1 | //! Frame a stream of bytes based on a length prefix |
2 | //! |
3 | //! Many protocols delimit their frames by prefacing frame data with a |
4 | //! frame head that specifies the length of the frame. The |
5 | //! `length_delimited` module provides utilities for handling the length |
6 | //! based framing. This allows the consumer to work with entire frames |
7 | //! without having to worry about buffering or other framing logic. |
8 | //! |
9 | //! # Getting started |
10 | //! |
11 | //! If implementing a protocol from scratch, using length delimited framing |
12 | //! is an easy way to get started. [`LengthDelimitedCodec::new()`] will |
13 | //! return a length delimited codec using default configuration values. |
14 | //! This can then be used to construct a framer to adapt a full-duplex |
15 | //! byte stream into a stream of frames. |
16 | //! |
17 | //! ``` |
18 | //! use tokio::io::{AsyncRead, AsyncWrite}; |
19 | //! use tokio_util::codec::{Framed, LengthDelimitedCodec}; |
20 | //! |
21 | //! fn bind_transport<T: AsyncRead + AsyncWrite>(io: T) |
22 | //! -> Framed<T, LengthDelimitedCodec> |
23 | //! { |
24 | //! Framed::new(io, LengthDelimitedCodec::new()) |
25 | //! } |
26 | //! # pub fn main() {} |
27 | //! ``` |
28 | //! |
29 | //! The returned transport implements `Sink + Stream` for `BytesMut`. It |
30 | //! encodes the frame with a big-endian `u32` header denoting the frame |
31 | //! payload length: |
32 | //! |
33 | //! ```text |
34 | //! +----------+--------------------------------+ |
35 | //! | len: u32 | frame payload | |
36 | //! +----------+--------------------------------+ |
37 | //! ``` |
38 | //! |
39 | //! Specifically, given the following: |
40 | //! |
41 | //! ``` |
42 | //! use tokio::io::{AsyncRead, AsyncWrite}; |
43 | //! use tokio_util::codec::{Framed, LengthDelimitedCodec}; |
44 | //! |
45 | //! use futures::SinkExt; |
46 | //! use bytes::Bytes; |
47 | //! |
48 | //! async fn write_frame<T>(io: T) -> Result<(), Box<dyn std::error::Error>> |
49 | //! where |
50 | //! T: AsyncRead + AsyncWrite + Unpin, |
51 | //! { |
52 | //! let mut transport = Framed::new(io, LengthDelimitedCodec::new()); |
53 | //! let frame = Bytes::from("hello world" ); |
54 | //! |
55 | //! transport.send(frame).await?; |
56 | //! Ok(()) |
57 | //! } |
58 | //! ``` |
59 | //! |
60 | //! The encoded frame will look like this: |
61 | //! |
62 | //! ```text |
63 | //! +---- len: u32 ----+---- data ----+ |
64 | //! | \x00\x00\x00\x0b | hello world | |
65 | //! +------------------+--------------+ |
66 | //! ``` |
67 | //! |
68 | //! # Decoding |
69 | //! |
70 | //! [`FramedRead`] adapts an [`AsyncRead`] into a `Stream` of [`BytesMut`], |
71 | //! such that each yielded [`BytesMut`] value contains the contents of an |
72 | //! entire frame. There are many configuration parameters enabling |
73 | //! [`FramedRead`] to handle a wide range of protocols. Here are some |
74 | //! examples that will cover the various options at a high level. |
75 | //! |
76 | //! ## Example 1 |
77 | //! |
78 | //! The following will parse a `u16` length field at offset 0, omitting the |
79 | //! frame head in the yielded `BytesMut`. |
80 | //! |
81 | //! ``` |
82 | //! # use tokio_stream::StreamExt; |
83 | //! # use tokio_util::codec::LengthDelimitedCodec; |
84 | //! # #[tokio::main] |
85 | //! # async fn main() { |
86 | //! # let io: &[u8] = b" \x00\x0BHello world" ; |
87 | //! let mut reader = LengthDelimitedCodec::builder() |
88 | //! .length_field_offset(0) // default value |
89 | //! .length_field_type::<u16>() |
90 | //! .length_adjustment(0) // default value |
91 | //! .new_read(io); |
92 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
93 | //! # assert_eq!(res, b"Hello world" ); |
94 | //! # } |
95 | //! ``` |
96 | //! |
97 | //! The following frame will be decoded as such: |
98 | //! |
99 | //! ```text |
100 | //! INPUT DECODED |
101 | //! +-- len ---+--- Payload ---+ +--- Payload ---+ |
102 | //! | \x00\x0B | Hello world | --> | Hello world | |
103 | //! +----------+---------------+ +---------------+ |
104 | //! ``` |
105 | //! |
106 | //! The value of the length field is 11 (`\x0B`) which represents the length |
107 | //! of the payload, `hello world`. By default, [`FramedRead`] assumes that |
108 | //! the length field represents the number of bytes that **follows** the |
109 | //! length field. Thus, the entire frame has a length of 13: 2 bytes for the |
110 | //! frame head + 11 bytes for the payload. |
111 | //! |
112 | //! ## Example 2 |
113 | //! |
114 | //! The following will parse a `u16` length field at offset 0, including the |
115 | //! frame head in the yielded `BytesMut`. |
116 | //! |
117 | //! ``` |
118 | //! # use tokio_stream::StreamExt; |
119 | //! # use tokio_util::codec::LengthDelimitedCodec; |
120 | //! # #[tokio::main] |
121 | //! # async fn main() { |
122 | //! # let io: &[u8] = b" \x00\x0BHello world" ; |
123 | //! let mut reader = LengthDelimitedCodec::builder() |
124 | //! .length_field_offset(0) // default value |
125 | //! .length_field_type::<u16>() |
126 | //! .length_adjustment(2) // Add head size to length |
127 | //! .num_skip(0) // Do NOT skip the head |
128 | //! .new_read(io); |
129 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
130 | //! # assert_eq!(res, b" \x00\x0BHello world" ); |
131 | //! # } |
132 | //! ``` |
133 | //! |
134 | //! The following frame will be decoded as such: |
135 | //! |
136 | //! ```text |
137 | //! INPUT DECODED |
138 | //! +-- len ---+--- Payload ---+ +-- len ---+--- Payload ---+ |
139 | //! | \x00\x0B | Hello world | --> | \x00\x0B | Hello world | |
140 | //! +----------+---------------+ +----------+---------------+ |
141 | //! ``` |
142 | //! |
143 | //! This is similar to the first example, the only difference is that the |
144 | //! frame head is **included** in the yielded `BytesMut` value. To achieve |
145 | //! this, we need to add the header size to the length with `length_adjustment`, |
146 | //! and set `num_skip` to `0` to prevent skipping the head. |
147 | //! |
148 | //! ## Example 3 |
149 | //! |
150 | //! The following will parse a `u16` length field at offset 0, omitting the |
151 | //! frame head in the yielded `BytesMut`. In this case, the length field |
152 | //! **includes** the frame head length. |
153 | //! |
154 | //! ``` |
155 | //! # use tokio_stream::StreamExt; |
156 | //! # use tokio_util::codec::LengthDelimitedCodec; |
157 | //! # #[tokio::main] |
158 | //! # async fn main() { |
159 | //! # let io: &[u8] = b" \x00\x0DHello world" ; |
160 | //! let mut reader = LengthDelimitedCodec::builder() |
161 | //! .length_field_offset(0) // default value |
162 | //! .length_field_type::<u16>() |
163 | //! .length_adjustment(-2) // size of head |
164 | //! .new_read(io); |
165 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
166 | //! # assert_eq!(res, b"Hello world" ); |
167 | //! # } |
168 | //! ``` |
169 | //! |
170 | //! The following frame will be decoded as such: |
171 | //! |
172 | //! ```text |
173 | //! INPUT DECODED |
174 | //! +-- len ---+--- Payload ---+ +--- Payload ---+ |
175 | //! | \x00\x0D | Hello world | --> | Hello world | |
176 | //! +----------+---------------+ +---------------+ |
177 | //! ``` |
178 | //! |
179 | //! In most cases, the length field represents the length of the payload |
180 | //! only, as shown in the previous examples. However, in some protocols the |
181 | //! length field represents the length of the whole frame, including the |
182 | //! head. In such cases, we specify a negative `length_adjustment` to adjust |
183 | //! the value provided in the frame head to represent the payload length. |
184 | //! |
185 | //! ## Example 4 |
186 | //! |
187 | //! The following will parse a 3 byte length field at offset 0 in a 5 byte |
188 | //! frame head, including the frame head in the yielded `BytesMut`. |
189 | //! |
190 | //! ``` |
191 | //! # use tokio_stream::StreamExt; |
192 | //! # use tokio_util::codec::LengthDelimitedCodec; |
193 | //! # #[tokio::main] |
194 | //! # async fn main() { |
195 | //! # let io: &[u8] = b" \x00\x00\x0B\xCA\xFEHello world" ; |
196 | //! let mut reader = LengthDelimitedCodec::builder() |
197 | //! .length_field_offset(0) // default value |
198 | //! .length_field_length(3) |
199 | //! .length_adjustment(3 + 2) // len field and remaining head |
200 | //! .num_skip(0) |
201 | //! .new_read(io); |
202 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
203 | //! # assert_eq!(res, b" \x00\x00\x0B\xCA\xFEHello world" ); |
204 | //! # } |
205 | //! ``` |
206 | //! |
207 | //! The following frame will be decoded as such: |
208 | //! |
209 | //! ```text |
210 | //! INPUT |
211 | //! +---- len -----+- head -+--- Payload ---+ |
212 | //! | \x00\x00\x0B | \xCAFE | Hello world | |
213 | //! +--------------+--------+---------------+ |
214 | //! |
215 | //! DECODED |
216 | //! +---- len -----+- head -+--- Payload ---+ |
217 | //! | \x00\x00\x0B | \xCAFE | Hello world | |
218 | //! +--------------+--------+---------------+ |
219 | //! ``` |
220 | //! |
221 | //! A more advanced example that shows a case where there is extra frame |
222 | //! head data between the length field and the payload. In such cases, it is |
223 | //! usually desirable to include the frame head as part of the yielded |
224 | //! `BytesMut`. This lets consumers of the length delimited framer to |
225 | //! process the frame head as needed. |
226 | //! |
227 | //! The positive `length_adjustment` value lets `FramedRead` factor in the |
228 | //! additional head into the frame length calculation. |
229 | //! |
230 | //! ## Example 5 |
231 | //! |
232 | //! The following will parse a `u16` length field at offset 1 of a 4 byte |
233 | //! frame head. The first byte and the length field will be omitted from the |
234 | //! yielded `BytesMut`, but the trailing 2 bytes of the frame head will be |
235 | //! included. |
236 | //! |
237 | //! ``` |
238 | //! # use tokio_stream::StreamExt; |
239 | //! # use tokio_util::codec::LengthDelimitedCodec; |
240 | //! # #[tokio::main] |
241 | //! # async fn main() { |
242 | //! # let io: &[u8] = b" \xCA\x00\x0B\xFEHello world" ; |
243 | //! let mut reader = LengthDelimitedCodec::builder() |
244 | //! .length_field_offset(1) // length of hdr1 |
245 | //! .length_field_type::<u16>() |
246 | //! .length_adjustment(1) // length of hdr2 |
247 | //! .num_skip(3) // length of hdr1 + LEN |
248 | //! .new_read(io); |
249 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
250 | //! # assert_eq!(res, b" \xFEHello world" ); |
251 | //! # } |
252 | //! ``` |
253 | //! |
254 | //! The following frame will be decoded as such: |
255 | //! |
256 | //! ```text |
257 | //! INPUT |
258 | //! +- hdr1 -+-- len ---+- hdr2 -+--- Payload ---+ |
259 | //! | \xCA | \x00\x0B | \xFE | Hello world | |
260 | //! +--------+----------+--------+---------------+ |
261 | //! |
262 | //! DECODED |
263 | //! +- hdr2 -+--- Payload ---+ |
264 | //! | \xFE | Hello world | |
265 | //! +--------+---------------+ |
266 | //! ``` |
267 | //! |
268 | //! The length field is situated in the middle of the frame head. In this |
269 | //! case, the first byte in the frame head could be a version or some other |
270 | //! identifier that is not needed for processing. On the other hand, the |
271 | //! second half of the head is needed. |
272 | //! |
273 | //! `length_field_offset` indicates how many bytes to skip before starting |
274 | //! to read the length field. `length_adjustment` is the number of bytes to |
275 | //! skip starting at the end of the length field. In this case, it is the |
276 | //! second half of the head. |
277 | //! |
278 | //! ## Example 6 |
279 | //! |
280 | //! The following will parse a `u16` length field at offset 1 of a 4 byte |
281 | //! frame head. The first byte and the length field will be omitted from the |
282 | //! yielded `BytesMut`, but the trailing 2 bytes of the frame head will be |
283 | //! included. In this case, the length field **includes** the frame head |
284 | //! length. |
285 | //! |
286 | //! ``` |
287 | //! # use tokio_stream::StreamExt; |
288 | //! # use tokio_util::codec::LengthDelimitedCodec; |
289 | //! # #[tokio::main] |
290 | //! # async fn main() { |
291 | //! # let io: &[u8] = b" \xCA\x00\x0F\xFEHello world" ; |
292 | //! let mut reader = LengthDelimitedCodec::builder() |
293 | //! .length_field_offset(1) // length of hdr1 |
294 | //! .length_field_type::<u16>() |
295 | //! .length_adjustment(-3) // length of hdr1 + LEN, negative |
296 | //! .num_skip(3) |
297 | //! .new_read(io); |
298 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
299 | //! # assert_eq!(res, b" \xFEHello world" ); |
300 | //! # } |
301 | //! ``` |
302 | //! |
303 | //! The following frame will be decoded as such: |
304 | //! |
305 | //! ```text |
306 | //! INPUT |
307 | //! +- hdr1 -+-- len ---+- hdr2 -+--- Payload ---+ |
308 | //! | \xCA | \x00\x0F | \xFE | Hello world | |
309 | //! +--------+----------+--------+---------------+ |
310 | //! |
311 | //! DECODED |
312 | //! +- hdr2 -+--- Payload ---+ |
313 | //! | \xFE | Hello world | |
314 | //! +--------+---------------+ |
315 | //! ``` |
316 | //! |
317 | //! Similar to the example above, the difference is that the length field |
318 | //! represents the length of the entire frame instead of just the payload. |
319 | //! The length of `hdr1` and `len` must be counted in `length_adjustment`. |
320 | //! Note that the length of `hdr2` does **not** need to be explicitly set |
321 | //! anywhere because it already is factored into the total frame length that |
322 | //! is read from the byte stream. |
323 | //! |
324 | //! ## Example 7 |
325 | //! |
326 | //! The following will parse a 3 byte length field at offset 0 in a 4 byte |
327 | //! frame head, excluding the 4th byte from the yielded `BytesMut`. |
328 | //! |
329 | //! ``` |
330 | //! # use tokio_stream::StreamExt; |
331 | //! # use tokio_util::codec::LengthDelimitedCodec; |
332 | //! # #[tokio::main] |
333 | //! # async fn main() { |
334 | //! # let io: &[u8] = b" \x00\x00\x0B\xFFHello world" ; |
335 | //! let mut reader = LengthDelimitedCodec::builder() |
336 | //! .length_field_offset(0) // default value |
337 | //! .length_field_length(3) |
338 | //! .length_adjustment(0) // default value |
339 | //! .num_skip(4) // skip the first 4 bytes |
340 | //! .new_read(io); |
341 | //! # let res = reader.next().await.unwrap().unwrap().to_vec(); |
342 | //! # assert_eq!(res, b"Hello world" ); |
343 | //! # } |
344 | //! ``` |
345 | //! |
346 | //! The following frame will be decoded as such: |
347 | //! |
348 | //! ```text |
349 | //! INPUT DECODED |
350 | //! +------- len ------+--- Payload ---+ +--- Payload ---+ |
351 | //! | \x00\x00\x0B\xFF | Hello world | => | Hello world | |
352 | //! +------------------+---------------+ +---------------+ |
353 | //! ``` |
354 | //! |
355 | //! A simple example where there are unused bytes between the length field |
356 | //! and the payload. |
357 | //! |
358 | //! # Encoding |
359 | //! |
360 | //! [`FramedWrite`] adapts an [`AsyncWrite`] into a `Sink` of [`BytesMut`], |
361 | //! such that each submitted [`BytesMut`] is prefaced by a length field. |
362 | //! There are fewer configuration options than [`FramedRead`]. Given |
363 | //! protocols that have more complex frame heads, an encoder should probably |
364 | //! be written by hand using [`Encoder`]. |
365 | //! |
366 | //! Here is a simple example, given a `FramedWrite` with the following |
367 | //! configuration: |
368 | //! |
369 | //! ``` |
370 | //! # use tokio::io::AsyncWrite; |
371 | //! # use tokio_util::codec::LengthDelimitedCodec; |
372 | //! # fn write_frame<T: AsyncWrite>(io: T) { |
373 | //! # let _ = |
374 | //! LengthDelimitedCodec::builder() |
375 | //! .length_field_type::<u16>() |
376 | //! .new_write(io); |
377 | //! # } |
378 | //! # pub fn main() {} |
379 | //! ``` |
380 | //! |
381 | //! A payload of `hello world` will be encoded as: |
382 | //! |
383 | //! ```text |
384 | //! +- len: u16 -+---- data ----+ |
385 | //! | \x00\x0b | hello world | |
386 | //! +------------+--------------+ |
387 | //! ``` |
388 | //! |
389 | //! [`LengthDelimitedCodec::new()`]: method@LengthDelimitedCodec::new |
390 | //! [`FramedRead`]: struct@FramedRead |
391 | //! [`FramedWrite`]: struct@FramedWrite |
392 | //! [`AsyncRead`]: trait@tokio::io::AsyncRead |
393 | //! [`AsyncWrite`]: trait@tokio::io::AsyncWrite |
394 | //! [`Encoder`]: trait@Encoder |
395 | //! [`BytesMut`]: bytes::BytesMut |
396 | |
397 | use crate::codec::{Decoder, Encoder, Framed, FramedRead, FramedWrite}; |
398 | |
399 | use tokio::io::{AsyncRead, AsyncWrite}; |
400 | |
401 | use bytes::{Buf, BufMut, Bytes, BytesMut}; |
402 | use std::error::Error as StdError; |
403 | use std::io::{self, Cursor}; |
404 | use std::{cmp, fmt, mem}; |
405 | |
406 | /// Configure length delimited `LengthDelimitedCodec`s. |
407 | /// |
408 | /// `Builder` enables constructing configured length delimited codecs. Note |
409 | /// that not all configuration settings apply to both encoding and decoding. See |
410 | /// the documentation for specific methods for more detail. |
411 | /// |
412 | /// Note that the if the value of [`Builder::max_frame_length`] becomes larger than |
413 | /// what can actually fit in [`Builder::length_field_length`], it will be clipped to |
414 | /// the maximum value that can fit. |
415 | #[derive (Debug, Clone, Copy)] |
416 | pub struct Builder { |
417 | // Maximum frame length |
418 | max_frame_len: usize, |
419 | |
420 | // Number of bytes representing the field length |
421 | length_field_len: usize, |
422 | |
423 | // Number of bytes in the header before the length field |
424 | length_field_offset: usize, |
425 | |
426 | // Adjust the length specified in the header field by this amount |
427 | length_adjustment: isize, |
428 | |
429 | // Total number of bytes to skip before reading the payload, if not set, |
430 | // `length_field_len + length_field_offset` |
431 | num_skip: Option<usize>, |
432 | |
433 | // Length field byte order (little or big endian) |
434 | length_field_is_big_endian: bool, |
435 | } |
436 | |
437 | /// An error when the number of bytes read is more than max frame length. |
438 | pub struct LengthDelimitedCodecError { |
439 | _priv: (), |
440 | } |
441 | |
442 | /// A codec for frames delimited by a frame head specifying their lengths. |
443 | /// |
444 | /// This allows the consumer to work with entire frames without having to worry |
445 | /// about buffering or other framing logic. |
446 | /// |
447 | /// See [module level] documentation for more detail. |
448 | /// |
449 | /// [module level]: index.html |
450 | #[derive (Debug, Clone)] |
451 | pub struct LengthDelimitedCodec { |
452 | // Configuration values |
453 | builder: Builder, |
454 | |
455 | // Read state |
456 | state: DecodeState, |
457 | } |
458 | |
459 | #[derive (Debug, Clone, Copy)] |
460 | enum DecodeState { |
461 | Head, |
462 | Data(usize), |
463 | } |
464 | |
465 | // ===== impl LengthDelimitedCodec ====== |
466 | |
467 | impl LengthDelimitedCodec { |
468 | /// Creates a new `LengthDelimitedCodec` with the default configuration values. |
469 | pub fn new() -> Self { |
470 | Self { |
471 | builder: Builder::new(), |
472 | state: DecodeState::Head, |
473 | } |
474 | } |
475 | |
476 | /// Creates a new length delimited codec builder with default configuration |
477 | /// values. |
478 | pub fn builder() -> Builder { |
479 | Builder::new() |
480 | } |
481 | |
482 | /// Returns the current max frame setting |
483 | /// |
484 | /// This is the largest size this codec will accept from the wire. Larger |
485 | /// frames will be rejected. |
486 | pub fn max_frame_length(&self) -> usize { |
487 | self.builder.max_frame_len |
488 | } |
489 | |
490 | /// Updates the max frame setting. |
491 | /// |
492 | /// The change takes effect the next time a frame is decoded. In other |
493 | /// words, if a frame is currently in process of being decoded with a frame |
494 | /// size greater than `val` but less than the max frame length in effect |
495 | /// before calling this function, then the frame will be allowed. |
496 | pub fn set_max_frame_length(&mut self, val: usize) { |
497 | self.builder.max_frame_length(val); |
498 | } |
499 | |
500 | fn decode_head(&mut self, src: &mut BytesMut) -> io::Result<Option<usize>> { |
501 | let head_len = self.builder.num_head_bytes(); |
502 | let field_len = self.builder.length_field_len; |
503 | |
504 | if src.len() < head_len { |
505 | // Not enough data |
506 | return Ok(None); |
507 | } |
508 | |
509 | let n = { |
510 | let mut src = Cursor::new(&mut *src); |
511 | |
512 | // Skip the required bytes |
513 | src.advance(self.builder.length_field_offset); |
514 | |
515 | // match endianness |
516 | let n = if self.builder.length_field_is_big_endian { |
517 | src.get_uint(field_len) |
518 | } else { |
519 | src.get_uint_le(field_len) |
520 | }; |
521 | |
522 | if n > self.builder.max_frame_len as u64 { |
523 | return Err(io::Error::new( |
524 | io::ErrorKind::InvalidData, |
525 | LengthDelimitedCodecError { _priv: () }, |
526 | )); |
527 | } |
528 | |
529 | // The check above ensures there is no overflow |
530 | let n = n as usize; |
531 | |
532 | // Adjust `n` with bounds checking |
533 | let n = if self.builder.length_adjustment < 0 { |
534 | n.checked_sub(-self.builder.length_adjustment as usize) |
535 | } else { |
536 | n.checked_add(self.builder.length_adjustment as usize) |
537 | }; |
538 | |
539 | // Error handling |
540 | match n { |
541 | Some(n) => n, |
542 | None => { |
543 | return Err(io::Error::new( |
544 | io::ErrorKind::InvalidInput, |
545 | "provided length would overflow after adjustment" , |
546 | )); |
547 | } |
548 | } |
549 | }; |
550 | |
551 | src.advance(self.builder.get_num_skip()); |
552 | |
553 | // Ensure that the buffer has enough space to read the incoming |
554 | // payload |
555 | src.reserve(n.saturating_sub(src.len())); |
556 | |
557 | Ok(Some(n)) |
558 | } |
559 | |
560 | fn decode_data(&self, n: usize, src: &mut BytesMut) -> Option<BytesMut> { |
561 | // At this point, the buffer has already had the required capacity |
562 | // reserved. All there is to do is read. |
563 | if src.len() < n { |
564 | return None; |
565 | } |
566 | |
567 | Some(src.split_to(n)) |
568 | } |
569 | } |
570 | |
571 | impl Decoder for LengthDelimitedCodec { |
572 | type Item = BytesMut; |
573 | type Error = io::Error; |
574 | |
575 | fn decode(&mut self, src: &mut BytesMut) -> io::Result<Option<BytesMut>> { |
576 | let n = match self.state { |
577 | DecodeState::Head => match self.decode_head(src)? { |
578 | Some(n) => { |
579 | self.state = DecodeState::Data(n); |
580 | n |
581 | } |
582 | None => return Ok(None), |
583 | }, |
584 | DecodeState::Data(n) => n, |
585 | }; |
586 | |
587 | match self.decode_data(n, src) { |
588 | Some(data) => { |
589 | // Update the decode state |
590 | self.state = DecodeState::Head; |
591 | |
592 | // Make sure the buffer has enough space to read the next head |
593 | src.reserve(self.builder.num_head_bytes().saturating_sub(src.len())); |
594 | |
595 | Ok(Some(data)) |
596 | } |
597 | None => Ok(None), |
598 | } |
599 | } |
600 | } |
601 | |
602 | impl Encoder<Bytes> for LengthDelimitedCodec { |
603 | type Error = io::Error; |
604 | |
605 | fn encode(&mut self, data: Bytes, dst: &mut BytesMut) -> Result<(), io::Error> { |
606 | let n = data.len(); |
607 | |
608 | if n > self.builder.max_frame_len { |
609 | return Err(io::Error::new( |
610 | io::ErrorKind::InvalidInput, |
611 | LengthDelimitedCodecError { _priv: () }, |
612 | )); |
613 | } |
614 | |
615 | // Adjust `n` with bounds checking |
616 | let n = if self.builder.length_adjustment < 0 { |
617 | n.checked_add(-self.builder.length_adjustment as usize) |
618 | } else { |
619 | n.checked_sub(self.builder.length_adjustment as usize) |
620 | }; |
621 | |
622 | let n = n.ok_or_else(|| { |
623 | io::Error::new( |
624 | io::ErrorKind::InvalidInput, |
625 | "provided length would overflow after adjustment" , |
626 | ) |
627 | })?; |
628 | |
629 | // Reserve capacity in the destination buffer to fit the frame and |
630 | // length field (plus adjustment). |
631 | dst.reserve(self.builder.length_field_len + n); |
632 | |
633 | if self.builder.length_field_is_big_endian { |
634 | dst.put_uint(n as u64, self.builder.length_field_len); |
635 | } else { |
636 | dst.put_uint_le(n as u64, self.builder.length_field_len); |
637 | } |
638 | |
639 | // Write the frame to the buffer |
640 | dst.extend_from_slice(&data[..]); |
641 | |
642 | Ok(()) |
643 | } |
644 | } |
645 | |
646 | impl Default for LengthDelimitedCodec { |
647 | fn default() -> Self { |
648 | Self::new() |
649 | } |
650 | } |
651 | |
652 | // ===== impl Builder ===== |
653 | |
654 | mod builder { |
655 | /// Types that can be used with `Builder::length_field_type`. |
656 | pub trait LengthFieldType {} |
657 | |
658 | impl LengthFieldType for u8 {} |
659 | impl LengthFieldType for u16 {} |
660 | impl LengthFieldType for u32 {} |
661 | impl LengthFieldType for u64 {} |
662 | |
663 | #[cfg (any( |
664 | target_pointer_width = "16" , |
665 | target_pointer_width = "32" , |
666 | target_pointer_width = "64" , |
667 | ))] |
668 | impl LengthFieldType for usize {} |
669 | } |
670 | |
671 | impl Builder { |
672 | /// Creates a new length delimited codec builder with default configuration |
673 | /// values. |
674 | /// |
675 | /// # Examples |
676 | /// |
677 | /// ``` |
678 | /// # use tokio::io::AsyncRead; |
679 | /// use tokio_util::codec::LengthDelimitedCodec; |
680 | /// |
681 | /// # fn bind_read<T: AsyncRead>(io: T) { |
682 | /// LengthDelimitedCodec::builder() |
683 | /// .length_field_offset(0) |
684 | /// .length_field_type::<u16>() |
685 | /// .length_adjustment(0) |
686 | /// .num_skip(0) |
687 | /// .new_read(io); |
688 | /// # } |
689 | /// # pub fn main() {} |
690 | /// ``` |
691 | pub fn new() -> Builder { |
692 | Builder { |
693 | // Default max frame length of 8MB |
694 | max_frame_len: 8 * 1_024 * 1_024, |
695 | |
696 | // Default byte length of 4 |
697 | length_field_len: 4, |
698 | |
699 | // Default to the header field being at the start of the header. |
700 | length_field_offset: 0, |
701 | |
702 | length_adjustment: 0, |
703 | |
704 | // Total number of bytes to skip before reading the payload, if not set, |
705 | // `length_field_len + length_field_offset` |
706 | num_skip: None, |
707 | |
708 | // Default to reading the length field in network (big) endian. |
709 | length_field_is_big_endian: true, |
710 | } |
711 | } |
712 | |
713 | /// Read the length field as a big endian integer |
714 | /// |
715 | /// This is the default setting. |
716 | /// |
717 | /// This configuration option applies to both encoding and decoding. |
718 | /// |
719 | /// # Examples |
720 | /// |
721 | /// ``` |
722 | /// # use tokio::io::AsyncRead; |
723 | /// use tokio_util::codec::LengthDelimitedCodec; |
724 | /// |
725 | /// # fn bind_read<T: AsyncRead>(io: T) { |
726 | /// LengthDelimitedCodec::builder() |
727 | /// .big_endian() |
728 | /// .new_read(io); |
729 | /// # } |
730 | /// # pub fn main() {} |
731 | /// ``` |
732 | pub fn big_endian(&mut self) -> &mut Self { |
733 | self.length_field_is_big_endian = true; |
734 | self |
735 | } |
736 | |
737 | /// Read the length field as a little endian integer |
738 | /// |
739 | /// The default setting is big endian. |
740 | /// |
741 | /// This configuration option applies to both encoding and decoding. |
742 | /// |
743 | /// # Examples |
744 | /// |
745 | /// ``` |
746 | /// # use tokio::io::AsyncRead; |
747 | /// use tokio_util::codec::LengthDelimitedCodec; |
748 | /// |
749 | /// # fn bind_read<T: AsyncRead>(io: T) { |
750 | /// LengthDelimitedCodec::builder() |
751 | /// .little_endian() |
752 | /// .new_read(io); |
753 | /// # } |
754 | /// # pub fn main() {} |
755 | /// ``` |
756 | pub fn little_endian(&mut self) -> &mut Self { |
757 | self.length_field_is_big_endian = false; |
758 | self |
759 | } |
760 | |
761 | /// Read the length field as a native endian integer |
762 | /// |
763 | /// The default setting is big endian. |
764 | /// |
765 | /// This configuration option applies to both encoding and decoding. |
766 | /// |
767 | /// # Examples |
768 | /// |
769 | /// ``` |
770 | /// # use tokio::io::AsyncRead; |
771 | /// use tokio_util::codec::LengthDelimitedCodec; |
772 | /// |
773 | /// # fn bind_read<T: AsyncRead>(io: T) { |
774 | /// LengthDelimitedCodec::builder() |
775 | /// .native_endian() |
776 | /// .new_read(io); |
777 | /// # } |
778 | /// # pub fn main() {} |
779 | /// ``` |
780 | pub fn native_endian(&mut self) -> &mut Self { |
781 | if cfg!(target_endian = "big" ) { |
782 | self.big_endian() |
783 | } else { |
784 | self.little_endian() |
785 | } |
786 | } |
787 | |
788 | /// Sets the max frame length in bytes |
789 | /// |
790 | /// This configuration option applies to both encoding and decoding. The |
791 | /// default value is 8MB. |
792 | /// |
793 | /// When decoding, the length field read from the byte stream is checked |
794 | /// against this setting **before** any adjustments are applied. When |
795 | /// encoding, the length of the submitted payload is checked against this |
796 | /// setting. |
797 | /// |
798 | /// When frames exceed the max length, an `io::Error` with the custom value |
799 | /// of the `LengthDelimitedCodecError` type will be returned. |
800 | /// |
801 | /// # Examples |
802 | /// |
803 | /// ``` |
804 | /// # use tokio::io::AsyncRead; |
805 | /// use tokio_util::codec::LengthDelimitedCodec; |
806 | /// |
807 | /// # fn bind_read<T: AsyncRead>(io: T) { |
808 | /// LengthDelimitedCodec::builder() |
809 | /// .max_frame_length(8 * 1024 * 1024) |
810 | /// .new_read(io); |
811 | /// # } |
812 | /// # pub fn main() {} |
813 | /// ``` |
814 | pub fn max_frame_length(&mut self, val: usize) -> &mut Self { |
815 | self.max_frame_len = val; |
816 | self |
817 | } |
818 | |
819 | /// Sets the unsigned integer type used to represent the length field. |
820 | /// |
821 | /// The default type is [`u32`]. The max type is [`u64`] (or [`usize`] on |
822 | /// 64-bit targets). |
823 | /// |
824 | /// # Examples |
825 | /// |
826 | /// ``` |
827 | /// # use tokio::io::AsyncRead; |
828 | /// use tokio_util::codec::LengthDelimitedCodec; |
829 | /// |
830 | /// # fn bind_read<T: AsyncRead>(io: T) { |
831 | /// LengthDelimitedCodec::builder() |
832 | /// .length_field_type::<u32>() |
833 | /// .new_read(io); |
834 | /// # } |
835 | /// # pub fn main() {} |
836 | /// ``` |
837 | /// |
838 | /// Unlike [`Builder::length_field_length`], this does not fail at runtime |
839 | /// and instead produces a compile error: |
840 | /// |
841 | /// ```compile_fail |
842 | /// # use tokio::io::AsyncRead; |
843 | /// # use tokio_util::codec::LengthDelimitedCodec; |
844 | /// # fn bind_read<T: AsyncRead>(io: T) { |
845 | /// LengthDelimitedCodec::builder() |
846 | /// .length_field_type::<u128>() |
847 | /// .new_read(io); |
848 | /// # } |
849 | /// # pub fn main() {} |
850 | /// ``` |
851 | pub fn length_field_type<T: builder::LengthFieldType>(&mut self) -> &mut Self { |
852 | self.length_field_length(mem::size_of::<T>()) |
853 | } |
854 | |
855 | /// Sets the number of bytes used to represent the length field |
856 | /// |
857 | /// The default value is `4`. The max value is `8`. |
858 | /// |
859 | /// This configuration option applies to both encoding and decoding. |
860 | /// |
861 | /// # Examples |
862 | /// |
863 | /// ``` |
864 | /// # use tokio::io::AsyncRead; |
865 | /// use tokio_util::codec::LengthDelimitedCodec; |
866 | /// |
867 | /// # fn bind_read<T: AsyncRead>(io: T) { |
868 | /// LengthDelimitedCodec::builder() |
869 | /// .length_field_length(4) |
870 | /// .new_read(io); |
871 | /// # } |
872 | /// # pub fn main() {} |
873 | /// ``` |
874 | pub fn length_field_length(&mut self, val: usize) -> &mut Self { |
875 | assert!(val > 0 && val <= 8, "invalid length field length" ); |
876 | self.length_field_len = val; |
877 | self |
878 | } |
879 | |
880 | /// Sets the number of bytes in the header before the length field |
881 | /// |
882 | /// This configuration option only applies to decoding. |
883 | /// |
884 | /// # Examples |
885 | /// |
886 | /// ``` |
887 | /// # use tokio::io::AsyncRead; |
888 | /// use tokio_util::codec::LengthDelimitedCodec; |
889 | /// |
890 | /// # fn bind_read<T: AsyncRead>(io: T) { |
891 | /// LengthDelimitedCodec::builder() |
892 | /// .length_field_offset(1) |
893 | /// .new_read(io); |
894 | /// # } |
895 | /// # pub fn main() {} |
896 | /// ``` |
897 | pub fn length_field_offset(&mut self, val: usize) -> &mut Self { |
898 | self.length_field_offset = val; |
899 | self |
900 | } |
901 | |
902 | /// Delta between the payload length specified in the header and the real |
903 | /// payload length |
904 | /// |
905 | /// # Examples |
906 | /// |
907 | /// ``` |
908 | /// # use tokio::io::AsyncRead; |
909 | /// use tokio_util::codec::LengthDelimitedCodec; |
910 | /// |
911 | /// # fn bind_read<T: AsyncRead>(io: T) { |
912 | /// LengthDelimitedCodec::builder() |
913 | /// .length_adjustment(-2) |
914 | /// .new_read(io); |
915 | /// # } |
916 | /// # pub fn main() {} |
917 | /// ``` |
918 | pub fn length_adjustment(&mut self, val: isize) -> &mut Self { |
919 | self.length_adjustment = val; |
920 | self |
921 | } |
922 | |
923 | /// Sets the number of bytes to skip before reading the payload |
924 | /// |
925 | /// Default value is `length_field_len + length_field_offset` |
926 | /// |
927 | /// This configuration option only applies to decoding |
928 | /// |
929 | /// # Examples |
930 | /// |
931 | /// ``` |
932 | /// # use tokio::io::AsyncRead; |
933 | /// use tokio_util::codec::LengthDelimitedCodec; |
934 | /// |
935 | /// # fn bind_read<T: AsyncRead>(io: T) { |
936 | /// LengthDelimitedCodec::builder() |
937 | /// .num_skip(4) |
938 | /// .new_read(io); |
939 | /// # } |
940 | /// # pub fn main() {} |
941 | /// ``` |
942 | pub fn num_skip(&mut self, val: usize) -> &mut Self { |
943 | self.num_skip = Some(val); |
944 | self |
945 | } |
946 | |
947 | /// Create a configured length delimited `LengthDelimitedCodec` |
948 | /// |
949 | /// # Examples |
950 | /// |
951 | /// ``` |
952 | /// use tokio_util::codec::LengthDelimitedCodec; |
953 | /// # pub fn main() { |
954 | /// LengthDelimitedCodec::builder() |
955 | /// .length_field_offset(0) |
956 | /// .length_field_type::<u16>() |
957 | /// .length_adjustment(0) |
958 | /// .num_skip(0) |
959 | /// .new_codec(); |
960 | /// # } |
961 | /// ``` |
962 | pub fn new_codec(&self) -> LengthDelimitedCodec { |
963 | let mut builder = *self; |
964 | |
965 | builder.adjust_max_frame_len(); |
966 | |
967 | LengthDelimitedCodec { |
968 | builder, |
969 | state: DecodeState::Head, |
970 | } |
971 | } |
972 | |
973 | /// Create a configured length delimited `FramedRead` |
974 | /// |
975 | /// # Examples |
976 | /// |
977 | /// ``` |
978 | /// # use tokio::io::AsyncRead; |
979 | /// use tokio_util::codec::LengthDelimitedCodec; |
980 | /// |
981 | /// # fn bind_read<T: AsyncRead>(io: T) { |
982 | /// LengthDelimitedCodec::builder() |
983 | /// .length_field_offset(0) |
984 | /// .length_field_type::<u16>() |
985 | /// .length_adjustment(0) |
986 | /// .num_skip(0) |
987 | /// .new_read(io); |
988 | /// # } |
989 | /// # pub fn main() {} |
990 | /// ``` |
991 | pub fn new_read<T>(&self, upstream: T) -> FramedRead<T, LengthDelimitedCodec> |
992 | where |
993 | T: AsyncRead, |
994 | { |
995 | FramedRead::new(upstream, self.new_codec()) |
996 | } |
997 | |
998 | /// Create a configured length delimited `FramedWrite` |
999 | /// |
1000 | /// # Examples |
1001 | /// |
1002 | /// ``` |
1003 | /// # use tokio::io::AsyncWrite; |
1004 | /// # use tokio_util::codec::LengthDelimitedCodec; |
1005 | /// # fn write_frame<T: AsyncWrite>(io: T) { |
1006 | /// LengthDelimitedCodec::builder() |
1007 | /// .length_field_type::<u16>() |
1008 | /// .new_write(io); |
1009 | /// # } |
1010 | /// # pub fn main() {} |
1011 | /// ``` |
1012 | pub fn new_write<T>(&self, inner: T) -> FramedWrite<T, LengthDelimitedCodec> |
1013 | where |
1014 | T: AsyncWrite, |
1015 | { |
1016 | FramedWrite::new(inner, self.new_codec()) |
1017 | } |
1018 | |
1019 | /// Create a configured length delimited `Framed` |
1020 | /// |
1021 | /// # Examples |
1022 | /// |
1023 | /// ``` |
1024 | /// # use tokio::io::{AsyncRead, AsyncWrite}; |
1025 | /// # use tokio_util::codec::LengthDelimitedCodec; |
1026 | /// # fn write_frame<T: AsyncRead + AsyncWrite>(io: T) { |
1027 | /// # let _ = |
1028 | /// LengthDelimitedCodec::builder() |
1029 | /// .length_field_type::<u16>() |
1030 | /// .new_framed(io); |
1031 | /// # } |
1032 | /// # pub fn main() {} |
1033 | /// ``` |
1034 | pub fn new_framed<T>(&self, inner: T) -> Framed<T, LengthDelimitedCodec> |
1035 | where |
1036 | T: AsyncRead + AsyncWrite, |
1037 | { |
1038 | Framed::new(inner, self.new_codec()) |
1039 | } |
1040 | |
1041 | fn num_head_bytes(&self) -> usize { |
1042 | let num = self.length_field_offset + self.length_field_len; |
1043 | cmp::max(num, self.num_skip.unwrap_or(0)) |
1044 | } |
1045 | |
1046 | fn get_num_skip(&self) -> usize { |
1047 | self.num_skip |
1048 | .unwrap_or(self.length_field_offset + self.length_field_len) |
1049 | } |
1050 | |
1051 | fn adjust_max_frame_len(&mut self) { |
1052 | // Calculate the maximum number that can be represented using `length_field_len` bytes. |
1053 | let max_number = match 1u64.checked_shl((8 * self.length_field_len) as u32) { |
1054 | Some(shl) => shl - 1, |
1055 | None => u64::MAX, |
1056 | }; |
1057 | |
1058 | let max_allowed_len = max_number.saturating_add_signed(self.length_adjustment as i64); |
1059 | |
1060 | if self.max_frame_len as u64 > max_allowed_len { |
1061 | self.max_frame_len = usize::try_from(max_allowed_len).unwrap_or(usize::MAX); |
1062 | } |
1063 | } |
1064 | } |
1065 | |
1066 | impl Default for Builder { |
1067 | fn default() -> Self { |
1068 | Self::new() |
1069 | } |
1070 | } |
1071 | |
1072 | // ===== impl LengthDelimitedCodecError ===== |
1073 | |
1074 | impl fmt::Debug for LengthDelimitedCodecError { |
1075 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
1076 | f.debug_struct(name:"LengthDelimitedCodecError" ).finish() |
1077 | } |
1078 | } |
1079 | |
1080 | impl fmt::Display for LengthDelimitedCodecError { |
1081 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
1082 | f.write_str(data:"frame size too big" ) |
1083 | } |
1084 | } |
1085 | |
1086 | impl StdError for LengthDelimitedCodecError {} |
1087 | |