1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
use bytes::{BufMut, BytesMut};
use std::{cmp, io, str, usize};
use tokio_io::_tokio_codec::{Decoder, Encoder};

/// A simple `Codec` implementation that splits up data into lines.
#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
pub struct LinesCodec {
    // Stored index of the next index to examine for a `\n` character.
    // This is used to optimize searching.
    // For example, if `decode` was called with `abc`, it would hold `3`,
    // because that is the next index to examine.
    // The next time `decode` is called with `abcde\n`, the method will
    // only look at `de\n` before returning.
    next_index: usize,

    /// The maximum length for a given line. If `usize::MAX`, lines will be
    /// read until a `\n` character is reached.
    max_length: usize,

    /// Are we currently discarding the remainder of a line which was over
    /// the length limit?
    is_discarding: bool,
}

impl LinesCodec {
    /// Returns a `LinesCodec` for splitting up data into lines.
    ///
    /// # Note
    ///
    /// The returned `LinesCodec` will not have an upper bound on the length
    /// of a buffered line. See the documentation for [`new_with_max_length`]
    /// for information on why this could be a potential security risk.
    ///
    /// [`new_with_max_length`]: #method.new_with_max_length
    pub fn new() -> LinesCodec {
        LinesCodec {
            next_index: 0,
            max_length: usize::MAX,
            is_discarding: false,
        }
    }

    /// Returns a `LinesCodec` with a maximum line length limit.
    ///
    /// If this is set, calls to `LinesCodec::decode` will return a
    /// [`LengthError`] when a line exceeds the length limit. Subsequent calls
    /// will discard up to `limit` bytes from that line until a newline
    /// character is reached, returning `None` until the line over the limit
    /// has been fully discarded. After that point, calls to `decode` will
    /// function as normal.
    ///
    /// # Note
    ///
    /// Setting a length limit is highly recommended for any `LinesCodec` which
    /// will be exposed to untrusted input. Otherwise, the size of the buffer
    /// that holds the line currently being read is unbounded. An attacker could
    /// exploit this unbounded buffer by sending an unbounded amount of input
    /// without any `\n` characters, causing unbounded memory consumption.
    ///
    /// [`LengthError`]: ../struct.LengthError
    pub fn new_with_max_length(max_length: usize) -> Self {
        LinesCodec {
            max_length,
            ..LinesCodec::new()
        }
    }

    /// Returns the maximum line length when decoding.
    ///
    /// ```
    /// use std::usize;
    /// use tokio_codec::LinesCodec;
    ///
    /// let codec = LinesCodec::new();
    /// assert_eq!(codec.max_length(), usize::MAX);
    /// ```
    /// ```
    /// use tokio_codec::LinesCodec;
    ///
    /// let codec = LinesCodec::new_with_max_length(256);
    /// assert_eq!(codec.max_length(), 256);
    /// ```
    pub fn max_length(&self) -> usize {
        self.max_length
    }

    fn discard(&mut self, newline_offset: Option<usize>, read_to: usize, buf: &mut BytesMut) {
        let discard_to = if let Some(offset) = newline_offset {
            // If we found a newline, discard up to that offset and
            // then stop discarding. On the next iteration, we'll try
            // to read a line normally.
            self.is_discarding = false;
            offset + self.next_index + 1
        } else {
            // Otherwise, we didn't find a newline, so we'll discard
            // everything we read. On the next iteration, we'll continue
            // discarding up to max_len bytes unless we find a newline.
            read_to
        };
        buf.advance(discard_to);
        self.next_index = 0;
    }
}

fn utf8(buf: &[u8]) -> Result<&str, io::Error> {
    str::from_utf8(buf)
        .map_err(|_| io::Error::new(io::ErrorKind::InvalidData, "Unable to decode input as UTF8"))
}

fn without_carriage_return(s: &[u8]) -> &[u8] {
    if let Some(&b'\r') = s.last() {
        &s[..s.len() - 1]
    } else {
        s
    }
}

impl Decoder for LinesCodec {
    type Item = String;
    // TODO: in the next breaking change, this should be changed to a custom
    // error type that indicates the "max length exceeded" condition better.
    type Error = io::Error;

    fn decode(&mut self, buf: &mut BytesMut) -> Result<Option<String>, io::Error> {
        loop {
            // Determine how far into the buffer we'll search for a newline. If
            // there's no max_length set, we'll read to the end of the buffer.
            let read_to = cmp::min(self.max_length.saturating_add(1), buf.len());

            let newline_offset = buf[self.next_index..read_to]
                .iter()
                .position(|b| *b == b'\n');

            if self.is_discarding {
                self.discard(newline_offset, read_to, buf);
            } else {
                return if let Some(offset) = newline_offset {
                    // Found a line!
                    let newline_index = offset + self.next_index;
                    self.next_index = 0;
                    let line = buf.split_to(newline_index + 1);
                    let line = &line[..line.len() - 1];
                    let line = without_carriage_return(line);
                    let line = utf8(line)?;

                    Ok(Some(line.to_string()))
                } else if buf.len() > self.max_length {
                    // Reached the maximum length without finding a
                    // newline, return an error and start discarding on the
                    // next call.
                    self.is_discarding = true;
                    Err(io::Error::new(
                        io::ErrorKind::Other,
                        "line length limit exceeded",
                    ))
                } else {
                    // We didn't find a line or reach the length limit, so the next
                    // call will resume searching at the current offset.
                    self.next_index = read_to;
                    Ok(None)
                };
            }
        }
    }

    fn decode_eof(&mut self, buf: &mut BytesMut) -> Result<Option<String>, io::Error> {
        Ok(match self.decode(buf)? {
            Some(frame) => Some(frame),
            None => {
                // No terminating newline - return remaining data, if any
                if buf.is_empty() || buf == &b"\r"[..] {
                    None
                } else {
                    let line = buf.take();
                    let line = without_carriage_return(&line);
                    let line = utf8(line)?;
                    self.next_index = 0;
                    Some(line.to_string())
                }
            }
        })
    }
}

impl Encoder for LinesCodec {
    type Item = String;
    type Error = io::Error;

    fn encode(&mut self, line: String, buf: &mut BytesMut) -> Result<(), io::Error> {
        buf.reserve(line.len() + 1);
        buf.put(line);
        buf.put_u8(b'\n');
        Ok(())
    }
}