ */
+static inline int can_use_console(unsigned int cpu)
+{
+       return cpu_online(cpu) || have_callable_console();
+}
+
+/*
+ * Try to get console ownership to actually show the kernel
+ * messages from a 'printk'. Return true (and with the
+ * console_lock held, and 'console_locked' set) if it
+ * is successful, false otherwise.
+ *
+ * This gets called with the 'logbuf_lock' spinlock held and
+ * interrupts disabled. It should return with 'lockbuf_lock'
+ * released but interrupts still disabled.
+ */
+static int console_trylock_for_printk(unsigned int cpu)
+       __releases(&logbuf_lock)
+{
+       int retval = 0, wake = 0;
+
+       if (console_trylock()) {
+               retval = 1;
+
+               /*
+                * If we can't use the console, we need to release
+                * the console semaphore by hand to avoid flushing
+                * the buffer. We need to hold the console semaphore
+                * in order to do this test safely.
+                */
+               if (!can_use_console(cpu)) {
+                       console_locked = 0;
+                       wake = 1;
+                       retval = 0;
+               }
+       }
+       logbuf_cpu = UINT_MAX;
+       raw_spin_unlock(&logbuf_lock);
+       if (wake)
+               up(&console_sem);
+       return retval;
+}
+
+int printk_delay_msec __read_mostly;
+
+static inline void printk_delay(void)
+{
+       if (unlikely(printk_delay_msec)) {
+               int m = printk_delay_msec;
+
+               while (m--) {
+                       mdelay(1);
+                       touch_nmi_watchdog();
+               }
+       }
+}
+
+/*
+ * Continuation lines are buffered, and not committed to the record buffer
+ * until the line is complete, or a race forces it. The line fragments
+ * though, are printed immediately to the consoles to ensure everything has
+ * reached the console in case of a kernel crash.
+ */
+static struct cont {
+       char buf[LOG_LINE_MAX];
+       size_t len;                     /* length == 0 means unused buffer */
+       size_t cons;                    /* bytes written to console */
+       struct task_struct *owner;      /* task of first print*/
+       u64 ts_nsec;                    /* time of first print */
+       u8 level;                       /* log level of first message */
+       u8 facility;                    /* log level of first message */
+       enum log_flags flags;           /* prefix, newline flags */
+       bool flushed:1;                 /* buffer sealed and committed */
+} cont;
+
+static void cont_flush(enum log_flags flags)
+{
+       if (cont.flushed)
+               return;
+       if (cont.len == 0)
+               return;
+
+       if (cont.cons) {
+               /*
+                * If a fragment of this line was directly flushed to the
+                * console; wait for the console to pick up the rest of the
+                * line. LOG_NOCONS suppresses a duplicated output.
+                */
+               log_store(cont.facility, cont.level, flags | LOG_NOCONS,
+                         cont.ts_nsec, NULL, 0, cont.buf, cont.len);
+               cont.flags = flags;
+               cont.flushed = true;
+       } else {
+               /*
+                * If no fragment of this line ever reached the console,
+                * just submit it to the store and free the buffer.
+                */
+               log_store(cont.facility, cont.level, flags, 0,
+                         NULL, 0, cont.buf, cont.len);
+               cont.len = 0;
+       }
+}
+
+static bool cont_add(int facility, int level, const char *text, size_t len)
+{
+       if (cont.len && cont.flushed)
+               return false;
+
+       if (cont.len + len > sizeof(cont.buf)) {
+               /* the line gets too long, split it up in separate records */
+               cont_flush(LOG_CONT);
+               return false;
+       }
+
+       if (!cont.len) {
+               cont.facility = facility;
+               cont.level = level;
+               cont.owner = current;
+               cont.ts_nsec = local_clock();
+               cont.flags = 0;
+               cont.cons = 0;
+               cont.flushed = false;
+       }
+
+       memcpy(cont.buf + cont.len, text, len);
+       cont.len += len;
+
+       if (cont.len > (sizeof(cont.buf) * 80) / 100)
+               cont_flush(LOG_CONT);
+
+       return true;
+}
+
+static size_t cont_print_text(char *text, size_t size)
+{
+       size_t textlen = 0;
+       size_t len;
+
+       if (cont.cons == 0 && (console_prev & LOG_NEWLINE)) {
+               textlen += print_time(cont.ts_nsec, text);
+               size -= textlen;
+       }
+
+       len = cont.len - cont.cons;
+       if (len > 0) {
+               if (len+1 > size)
+                       len = size-1;
+               memcpy(text + textlen, cont.buf + cont.cons, len);
+               textlen += len;
+               cont.cons = cont.len;
+       }
+
+       if (cont.flushed) {
+               if (cont.flags & LOG_NEWLINE)
+                       text[textlen++] = '\n';
+               /* got everything, release buffer */
+               cont.len = 0;
+       }
+       return textlen;
+}
+
+asmlinkage int vprintk_emit(int facility, int level,
+                           const char *dict, size_t dictlen,
+                           const char *fmt, va_list args)
+{
+       static int recursion_bug;
+       static char textbuf[LOG_LINE_MAX];
+       char *text = textbuf;
+       size_t text_len;
+       enum log_flags lflags = 0;
+       unsigned long flags;
+       int this_cpu;
+       int printed_len = 0;
+
+       boot_delay_msec(level);
+       printk_delay();
+
+       /* This stops the holder of console_sem just where we want him */
+       local_irq_save(flags);
+       this_cpu = smp_processor_id();
+
+       /*
+        * Ouch, printk recursed into itself!
+        */
+       if (unlikely(logbuf_cpu == this_cpu)) {
+               /*
+                * If a crash is occurring during printk() on this CPU,
+                * then try to get the crash message out but make sure
+                * we can't deadlock. Otherwise just return to avoid the
+                * recursion and return - but flag the recursion so that
+                * it can be printed at the next appropriate moment:
+                */
+               if (!oops_in_progress && !lockdep_recursing(current)) {
+                       recursion_bug = 1;
+                       goto out_restore_irqs;
+               }
+               zap_locks();
+       }
+
+       lockdep_off();
+       raw_spin_lock(&logbuf_lock);
+       logbuf_cpu = this_cpu;
+
+       if (recursion_bug) {
+               static const char recursion_msg[] =
+                       "BUG: recent printk recursion!";
+
+               recursion_bug = 0;
+               printed_len += strlen(recursion_msg);
+               /* emit KERN_CRIT message */
+               log_store(0, 2, LOG_PREFIX|LOG_NEWLINE, 0,
+                         NULL, 0, recursion_msg, printed_len);
+       }
+
+       /*
+        * The printf needs to come first; we need the syslog
+        * prefix which might be passed-in as a parameter.
+        */
+       text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
+
+       /* mark and strip a trailing newline */
+       if (text_len && text[text_len-1] == '\n') {
+               text_len--;
+               lflags |= LOG_NEWLINE;
+       }
+
+       /* strip kernel syslog prefix and extract log level or control flags */
+       if (facility == 0) {
+               int kern_level = printk_get_level(text);
+
+               if (kern_level) {
+                       const char *end_of_header = printk_skip_level(text);
+                       switch (kern_level) {
+                       case '0' ... '7':
+                               if (level == -1)
+                                       level = kern_level - '0';
+                       case 'd':       /* KERN_DEFAULT */
+                               lflags |= LOG_PREFIX;
+                       case 'c':       /* KERN_CONT */
+                               break;
+                       }
+                       text_len -= end_of_header - text;
+                       text = (char *)end_of_header;
+               }
+       }
+
+       if (level == -1)
+               level = default_message_loglevel;
+
+       if (dict)
+               lflags |= LOG_PREFIX|LOG_NEWLINE;
+
+       if (!(lflags & LOG_NEWLINE)) {
+               /*
+                * Flush the conflicting buffer. An earlier newline was missing,
+                * or another task also prints continuation lines.
+                */
+               if (cont.len && (lflags & LOG_PREFIX || cont.owner != current))
+                       cont_flush(LOG_NEWLINE);
+
+               /* buffer line if possible, otherwise store it right away */
+               if (!cont_add(facility, level, text, text_len))
+                       log_store(facility, level, lflags | LOG_CONT, 0,
+                                 dict, dictlen, text, text_len);
+       } else {
+               bool stored = false;
+
+               /*
+                * If an earlier newline was missing and it was the same task,
+                * either merge it with the current buffer and flush, or if
+                * there was a race with interrupts (prefix == true) then just
+                * flush it out and store this line separately.
+                */
+               if (cont.len && cont.owner == current) {
+                       if (!(lflags & LOG_PREFIX))
+                               stored = cont_add(facility, level, text, text_len);
+                       cont_flush(LOG_NEWLINE);
+               }
+
+               if (!stored)
+                       log_store(facility, level, lflags, 0,
+                                 dict, dictlen, text, text_len);
+       }
+       printed_len += text_len;
+
+       /*
+        * Try to acquire and then immediately release the console semaphore.
+        * The release will print out buffers and wake up /dev/kmsg and syslog()
+        * users.
+        *
+        * The console_trylock_for_printk() function will release 'logbuf_lock'
+        * regardless of whether it actually gets the console semaphore or not.
+        */
+       if (console_trylock_for_printk(this_cpu))
+               console_unlock();
+
+       lockdep_on();
+out_restore_irqs:
+       local_irq_restore(flags);
+
+       return printed_len;
+}
+EXPORT_SYMBOL(vprintk_emit);
+
+asmlinkage int vprintk(const char *fmt, va_list args)
+{
+       return vprintk_emit(0, -1, NULL, 0, fmt, args);
+}
+EXPORT_SYMBOL(vprintk);
+
+asmlinkage int printk_emit(int facility, int level,
+                          const char *dict, size_t dictlen,
+                          const char *fmt, ...)
+{
+       va_list args;
+       int r;
+
+       va_start(args, fmt);
+       r = vprintk_emit(facility, level, dict, dictlen, fmt, args);
+       va_end(args);
+
+       return r;
+}
+EXPORT_SYMBOL(printk_emit);
+
+/**
+ * printk - print a kernel message
+ * @fmt: format string
+ *
+ * This is printk(). It can be called from any context. We want it to work.
+ *
+ * We try to grab the console_lock. If we succeed, it's easy - we log the
+ * output and call the console drivers.  If we fail to get the semaphore, we
+ * place the output into the log buffer and return. The current holder of
+ * the console_sem will notice the new output in console_unlock(); and will
+ * send it to the consoles before releasing the lock.
+ *
+ * One effect of this deferred printing is that code which calls printk() and
+ * then changes console_loglevel may break. This is because console_loglevel
+ * is inspected when the actual printing occurs.
+ *
+ * See also:
+ * printf(3)
+ *
+ * See the vsnprintf() documentation for format string extensions over C99.
+ */
+asmlinkage int printk(const char *fmt, ...)
+{
+       va_list args;
+       int r;
+
+#ifdef CONFIG_KGDB_KDB
+       if (unlikely(kdb_trap_printk)) {
+               va_start(args, fmt);
+               r = vkdb_printf(fmt, args);
+               va_end(args);
+               return r;
+       }
+#endif
+       va_start(args, fmt);
+       r = vprintk_emit(0, -1, NULL, 0, fmt, args);
+       va_end(args);
+
+       return r;
+}
+EXPORT_SYMBOL(printk);
+
+#else /* CONFIG_PRINTK */
+
+#define LOG_LINE_MAX           0
+#define PREFIX_MAX             0
+#define LOG_LINE_MAX 0
+static u64 syslog_seq;
+static u32 syslog_idx;
+static u64 console_seq;
+static u32 console_idx;
+static enum log_flags syslog_prev;
+static u64 log_first_seq;
+static u32 log_first_idx;
+static u64 log_next_seq;
+static enum log_flags console_prev;
+static struct cont {
+       size_t len;
+       size_t cons;
+       u8 level;
+       bool flushed:1;
+} cont;
+static struct log *log_from_idx(u32 idx) { return NULL; }
+static u32 log_next(u32 idx) { return 0; }
+static void call_console_drivers(int level, const char *text, size_t len) {}
+static size_t msg_print_text(const struct log *msg, enum log_flags prev,
+                            bool syslog, char *buf, size_t size) { return 0; }
+static size_t cont_print_text(char *text, size_t size) { return 0; }
+
+#endif /* CONFIG_PRINTK */
+
+#ifdef CONFIG_EARLY_PRINTK
+struct console *early_console;
+
+void early_vprintk(const char *fmt, va_list ap)
+{
+       if (early_console) {
+               char buf[512];
+               int n = vscnprintf(buf, sizeof(buf), fmt, ap);
+
+               early_console->write(early_console, buf, n);
+       }
+}
+
+asmlinkage void early_printk(const char *fmt, ...)
+{
+       va_list ap;
+
+       va_start(ap, fmt);
+       early_vprintk(fmt, ap);
+       va_end(ap);
+}
+#endif
+
+static int __add_preferred_console(char *name, int idx, char *options,
+                                  char *brl_options)
+{
+       struct console_cmdline *c;
+       int i;
+
+       /*
+        *      See if this tty is not yet registered, and
+        *      if we have a slot free.
+        */
+       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
+               if (strcmp(console_cmdline[i].name, name) == 0 &&
+                         console_cmdline[i].index == idx) {
+                               if (!brl_options)
+                                       selected_console = i;
+                               return 0;
+               }
+       if (i == MAX_CMDLINECONSOLES)
+               return -E2BIG;
+       if (!brl_options)
+               selected_console = i;
+       c = &console_cmdline[i];
+       strlcpy(c->name, name, sizeof(c->name));
+       c->options = options;
+#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
+       c->brl_options = brl_options;
+#endif
+       c->index = idx;
+       return 0;
+}
+/*
+ * Set up a list of consoles.  Called from init/main.c
+ */
+static int __init console_setup(char *str)
+{
+       char buf[sizeof(console_cmdline[0].name) + 4]; /* 4 for index */
+       char *s, *options, *brl_options = NULL;
+       int idx;
+
+#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
+       if (!memcmp(str, "brl,", 4)) {
+               brl_options = "";
+               str += 4;
+       } else if (!memcmp(str, "brl=", 4)) {
+               brl_options = str + 4;
+               str = strchr(brl_options, ',');
+               if (!str) {
+                       printk(KERN_ERR "need port name after brl=\n");
+                       return 1;
+               }
+               *(str++) = 0;
+       }
+#endif
+
+       /*
+        * Decode str into name, index, options.
+        */
+       if (str[0] >= '0' && str[0] <= '9') {
+               strcpy(buf, "ttyS");
+               strncpy(buf + 4, str, sizeof(buf) - 5);
+       } else {
+               strncpy(buf, str, sizeof(buf) - 1);
+       }
+       buf[sizeof(buf) - 1] = 0;
+       if ((options = strchr(str, ',')) != NULL)
+               *(options++) = 0;
+#ifdef __sparc__
+       if (!strcmp(str, "ttya"))
+               strcpy(buf, "ttyS0");
+       if (!strcmp(str, "ttyb"))
+               strcpy(buf, "ttyS1");
+#endif
+       for (s = buf; *s; s++)
+               if ((*s >= '0' && *s <= '9') || *s == ',')
+                       break;
+       idx = simple_strtoul(s, NULL, 10);
+       *s = 0;
+
+       __add_preferred_console(buf, idx, options, brl_options);
+       console_set_on_cmdline = 1;
+       return 1;
+}
+__setup("console=", console_setup);
+
+/**
+ * add_preferred_console - add a device to the list of preferred consoles.
+ * @name: device name
+ * @idx: device index
+ * @options: options for this console
+ *
+ * The last preferred console added will be used for kernel messages
+ * and stdin/out/err for init.  Normally this is used by console_setup
+ * above to handle user-supplied console arguments; however it can also
+ * be used by arch-specific code either to override the user or more
+ * commonly to provide a default console (ie from PROM variables) when
+ * the user has not supplied one.
+ */
+int add_preferred_console(char *name, int idx, char *options)
+{
+       return __add_preferred_console(name, idx, options, NULL);
+}
+
+int update_console_cmdline(char *name, int idx, char *name_new, int idx_new, char *options)
+{
+       struct console_cmdline *c;
+       int i;
+
+       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0]; i++)
+               if (strcmp(console_cmdline[i].name, name) == 0 &&
+                         console_cmdline[i].index == idx) {
+                               c = &console_cmdline[i];
+                               strlcpy(c->name, name_new, sizeof(c->name));
+                               c->name[sizeof(c->name) - 1] = 0;
+                               c->options = options;
+                               c->index = idx_new;
+                               return i;
+               }
+       /* not found */
+       return -1;
+}
+
+bool console_suspend_enabled = 1;
+EXPORT_SYMBOL(console_suspend_enabled);
+
+static int __init console_suspend_disable(char *str)
+{
+       console_suspend_enabled = 0;
+       return 1;
+}
+__setup("no_console_suspend", console_suspend_disable);
+module_param_named(console_suspend, console_suspend_enabled,
+               bool, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(console_suspend, "suspend console during suspend"
+       " and hibernate operations");
+
+/**
+ * suspend_console - suspend the console subsystem
+ *
+ * This disables printk() while we go into suspend states
+ */
+void suspend_console(void)
+{
+       if (!console_suspend_enabled)
+               return;
+       printk("Suspending console(s) (use no_console_suspend to debug)\n");
+       console_lock();
+       console_suspended = 1;
+       up(&console_sem);
+}
+
+void resume_console(void)
+{
+       if (!console_suspend_enabled)
+               return;
+       down(&console_sem);
+       console_suspended = 0;
+       console_unlock();
+}
+
+/**
+ * console_cpu_notify - print deferred console messages after CPU hotplug
+ * @self: notifier struct
+ * @action: CPU hotplug event
+ * @hcpu: unused
+ *
+ * If printk() is called from a CPU that is not online yet, the messages
+ * will be spooled but will not show up on the console.  This function is
+ * called when a new CPU comes online (or fails to come up), and ensures
+ * that any such output gets printed.
+ */
+static int console_cpu_notify(struct notifier_block *self,
+       unsigned long action, void *hcpu)
+{
+       switch (action) {
+       case CPU_ONLINE:
+       case CPU_DEAD:
+       case CPU_DOWN_FAILED:
+       case CPU_UP_CANCELED:
+               console_lock();
+               console_unlock();
+       }
+       return NOTIFY_OK;
+}
+
+/**
+ * console_lock - lock the console system for exclusive use.
+ *
+ * Acquires a lock which guarantees that the caller has
+ * exclusive access to the console system and the console_drivers list.
+ *
+ * Can sleep, returns nothing.
+ */
+void console_lock(void)
+{
+       might_sleep();
+
+       down(&console_sem);
+       if (console_suspended)
+               return;
+       console_locked = 1;
+       console_may_schedule = 1;
+       mutex_acquire(&console_lock_dep_map, 0, 0, _RET_IP_);
+}
+EXPORT_SYMBOL(console_lock);
+
+/**
+ * console_trylock - try to lock the console system for exclusive use.
+ *
+ * Tried to acquire a lock which guarantees that the caller has
+ * exclusive access to the console system and the console_drivers list.
+ *
+ * returns 1 on success, and 0 on failure to acquire the lock.
+ */
+int console_trylock(void)
+{
+       if (down_trylock(&console_sem))
+               return 0;
+       if (console_suspended) {
+               up(&console_sem);
+               return 0;
+       }
+       console_locked = 1;
+       console_may_schedule = 0;
+       mutex_acquire(&console_lock_dep_map, 0, 1, _RET_IP_);
+       return 1;
+}
+EXPORT_SYMBOL(console_trylock);
+
+int is_console_locked(void)
+{
+       return console_locked;
+}
+
+static void console_cont_flush(char *text, size_t size)
+{
+       unsigned long flags;
+       size_t len;
+
+       raw_spin_lock_irqsave(&logbuf_lock, flags);
+
+       if (!cont.len)
+               goto out;
+
+       /*
+        * We still queue earlier records, likely because the console was
+        * busy. The earlier ones need to be printed before this one, we
+        * did not flush any fragment so far, so just let it queue up.
+        */
+       if (console_seq < log_next_seq && !cont.cons)
+               goto out;
+
+       len = cont_print_text(text, size);
+       raw_spin_unlock(&logbuf_lock);
+       stop_critical_timings();
+       call_console_drivers(cont.level, text, len);
+       start_critical_timings();
+       local_irq_restore(flags);
+       return;
+out:
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+}
+
+/**
+ * console_unlock - unlock the console system
+ *
+ * Releases the console_lock which the caller holds on the console system
+ * and the console driver list.
+ *
+ * While the console_lock was held, console output may have been buffered
+ * by printk().  If this is the case, console_unlock(); emits
+ * the output prior to releasing the lock.
+ *
+ * If there is output waiting, we wake /dev/kmsg and syslog() users.
+ *
+ * console_unlock(); may be called from any context.
+ */
+void console_unlock(void)
+{
+       static char text[LOG_LINE_MAX + PREFIX_MAX];
+       static u64 seen_seq;
+       unsigned long flags;
+       bool wake_klogd = false;
+       bool retry;
+
+       if (console_suspended) {
+               up(&console_sem);
+               return;
+       }
+
+       console_may_schedule = 0;
+
+       /* flush buffered message fragment immediately to console */
+       console_cont_flush(text, sizeof(text));
+again:
+       for (;;) {
+               struct log *msg;
+               size_t len;
+               int level;
+
+               raw_spin_lock_irqsave(&logbuf_lock, flags);
+               if (seen_seq != log_next_seq) {
+                       wake_klogd = true;
+                       seen_seq = log_next_seq;
+               }
+
+               if (console_seq < log_first_seq) {
+                       /* messages are gone, move to first one */
+                       console_seq = log_first_seq;
+                       console_idx = log_first_idx;
+                       console_prev = 0;
+               }
+skip:
+               if (console_seq == log_next_seq)
+                       break;
+
+               msg = log_from_idx(console_idx);
+               if (msg->flags & LOG_NOCONS) {
+                       /*
+                        * Skip record we have buffered and already printed
+                        * directly to the console when we received it.
+                        */
+                       console_idx = log_next(console_idx);
+                       console_seq++;
+                       /*
+                        * We will get here again when we register a new
+                        * CON_PRINTBUFFER console. Clear the flag so we
+                        * will properly dump everything later.
+                        */
+                       msg->flags &= ~LOG_NOCONS;
+                       console_prev = msg->flags;
+                       goto skip;
+               }
+
+               level = msg->level;
+               len = msg_print_text(msg, console_prev, false,
+                                    text, sizeof(text));
+               console_idx = log_next(console_idx);
+               console_seq++;
+               console_prev = msg->flags;
+               raw_spin_unlock(&logbuf_lock);
+
+               stop_critical_timings();        /* don't trace print latency */
+               call_console_drivers(level, text, len);
+               start_critical_timings();
+               local_irq_restore(flags);
+       }
+       console_locked = 0;
+       mutex_release(&console_lock_dep_map, 1, _RET_IP_);
+
+       /* Release the exclusive_console once it is used */
+       if (unlikely(exclusive_console))
+               exclusive_console = NULL;
+
+       raw_spin_unlock(&logbuf_lock);
+
+       up(&console_sem);
+
+       /*
+        * Someone could have filled up the buffer again, so re-check if there's
+        * something to flush. In case we cannot trylock the console_sem again,
+        * there's a new owner and the console_unlock() from them will do the
+        * flush, no worries.
+        */
+       raw_spin_lock(&logbuf_lock);
+       retry = console_seq != log_next_seq;
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+
+       if (retry && console_trylock())
+               goto again;
+
+       if (wake_klogd)
+               wake_up_klogd();
+}
+EXPORT_SYMBOL(console_unlock);
+
+/**
+ * console_conditional_schedule - yield the CPU if required
+ *
+ * If the console code is currently allowed to sleep, and
+ * if this CPU should yield the CPU to another task, do
+ * so here.
+ *
+ * Must be called within console_lock();.
+ */
+void __sched console_conditional_schedule(void)
+{
+       if (console_may_schedule)
+               cond_resched();
+}
+EXPORT_SYMBOL(console_conditional_schedule);
+
+void console_unblank(void)
+{
+       struct console *c;
+
+       /*
+        * console_unblank can no longer be called in interrupt context unless
+        * oops_in_progress is set to 1..
+        */
+       if (oops_in_progress) {
+               if (down_trylock(&console_sem) != 0)
+                       return;
+       } else
+               console_lock();
+
+       console_locked = 1;
+       console_may_schedule = 0;
+       for_each_console(c)
+               if ((c->flags & CON_ENABLED) && c->unblank)
+                       c->unblank();
+       console_unlock();
+}
+
+/*
+ * Return the console tty driver structure and its associated index
+ */
+struct tty_driver *console_device(int *index)
+{
+       struct console *c;
+       struct tty_driver *driver = NULL;
+
+       console_lock();
+       for_each_console(c) {
+               if (!c->device)
+                       continue;
+               driver = c->device(c, index);
+               if (driver)
+                       break;
+       }
+       console_unlock();
+       return driver;
+}
+
+/*
+ * Prevent further output on the passed console device so that (for example)
+ * serial drivers can disable console output before suspending a port, and can
+ * re-enable output afterwards.
+ */
+void console_stop(struct console *console)
+{
+       console_lock();
+       console->flags &= ~CON_ENABLED;
+       console_unlock();
+}
+EXPORT_SYMBOL(console_stop);
+
+void console_start(struct console *console)
+{
+       console_lock();
+       console->flags |= CON_ENABLED;
+       console_unlock();
+}
+EXPORT_SYMBOL(console_start);
+
+static int __read_mostly keep_bootcon;
+
+static int __init keep_bootcon_setup(char *str)
+{
+       keep_bootcon = 1;
+       printk(KERN_INFO "debug: skip boot console de-registration.\n");
+
+       return 0;
+}
+
+early_param("keep_bootcon", keep_bootcon_setup);
+
+/*
+ * The console driver calls this routine during kernel initialization
+ * to register the console printing procedure with printk() and to
+ * print any messages that were printed by the kernel before the
+ * console driver was initialized.
+ *
+ * This can happen pretty early during the boot process (because of
+ * early_printk) - sometimes before setup_arch() completes - be careful
+ * of what kernel features are used - they may not be initialised yet.
+ *
+ * There are two types of consoles - bootconsoles (early_printk) and
+ * "real" consoles (everything which is not a bootconsole) which are
+ * handled differently.
+ *  - Any number of bootconsoles can be registered at any time.
+ *  - As soon as a "real" console is registered, all bootconsoles
+ *    will be unregistered automatically.
+ *  - Once a "real" console is registered, any attempt to register a
+ *    bootconsoles will be rejected
+ */
+void register_console(struct console *newcon)
+{
+       int i;
+       unsigned long flags;
+       struct console *bcon = NULL;
+
+       /*
+        * before we register a new CON_BOOT console, make sure we don't
+        * already have a valid console
+        */
+       if (console_drivers && newcon->flags & CON_BOOT) {
+               /* find the last or real console */
+               for_each_console(bcon) {
+                       if (!(bcon->flags & CON_BOOT)) {
+                               printk(KERN_INFO "Too late to register bootconsole %s%d\n",
+                                       newcon->name, newcon->index);
+                               return;
+                       }
+               }
+       }
+
+       if (console_drivers && console_drivers->flags & CON_BOOT)
+               bcon = console_drivers;
+
+       if (preferred_console < 0 || bcon || !console_drivers)
+               preferred_console = selected_console;
+
+       if (newcon->early_setup)
+               newcon->early_setup();
+
+       /*
+        *      See if we want to use this console driver. If we
+        *      didn't select a console we take the first one
+        *      that registers here.
+        */
+       if (preferred_console < 0) {
+               if (newcon->index < 0)
+                       newcon->index = 0;
+               if (newcon->setup == NULL ||
+                   newcon->setup(newcon, NULL) == 0) {
+                       newcon->flags |= CON_ENABLED;
+                       if (newcon->device) {
+                               newcon->flags |= CON_CONSDEV;
+                               preferred_console = 0;
+                       }
+               }
+       }
+
+       /*
+        *      See if this console matches one we selected on
+        *      the command line.
+        */
+       for (i = 0; i < MAX_CMDLINECONSOLES && console_cmdline[i].name[0];
+                       i++) {
+               if (strcmp(console_cmdline[i].name, newcon->name) != 0)
+                       continue;
+               if (newcon->index >= 0 &&
+                   newcon->index != console_cmdline[i].index)
+                       continue;
+               if (newcon->index < 0)
+                       newcon->index = console_cmdline[i].index;
+#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
+               if (console_cmdline[i].brl_options) {
+                       newcon->flags |= CON_BRL;
+                       braille_register_console(newcon,
+                                       console_cmdline[i].index,
+                                       console_cmdline[i].options,
+                                       console_cmdline[i].brl_options);
+                       return;
+               }
+#endif
+               if (newcon->setup &&
+                   newcon->setup(newcon, console_cmdline[i].options) != 0)
+                       break;
+               newcon->flags |= CON_ENABLED;
+               newcon->index = console_cmdline[i].index;
+               if (i == selected_console) {
+                       newcon->flags |= CON_CONSDEV;
+                       preferred_console = selected_console;
+               }
+               break;
+       }
+
+       if (!(newcon->flags & CON_ENABLED))
+               return;
+
+       /*
+        * If we have a bootconsole, and are switching to a real console,
+        * don't print everything out again, since when the boot console, and
+        * the real console are the same physical device, it's annoying to
+        * see the beginning boot messages twice
+        */
+       if (bcon && ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV))
+               newcon->flags &= ~CON_PRINTBUFFER;
+
+       /*
+        *      Put this console in the list - keep the
+        *      preferred driver at the head of the list.
+        */
+       console_lock();
+       if ((newcon->flags & CON_CONSDEV) || console_drivers == NULL) {
+               newcon->next = console_drivers;
+               console_drivers = newcon;
+               if (newcon->next)
+                       newcon->next->flags &= ~CON_CONSDEV;
+       } else {
+               newcon->next = console_drivers->next;
+               console_drivers->next = newcon;
+       }
+       if (newcon->flags & CON_PRINTBUFFER) {
+               /*
+                * console_unlock(); will print out the buffered messages
+                * for us.
+                */
+               raw_spin_lock_irqsave(&logbuf_lock, flags);
+               console_seq = syslog_seq;
+               console_idx = syslog_idx;
+               console_prev = syslog_prev;
+               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+               /*
+                * We're about to replay the log buffer.  Only do this to the
+                * just-registered console to avoid excessive message spam to
+                * the already-registered consoles.
+                */
+               exclusive_console = newcon;
+       }
+       console_unlock();
+       console_sysfs_notify();
+
+       /*
+        * By unregistering the bootconsoles after we enable the real console
+        * we get the "console xxx enabled" message on all the consoles -
+        * boot consoles, real consoles, etc - this is to ensure that end
+        * users know there might be something in the kernel's log buffer that
+        * went to the bootconsole (that they do not see on the real console)
+        */
+       if (bcon &&
+           ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
+           !keep_bootcon) {
+               /* we need to iterate through twice, to make sure we print
+                * everything out, before we unregister the console(s)
+                */
+               printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
+                       newcon->name, newcon->index);
+               for_each_console(bcon)
+                       if (bcon->flags & CON_BOOT)
+                               unregister_console(bcon);
+       } else {
+               printk(KERN_INFO "%sconsole [%s%d] enabled\n",
+                       (newcon->flags & CON_BOOT) ? "boot" : "" ,
+                       newcon->name, newcon->index);
+       }
+}
+EXPORT_SYMBOL(register_console);
+
+int unregister_console(struct console *console)
+{
+        struct console *a, *b;
+       int res = 1;
+
+#ifdef CONFIG_A11Y_BRAILLE_CONSOLE
+       if (console->flags & CON_BRL)
+               return braille_unregister_console(console);
+#endif
+
+       console_lock();
+       if (console_drivers == console) {
+               console_drivers=console->next;
+               res = 0;
+       } else if (console_drivers) {
+               for (a=console_drivers->next, b=console_drivers ;
+                    a; b=a, a=b->next) {
+                       if (a == console) {
+                               b->next = a->next;
+                               res = 0;
+                               break;
+                       }
+               }
+       }
+
+       /*
+        * If this isn't the last console and it has CON_CONSDEV set, we
+        * need to set it on the next preferred console.
+        */
+       if (console_drivers != NULL && console->flags & CON_CONSDEV)
+               console_drivers->flags |= CON_CONSDEV;
+
+       console_unlock();
+       console_sysfs_notify();
+       return res;
+}
+EXPORT_SYMBOL(unregister_console);
+
+static int __init printk_late_init(void)
+{
+       struct console *con;
+
+       for_each_console(con) {
+               if (!keep_bootcon && con->flags & CON_BOOT) {
+                       printk(KERN_INFO "turn off boot console %s%d\n",
+                               con->name, con->index);
+                       unregister_console(con);
+               }
+       }
+       hotcpu_notifier(console_cpu_notify, 0);
+       return 0;
+}
+late_initcall(printk_late_init);
+
+#if defined CONFIG_PRINTK
+/*
+ * Delayed printk version, for scheduler-internal messages:
+ */
+#define PRINTK_BUF_SIZE                512
+
+#define PRINTK_PENDING_WAKEUP  0x01
+#define PRINTK_PENDING_SCHED   0x02
+
+static DEFINE_PER_CPU(int, printk_pending);
+static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
+
+static void wake_up_klogd_work_func(struct irq_work *irq_work)
+{
+       int pending = __this_cpu_xchg(printk_pending, 0);
+
+       if (pending & PRINTK_PENDING_SCHED) {
+               char *buf = __get_cpu_var(printk_sched_buf);
+               printk(KERN_WARNING "[sched_delayed] %s", buf);
+       }
+
+       if (pending & PRINTK_PENDING_WAKEUP)
+               wake_up_interruptible(&log_wait);
+}
+
+static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) = {
+       .func = wake_up_klogd_work_func,
+       .flags = IRQ_WORK_LAZY,
+};
+
+void wake_up_klogd(void)
+{
+       preempt_disable();
+       if (waitqueue_active(&log_wait)) {
+               this_cpu_or(printk_pending, PRINTK_PENDING_WAKEUP);
+               irq_work_queue(&__get_cpu_var(wake_up_klogd_work));
+       }
+       preempt_enable();
+}
+
+int printk_sched(const char *fmt, ...)
+{
+       unsigned long flags;
+       va_list args;
+       char *buf;
+       int r;
+
+       local_irq_save(flags);
+       buf = __get_cpu_var(printk_sched_buf);
+
+       va_start(args, fmt);
+       r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
+       va_end(args);
+
+       __this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
+       irq_work_queue(&__get_cpu_var(wake_up_klogd_work));
+       local_irq_restore(flags);
+
+       return r;
+}
+
+/*
+ * printk rate limiting, lifted from the networking subsystem.
+ *
+ * This enforces a rate limit: not more than 10 kernel messages
+ * every 5s to make a denial-of-service attack impossible.
+ */
+DEFINE_RATELIMIT_STATE(printk_ratelimit_state, 5 * HZ, 10);
+
+int __printk_ratelimit(const char *func)
+{
+       return ___ratelimit(&printk_ratelimit_state, func);
+}
+EXPORT_SYMBOL(__printk_ratelimit);
+
+/**
+ * printk_timed_ratelimit - caller-controlled printk ratelimiting
+ * @caller_jiffies: pointer to caller's state
+ * @interval_msecs: minimum interval between prints
+ *
+ * printk_timed_ratelimit() returns true if more than @interval_msecs
+ * milliseconds have elapsed since the last time printk_timed_ratelimit()
+ * returned true.
+ */
+bool printk_timed_ratelimit(unsigned long *caller_jiffies,
+                       unsigned int interval_msecs)
+{
+       if (*caller_jiffies == 0
+                       || !time_in_range(jiffies, *caller_jiffies,
+                                       *caller_jiffies
+                                       + msecs_to_jiffies(interval_msecs))) {
+               *caller_jiffies = jiffies;
+               return true;
+       }
+       return false;
+}
+EXPORT_SYMBOL(printk_timed_ratelimit);
+
+static DEFINE_SPINLOCK(dump_list_lock);
+static LIST_HEAD(dump_list);
+
+/**
+ * kmsg_dump_register - register a kernel log dumper.
+ * @dumper: pointer to the kmsg_dumper structure
+ *
+ * Adds a kernel log dumper to the system. The dump callback in the
+ * structure will be called when the kernel oopses or panics and must be
+ * set. Returns zero on success and %-EINVAL or %-EBUSY otherwise.
+ */
+int kmsg_dump_register(struct kmsg_dumper *dumper)
+{
+       unsigned long flags;
+       int err = -EBUSY;
+
+       /* The dump callback needs to be set */
+       if (!dumper->dump)
+               return -EINVAL;
+
+       spin_lock_irqsave(&dump_list_lock, flags);
+       /* Don't allow registering multiple times */
+       if (!dumper->registered) {
+               dumper->registered = 1;
+               list_add_tail_rcu(&dumper->list, &dump_list);
+               err = 0;
+       }
+       spin_unlock_irqrestore(&dump_list_lock, flags);
+
+       return err;
+}
+EXPORT_SYMBOL_GPL(kmsg_dump_register);
+
+/**
+ * kmsg_dump_unregister - unregister a kmsg dumper.
+ * @dumper: pointer to the kmsg_dumper structure
+ *
+ * Removes a dump device from the system. Returns zero on success and
+ * %-EINVAL otherwise.
+ */
+int kmsg_dump_unregister(struct kmsg_dumper *dumper)
+{
+       unsigned long flags;
+       int err = -EINVAL;
+
+       spin_lock_irqsave(&dump_list_lock, flags);
+       if (dumper->registered) {
+               dumper->registered = 0;
+               list_del_rcu(&dumper->list);
+               err = 0;
+       }
+       spin_unlock_irqrestore(&dump_list_lock, flags);
+       synchronize_rcu();
+
+       return err;
+}
+EXPORT_SYMBOL_GPL(kmsg_dump_unregister);
+
+static bool always_kmsg_dump;
+module_param_named(always_kmsg_dump, always_kmsg_dump, bool, S_IRUGO | S_IWUSR);
+
+/**
+ * kmsg_dump - dump kernel log to kernel message dumpers.
+ * @reason: the reason (oops, panic etc) for dumping
+ *
+ * Call each of the registered dumper's dump() callback, which can
+ * retrieve the kmsg records with kmsg_dump_get_line() or
+ * kmsg_dump_get_buffer().
+ */
+void kmsg_dump(enum kmsg_dump_reason reason)
+{
+       struct kmsg_dumper *dumper;
+       unsigned long flags;
+
+       if ((reason > KMSG_DUMP_OOPS) && !always_kmsg_dump)
+               return;
+
+       rcu_read_lock();
+       list_for_each_entry_rcu(dumper, &dump_list, list) {
+               if (dumper->max_reason && reason > dumper->max_reason)
+                       continue;
+
+               /* initialize iterator with data about the stored records */
+               dumper->active = true;
+
+               raw_spin_lock_irqsave(&logbuf_lock, flags);
+               dumper->cur_seq = clear_seq;
+               dumper->cur_idx = clear_idx;
+               dumper->next_seq = log_next_seq;
+               dumper->next_idx = log_next_idx;
+               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+
+               /* invoke dumper which will iterate over records */
+               dumper->dump(dumper, reason);
+
+               /* reset iterator */
+               dumper->active = false;
+       }
+       rcu_read_unlock();
+}
+
+/**
+ * kmsg_dump_get_line_nolock - retrieve one kmsg log line (unlocked version)
+ * @dumper: registered kmsg dumper
+ * @syslog: include the "<4>" prefixes
+ * @line: buffer to copy the line to
+ * @size: maximum size of the buffer
+ * @len: length of line placed into buffer
+ *
+ * Start at the beginning of the kmsg buffer, with the oldest kmsg
+ * record, and copy one record into the provided buffer.
+ *
+ * Consecutive calls will return the next available record moving
+ * towards the end of the buffer with the youngest messages.
+ *
+ * A return value of FALSE indicates that there are no more records to
+ * read.
+ *
+ * The function is similar to kmsg_dump_get_line(), but grabs no locks.
+ */
+bool kmsg_dump_get_line_nolock(struct kmsg_dumper *dumper, bool syslog,
+                              char *line, size_t size, size_t *len)
+{
+       struct log *msg;
+       size_t l = 0;
+       bool ret = false;
+
+       if (!dumper->active)
+               goto out;
+
+       if (dumper->cur_seq < log_first_seq) {
+               /* messages are gone, move to first available one */
+               dumper->cur_seq = log_first_seq;
+               dumper->cur_idx = log_first_idx;
+       }
+
+       /* last entry */
+       if (dumper->cur_seq >= log_next_seq)
+               goto out;
+
+       msg = log_from_idx(dumper->cur_idx);
+       l = msg_print_text(msg, 0, syslog, line, size);
+
+       dumper->cur_idx = log_next(dumper->cur_idx);
+       dumper->cur_seq++;
+       ret = true;
+out:
+       if (len)
+               *len = l;
+       return ret;
+}
+
+/**
+ * kmsg_dump_get_line - retrieve one kmsg log line
+ * @dumper: registered kmsg dumper
+ * @syslog: include the "<4>" prefixes
+ * @line: buffer to copy the line to
+ * @size: maximum size of the buffer
+ * @len: length of line placed into buffer
+ *
+ * Start at the beginning of the kmsg buffer, with the oldest kmsg
+ * record, and copy one record into the provided buffer.
+ *
+ * Consecutive calls will return the next available record moving
+ * towards the end of the buffer with the youngest messages.
+ *
+ * A return value of FALSE indicates that there are no more records to
+ * read.
+ */
+bool kmsg_dump_get_line(struct kmsg_dumper *dumper, bool syslog,
+                       char *line, size_t size, size_t *len)
+{
+       unsigned long flags;
+       bool ret;
+
+       raw_spin_lock_irqsave(&logbuf_lock, flags);
+       ret = kmsg_dump_get_line_nolock(dumper, syslog, line, size, len);
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(kmsg_dump_get_line);
+
+/**
+ * kmsg_dump_get_buffer - copy kmsg log lines
+ * @dumper: registered kmsg dumper
+ * @syslog: include the "<4>" prefixes
+ * @buf: buffer to copy the line to
+ * @size: maximum size of the buffer
+ * @len: length of line placed into buffer
+ *
+ * Start at the end of the kmsg buffer and fill the provided buffer
+ * with as many of the the *youngest* kmsg records that fit into it.
+ * If the buffer is large enough, all available kmsg records will be
+ * copied with a single call.
+ *
+ * Consecutive calls will fill the buffer with the next block of
+ * available older records, not including the earlier retrieved ones.
+ *
+ * A return value of FALSE indicates that there are no more records to
+ * read.
+ */
+bool kmsg_dump_get_buffer(struct kmsg_dumper *dumper, bool syslog,
+                         char *buf, size_t size, size_t *len)
+{
+       unsigned long flags;
+       u64 seq;
+       u32 idx;
+       u64 next_seq;
+       u32 next_idx;
+       enum log_flags prev;
+       size_t l = 0;
+       bool ret = false;
+
+       if (!dumper->active)
+               goto out;
+
+       raw_spin_lock_irqsave(&logbuf_lock, flags);
+       if (dumper->cur_seq < log_first_seq) {
+               /* messages are gone, move to first available one */
+               dumper->cur_seq = log_first_seq;
+               dumper->cur_idx = log_first_idx;
+       }
+
+       /* last entry */
+       if (dumper->cur_seq >= dumper->next_seq) {
+               raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+               goto out;
+       }
+
+       /* calculate length of entire buffer */
+       seq = dumper->cur_seq;
+       idx = dumper->cur_idx;
+       prev = 0;
+       while (seq < dumper->next_seq) {
+               struct log *msg = log_from_idx(idx);
+
+               l += msg_print_text(msg, prev, true, NULL, 0);
+               idx = log_next(idx);
+               seq++;
+               prev = msg->flags;
+       }
+
+       /* move first record forward until length fits into the buffer */
+       seq = dumper->cur_seq;
+       idx = dumper->cur_idx;
+       prev = 0;
+       while (l > size && seq < dumper->next_seq) {
+               struct log *msg = log_from_idx(idx);
+
+               l -= msg_print_text(msg, prev, true, NULL, 0);
+               idx = log_next(idx);
+               seq++;
+               prev = msg->flags;
+       }
+
+       /* last message in next interation */
+       next_seq = seq;
+       next_idx = idx;
+
+       l = 0;
+       prev = 0;
+       while (seq < dumper->next_seq) {
+               struct log *msg = log_from_idx(idx);
+
+               l += msg_print_text(msg, prev, syslog, buf + l, size - l);
+               idx = log_next(idx);
+               seq++;
+               prev = msg->flags;
+       }
+
+       dumper->next_seq = next_seq;
+       dumper->next_idx = next_idx;
+       ret = true;
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+out:
+       if (len)
+               *len = l;
+       return ret;
+}
+EXPORT_SYMBOL_GPL(kmsg_dump_get_buffer);
+
+/**
+ * kmsg_dump_rewind_nolock - reset the interator (unlocked version)
+ * @dumper: registered kmsg dumper
+ *
+ * Reset the dumper's iterator so that kmsg_dump_get_line() and
+ * kmsg_dump_get_buffer() can be called again and used multiple
+ * times within the same dumper.dump() callback.
+ *
+ * The function is similar to kmsg_dump_rewind(), but grabs no locks.
+ */
+void kmsg_dump_rewind_nolock(struct kmsg_dumper *dumper)
+{
+       dumper->cur_seq = clear_seq;
+       dumper->cur_idx = clear_idx;
+       dumper->next_seq = log_next_seq;
+       dumper->next_idx = log_next_idx;
+}
+
+/**
+ * kmsg_dump_rewind - reset the interator
+ * @dumper: registered kmsg dumper
+ *
+ * Reset the dumper's iterator so that kmsg_dump_get_line() and
+ * kmsg_dump_get_buffer() can be called again and used multiple
+ * times within the same dumper.dump() callback.
+ */
+void kmsg_dump_rewind(struct kmsg_dumper *dumper)
+{
+       unsigned long flags;
+
+       raw_spin_lock_irqsave(&logbuf_lock, flags);
+       kmsg_dump_rewind_nolock(dumper);
+       raw_spin_unlock_irqrestore(&logbuf_lock, flags);
+}
+EXPORT_SYMBOL_GPL(kmsg_dump_rewind);
+
+static char dump_stack_arch_desc_str[128];
+
+/**
+ * dump_stack_set_arch_desc - set arch-specific str to show with task dumps
+ * @fmt: printf-style format string
+ * @...: arguments for the format string
+ *
+ * The configured string will be printed right after utsname during task
+ * dumps.  Usually used to add arch-specific system identifiers.  If an
+ * arch wants to make use of such an ID string, it should initialize this
+ * as soon as possible during boot.
+ */
+void __init dump_stack_set_arch_desc(const char *fmt, ...)
+{
+       va_list args;
+
+       va_start(args, fmt);
+       vsnprintf(dump_stack_arch_desc_str, sizeof(dump_stack_arch_desc_str),
+                 fmt, args);
+       va_end(args);
+}
+
+/**
+ * dump_stack_print_info - print generic debug info for dump_stack()
+ * @log_lvl: log level
+ *
+ * Arch-specific dump_stack() implementations can use this function to
+ * print out the same debug information as the generic dump_stack().
+ */
+void dump_stack_print_info(const char *log_lvl)
+{
+       printk("%sCPU: %d PID: %d Comm: %.20s %s %s %.*s\n",
+              log_lvl, raw_smp_processor_id(), current->pid, current->comm,
+              print_tainted(), init_utsname()->release,
+              (int)strcspn(init_utsname()->version, " "),
+              init_utsname()->version);
+
+       if (dump_stack_arch_desc_str[0] != '\0')
+               printk("%sHardware name: %s\n",
+                      log_lvl, dump_stack_arch_desc_str);
+
+       print_worker_info(log_lvl, current);
+}
+
+/**
+ * show_regs_print_info - print generic debug info for show_regs()
+ * @log_lvl: log level
+ *
+ * show_regs() implementations can use this function to print out generic
+ * debug information.
+ */
+void show_regs_print_info(const char *log_lvl)
+{
+       dump_stack_print_info(log_lvl);
+
+       printk("%stask: %p ti: %p task.ti: %p\n",
+              log_lvl, current, current_thread_info(),
+              task_thread_info(current));
+}
+
+#endif