hickory_proto/serialize/txt/
zone.rs

1// Copyright 2015-2023 Benjamin Fry <benjaminfry@me.com>
2//
3// Licensed under the Apache License, Version 2.0, <LICENSE-APACHE or
4// https://apache.org/licenses/LICENSE-2.0> or the MIT license <LICENSE-MIT or
5// https://opensource.org/licenses/MIT>, at your option. This file may not be
6// copied, modified, or distributed except according to those terms.
7
8use alloc::{
9    borrow::Cow,
10    collections::BTreeMap,
11    string::{String, ToString},
12    vec::Vec,
13};
14use core::{mem, str::FromStr};
15use std::{
16    fs,
17    path::{Path, PathBuf},
18};
19
20use crate::{
21    rr::{DNSClass, LowerName, Name, RData, Record, RecordSet, RecordType, RrKey},
22    serialize::txt::{
23        ParseError, ParseErrorKind, ParseResult,
24        parse_rdata::RDataParser,
25        zone_lex::{Lexer, Token},
26    },
27};
28
29/// ```text
30/// 5. ZONE FILES
31///
32/// Zone files are text files that contain RRs in text form.  Since the
33/// contents of a zone can be expressed in the form of a list of RRs a
34/// Zone File is most often used to define a zone, though it can be used
35/// to list a cache's contents.  Hence, this section first discusses the
36/// format of RRs in a Zone File, and then the special considerations when
37/// a Zone File is used to create a zone in some name server.
38///
39/// 5.1. Format
40///
41/// The format of these files is a sequence of entries.  Entries are
42/// predominantly line-oriented, though parentheses can be used to continue
43/// a list of items across a line boundary, and text literals can contain
44/// CRLF within the text.  Any combination of tabs and spaces act as a
45/// delimiter between the separate items that make up an entry.  The end of
46/// any line in the Zone File can end with a comment.  The comment starts
47/// with a ";" (semicolon).
48///
49/// The following entries are defined:
50///
51///     <blank>[<comment>]
52///
53///     $ORIGIN <domain-name> [<comment>]
54///
55///     $INCLUDE <file-name> [<domain-name>] [<comment>]
56///
57///     <domain-name><rr> [<comment>]
58///
59///     <blank><rr> [<comment>]
60///
61/// Blank lines, with or without comments, are allowed anywhere in the file.
62///
63/// Two control entries are defined: $ORIGIN and $INCLUDE.  $ORIGIN is
64/// followed by a domain name, and resets the current origin for relative
65/// domain names to the stated name.  $INCLUDE inserts the named file into
66/// the current file, and may optionally specify a domain name that sets the
67/// relative domain name origin for the included file.  $INCLUDE may also
68/// have a comment.  Note that a $INCLUDE entry never changes the relative
69/// origin of the parent file, regardless of changes to the relative origin
70/// made within the included file.
71///
72/// The last two forms represent RRs.  If an entry for an RR begins with a
73/// blank, then the RR is assumed to be owned by the last stated owner.  If
74/// an RR entry begins with a <domain-name>, then the owner name is reset.
75///
76/// <rr> contents take one of the following forms:
77///
78///     [<TTL>] [<class>] <type> <RDATA>
79///
80///     [<class>] [<TTL>] <type> <RDATA>
81///
82/// The RR begins with optional TTL and class fields, followed by a type and
83/// RDATA field appropriate to the type and class.  Class and type use the
84/// standard mnemonics, TTL is a decimal integer.  Omitted class and TTL
85/// values are default to the last explicitly stated values.  Since type and
86/// class mnemonics are disjoint, the parse is unique.  (Note that this
87/// order is different from the order used in examples and the order used in
88/// the actual RRs; the given order allows easier parsing and defaulting.)
89///
90/// <domain-name>s make up a large share of the data in the Zone File.
91/// The labels in the domain name are expressed as character strings and
92/// separated by dots.  Quoting conventions allow arbitrary characters to be
93/// stored in domain names.  Domain names that end in a dot are called
94/// absolute, and are taken as complete.  Domain names which do not end in a
95/// dot are called relative; the actual domain name is the concatenation of
96/// the relative part with an origin specified in a $ORIGIN, $INCLUDE, or as
97/// an argument to the Zone File loading routine.  A relative name is an
98/// error when no origin is available.
99///
100/// <character-string> is expressed in one or two ways: as a contiguous set
101/// of characters without interior spaces, or as a string beginning with a "
102/// and ending with a ".  Inside a " delimited string any character can
103/// occur, except for a " itself, which must be quoted using \ (back slash).
104///
105/// Because these files are text files several special encodings are
106/// necessary to allow arbitrary data to be loaded.  In particular:
107///
108///                 of the root.
109///
110/// @               A free standing @ is used to denote the current origin.
111///
112/// \X              where X is any character other than a digit (0-9), is
113///                 used to quote that character so that its special meaning
114///                 does not apply.  For example, "\." can be used to place
115///                 a dot character in a label.
116///
117/// \DDD            where each D is a digit is the octet corresponding to
118///                 the decimal number described by DDD.  The resulting
119///                 octet is assumed to be text and is not checked for
120///                 special meaning.
121///
122/// ( )             Parentheses are used to group data that crosses a line
123///                 boundary.  In effect, line terminations are not
124///                 recognized within parentheses.
125///
126/// ;               Semicolon is used to start a comment; the remainder of
127///                 the line is ignored.
128/// ```
129pub struct Parser<'a> {
130    lexers: Vec<(Lexer<'a>, Option<PathBuf>)>,
131    origin: Option<Name>,
132}
133
134impl<'a> Parser<'a> {
135    /// Returns a new Zone file parser
136    ///
137    /// The `path` argument's parent directory is used to resolve relative `$INCLUDE` paths.
138    /// Relative `$INCLUDE` paths will yield an error if `path` is `None`.
139    pub fn new(
140        input: impl Into<Cow<'a, str>>,
141        path: Option<PathBuf>,
142        origin: Option<Name>,
143    ) -> Self {
144        Self {
145            lexers: vec![(Lexer::new(input), path)],
146            origin,
147        }
148    }
149
150    /// Parse a file from the Lexer
151    ///
152    /// # Return
153    ///
154    /// A pair of the Zone origin name and a map of all Keys to RecordSets
155    pub fn parse(mut self) -> ParseResult<(Name, BTreeMap<RrKey, RecordSet>)> {
156        let mut origin = self.origin;
157        let mut records: BTreeMap<RrKey, RecordSet> = BTreeMap::new();
158        let mut class: DNSClass = DNSClass::IN;
159        let mut current_name: Option<Name> = None;
160        let mut rtype: Option<RecordType> = None;
161        let mut ttl: Option<u32> = None;
162        let mut state = State::StartLine;
163        let mut stack = self.lexers.len();
164
165        'outer: while let Some((lexer, path)) = self.lexers.last_mut() {
166            while let Some(t) = lexer.next_token()? {
167                state = match state {
168                    State::StartLine => {
169                        // current_name is not reset on the next line b/c it might be needed from the previous
170                        rtype = None;
171
172                        match t {
173                            // if Dollar, then $INCLUDE or $ORIGIN
174                            Token::Include => State::Include(None),
175                            Token::Origin => State::Origin,
176                            Token::Ttl => State::Ttl,
177
178                            // if CharData, then Name then ttl_class_type
179                            Token::CharData(data) => {
180                                current_name = Some(Name::parse(&data, origin.as_ref())?);
181                                State::TtlClassType
182                            }
183
184                            // @ is a placeholder for specifying the current origin
185                            Token::At => {
186                                current_name.clone_from(&origin); // TODO a COW or RC would reduce copies...
187                                State::TtlClassType
188                            }
189
190                            // if blank, then nothing or ttl_class_type
191                            Token::Blank => State::TtlClassType,
192                            Token::EOL => State::StartLine, // probably a comment
193                            _ => return Err(ParseErrorKind::UnexpectedToken(t).into()),
194                        }
195                    }
196                    State::Ttl => match t {
197                        Token::CharData(data) => {
198                            ttl = Some(Self::parse_time(&data)?);
199                            State::StartLine
200                        }
201                        _ => return Err(ParseErrorKind::UnexpectedToken(t).into()),
202                    },
203                    State::Origin => {
204                        match t {
205                            Token::CharData(data) => {
206                                // TODO an origin was specified, should this be legal? definitely confusing...
207                                origin = Some(Name::parse(&data, None)?);
208                                State::StartLine
209                            }
210                            _ => return Err(ParseErrorKind::UnexpectedToken(t).into()),
211                        }
212                    }
213                    State::Include(include_path) => match (t, include_path) {
214                        (Token::CharData(data), None) => State::Include(Some(data)),
215                        (Token::EOL, Some(include_path)) => {
216                            // RFC1035 (section 5) does not specify how filename for $INCLUDE
217                            // should be resolved into file path. The underlying code implements the
218                            // following:
219                            // * if the path is absolute (relies on Path::is_absolute), it uses normalized path
220                            // * otherwise, it joins the path with parent root of the current file
221                            //
222                            // TODO: Inlining files specified using non-relative path might potentially introduce
223                            // security issue in some cases (e.g. when working with zone files from untrusted sources)
224                            // and should probably be configurable by user.
225
226                            if stack > MAX_INCLUDE_LEVEL {
227                                return Err(ParseErrorKind::Message(
228                                    "Max depth level for nested $INCLUDE is reached",
229                                )
230                                .into());
231                            }
232
233                            let include = Path::new(&include_path);
234                            let include = match (include.is_absolute(), path) {
235                                (true, _) => include.to_path_buf(),
236                                (false, Some(path)) => path
237                                    .parent()
238                                    .expect("file has to have parent folder")
239                                    .join(include),
240                                (false, None) => {
241                                    return Err(ParseErrorKind::Message(
242                                        "Relative $INCLUDE is not supported",
243                                    )
244                                    .into());
245                                }
246                            };
247
248                            let input = fs::read_to_string(&include)?;
249                            let lexer = Lexer::new(input);
250                            self.lexers.push((lexer, Some(include)));
251                            stack += 1;
252                            state = State::StartLine;
253                            continue 'outer;
254                        }
255                        (Token::CharData(_), Some(_)) => {
256                            return Err(ParseErrorKind::Message(
257                                "Domain name for $INCLUDE is not supported",
258                            )
259                            .into());
260                        }
261                        (t, _) => {
262                            return Err(ParseErrorKind::UnexpectedToken(t).into());
263                        }
264                    },
265                    State::TtlClassType => {
266                        match t {
267                            // if number, TTL
268                            // Token::Number(num) => ttl = Some(*num),
269                            // One of Class or Type (these cannot be overlapping!)
270                            Token::CharData(mut data) => {
271                                // if it's a number it's a ttl
272                                let result: ParseResult<u32> = Self::parse_time(&data);
273                                if result.is_ok() {
274                                    ttl = result.ok();
275                                    State::TtlClassType // hm, should this go to just ClassType?
276                                } else {
277                                    // if can parse DNSClass, then class
278                                    data.make_ascii_uppercase();
279                                    let result = DNSClass::from_str(&data);
280                                    if let Ok(parsed) = result {
281                                        class = parsed;
282                                        State::TtlClassType
283                                    } else {
284                                        // if can parse RecordType, then RecordType
285                                        rtype = Some(RecordType::from_str(&data)?);
286                                        State::Record(vec![])
287                                    }
288                                }
289                            }
290                            // could be nothing if started with blank and is a comment, i.e. EOL
291                            Token::EOL => {
292                                State::StartLine // next line
293                            }
294                            _ => return Err(ParseErrorKind::UnexpectedToken(t).into()),
295                        }
296                    }
297                    State::Record(record_parts) => {
298                        // b/c of ownership rules, perhaps, just collect all the RData components as a list of
299                        //  tokens to pass into the processor
300                        match t {
301                            Token::EOL => {
302                                Self::flush_record(
303                                    record_parts,
304                                    &origin,
305                                    &current_name,
306                                    rtype,
307                                    &mut ttl,
308                                    class,
309                                    &mut records,
310                                )?;
311                                State::StartLine
312                            }
313                            Token::CharData(part) => {
314                                let mut record_parts = record_parts;
315                                record_parts.push(part);
316                                State::Record(record_parts)
317                            }
318                            // TODO: we should not tokenize the list...
319                            Token::List(list) => {
320                                let mut record_parts = record_parts;
321                                record_parts.extend(list);
322                                State::Record(record_parts)
323                            }
324                            _ => return Err(ParseErrorKind::UnexpectedToken(t).into()),
325                        }
326                    }
327                };
328            }
329
330            // Extra flush at the end for the case of missing endline
331            if let State::Record(record_parts) = mem::replace(&mut state, State::StartLine) {
332                Self::flush_record(
333                    record_parts,
334                    &origin,
335                    &current_name,
336                    rtype,
337                    &mut ttl,
338                    class,
339                    &mut records,
340                )?;
341            }
342
343            stack -= 1;
344            self.lexers.pop();
345        }
346
347        //
348        // build the Authority and return.
349        let origin = origin.ok_or_else(|| {
350            ParseError::from(ParseErrorKind::Message("$ORIGIN was not specified"))
351        })?;
352        Ok((origin, records))
353    }
354
355    fn flush_record(
356        record_parts: Vec<String>,
357        origin: &Option<Name>,
358        current_name: &Option<Name>,
359        rtype: Option<RecordType>,
360        ttl: &mut Option<u32>,
361        class: DNSClass,
362        records: &mut BTreeMap<RrKey, RecordSet>,
363    ) -> ParseResult<()> {
364        // call out to parsers for difference record types
365        // all tokens as part of the Record should be chardata...
366        let rtype = rtype.ok_or_else(|| {
367            ParseError::from(ParseErrorKind::Message("record type not specified"))
368        })?;
369        let rdata = RData::parse(
370            rtype,
371            record_parts.iter().map(AsRef::as_ref),
372            origin.as_ref(),
373        )?;
374
375        // verify that we have everything we need for the record
376        // TODO COW or RC would reduce mem usage, perhaps Name should have an intern()...
377        //  might want to wait until RC.weak() stabilizes, as that would be needed for global
378        //  memory where you want
379        let name = current_name.clone().ok_or_else(|| {
380            ParseError::from(ParseErrorKind::Message("record name not specified"))
381        })?;
382
383        // slightly annoying, need to grab the TTL, then move rdata into the record,
384        //  then check the Type again and have custom add logic.
385        let set_ttl = match rtype {
386            RecordType::SOA => {
387                // TTL for the SOA is set internally...
388                // expire is for the SOA, minimum is default for records
389                if let RData::SOA(soa) = &rdata {
390                    // TODO, this looks wrong, get_expire() should be get_minimum(), right?
391                    let set_ttl = soa.expire() as u32; // the spec seems a little inaccurate with u32 and i32
392                    if ttl.is_none() {
393                        *ttl = Some(soa.minimum());
394                    } // TODO: should this only set it if it's not set?
395                    set_ttl
396                } else {
397                    let msg = format!("Invalid RData here, expected SOA: {rdata:?}");
398                    return ParseResult::Err(ParseError::from(ParseErrorKind::Msg(msg)));
399                }
400            }
401            _ => ttl.ok_or_else(|| {
402                ParseError::from(ParseErrorKind::Message("record ttl not specified"))
403            })?,
404        };
405
406        // TODO: validate record, e.g. the name of SRV record allows _ but others do not.
407
408        // move the rdata into record...
409        let mut record = Record::from_rdata(name, set_ttl, rdata);
410        record.set_dns_class(class);
411
412        // add to the map
413        let key = RrKey::new(LowerName::new(record.name()), record.record_type());
414        match rtype {
415            RecordType::SOA => {
416                let set = record.into();
417                if records.insert(key, set).is_some() {
418                    return Err(ParseErrorKind::Message("SOA is already specified").into());
419                }
420            }
421            _ => {
422                // add a Vec if it's not there, then add the record to the list
423                let set = records.entry(key).or_insert_with(|| {
424                    RecordSet::new(record.name().clone(), record.record_type(), 0)
425                });
426                set.insert(record, 0);
427            }
428        }
429        Ok(())
430    }
431
432    /// parses the string following the rules from:
433    ///  <https://tools.ietf.org/html/rfc2308> (NXCaching RFC) and
434    ///  <https://www.zytrax.com/books/dns/apa/time.html>
435    ///
436    /// default is seconds
437    /// #s = seconds = # x 1 seconds (really!)
438    /// #m = minutes = # x 60 seconds
439    /// #h = hours   = # x 3600 seconds
440    /// #d = day     = # x 86400 seconds
441    /// #w = week    = # x 604800 seconds
442    ///
443    /// returns the result of the parsing or and error
444    ///
445    /// # Example
446    /// ```
447    /// use hickory_proto::serialize::txt::Parser;
448    ///
449    /// assert_eq!(Parser::parse_time("0").unwrap(),  0);
450    /// assert!(Parser::parse_time("s").is_err());
451    /// assert!(Parser::parse_time("").is_err());
452    /// assert_eq!(Parser::parse_time("0s").unwrap(), 0);
453    /// assert_eq!(Parser::parse_time("1").unwrap(),  1);
454    /// assert_eq!(Parser::parse_time("1S").unwrap(), 1);
455    /// assert_eq!(Parser::parse_time("1s").unwrap(), 1);
456    /// assert_eq!(Parser::parse_time("1M").unwrap(), 60);
457    /// assert_eq!(Parser::parse_time("1m").unwrap(), 60);
458    /// assert_eq!(Parser::parse_time("1H").unwrap(), 3600);
459    /// assert_eq!(Parser::parse_time("1h").unwrap(), 3600);
460    /// assert_eq!(Parser::parse_time("1D").unwrap(), 86400);
461    /// assert_eq!(Parser::parse_time("1d").unwrap(), 86400);
462    /// assert_eq!(Parser::parse_time("1W").unwrap(), 604800);
463    /// assert_eq!(Parser::parse_time("1w").unwrap(), 604800);
464    /// assert_eq!(Parser::parse_time("1s2d3w4h2m").unwrap(), 1+2*86400+3*604800+4*3600+2*60);
465    /// assert_eq!(Parser::parse_time("3w3w").unwrap(), 3*604800+3*604800);
466    /// assert!(Parser::parse_time("7102w").is_err());
467    /// ```
468    pub fn parse_time(ttl_str: &str) -> ParseResult<u32> {
469        if ttl_str.is_empty() {
470            return Err(ParseErrorKind::ParseTime(ttl_str.to_string()).into());
471        }
472
473        let (mut state, mut value) = (None, 0_u32);
474        for (i, c) in ttl_str.chars().enumerate() {
475            let start = match (state, c) {
476                (None, '0'..='9') => {
477                    state = Some(i);
478                    continue;
479                }
480                (Some(_), '0'..='9') => continue,
481                (Some(start), 'S' | 's' | 'M' | 'm' | 'H' | 'h' | 'D' | 'd' | 'W' | 'w') => start,
482                _ => return Err(ParseErrorKind::ParseTime(ttl_str.to_string()).into()),
483            };
484
485            // All allowed chars are ASCII, so using char indexes to slice &[u8] is OK
486            let number = u32::from_str(&ttl_str[start..i])
487                .map_err(|_| ParseErrorKind::ParseTime(ttl_str.to_string()))?;
488
489            let multiplier = match c {
490                'S' | 's' => 1,
491                'M' | 'm' => 60,
492                'H' | 'h' => 3_600,
493                'D' | 'd' => 86_400,
494                'W' | 'w' => 604_800,
495                _ => unreachable!(),
496            };
497
498            value = number
499                .checked_mul(multiplier)
500                .and_then(|add| value.checked_add(add))
501                .ok_or_else(|| ParseErrorKind::ParseTime(ttl_str.to_string()))?;
502
503            state = None;
504        }
505
506        if let Some(start) = state {
507            // All allowed chars are ASCII, so using char indexes to slice &[u8] is OK
508            let number = u32::from_str(&ttl_str[start..])
509                .map_err(|_| ParseErrorKind::ParseTime(ttl_str.to_string()))?;
510            value = value
511                .checked_add(number)
512                .ok_or_else(|| ParseErrorKind::ParseTime(ttl_str.to_string()))?;
513        }
514
515        Ok(value)
516    }
517}
518
519#[allow(unused)]
520enum State {
521    StartLine,    // start of line, @, $<WORD>, Name, Blank
522    TtlClassType, // [<TTL>] [<class>] <type>,
523    Ttl,          // $TTL <time>
524    Record(Vec<String>),
525    Include(Option<String>), // $INCLUDE <filename>
526    Origin,
527}
528
529/// Max traversal depth for $INCLUDE files
530const MAX_INCLUDE_LEVEL: usize = 256;
531
532#[cfg(test)]
533mod tests {
534    use alloc::string::ToString;
535
536    use super::*;
537
538    #[test]
539    #[allow(clippy::uninlined_format_args)]
540    fn test_zone_parse() {
541        let domain = Name::from_str("parameter.origin.org.").unwrap();
542
543        let zone_data = r#"$ORIGIN parsed.zone.origin.org.
544 faulty-record-type 60 IN A 1.2.3.4
545"#;
546
547        let result = Parser::new(zone_data, None, Some(domain)).parse();
548        assert!(
549            result.is_err()
550                & result
551                    .as_ref()
552                    .unwrap_err()
553                    .to_string()
554                    .contains("FAULTY-RECORD-TYPE"),
555            "unexpected success: {:#?}",
556            result
557        );
558    }
559}