community.general/plugins/action/system/iptables_state.py

# Copyright: (c) 2020, quidame <quidame@poivron.org>
# GNU General Public License v3.0+ (see COPYING or https://www.gnu.org/licenses/gpl-3.0.txt)

from __future__ import absolute_import, division, print_function
__metaclass__ = type

import time

from ansible.plugins.action import ActionBase
from ansible.errors import AnsibleActionFail, AnsibleConnectionFailure
from ansible.utils.vars import merge_hash
from ansible.utils.display import Display

display = Display()


class ActionModule(ActionBase):

    # Keep internal params away from user interactions
    _VALID_ARGS = frozenset(('path', 'state', 'table', 'noflush', 'counters', 'modprobe', 'ip_version', 'wait'))
    DEFAULT_SUDOABLE = True

    MSG_ERROR__ASYNC_AND_POLL_NOT_ZERO = (
        "This module doesn't support async>0 and poll>0 when its 'state' param "
        "is set to 'restored'. To enable its rollback feature (that needs the "
        "module to run asynchronously on the remote), please set task attribute "
        "'poll' (=%s) to 0, and 'async' (=%s) to a value >2 and not greater than "
        "'ansible_timeout' (=%s) (recommended).")
    MSG_WARNING__NO_ASYNC_IS_NO_ROLLBACK = (
        "Attempts to restore iptables state without rollback in case of mistake "
        "may lead the ansible controller to loose access to the hosts and never "
        "regain it before fixing firewall rules through a serial console, or any "
        "other way except SSH. Please set task attribute 'poll' (=%s) to 0, and "
        "'async' (=%s) to a value >2 and not greater than 'ansible_timeout' (=%s) "
        "(recommended).")
    MSG_WARNING__ASYNC_GREATER_THAN_TIMEOUT = (
        "You attempt to restore iptables state with rollback in case of mistake, "
        "but with settings that will lead this rollback to happen AFTER that the "
        "controller will reach its own timeout. Please set task attribute 'poll' "
        "(=%s) to 0, and 'async' (=%s) to a value >2 and not greater than "
        "'ansible_timeout' (=%s) (recommended).")

    def _async_result(self, async_status_args, task_vars, timeout):
        '''
        Retrieve results of the asynchonous task, and display them in place of
        the async wrapper results (those with the ansible_job_id key).
        '''
        async_status = self._task.copy()
        async_status.args = async_status_args
        async_status.action = 'ansible.builtin.async_status'
        async_status.async_val = 0
        async_action = self._shared_loader_obj.action_loader.get(
            async_status.action, task=async_status, connection=self._connection,
            play_context=self._play_context, loader=self._loader, templar=self._templar,
            shared_loader_obj=self._shared_loader_obj)

        if async_status.args['mode'] == 'cleanup':
            return async_action.run(task_vars=task_vars)

        # At least one iteration is required, even if timeout is 0.
        for dummy in range(max(1, timeout)):
            async_result = async_action.run(task_vars=task_vars)
            if async_result.get('finished', 0) == 1:
                break
            time.sleep(min(1, timeout))

        return async_result

    def run(self, tmp=None, task_vars=None):

        self._supports_check_mode = True
        self._supports_async = True

        result = super(ActionModule, self).run(tmp, task_vars)
        del tmp  # tmp no longer has any effect

        if not result.get('skipped'):

            # FUTURE: better to let _execute_module calculate this internally?
            wrap_async = self._task.async_val and not self._connection.has_native_async

            # Set short names for values we'll have to compare or reuse
            task_poll = self._task.poll
            task_async = self._task.async_val
            check_mode = self._play_context.check_mode
            max_timeout = self._connection._play_context.timeout
            module_args = self._task.args

            if module_args.get('state', None) == 'restored':
                if not wrap_async:
                    if not check_mode:
                        display.warning(self.MSG_WARNING__NO_ASYNC_IS_NO_ROLLBACK % (
                            task_poll,
                            task_async,
                            max_timeout))
                elif task_poll:
                    raise AnsibleActionFail(self.MSG_ERROR__ASYNC_AND_POLL_NOT_ZERO % (
                        task_poll,
                        task_async,
                        max_timeout))
                else:
                    if task_async > max_timeout and not check_mode:
                        display.warning(self.MSG_WARNING__ASYNC_GREATER_THAN_TIMEOUT % (
                            task_poll,
                            task_async,
                            max_timeout))

                    # inject the async directory based on the shell option into the
                    # module args
                    async_dir = self.get_shell_option('async_dir', default="~/.ansible_async")

                    # Bind the loop max duration to consistent values on both
                    # remote and local sides (if not the same, make the loop
                    # longer on the controller); and set a backup file path.
                    module_args['_timeout'] = task_async
                    module_args['_back'] = '%s/iptables.state' % async_dir
                    async_status_args = dict(mode='status')
                    confirm_cmd = 'rm -f %s' % module_args['_back']
                    starter_cmd = 'touch %s.starter' % module_args['_back']
                    remaining_time = max(task_async, max_timeout)

            # do work!
            result = merge_hash(result, self._execute_module(module_args=module_args, task_vars=task_vars, wrap_async=wrap_async))

            # Then the 3-steps "go ahead or rollback":
            # 1. Catch early errors of the module (in asynchronous task) if any.
            #    Touch a file on the target to signal the module to process now.
            # 2. Reset connection to ensure a persistent one will not be reused.
            # 3. Confirm the restored state by removing the backup on the remote.
            #    Retrieve the results of the asynchronous task to return them.
            if '_back' in module_args:
                async_status_args['jid'] = result.get('ansible_job_id', None)
                if async_status_args['jid'] is None:
                    raise AnsibleActionFail("Unable to get 'ansible_job_id'.")

                # Catch early errors due to missing mandatory option, bad
                # option type/value, missing required system command, etc.
                result = merge_hash(result, self._async_result(async_status_args, task_vars, 0))

                # The module is aware to not process the main iptables-restore
                # command before finding (and deleting) the 'starter' cookie on
                # the host, so the previous query will not reach ssh timeout.
                dummy = self._low_level_execute_command(starter_cmd, sudoable=self.DEFAULT_SUDOABLE)

                # As the main command is not yet executed on the target, here
                # 'finished' means 'failed before main command be executed'.
                if not result['finished']:
                    try:
                        self._connection.reset()
                    except AttributeError:
                        pass

                    for dummy in range(max_timeout):
                        time.sleep(1)
                        remaining_time -= 1
                        # - AnsibleConnectionFailure covers rejected requests (i.e.
                        #   by rules with '--jump REJECT')
                        # - ansible_timeout is able to cover dropped requests (due
                        #   to a rule or policy DROP) if not lower than async_val.
                        try:
                            dummy = self._low_level_execute_command(confirm_cmd, sudoable=self.DEFAULT_SUDOABLE)
                            break
                        except AnsibleConnectionFailure:
                            continue

                    result = merge_hash(result, self._async_result(async_status_args, task_vars, remaining_time))

                # Cleanup async related stuff and internal params
                for key in ('ansible_job_id', 'results_file', 'started', 'finished'):
                    if result.get(key):
                        del result[key]

                if result.get('invocation', {}).get('module_args'):
                    for key in ('_back', '_timeout', '_async_dir', 'jid'):
                        if result['invocation']['module_args'].get(key):
                            del result['invocation']['module_args'][key]

                async_status_args['mode'] = 'cleanup'
                dummy = self._async_result(async_status_args, task_vars, 0)

        if not wrap_async:
            # remove a temporary path we created
            self._remove_tmp_path(self._connection._shell.tmpdir)

        return result