python-maas-provisioningserver 1.5.4+bzr2294-0ubuntu1.2 / usr / lib / python2.7 / dist-packages / provisioningserver / utils / __init__.py

# Copyright 2012-2014 Canonical Ltd.  This software is licensed under the
# GNU Affero General Public License version 3 (see the file LICENSE).

"""Utilities for the provisioning server."""

from __future__ import (
    absolute_import,
    print_function,
    unicode_literals,
    )

str = None

__metaclass__ = type
__all__ = [
    "ActionScript",
    "atomic_write",
    "deferred",
    "filter_dict",
    "find_ip_via_arp",
    "import_settings",
    "incremental_write",
    "locate_config",
    "MainScript",
    "ensure_dir",
    "parse_key_value_file",
    "read_text_file",
    "ShellTemplate",
    "sudo_write_file",
    "write_custom_config_section",
    "write_text_file",
    ]

from argparse import ArgumentParser
import codecs
from contextlib import contextmanager
import errno
from functools import wraps
import logging
import os
from os import fdopen
from os.path import isdir
from pipes import quote
from shutil import rmtree
import signal
import string
import subprocess
from subprocess import (
    CalledProcessError,
    PIPE,
    Popen,
    )
import sys
import tempfile
from time import time

from crochet import run_in_reactor
from lockfile import FileLock
from lxml import etree
import netifaces
import tempita
from twisted.internet.defer import maybeDeferred
from twisted.python.threadable import isInIOThread

# A table suitable for use with str.translate() to replace each
# non-printable and non-ASCII character in a byte string with a question
# mark, mimicking the "replace" strategy when encoding and decoding.
non_printable_replace_table = b"".join(
    chr(i) if chr(i) in string.printable else b"?"
    for i in xrange(0xff + 0x01))


class ExternalProcessError(CalledProcessError):
    """Raised when there's a problem calling an external command.

    Unlike `CalledProcessError`:

    - `__str__()` returns a string containing the output of the failed
      external process, if available. All non-printable and non-ASCII
      characters are filtered out, replaced by question marks.

    - `__unicode__()` is defined, and tries to return something
      analagous to `__str__()` but keeping in valid unicode characters
      from the error message.

    """

    @staticmethod
    def _to_unicode(string):
        if isinstance(string, bytes):
            return string.decode("ascii", "replace")
        else:
            return unicode(string)

    @staticmethod
    def _to_ascii(string, table=non_printable_replace_table):
        if isinstance(string, unicode):
            return string.encode("ascii", "replace")
        else:
            return bytes(string).translate(table)

    def __unicode__(self):
        cmd = u" ".join(quote(self._to_unicode(part)) for part in self.cmd)
        output = self._to_unicode(self.output)
        return u"Command `%s` returned non-zero exit status %d:\n%s" % (
            cmd, self.returncode, output)

    def __str__(self):
        cmd = b" ".join(quote(self._to_ascii(part)) for part in self.cmd)
        output = self._to_ascii(self.output)
        return b"Command `%s` returned non-zero exit status %d:\n%s" % (
            cmd, self.returncode, output)


def call_and_check(command, *args, **kwargs):
    """A wrapper around subprocess.check_call().

    When an error occurs, raise an ExternalProcessError.
    """
    try:
        return subprocess.check_call(command, *args, **kwargs)
    except subprocess.CalledProcessError as error:
        error.__class__ = ExternalProcessError
        raise


def call_capture_and_check(command, *args, **kwargs):
    """A wrapper around subprocess.check_output().

    When an error occurs, raise an ExternalProcessError.
    """
    try:
        return subprocess.check_output(command, *args, **kwargs)
    except subprocess.CalledProcessError as error:
        error.__class__ = ExternalProcessError
        raise


def locate_config(*path):
    """Return the location of a given config file or directory.

    Defaults to `/etc/maas` (followed by any further path elements you
    specify), but can be overridden using the `MAAS_CONFIG_DIR` environment
    variable.  (When running from a branch, this variable will point to the
    `etc/maas` inside the branch.)

    The result is absolute and normalized.
    """
    # Check for MAAS_CONFIG_DIR.  Count empty string as "not set."
    env_setting = os.getenv('MAAS_CONFIG_DIR', '')
    if env_setting == '':
        # Running from installed package.  Config is in /etc/maas.
        config_dir = '/etc/maas'
    else:
        # Running from branch or other customized setup.  Config is at
        # $MAAS_CONFIG_DIR/etc/maas.
        config_dir = env_setting

    return os.path.abspath(os.path.join(config_dir, *path))


def find_settings(whence):
    """Return settings from `whence`, which is assumed to be a module."""
    # XXX 2012-10-11 JeroenVermeulen, bug=1065456: Put this in a shared
    # location.  It's currently duplicated from elsewhere.
    return {
        name: value
        for name, value in vars(whence).items()
        if not name.startswith("_")
        }


def import_settings(whence):
    """Import settings from `whence` into the caller's global scope."""
    # XXX 2012-10-11 JeroenVermeulen, bug=1065456: Put this in a shared
    # location.  It's currently duplicated from elsewhere.
    source = find_settings(whence)
    target = sys._getframe(1).f_globals
    target.update(source)


def deferred(func):
    """Decorates a function to ensure that it always returns a `Deferred`.

    This also serves a secondary documentation purpose; functions decorated
    with this are readily identifiable as asynchronous.
    """
    @wraps(func)
    def wrapper(*args, **kwargs):
        return maybeDeferred(func, *args, **kwargs)
    return wrapper


def asynchronous(func):
    """Decorates a function to ensure that it always runs in the reactor.

    If the wrapper is called from the reactor thread, it will call
    straight through to the wrapped function. It will not be wrapped by
    `maybeDeferred` for example.

    If the wrapper is called from another thread, it will return a
    :class:`crochet.EventualResult`, as if it had been decorated with
    `crochet.run_in_reactor`.

    This also serves a secondary documentation purpose; functions decorated
    with this are readily identifiable as asynchronous.
    """
    func_in_reactor = run_in_reactor(func)

    @wraps(func)
    def wrapper(*args, **kwargs):
        if isInIOThread():
            return func(*args, **kwargs)
        else:
            return func_in_reactor(*args, **kwargs)
    return wrapper


def synchronous(func):
    """Decorator to ensure that `func` never runs in the reactor thread.

    If the wrapped function is called from the reactor thread, this will
    raise a :class:`AssertionError`, implying that this is a programming
    error. Calls from outside the reactor will proceed unaffected.

    There is an asymettry with the `asynchronous` decorator. The reason
    is that it is essential to be aware when `deferToThread()` is being
    used, so that in-reactor code knows to synchronise with it, to add a
    callback to the :class:`Deferred` that it returns, for example. The
    expectation with `asynchronous` is that the return value is always
    important, and will be appropriate to the environment in which it is
    utilised.

    This also serves a secondary documentation purpose; functions decorated
    with this are readily identifiable as synchronous, or blocking.

    :raises AssertionError: When called inside the reactor thread.
    """
    @wraps(func)
    def wrapper(*args, **kwargs):
        if isInIOThread():
            raise AssertionError(
                "Function %s(...) must not be called in the "
                "reactor thread." % func.__name__)
        else:
            return func(*args, **kwargs)
    return wrapper


def filter_dict(dictionary, desired_keys):
    """Return a version of `dictionary` restricted to `desired_keys`.

    This is like a set union, except the values from `dictionary` come along.
    (Actually `desired_keys` can be a `dict`, but its values will be ignored).
    """
    return {
        key: value
        for key, value in dictionary.items()
        if key in desired_keys
    }


def _write_temp_file(content, filename):
    """Write the given `content` in a temporary file next to `filename`."""
    # Write the file to a temporary place (next to the target destination,
    # to ensure that it is on the same filesystem).
    directory = os.path.dirname(filename)
    prefix = ".%s." % os.path.basename(filename)
    suffix = ".tmp"
    try:
        temp_fd, temp_file = tempfile.mkstemp(
            dir=directory, suffix=suffix, prefix=prefix)
    except OSError, error:
        if error.filename is None:
            error.filename = os.path.join(
                directory, prefix + "XXXXXX" + suffix)
        raise
    else:
        with os.fdopen(temp_fd, "wb") as f:
            f.write(content)
            # Finish writing this file to the filesystem, and then, tell the
            # filesystem to push it down onto persistent storage.  This
            # prevents a nasty hazard in aggressively optimized filesystems
            # where you replace an old but consistent file with a new one that
            # is still in cache, and lose power before the new file can be made
            # fully persistent.
            # This was a particular problem with ext4 at one point; it may
            # still be.
            f.flush()
            os.fsync(f)
        return temp_file


def atomic_write(content, filename, overwrite=True, mode=0600):
    """Write `content` into the file `filename` in an atomic fashion.

    This requires write permissions to the directory that `filename` is in.
    It creates a temporary file in the same directory (so that it will be
    on the same filesystem as the destination) and then renames it to
    replace the original, if any.  Such a rename is atomic in POSIX.

    :param overwrite: Overwrite `filename` if it already exists?  Default
        is True.
    :param mode: Access permissions for the file, if written.
    """
    temp_file = _write_temp_file(content, filename)
    os.chmod(temp_file, mode)
    try:
        if overwrite:
            os.rename(temp_file, filename)
        else:
            lock = FileLock(filename)
            lock.acquire()
            try:
                if not os.path.isfile(filename):
                    os.rename(temp_file, filename)
            finally:
                lock.release()
    finally:
        if os.path.isfile(temp_file):
            os.remove(temp_file)


def incremental_write(content, filename, mode=0600):
    """Write the given `content` into the file `filename` and
    increment the modification time by 1 sec.

    :param mode: Access permissions for the file.
    """
    old_mtime = get_mtime(filename)
    atomic_write(content, filename, mode=mode)
    new_mtime = pick_new_mtime(old_mtime)
    os.utime(filename, (new_mtime, new_mtime))


def get_mtime(filename):
    """Return a file's modification time, or None if it does not exist."""
    try:
        return os.stat(filename).st_mtime
    except OSError as e:
        if e.errno == errno.ENOENT:
            # File does not exist.  Be helpful, return None.
            return None
        else:
            # Other failure.  The caller will want to know.
            raise


def pick_new_mtime(old_mtime=None, starting_age=1000):
    """Choose a new modification time for a file that needs it updated.

    This function is used to manage the modification time of files
    for which we need to see an increment in the modification time
    each time the file is modified.  This is the case for DNS zone
    files which only get properly reloaded if BIND sees that the
    modification time is > to the time it has in its database.

    Modification time can have a resolution as low as one second in
    some relevant environments (we have observed this with ext3).
    To produce mtime changes regardless, we set a file's modification
    time in the past when it is first written, and
    increment it by 1 second on each subsequent write.

    However we also want to be careful not to set the modification time
    in the future, mostly because BIND does not deal with that very
    well.

    :param old_mtime: File's previous modification time, as a number
        with a unity of one second, or None if it did not previously
        exist.
    :param starting_age: If the file did not exist previously, set its
        modification time this many seconds in the past.
    """
    now = time()
    if old_mtime is None:
        # File is new.  Set modification time in the past to have room for
        # sub-second modifications.
        return now - starting_age
    elif old_mtime + 1 <= now:
        # There is room to increment the file's mtime by one second
        # without ending up in the future.
        return old_mtime + 1
    else:
        # We can't increase the file's modification time.  Give up and
        # return the previous modification time.
        return old_mtime


def split_lines(input, separator):
    """Split each item from `input` into a key/value pair."""
    return (line.split(separator, 1) for line in input if line.strip() != '')


def strip_pairs(input):
    """Strip whitespace of each key/value pair in input."""
    return ((key.strip(), value.strip()) for (key, value) in input)


def parse_key_value_file(file_name, separator=":"):
    """Parse a text file into a dict of key/value pairs.

    Use this for simple key:value or key=value files. There are no
    sections, as required for python's ConfigParse. Whitespace and empty
    lines are ignored.

    :param file_name: Name of file to parse.
    :param separator: The text that separates each key from its value.
    """
    with open(file_name, 'rb') as input:
        return dict(strip_pairs(split_lines(input, separator)))


# Header and footer comments for MAAS custom config sections, as managed
# by write_custom_config_section.
maas_custom_config_markers = (
    "## Begin MAAS settings.  Do not edit; MAAS will overwrite this section.",
    "## End MAAS settings.",
    )


def find_list_item(item, in_list, starting_at=0):
    """Return index of `item` in `in_list`, or None if not found."""
    try:
        return in_list.index(item, starting_at)
    except ValueError:
        return None


def write_custom_config_section(original_text, custom_section):
    """Insert or replace a custom section in a configuration file's text.

    This allows you to rewrite configuration files that are not owned by
    MAAS, but where MAAS will have one section for its own settings.  It
    doesn't read or write any files; this is a pure text operation.

    Appends `custom_section` to the end of `original_text` if there was no
    custom MAAS section yet.  Otherwise, replaces the existing custom MAAS
    section with `custom_section`.  Returns the new text.

    Assumes that the configuration file's format accepts lines starting with
    hash marks (#) as comments.  The custom section will be bracketed by
    special marker comments that make it clear that MAAS wrote the section
    and it should not be edited by hand.

    :param original_text: The config file's current text.
    :type original_text: unicode
    :param custom_section: Custom config section to insert.
    :type custom_section: unicode
    :return: New config file text.
    :rtype: unicode
    """
    header, footer = maas_custom_config_markers
    lines = original_text.splitlines()
    header_index = find_list_item(header, lines)
    if header_index is not None:
        footer_index = find_list_item(footer, lines, header_index)
        if footer_index is None:
            # There's a header but no footer.  Pretend we didn't see the
            # header; just append a new custom section at the end.  Any
            # subsequent rewrite will replace the part starting at the
            # header and ending at the header we will add here.  At that
            # point there will be no trace of the strange situation
            # left.
            header_index = None

    if header_index is None:
        # There was no MAAS custom section in this file.  Append it at
        # the end.
        lines += [
            header,
            custom_section,
            footer,
            ]
    else:
        # There is a MAAS custom section in the file.  Replace it.
        lines = (
            lines[:(header_index + 1)] +
            [custom_section] +
            lines[footer_index:])

    return '\n'.join(lines) + '\n'


def sudo_write_file(filename, contents, encoding='utf-8', mode=0644):
    """Write (or overwrite) file as root.  USE WITH EXTREME CARE.

    Runs an atomic update using non-interactive `sudo`.  This will fail if
    it needs to prompt for a password.
    """
    raw_contents = contents.encode(encoding)
    command = [
        'sudo', '-n', 'maas-provision', 'atomic-write',
        '--filename', filename,
        '--mode', oct(mode),
        ]
    proc = Popen(command, stdin=PIPE)
    stdout, stderr = proc.communicate(raw_contents)
    if proc.returncode != 0:
        raise ExternalProcessError(proc.returncode, command, stderr)


class Safe:
    """An object that is safe to render as-is."""

    __slots__ = ("value",)

    def __init__(self, value):
        self.value = value

    def __repr__(self):
        return "<%s %r>" % (
            self.__class__.__name__, self.value)


def escape_py_literal(string):
    """Escape and quote a string for use as a python literal."""
    return repr(string).decode('ascii')


class ShellTemplate(tempita.Template):
    """A Tempita template specialised for writing shell scripts.

    By default, substitutions will be escaped using `pipes.quote`, unless
    they're marked as safe. This can be done using Tempita's filter syntax::

      {{foobar|safe}}

    or as a plain Python expression::

      {{safe(foobar)}}

    """

    default_namespace = dict(
        tempita.Template.default_namespace,
        safe=Safe)

    def _repr(self, value, pos):
        """Shell-quote the value by default."""
        rep = super(ShellTemplate, self)._repr
        if isinstance(value, Safe):
            return rep(value.value, pos)
        else:
            return quote(rep(value, pos))


class ActionScript:
    """A command-line script that follows a command+verb pattern."""

    def __init__(self, description):
        super(ActionScript, self).__init__()
        # See http://docs.python.org/release/2.7/library/argparse.html.
        self.parser = ArgumentParser(description=description)
        self.subparsers = self.parser.add_subparsers(title="actions")

    @staticmethod
    def setup():
        # Ensure stdout and stderr are line-bufferred.
        sys.stdout = fdopen(sys.stdout.fileno(), "ab", 1)
        sys.stderr = fdopen(sys.stderr.fileno(), "ab", 1)
        # Run the SIGINT handler on SIGTERM; `svc -d` sends SIGTERM.
        signal.signal(signal.SIGTERM, signal.default_int_handler)

    def register(self, name, handler, *args, **kwargs):
        """Register an action for the given name.

        :param name: The name of the action.
        :param handler: An object, a module for example, that has `run` and
            `add_arguments` callables. The docstring of the `run` callable is
            used as the help text for the newly registered action.
        :param args: Additional positional arguments for the subparser_.
        :param kwargs: Additional named arguments for the subparser_.

        .. _subparser:
          http://docs.python.org/
            release/2.7/library/argparse.html#sub-commands
        """
        parser = self.subparsers.add_parser(
            name, *args, help=handler.run.__doc__, **kwargs)
        parser.set_defaults(handler=handler)
        handler.add_arguments(parser)
        return parser

    def execute(self, argv=None):
        """Execute this action.

        This is intended for in-process invocation of an action, though it may
        still raise L{SystemExit}. The L{__call__} method is intended for when
        this object is executed as a script proper.
        """
        args = self.parser.parse_args(argv)
        args.handler.run(args)

    def __call__(self, argv=None):
        try:
            self.setup()
            self.execute(argv)
        except CalledProcessError as error:
            # Print error.cmd and error.output too?
            raise SystemExit(error.returncode)
        except KeyboardInterrupt:
            raise SystemExit(1)
        else:
            raise SystemExit(0)


class MainScript(ActionScript):
    """An `ActionScript` that always accepts a `--config-file` option.

    The `--config-file` option defaults to the value of
    `MAAS_PROVISIONING_SETTINGS` in the process's environment, or absent
    that, `$MAAS_CONFIG_DIR/pserv.yaml` (normally /etc/maas/pserv.yaml for
    packaged installations, or when running from branch, the equivalent
    inside that branch).
    """

    def __init__(self, description):
        # Avoid circular imports.
        from provisioningserver.config import Config

        super(MainScript, self).__init__(description)
        self.parser.add_argument(
            "-c", "--config-file", metavar="FILENAME",
            help="Configuration file to load [%(default)s].",
            default=Config.DEFAULT_FILENAME)


class AtomicWriteScript:
    """Wrap the atomic_write function turning it into an ActionScript.

    To use:
    >>> main = MainScript(atomic_write.__doc__)
    >>> main.register("myscriptname", AtomicWriteScript)
    >>> main()
    """

    @staticmethod
    def add_arguments(parser):
        """Initialise options for writing files atomically.

        :param parser: An instance of :class:`ArgumentParser`.
        """
        parser.add_argument(
            "--no-overwrite", action="store_true", required=False,
            default=False, help="Don't overwrite file if it exists")
        parser.add_argument(
            "--filename", action="store", required=True, help=(
                "The name of the file in which to store contents of stdin"))
        parser.add_argument(
            "--mode", action="store", required=False, default=None, help=(
                "They permissions to set on the file. If not set "
                "will be r/w only to owner"))

    @staticmethod
    def run(args):
        """Take content from stdin and write it atomically to a file."""
        content = sys.stdin.read()
        if args.mode is not None:
            mode = int(args.mode, 8)
        else:
            mode = 0600
        atomic_write(
            content, args.filename, overwrite=not args.no_overwrite,
            mode=mode)


def get_all_interface_addresses():
    """For each network interface, yield its IPv4 address."""
    for interface in netifaces.interfaces():
        addresses = netifaces.ifaddresses(interface)
        if netifaces.AF_INET in addresses:
            for inet_address in addresses[netifaces.AF_INET]:
                if "addr" in inet_address:
                    yield inet_address["addr"]


def ensure_dir(path):
    """Do the equivalent of `mkdir -p`, creating `path` if it didn't exist."""
    try:
        os.makedirs(path)
    except OSError as e:
        if e.errno != errno.EEXIST:
            raise
        if not isdir(path):
            # Path exists, but isn't a directory.
            raise
        # Otherwise, the error is that the directory already existed.
        # Which is actually success.


@contextmanager
def tempdir(suffix=b'', prefix=b'maas-', location=None):
    """Context manager: temporary directory.

    Creates a temporary directory (yielding its path, as `unicode`), and
    cleans it up again when exiting the context.

    The directory will be readable, writable, and searchable only to the
    system user who creates it.

    >>> with tempdir() as playground:
    ...     my_file = os.path.join(playground, "my-file")
    ...     with open(my_file, 'wb') as handle:
    ...         handle.write(b"Hello.\n")
    ...     files = os.listdir(playground)
    >>> files
    [u'my-file']
    >>> os.path.isdir(playground)
    False
    """
    path = tempfile.mkdtemp(suffix, prefix, location)
    if isinstance(path, bytes):
        path = path.decode(sys.getfilesystemencoding())
    assert isinstance(path, unicode)
    try:
        yield path
    finally:
        rmtree(path, ignore_errors=True)


def read_text_file(path, encoding='utf-8'):
    """Read and decode the text file at the given path."""
    with codecs.open(path, encoding=encoding) as infile:
        return infile.read()


def write_text_file(path, text, encoding='utf-8'):
    """Write the given unicode text to the given file path.

    If the file existed, it will be overwritten.
    """
    with codecs.open(path, 'w', encoding) as outfile:
        outfile.write(text)


def is_compiled_xpath(xpath):
    """Is `xpath` a compiled expression?"""
    return isinstance(xpath, etree.XPath)


def is_compiled_doc(doc):
    """Is `doc` a compiled XPath document evaluator?"""
    return isinstance(doc, etree.XPathDocumentEvaluator)


def match_xpath(xpath, doc):
    """Return a match of expression `xpath` against document `doc`.

    :type xpath: Either `unicode` or `etree.XPath`
    :type doc: Either `etree._ElementTree` or `etree.XPathDocumentEvaluator`

    :rtype: bool
    """
    is_xpath_compiled = is_compiled_xpath(xpath)
    is_doc_compiled = is_compiled_doc(doc)

    if is_xpath_compiled and is_doc_compiled:
        return doc(xpath.path)
    elif is_xpath_compiled:
        return xpath(doc)
    elif is_doc_compiled:
        return doc(xpath)
    else:
        return doc.xpath(xpath)


def try_match_xpath(xpath, doc, logger=logging):
    """See if the XPath expression matches the given XML document.

    Invalid XPath expressions are logged, and are returned as a
    non-match.

    :type xpath: Either `unicode` or `etree.XPath`
    :type doc: Either `etree._ElementTree` or `etree.XPathDocumentEvaluator`

    :rtype: bool
    """
    try:
        # Evaluating an XPath expression against a document with LXML
        # can return a list or a string, and perhaps other types.
        # Casting the return value into a boolean context appears to
        # be the most reliable way of detecting a match.
        return bool(match_xpath(xpath, doc))
    except etree.XPathEvalError:
        # Get a plaintext version of `xpath`.
        expr = xpath.path if is_compiled_xpath(xpath) else xpath
        logger.exception("Invalid expression: %s", expr)
        return False


def classify(func, subjects):
    """Classify `subjects` according to `func`.

    Splits `subjects` into two lists: one for those which `func`
    returns a truth-like value, and one for the others.

    :param subjects: An iterable of `(ident, subject)` tuples, where
        `subject` is an argument that can be passed to `func` for
        classification.
    :param func: A function that takes a single argument.

    :return: A ``(matched, other)`` tuple, where ``matched`` and
        ``other`` are `list`s of `ident` values; `subject` values are
        not returned.
    """
    matched, other = [], []
    for ident, subject in subjects:
        bucket = matched if func(subject) else other
        bucket.append(ident)
    return matched, other


def find_ip_via_arp(mac):
    """Find the IP address for `mac` by reading the output of arp -n.

    Returns `None` if the MAC is not found.

    We do this because we aren't necessarily the only DHCP server on the
    network, so we can't check our own leases file and be guaranteed to find an
    IP that matches.

    :param mac: The mac address, e.g. '1c:6f:65:d5:56:98'.
    """

    output = call_capture_and_check(['arp', '-n']).split('\n')

    for line in output:
        columns = line.split()
        if len(columns) == 5 and columns[2] == mac:
            return columns[0]
    return None


def find_mac_via_arp(ip):
    """Find the MAC address for `ip` by reading the output of arp -n.

    Returns `None` if the IP is not found.

    We do this because we aren't necessarily the only DHCP server on the
    network, so we can't check our own leases file and be guaranteed to find an
    IP that matches.

    :param ip: The ip address, e.g. '192.168.1.1'.
    """

    output = call_capture_and_check(['arp', '-n']).split('\n')

    for line in sorted(output):
        columns = line.split()
        if len(columns) == 5 and columns[0] == ip:
            return columns[2]
    return None