path: root/systemd/journal.py

#  -*- Mode: python; coding:utf-8; indent-tabs-mode: nil -*- */
#
#
#  Copyright 2012 David Strauss <david@davidstrauss.net>
#  Copyright 2012 Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl>
#  Copyright 2012 Marti Raudsepp <marti@juffo.org>
#
#  python-systemd is free software; you can redistribute it and/or modify it
#  under the terms of the GNU Lesser General Public License as published by
#  the Free Software Foundation; either version 2.1 of the License, or
#  (at your option) any later version.
#
#  python-systemd is distributed in the hope that it will be useful, but
#  WITHOUT ANY WARRANTY; without even the implied warranty of
#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
#  Lesser General Public License for more details.
#
#  You should have received a copy of the GNU Lesser General Public License
#  along with python-systemd; If not, see <http://www.gnu.org/licenses/>.

from __future__ import division

import sys as _sys
import datetime as _datetime
import uuid as _uuid
import traceback as _traceback
import os as _os
import logging as _logging
if _sys.version_info >= (3,3):
    from collections import ChainMap as _ChainMap
from syslog import (LOG_EMERG, LOG_ALERT, LOG_CRIT, LOG_ERR,
                    LOG_WARNING, LOG_NOTICE, LOG_INFO, LOG_DEBUG)
from ._journal import __version__, sendv, stream_fd
from ._reader import (_Reader, NOP, APPEND, INVALIDATE,
                      LOCAL_ONLY, RUNTIME_ONLY,
                      SYSTEM, SYSTEM_ONLY, CURRENT_USER,
                      OS_ROOT,
                      _get_catalog)
from . import id128 as _id128

if _sys.version_info >= (3,):
    from ._reader import Monotonic
else:
    Monotonic = tuple

def _convert_monotonic(m):
    return Monotonic((_datetime.timedelta(microseconds=m[0]),
                      _uuid.UUID(bytes=m[1])))

def _convert_source_monotonic(s):
    return _datetime.timedelta(microseconds=int(s))

def _convert_realtime(t):
    return _datetime.datetime.fromtimestamp(t / 1000000)

def _convert_timestamp(s):
    return _datetime.datetime.fromtimestamp(int(s) / 1000000)

def _convert_trivial(x):
    return x

if _sys.version_info >= (3,):
    def _convert_uuid(s):
        return _uuid.UUID(s.decode())
else:
    _convert_uuid = _uuid.UUID

DEFAULT_CONVERTERS = {
    'MESSAGE_ID': _convert_uuid,
    '_MACHINE_ID': _convert_uuid,
    '_BOOT_ID': _convert_uuid,
    'PRIORITY': int,
    'LEADER': int,
    'SESSION_ID': int,
    'USERSPACE_USEC': int,
    'INITRD_USEC': int,
    'KERNEL_USEC': int,
    '_UID': int,
    '_GID': int,
    '_PID': int,
    'SYSLOG_FACILITY': int,
    'SYSLOG_PID': int,
    '_AUDIT_SESSION': int,
    '_AUDIT_LOGINUID': int,
    '_SYSTEMD_SESSION': int,
    '_SYSTEMD_OWNER_UID': int,
    'CODE_LINE': int,
    'ERRNO': int,
    'EXIT_STATUS': int,
    '_SOURCE_REALTIME_TIMESTAMP': _convert_timestamp,
    '__REALTIME_TIMESTAMP': _convert_realtime,
    '_SOURCE_MONOTONIC_TIMESTAMP': _convert_source_monotonic,
    '__MONOTONIC_TIMESTAMP': _convert_monotonic,
    '__CURSOR': _convert_trivial,
    'COREDUMP': bytes,
    'COREDUMP_PID': int,
    'COREDUMP_UID': int,
    'COREDUMP_GID': int,
    'COREDUMP_SESSION': int,
    'COREDUMP_SIGNAL': int,
    'COREDUMP_TIMESTAMP': _convert_timestamp,
}

_IDENT_CHARACTER = set('ABCDEFGHIJKLMNOPQRTSUVWXYZ_0123456789')

def _valid_field_name(s):
    return not (set(s) - _IDENT_CHARACTER)

class Reader(_Reader):
    """Access systemd journal entries.

    Entries are subject to filtering and limits, see `add_match`, `this_boot`,
    `this_machine` functions and the `data_treshold` attribute.

    Note that in order to access the system journal, a non-root user must have
    the necessary privileges, see journalctl(1) for details.  Unprivileged users
    can access only their own journal.

    Example usage to print out all informational or higher level messages for
    systemd-udevd for this boot:

    >>> from systemd import journal
    >>> j = journal.Reader()
    >>> j.this_boot()
    >>> j.log_level(journal.LOG_INFO)
    >>> j.add_match(_SYSTEMD_UNIT="systemd-udevd.service")
    >>> for entry in j:                                 # doctest: +SKIP
    ...    print(entry['MESSAGE'])
    starting version ...

    See systemd.journal-fields(7) for more info on typical fields found in the
    journal.

    """
    def __init__(self, flags=None, path=None, files=None, converters=None):
        """Create a new Reader.

        Argument `flags` defines the open flags of the journal, which can be one
        of, or ORed combination of constants: LOCAL_ONLY (default) opens journal
        on local machine only; RUNTIME_ONLY opens only volatile journal files;
        and SYSTEM_ONLY opens only journal files of system services and the kernel.

        Argument `path` is the directory of journal files, either a file system
        path or a file descriptor. Note that `flags`, `path`, and `files` are
        exclusive.

        Argument `converters` is a dictionary which updates the
        DEFAULT_CONVERTERS to convert journal field values. Field names are used
        as keys into this dictionary. The values must be single argument
        functions, which take a `bytes` object and return a converted
        value. When there's no entry for a field name, then the default UTF-8
        decoding will be attempted. If the conversion fails with a ValueError,
        unconverted bytes object will be returned. (Note that ValueEror is a
        superclass of UnicodeDecodeError).

        Reader implements the context manager protocol: the journal will be
        closed when exiting the block.
        """
        if flags is None:
            if path is None and files is None:
                # This mimics journalctl behaviour of default to local journal only
                flags = LOCAL_ONLY
            else:
                flags = 0

        super(Reader, self).__init__(flags, path, files)
        if _sys.version_info >= (3,3):
            self.converters = _ChainMap()
            if converters is not None:
                self.converters.maps.append(converters)
            self.converters.maps.append(DEFAULT_CONVERTERS)
        else:
            self.converters = DEFAULT_CONVERTERS.copy()
            if converters is not None:
                self.converters.update(converters)

    def _convert_field(self, key, value):
        """Convert value using self.converters[key].

        If `key` is not present in self.converters, a standard unicode decoding
        will be attempted.  If the conversion (either key-specific or the
        default one) fails with a ValueError, the original bytes object will be
        returned.
        """
        convert = self.converters.get(key, bytes.decode)
        try:
            return convert(value)
        except ValueError:
            # Leave in default bytes
            return value

    def _convert_entry(self, entry):
        """Convert entire journal entry utilising _convert_field."""
        result = {}
        for key, value in entry.items():
            if isinstance(value, list):
                result[key] = [self._convert_field(key, val) for val in value]
            else:
                result[key] = self._convert_field(key, value)
        return result

    def __iter__(self):
        """Return self.

        Part of the iterator protocol.
        """
        return self

    def __next__(self):
        """Return the next entry in the journal.

        Returns self.get_next() or raises StopIteration.

        Part of the iterator protocol.
        """
        ans = self.get_next()
        if ans:
            return ans
        else:
            raise StopIteration()

    if _sys.version_info < (3,):
        next = __next__

    def add_match(self, *args, **kwargs):
        """Add one or more matches to the filter journal log entries.

        All matches of different field are combined in a logical AND, and
        matches of the same field are automatically combined in a logical OR.
        Matches can be passed as strings of form "FIELD=value", or keyword
        arguments FIELD="value".
        """
        args = list(args)
        args.extend(_make_line(key, val) for key, val in kwargs.items())
        for arg in args:
            super(Reader, self).add_match(arg)

    def get_next(self, skip=1):
        r"""Return the next log entry as a dictionary.

        Entries will be processed with converters specified during Reader
        creation.

        Optional skip value will return the `skip`\-th log entry.

        Currently a standard dictionary of fields is returned, but in the
        future this might be changed to a different mapping type, so the
        calling code should not make assumptions about a specific type.
        """
        if super(Reader, self)._next(skip):
            entry = super(Reader, self)._get_all()
            if entry:
                entry['__REALTIME_TIMESTAMP'] = self._get_realtime()
                entry['__MONOTONIC_TIMESTAMP'] = self._get_monotonic()
                entry['__CURSOR'] = self._get_cursor()
                return self._convert_entry(entry)
        return dict()

    def get_previous(self, skip=1):
        r"""Return the previous log entry.

        Equivalent to get_next(-skip).

        Optional skip value will return the -`skip`\-th log entry.

        Entries will be processed with converters specified during Reader
        creation.

        Currently a standard dictionary of fields is returned, but in the
        future this might be changed to a different mapping type, so the
        calling code should not make assumptions about a specific type.
        """
        return self.get_next(-skip)

    def query_unique(self, field):
        """Return a list of unique values appearing in the journal for the given
        `field`.

        Note this does not respect any journal matches.

        Entries will be processed with converters specified during
        Reader creation.
        """
        return set(self._convert_field(field, value)
                   for value in super(Reader, self).query_unique(field))

    def wait(self, timeout=None):
        """Wait for a change in the journal.

        `timeout` is the maximum time in seconds to wait, or None which
        means to wait forever.

        Returns one of NOP (no change), APPEND (new entries have been added to
        the end of the journal), or INVALIDATE (journal files have been added or
        removed).
        """
        us = -1 if timeout is None else int(timeout * 1000000)
        return super(Reader, self).wait(us)

    def seek_realtime(self, realtime):
        """Seek to a matching journal entry nearest to `timestamp` time.

        Argument `realtime` must be either an integer UNIX timestamp (in
        microseconds since the beginning of the UNIX epoch), or an float UNIX
        timestamp (in seconds since the beginning of the UNIX epoch), or a
        datetime.datetime instance. The integer form is deprecated.

        >>> import time
        >>> from systemd import journal

        >>> yesterday = time.time() - 24 * 60**2
        >>> j = journal.Reader()
        >>> j.seek_realtime(yesterday)
        """
        if isinstance(realtime, _datetime.datetime):
            realtime = int(float(realtime.strftime("%s.%f")) * 1000000)
        elif not isinstance(realtime, int):
            realtime = int(realtime * 1000000)
        return super(Reader, self).seek_realtime(realtime)

    def seek_monotonic(self, monotonic, bootid=None):
        """Seek to a matching journal entry nearest to `monotonic` time.

        Argument `monotonic` is a timestamp from boot in either seconds or a
        datetime.timedelta instance. Argument `bootid` is a string or UUID
        representing which boot the monotonic time is reference to. Defaults to
        current bootid.
        """
        if isinstance(monotonic, _datetime.timedelta):
            monotonic = monotonic.total_seconds()
        monotonic = int(monotonic * 1000000)
        if isinstance(bootid, _uuid.UUID):
            bootid = bootid.hex
        return super(Reader, self).seek_monotonic(monotonic, bootid)

    def log_level(self, level):
        """Set maximum log `level` by setting matches for PRIORITY.
        """
        if 0 <= level <= 7:
            for i in range(level+1):
                self.add_match(PRIORITY="%d" % i)
        else:
            raise ValueError("Log level must be 0 <= level <= 7")

    def messageid_match(self, messageid):
        """Add match for log entries with specified `messageid`.

        `messageid` can be string of hexadicimal digits or a UUID
        instance. Standard message IDs can be found in systemd.id128.

        Equivalent to add_match(MESSAGE_ID=`messageid`).
        """
        if isinstance(messageid, _uuid.UUID):
            messageid = messageid.hex
        self.add_match(MESSAGE_ID=messageid)

    def this_boot(self, bootid=None):
        """Add match for _BOOT_ID for current boot or the specified boot ID.

        If specified, bootid should be either a UUID or a 32 digit hex number.

        Equivalent to add_match(_BOOT_ID='bootid').
        """
        if bootid is None:
            bootid = _id128.get_boot().hex
        else:
            bootid = getattr(bootid, 'hex', bootid)
        self.add_match(_BOOT_ID=bootid)

    def this_machine(self, machineid=None):
        """Add match for _MACHINE_ID equal to the ID of this machine.

        If specified, machineid should be either a UUID or a 32 digit hex
        number.

        Equivalent to add_match(_MACHINE_ID='machineid').
        """
        if machineid is None:
            machineid = _id128.get_machine().hex
        else:
            machineid = getattr(machineid, 'hex', machineid)
        self.add_match(_MACHINE_ID=machineid)


def get_catalog(mid):
    """Return catalog entry for the specified ID.

    `mid` should be either a UUID or a 32 digit hex number.
    """
    if isinstance(mid, _uuid.UUID):
        mid = mid.hex
    return _get_catalog(mid)

def _make_line(field, value):
    if isinstance(value, bytes):
        return field.encode('utf-8') + b'=' + value
    elif isinstance(value, str):
        return field + '=' + value
    else:
        return field + '=' + str(value)

def send(MESSAGE, MESSAGE_ID=None,
         CODE_FILE=None, CODE_LINE=None, CODE_FUNC=None,
         **kwargs):
    r"""Send a message to the journal.

    >>> from systemd import journal
    >>> journal.send('Hello world')
    >>> journal.send('Hello, again, world', FIELD2='Greetings!')
    >>> journal.send('Binary message', BINARY=b'\xde\xad\xbe\xef')

    Value of the MESSAGE argument will be used for the MESSAGE= field. MESSAGE
    must be a string and will be sent as UTF-8 to the journal.

    MESSAGE_ID can be given to uniquely identify the type of message. It must be
    a string or a uuid.UUID object.

    CODE_LINE, CODE_FILE, and CODE_FUNC can be specified to identify the caller.
    Unless at least on of the three is given, values are extracted from the
    stack frame of the caller of send(). CODE_FILE and CODE_FUNC must be
    strings, CODE_LINE must be an integer.

    Additional fields for the journal entry can only be specified as keyword
    arguments. The payload can be either a string or bytes. A string will be
    sent as UTF-8, and bytes will be sent as-is to the journal.

    Other useful fields include PRIORITY, SYSLOG_FACILITY, SYSLOG_IDENTIFIER,
    SYSLOG_PID.
    """

    args = ['MESSAGE=' + MESSAGE]

    if MESSAGE_ID is not None:
        id = getattr(MESSAGE_ID, 'hex', MESSAGE_ID)
        args.append('MESSAGE_ID=' + id)

    if CODE_LINE == CODE_FILE == CODE_FUNC == None:
        CODE_FILE, CODE_LINE, CODE_FUNC = _traceback.extract_stack(limit=2)[0][:3]
    if CODE_FILE is not None:
        args.append('CODE_FILE=' + CODE_FILE)
    if CODE_LINE is not None:
        args.append('CODE_LINE={:d}'.format(CODE_LINE))
    if CODE_FUNC is not None:
        args.append('CODE_FUNC=' + CODE_FUNC)

    args.extend(_make_line(key, val) for key, val in kwargs.items())
    return sendv(*args)

def stream(identifier=None, priority=LOG_INFO, level_prefix=False):
    r"""Return a file object wrapping a stream to journal.

    Log messages written to this file as simple newline sepearted text strings
    are written to the journal.

    The file will be line buffered, so messages are actually sent after a
    newline character is written.

    >>> from systemd import journal
    >>> stream = journal.stream('myapp')                       # doctest: +SKIP
    >>> res = stream.write('message...\n')                     # doctest: +SKIP

    will produce the following message in the journal::

      PRIORITY=7
      SYSLOG_IDENTIFIER=myapp
      MESSAGE=message...

    If identifier is None, a suitable default based on sys.argv[0] will be used.

    This interface can be used conveniently with the print function:

    >>> from __future__ import print_function
    >>> stream = journal.stream()                              # doctest: +SKIP
    >>> print('message...', file=stream)                       # doctest: +SKIP

    priority is the syslog priority, one of `LOG_EMERG`, `LOG_ALERT`,
    `LOG_CRIT`, `LOG_ERR`, `LOG_WARNING`, `LOG_NOTICE`, `LOG_INFO`, `LOG_DEBUG`.

    level_prefix is a boolean. If true, kernel-style log priority level prefixes
    (such as '<1>') are interpreted. See sd-daemon(3) for more information.
    """

    if identifier is None:
        if not _sys.argv or not _sys.argv[0] or _sys.argv[0] == '-c':
            identifier = 'python'
        else:
            identifier = _sys.argv[0]

    fd = stream_fd(identifier, priority, level_prefix)
    return _os.fdopen(fd, 'w', 1)

class JournalHandler(_logging.Handler):
    """Journal handler class for the Python logging framework.

    Please see the Python logging module documentation for an overview:
    http://docs.python.org/library/logging.html.

    To create a custom logger whose messages go only to journal:

    >>> import logging
    >>> log = logging.getLogger('custom_logger_name')
    >>> log.propagate = False
    >>> log.addHandler(JournalHandler())
    >>> log.warning("Some message: %s", 'detail')

    Note that by default, message levels `INFO` and `DEBUG` are ignored by the
    logging framework. To enable those log levels:

    >>> log.setLevel(logging.DEBUG)

    To redirect all logging messages to journal regardless of where they come
    from, attach it to the root logger:

    >>> logging.root.addHandler(JournalHandler())

    For more complex configurations when using `dictConfig` or `fileConfig`,
    specify `systemd.journal.JournalHandler` as the handler class.  Only
    standard handler configuration options are supported: `level`, `formatter`,
    `filters`.

    To attach journal MESSAGE_ID, an extra field is supported:

    >>> import uuid
    >>> mid = uuid.UUID('0123456789ABCDEF0123456789ABCDEF')
    >>> log.warning("Message with ID", extra={'MESSAGE_ID': mid})

    Fields to be attached to all messages sent through this handler can be
    specified as keyword arguments. This probably makes sense only for
    SYSLOG_IDENTIFIER and similar fields which are constant for the whole
    program:

    >>> JournalHandler(SYSLOG_IDENTIFIER='my-cool-app')
    <systemd.journal.JournalHandler object at ...>

    The following journal fields will be sent: `MESSAGE`, `PRIORITY`,
    `THREAD_NAME`, `CODE_FILE`, `CODE_LINE`, `CODE_FUNC`, `LOGGER` (name as
    supplied to getLogger call), `MESSAGE_ID` (optional, see above),
    `SYSLOG_IDENTIFIER` (defaults to sys.argv[0]).
    """

    def __init__(self, level=_logging.NOTSET, **kwargs):
        super(JournalHandler, self).__init__(level)

        for name in kwargs:
            if not _valid_field_name(name):
                raise ValueError('Invalid field name: ' + name)
        if 'SYSLOG_IDENTIFIER' not in kwargs:
            kwargs['SYSLOG_IDENTIFIER'] = _sys.argv[0]

        self.send = kwargs.pop('SENDER_FUNCTION', send)
        self._extra = kwargs

    def emit(self, record):
        """Write `record` as a journal event.

        MESSAGE is taken from the message provided by the user, and PRIORITY,
        LOGGER, THREAD_NAME, CODE_{FILE,LINE,FUNC} fields are appended
        automatically. In addition, record.MESSAGE_ID will be used if present.
        """
        try:
            msg = self.format(record)
            pri = self.mapPriority(record.levelno)
            mid = getattr(record, 'MESSAGE_ID', None)
            extras = { k:str(v) for k,v in self._extra.items() }
            extras.update({
                k:str(v) for k,v in record.__dict__.items()
            })

            if record.exc_text:
                extras['EXCEPTION_TEXT'] = record.exc_text

            if record.exc_info:
                extras['EXCEPTION_INFO'] = record.exc_info

            if record.args:
                extras['CODE_ARGS'] = str(record.args)

            self.send(msg,
                      MESSAGE_ID=mid,
                      PRIORITY=format(pri),
                      LOGGER=record.name,
                      THREAD_NAME=record.threadName,
                      PROCESS_NAME=record.processName,
                      CODE_FILE=record.pathname,
                      CODE_LINE=record.lineno,
                      CODE_FUNC=record.funcName,
                      **extras)
        except Exception:
            self.handleError(record)

    @staticmethod
    def mapPriority(levelno):
        """Map logging levels to journald priorities.

        Since Python log level numbers are "sparse", we have to map numbers in
        between the standard levels too.
        """
        if levelno <= _logging.DEBUG:
            return LOG_DEBUG
        elif levelno <= _logging.INFO:
            return LOG_INFO
        elif levelno <= _logging.WARNING:
            return LOG_WARNING
        elif levelno <= _logging.ERROR:
            return LOG_ERR
        elif levelno <= _logging.CRITICAL:
            return LOG_CRIT
        else:
            return LOG_ALERT