Files
BanGUI/backend/app/utils/config_parser.py
Lukas 6e4797d71e feat: config file parser + writer utilities with full test coverage (Tasks 4.1, 4.2)
- Fail2BanConfigParser class: merge order, include directives (before/after),
  variable interpolation %(var)s, split_multiline, ordered_conf_files
- config_writer: write_local_override, remove_local_key, delete_local_file
  with atomic writes (os.replace), per-file threading locks, .local-only guard
- 79 tests in tests/test_utils/ (all passing)
- mypy --strict: 60 source files, 0 errors
- ruff: all checks passed
2026-03-13 19:38:03 +01:00

359 lines
13 KiB
Python

"""Fail2ban INI-style config parser with include and interpolation support.
Provides a :class:`Fail2BanConfigParser` class that wraps Python's
:class:`configparser.RawConfigParser` with fail2ban-specific behaviour:
- **Merge order**: ``.conf`` file first, then ``.local`` overlay, then ``*.d/``
directory overrides — each subsequent layer overwrites earlier values.
- **Include directives**: ``[INCLUDES]`` sections can specify ``before`` and
``after`` filenames. ``before`` is loaded at lower priority (loaded first),
``after`` at higher priority (loaded last). Both are resolved relative to
the directory of the including file. Circular includes and runaway recursion
are detected and logged.
- **Variable interpolation**: :meth:`interpolate` resolves ``%(variable)s``
references using the ``[DEFAULT]`` section, the ``[Init]`` section, and any
caller-supplied variables. Multiple passes handle nested references.
- **Multi-line values**: Handled transparently by ``configparser``; the
:meth:`split_multiline` helper further strips blank lines and ``#`` comments.
- **Comments**: ``configparser`` strips full-line ``#``/``;`` comments; inline
comments inside multi-line values are stripped by :meth:`split_multiline`.
All methods are synchronous. Call from async contexts via
:func:`asyncio.get_event_loop().run_in_executor`.
"""
from __future__ import annotations
import configparser
import re
from typing import TYPE_CHECKING
import structlog
if TYPE_CHECKING:
from pathlib import Path
log: structlog.stdlib.BoundLogger = structlog.get_logger()
# Compiled pattern that matches fail2ban-style %(variable_name)s references.
_INTERPOLATE_RE: re.Pattern[str] = re.compile(r"%\((\w+)\)s")
# Guard against infinite interpolation loops.
_MAX_INTERPOLATION_PASSES: int = 10
class Fail2BanConfigParser:
"""Parse fail2ban INI config files with include resolution and interpolation.
Typical usage for a ``filter.d/`` file::
parser = Fail2BanConfigParser(config_dir=Path("/etc/fail2ban"))
parser.read_with_overrides(Path("/etc/fail2ban/filter.d/sshd.conf"))
section = parser.section_dict("Definition")
failregex = parser.split_multiline(section.get("failregex", ""))
Args:
config_dir: Optional fail2ban configuration root directory. Used only
by :meth:`ordered_conf_files`; pass ``None`` if not needed.
max_include_depth: Maximum ``[INCLUDES]`` nesting depth before giving up.
"""
def __init__(
self,
config_dir: Path | None = None,
max_include_depth: int = 10,
) -> None:
self._config_dir = config_dir
self._max_include_depth = max_include_depth
self._parser: configparser.RawConfigParser = self._make_parser()
# Tracks resolved absolute paths to detect include cycles.
self._read_paths: set[Path] = set()
# ------------------------------------------------------------------
# Internal helpers
# ------------------------------------------------------------------
@staticmethod
def _make_parser() -> configparser.RawConfigParser:
"""Return a case-sensitive :class:`configparser.RawConfigParser`."""
parser = configparser.RawConfigParser(interpolation=None, strict=False)
# Keep original key casing (fail2ban is case-sensitive in option names).
parser.optionxform = str # type: ignore[assignment]
return parser
def _get_include(
self,
include_dir: Path,
tmp_parser: configparser.RawConfigParser,
key: str,
) -> Path | None:
"""Return the resolved path for an include directive, or ``None``."""
if not tmp_parser.has_section("INCLUDES"):
return None
if not tmp_parser.has_option("INCLUDES", key):
return None
raw = tmp_parser.get("INCLUDES", key).strip()
if not raw:
return None
return include_dir / raw
# ------------------------------------------------------------------
# Public interface — reading files
# ------------------------------------------------------------------
def read_file(self, path: Path, _depth: int = 0) -> None:
"""Read *path*, following ``[INCLUDES]`` ``before``/``after`` directives.
``before`` references are loaded before the current file (lower
priority); ``after`` references are loaded after (higher priority).
Circular includes are detected by tracking resolved absolute paths.
Args:
path: Config file to read.
_depth: Current include nesting depth. Internal parameter.
"""
if _depth > self._max_include_depth:
log.warning(
"include_depth_exceeded",
path=str(path),
max_depth=self._max_include_depth,
)
return
resolved = path.resolve()
if resolved in self._read_paths:
log.debug("include_cycle_detected", path=str(path))
return
try:
content = path.read_text(encoding="utf-8")
except OSError as exc:
log.warning("config_read_error", path=str(path), error=str(exc))
return
# Pre-scan for includes without yet committing to the main parser.
tmp = self._make_parser()
try:
tmp.read_string(content)
except configparser.Error as exc:
log.warning("config_parse_error", path=str(path), error=str(exc))
return
include_dir = path.parent
before_path = self._get_include(include_dir, tmp, "before")
after_path = self._get_include(include_dir, tmp, "after")
# Load ``before`` first (lower priority than current file).
if before_path is not None:
self.read_file(before_path, _depth=_depth + 1)
# Mark this path visited *before* merging to guard against cycles
# introduced by the ``after`` include referencing the same file.
self._read_paths.add(resolved)
# Merge current file into the accumulating parser.
try:
self._parser.read_string(content, source=str(path))
except configparser.Error as exc:
log.warning(
"config_parse_string_error", path=str(path), error=str(exc)
)
# Load ``after`` last (highest priority).
if after_path is not None:
self.read_file(after_path, _depth=_depth + 1)
def read_with_overrides(self, conf_path: Path) -> None:
"""Read *conf_path* and its ``.local`` override if it exists.
The ``.local`` file is read after the ``.conf`` file so its values
take precedence. Include directives inside each file are still honoured.
Args:
conf_path: Path to the ``.conf`` file. The corresponding
``.local`` is derived by replacing the suffix with ``.local``.
"""
self.read_file(conf_path)
local_path = conf_path.with_suffix(".local")
if local_path.is_file():
self.read_file(local_path)
# ------------------------------------------------------------------
# Public interface — querying parsed data
# ------------------------------------------------------------------
def sections(self) -> list[str]:
"""Return all section names (excludes the ``[DEFAULT]`` pseudo-section).
Returns:
Sorted list of section names present in the parsed files.
"""
return list(self._parser.sections())
def has_section(self, section: str) -> bool:
"""Return whether *section* exists in the parsed configuration.
Args:
section: Section name to check.
"""
return self._parser.has_section(section)
def get(self, section: str, key: str) -> str | None:
"""Return the raw value for *key* in *section*, or ``None``.
Args:
section: Section name.
key: Option name.
Returns:
Raw option value string, or ``None`` if not present.
"""
if self._parser.has_section(section) and self._parser.has_option(
section, key
):
return self._parser.get(section, key)
return None
def section_dict(
self,
section: str,
*,
skip: frozenset[str] | None = None,
) -> dict[str, str]:
"""Return all key-value pairs from *section* as a plain :class:`dict`.
Keys whose names start with ``__`` (configparser internals from
``DEFAULT`` inheritance) are always excluded.
Args:
section: Section name to read.
skip: Additional key names to exclude.
Returns:
Mapping of option name → raw value. Empty dict if section absent.
"""
if not self._parser.has_section(section):
return {}
drop: frozenset[str] = skip or frozenset()
return {
k: v
for k, v in self._parser.items(section)
if not k.startswith("__") and k not in drop
}
def defaults(self) -> dict[str, str]:
"""Return all ``[DEFAULT]`` section key-value pairs.
Returns:
Dict of default keys and their values.
"""
return dict(self._parser.defaults())
# ------------------------------------------------------------------
# Public interface — interpolation and helpers
# ------------------------------------------------------------------
def interpolate(
self,
value: str,
extra_vars: dict[str, str] | None = None,
) -> str:
"""Resolve ``%(variable)s`` references in *value*.
Variables are resolved in the following priority order (low → high):
1. ``[DEFAULT]`` section values.
2. ``[Init]`` section values (fail2ban action parameters).
3. *extra_vars* provided by the caller.
Multiple passes are performed to handle nested references (up to
:data:`_MAX_INTERPOLATION_PASSES` iterations). Unresolvable references
are left unchanged.
Args:
value: Raw string possibly containing ``%(name)s`` placeholders.
extra_vars: Optional caller-supplied variables (highest priority).
Returns:
String with ``%(name)s`` references substituted where possible.
"""
vars_: dict[str, str] = {}
vars_.update(self.defaults())
vars_.update(self.section_dict("Init"))
if extra_vars:
vars_.update(extra_vars)
def _sub(m: re.Match[str]) -> str:
return vars_.get(m.group(1), m.group(0))
result = value
for _ in range(_MAX_INTERPOLATION_PASSES):
new = _INTERPOLATE_RE.sub(_sub, result)
if new == result:
break
result = new
return result
@staticmethod
def split_multiline(raw: str) -> list[str]:
"""Split a multi-line INI value into individual non-blank lines.
Each line is stripped of surrounding whitespace. Lines that are empty
or that start with ``#`` (comments) are discarded.
Used for ``failregex``, ``ignoreregex``, ``action``, and ``logpath``
values which fail2ban allows to span multiple lines.
Args:
raw: Raw multi-line string from configparser.
Returns:
List of stripped, non-empty, non-comment strings.
"""
result: list[str] = []
for line in raw.splitlines():
stripped = line.strip()
if stripped and not stripped.startswith("#"):
result.append(stripped)
return result
# ------------------------------------------------------------------
# Class-level utility — file ordering
# ------------------------------------------------------------------
@classmethod
def ordered_conf_files(cls, config_dir: Path, base_name: str) -> list[Path]:
"""Return config files for *base_name* in fail2ban merge order.
Merge order (ascending priority — later entries override earlier):
1. ``{config_dir}/{base_name}.conf``
2. ``{config_dir}/{base_name}.local``
3. ``{config_dir}/{base_name}.d/*.conf`` (sorted alphabetically)
4. ``{config_dir}/{base_name}.d/*.local`` (sorted alphabetically)
Args:
config_dir: Fail2ban configuration root directory.
base_name: Config base name without extension (e.g. ``"jail"``).
Returns:
List of existing :class:`~pathlib.Path` objects in ascending
priority order (only files that actually exist are included).
"""
files: list[Path] = []
conf = config_dir / f"{base_name}.conf"
if conf.is_file():
files.append(conf)
local = config_dir / f"{base_name}.local"
if local.is_file():
files.append(local)
d_dir = config_dir / f"{base_name}.d"
if d_dir.is_dir():
files.extend(sorted(d_dir.glob("*.conf")))
files.extend(sorted(d_dir.glob("*.local")))
return files