1
0
Fork 0
mirror of https://github.com/actix/actix-web.git synced 2024-06-11 01:39:33 +00:00
actix-web/src/h1/decoder.rs

500 lines
17 KiB
Rust
Raw Normal View History

2018-04-29 05:20:32 +00:00
use std::{io, mem};
use bytes::{Bytes, BytesMut};
use futures::{Async, Poll};
use httparse;
use error::ParseError;
use http::header::{HeaderName, HeaderValue};
use http::{header, HttpTryFrom, Method, Uri, Version};
2018-10-05 03:02:10 +00:00
use request::{MessageFlags, Request, RequestPool};
2018-04-29 05:20:32 +00:00
use uri::Url;
const MAX_BUFFER_SIZE: usize = 131_072;
const MAX_HEADERS: usize = 96;
pub(crate) struct H1Decoder {
decoder: Option<EncodingDecoder>,
2018-10-04 23:22:00 +00:00
pool: &'static RequestPool,
2018-04-29 05:20:32 +00:00
}
2018-10-02 02:18:24 +00:00
#[derive(Debug)]
2018-10-05 03:02:10 +00:00
pub enum InMessage {
2018-10-04 23:22:00 +00:00
Message(Request),
MessageWithPayload(Request),
2018-04-29 05:20:32 +00:00
Chunk(Bytes),
Eof,
}
impl H1Decoder {
2018-10-04 23:22:00 +00:00
pub fn new(pool: &'static RequestPool) -> H1Decoder {
H1Decoder {
pool,
decoder: None,
}
2018-04-29 05:20:32 +00:00
}
2018-10-05 03:02:10 +00:00
pub fn decode(
&mut self, src: &mut BytesMut,
) -> Result<Option<InMessage>, ParseError> {
2018-04-29 05:20:32 +00:00
// read payload
if self.decoder.is_some() {
match self.decoder.as_mut().unwrap().decode(src)? {
2018-10-05 03:02:10 +00:00
Async::Ready(Some(bytes)) => return Ok(Some(InMessage::Chunk(bytes))),
2018-04-29 05:20:32 +00:00
Async::Ready(None) => {
self.decoder.take();
2018-10-05 03:02:10 +00:00
return Ok(Some(InMessage::Eof));
2018-04-29 05:20:32 +00:00
}
Async::NotReady => return Ok(None),
}
}
2018-10-04 23:22:00 +00:00
match self.parse_message(src)? {
2018-04-29 05:20:32 +00:00
Async::Ready((msg, decoder)) => {
2018-05-17 17:58:08 +00:00
self.decoder = decoder;
2018-10-04 23:22:00 +00:00
if self.decoder.is_some() {
2018-10-05 03:02:10 +00:00
Ok(Some(InMessage::MessageWithPayload(msg)))
2018-10-04 23:22:00 +00:00
} else {
2018-10-05 03:02:10 +00:00
Ok(Some(InMessage::Message(msg)))
2018-10-04 23:22:00 +00:00
}
2018-04-29 05:20:32 +00:00
}
Async::NotReady => {
if src.len() >= MAX_BUFFER_SIZE {
error!("MAX_BUFFER_SIZE unprocessed data reached, closing");
2018-10-04 23:22:00 +00:00
Err(ParseError::TooLarge)
2018-04-29 05:20:32 +00:00
} else {
Ok(None)
}
}
}
}
2018-10-04 23:22:00 +00:00
fn parse_message(
&self, buf: &mut BytesMut,
2018-06-25 04:58:04 +00:00
) -> Poll<(Request, Option<EncodingDecoder>), ParseError> {
2018-04-29 05:20:32 +00:00
// Parse http message
let mut has_upgrade = false;
let mut chunked = false;
let mut content_length = None;
let msg = {
2018-06-23 06:28:55 +00:00
// Unsafe: we read only this data only after httparse parses headers into.
// performance bump for pipeline benchmarks.
let mut headers: [HeaderIndex; MAX_HEADERS] =
2018-04-29 05:20:32 +00:00
unsafe { mem::uninitialized() };
let (len, method, path, version, headers_len) = {
2018-06-23 06:28:55 +00:00
let mut parsed: [httparse::Header; MAX_HEADERS] =
unsafe { mem::uninitialized() };
let mut req = httparse::Request::new(&mut parsed);
match req.parse(buf)? {
2018-04-29 05:20:32 +00:00
httparse::Status::Complete(len) => {
let method = Method::from_bytes(req.method.unwrap().as_bytes())
.map_err(|_| ParseError::Method)?;
let path = Url::new(Uri::try_from(req.path.unwrap())?);
let version = if req.version.unwrap() == 1 {
Version::HTTP_11
} else {
Version::HTTP_10
};
2018-06-23 06:28:55 +00:00
HeaderIndex::record(buf, req.headers, &mut headers);
2018-04-29 05:20:32 +00:00
(len, method, path, version, req.headers.len())
}
httparse::Status::Partial => return Ok(Async::NotReady),
}
};
let slice = buf.split_to(len).freeze();
// convert headers
2018-10-04 23:22:00 +00:00
let mut msg = RequestPool::get(self.pool);
2018-04-29 05:20:32 +00:00
{
2018-07-04 16:52:49 +00:00
let inner = msg.inner_mut();
2018-06-25 04:58:04 +00:00
inner
2018-05-01 05:04:24 +00:00
.flags
2018-06-25 04:58:04 +00:00
.get_mut()
2018-05-01 05:04:24 +00:00
.set(MessageFlags::KEEPALIVE, version != Version::HTTP_10);
2018-04-29 05:20:32 +00:00
2018-06-23 06:28:55 +00:00
for idx in headers[..headers_len].iter() {
if let Ok(name) =
HeaderName::from_bytes(&slice[idx.name.0..idx.name.1])
{
// Unsafe: httparse check header value for valid utf-8
2018-04-29 05:20:32 +00:00
let value = unsafe {
HeaderValue::from_shared_unchecked(
2018-06-23 06:28:55 +00:00
slice.slice(idx.value.0, idx.value.1),
2018-04-29 05:20:32 +00:00
)
};
match name {
header::CONTENT_LENGTH => {
if let Ok(s) = value.to_str() {
if let Ok(len) = s.parse::<u64>() {
2018-05-17 17:58:08 +00:00
content_length = Some(len);
2018-04-29 05:20:32 +00:00
} else {
debug!("illegal Content-Length: {:?}", len);
return Err(ParseError::Header);
}
} else {
debug!("illegal Content-Length: {:?}", len);
return Err(ParseError::Header);
}
}
// transfer-encoding
header::TRANSFER_ENCODING => {
if let Ok(s) = value.to_str() {
chunked = s.to_lowercase().contains("chunked");
} else {
return Err(ParseError::Header);
}
}
// connection keep-alive state
header::CONNECTION => {
2018-05-01 05:04:24 +00:00
let ka = if let Ok(conn) = value.to_str() {
2018-04-29 05:20:32 +00:00
if version == Version::HTTP_10
&& conn.contains("keep-alive")
{
true
} else {
2018-08-23 16:48:01 +00:00
version == Version::HTTP_11 && !(conn
.contains("close")
|| conn.contains("upgrade"))
2018-04-29 05:20:32 +00:00
}
} else {
false
};
2018-06-25 04:58:04 +00:00
inner.flags.get_mut().set(MessageFlags::KEEPALIVE, ka);
2018-04-29 05:20:32 +00:00
}
2018-07-05 07:16:16 +00:00
header::UPGRADE => {
has_upgrade = true;
}
2018-04-29 05:20:32 +00:00
_ => (),
}
2018-06-25 04:58:04 +00:00
inner.headers.append(name, value);
2018-04-29 05:20:32 +00:00
} else {
return Err(ParseError::Header);
}
}
2018-06-25 04:58:04 +00:00
inner.url = path;
inner.method = method;
inner.version = version;
2018-04-29 05:20:32 +00:00
}
msg
};
// https://tools.ietf.org/html/rfc7230#section-3.3.3
let decoder = if chunked {
// Chunked encoding
Some(EncodingDecoder::chunked())
} else if let Some(len) = content_length {
// Content-Length
Some(EncodingDecoder::length(len))
2018-06-25 04:58:04 +00:00
} else if has_upgrade || msg.inner.method == Method::CONNECT {
2018-04-29 05:20:32 +00:00
// upgrade(websocket) or connect
Some(EncodingDecoder::eof())
} else {
None
};
Ok(Async::Ready((msg, decoder)))
}
}
2018-06-23 06:28:55 +00:00
#[derive(Clone, Copy)]
pub(crate) struct HeaderIndex {
pub(crate) name: (usize, usize),
pub(crate) value: (usize, usize),
}
impl HeaderIndex {
pub(crate) fn record(
bytes: &[u8], headers: &[httparse::Header], indices: &mut [HeaderIndex],
) {
let bytes_ptr = bytes.as_ptr() as usize;
for (header, indices) in headers.iter().zip(indices.iter_mut()) {
let name_start = header.name.as_ptr() as usize - bytes_ptr;
let name_end = name_start + header.name.len();
indices.name = (name_start, name_end);
let value_start = header.value.as_ptr() as usize - bytes_ptr;
let value_end = value_start + header.value.len();
indices.value = (value_start, value_end);
}
}
}
2018-04-29 05:20:32 +00:00
/// Decoders to handle different Transfer-Encodings.
///
/// If a message body does not include a Transfer-Encoding, it *should*
/// include a Content-Length header.
#[derive(Debug, Clone, PartialEq)]
pub struct EncodingDecoder {
kind: Kind,
}
impl EncodingDecoder {
pub fn length(x: u64) -> EncodingDecoder {
EncodingDecoder {
kind: Kind::Length(x),
}
}
pub fn chunked() -> EncodingDecoder {
EncodingDecoder {
kind: Kind::Chunked(ChunkedState::Size, 0),
}
}
pub fn eof() -> EncodingDecoder {
EncodingDecoder {
kind: Kind::Eof(false),
}
}
}
#[derive(Debug, Clone, PartialEq)]
enum Kind {
/// A Reader used when a Content-Length header is passed with a positive
/// integer.
Length(u64),
/// A Reader used when Transfer-Encoding is `chunked`.
Chunked(ChunkedState, u64),
/// A Reader used for responses that don't indicate a length or chunked.
///
/// Note: This should only used for `Response`s. It is illegal for a
/// `Request` to be made with both `Content-Length` and
/// `Transfer-Encoding: chunked` missing, as explained from the spec:
///
/// > If a Transfer-Encoding header field is present in a response and
/// > the chunked transfer coding is not the final encoding, the
/// > message body length is determined by reading the connection until
/// > it is closed by the server. If a Transfer-Encoding header field
/// > is present in a request and the chunked transfer coding is not
/// > the final encoding, the message body length cannot be determined
/// > reliably; the server MUST respond with the 400 (Bad Request)
/// > status code and then close the connection.
Eof(bool),
}
#[derive(Debug, PartialEq, Clone)]
enum ChunkedState {
Size,
SizeLws,
Extension,
SizeLf,
Body,
BodyCr,
BodyLf,
EndCr,
EndLf,
End,
}
impl EncodingDecoder {
pub fn decode(&mut self, body: &mut BytesMut) -> Poll<Option<Bytes>, io::Error> {
match self.kind {
Kind::Length(ref mut remaining) => {
if *remaining == 0 {
Ok(Async::Ready(None))
} else {
if body.is_empty() {
return Ok(Async::NotReady);
}
let len = body.len() as u64;
let buf;
if *remaining > len {
buf = body.take().freeze();
*remaining -= len;
} else {
buf = body.split_to(*remaining as usize).freeze();
*remaining = 0;
}
trace!("Length read: {}", buf.len());
Ok(Async::Ready(Some(buf)))
}
}
Kind::Chunked(ref mut state, ref mut size) => {
loop {
let mut buf = None;
// advances the chunked state
*state = try_ready!(state.step(body, size, &mut buf));
if *state == ChunkedState::End {
trace!("End of chunked stream");
return Ok(Async::Ready(None));
}
if let Some(buf) = buf {
return Ok(Async::Ready(Some(buf)));
}
if body.is_empty() {
return Ok(Async::NotReady);
}
}
}
Kind::Eof(ref mut is_eof) => {
if *is_eof {
Ok(Async::Ready(None))
} else if !body.is_empty() {
Ok(Async::Ready(Some(body.take().freeze())))
} else {
Ok(Async::NotReady)
}
}
}
}
}
macro_rules! byte (
($rdr:ident) => ({
if $rdr.len() > 0 {
let b = $rdr[0];
$rdr.split_to(1);
b
} else {
return Ok(Async::NotReady)
}
})
);
impl ChunkedState {
fn step(
2018-05-17 19:20:20 +00:00
&self, body: &mut BytesMut, size: &mut u64, buf: &mut Option<Bytes>,
2018-04-29 05:20:32 +00:00
) -> Poll<ChunkedState, io::Error> {
use self::ChunkedState::*;
match *self {
Size => ChunkedState::read_size(body, size),
SizeLws => ChunkedState::read_size_lws(body),
Extension => ChunkedState::read_extension(body),
SizeLf => ChunkedState::read_size_lf(body, size),
Body => ChunkedState::read_body(body, size, buf),
BodyCr => ChunkedState::read_body_cr(body),
BodyLf => ChunkedState::read_body_lf(body),
EndCr => ChunkedState::read_end_cr(body),
EndLf => ChunkedState::read_end_lf(body),
End => Ok(Async::Ready(ChunkedState::End)),
}
}
fn read_size(rdr: &mut BytesMut, size: &mut u64) -> Poll<ChunkedState, io::Error> {
let radix = 16;
match byte!(rdr) {
b @ b'0'...b'9' => {
*size *= radix;
*size += u64::from(b - b'0');
}
b @ b'a'...b'f' => {
*size *= radix;
*size += u64::from(b + 10 - b'a');
}
b @ b'A'...b'F' => {
*size *= radix;
*size += u64::from(b + 10 - b'A');
}
b'\t' | b' ' => return Ok(Async::Ready(ChunkedState::SizeLws)),
b';' => return Ok(Async::Ready(ChunkedState::Extension)),
b'\r' => return Ok(Async::Ready(ChunkedState::SizeLf)),
_ => {
return Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk size line: Invalid Size",
));
}
}
Ok(Async::Ready(ChunkedState::Size))
}
fn read_size_lws(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
trace!("read_size_lws");
match byte!(rdr) {
// LWS can follow the chunk size, but no more digits can come
b'\t' | b' ' => Ok(Async::Ready(ChunkedState::SizeLws)),
b';' => Ok(Async::Ready(ChunkedState::Extension)),
b'\r' => Ok(Async::Ready(ChunkedState::SizeLf)),
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk size linear white space",
)),
}
}
fn read_extension(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\r' => Ok(Async::Ready(ChunkedState::SizeLf)),
_ => Ok(Async::Ready(ChunkedState::Extension)), // no supported extensions
}
}
fn read_size_lf(
2018-05-17 19:20:20 +00:00
rdr: &mut BytesMut, size: &mut u64,
2018-04-29 05:20:32 +00:00
) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\n' if *size > 0 => Ok(Async::Ready(ChunkedState::Body)),
b'\n' if *size == 0 => Ok(Async::Ready(ChunkedState::EndCr)),
2018-04-29 16:09:08 +00:00
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk size LF",
)),
2018-04-29 05:20:32 +00:00
}
}
fn read_body(
2018-05-17 19:20:20 +00:00
rdr: &mut BytesMut, rem: &mut u64, buf: &mut Option<Bytes>,
2018-04-29 05:20:32 +00:00
) -> Poll<ChunkedState, io::Error> {
trace!("Chunked read, remaining={:?}", rem);
let len = rdr.len() as u64;
if len == 0 {
Ok(Async::Ready(ChunkedState::Body))
} else {
let slice;
if *rem > len {
slice = rdr.take().freeze();
*rem -= len;
} else {
slice = rdr.split_to(*rem as usize).freeze();
*rem = 0;
}
*buf = Some(slice);
if *rem > 0 {
Ok(Async::Ready(ChunkedState::Body))
} else {
Ok(Async::Ready(ChunkedState::BodyCr))
}
}
}
fn read_body_cr(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\r' => Ok(Async::Ready(ChunkedState::BodyLf)),
2018-04-29 16:09:08 +00:00
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk body CR",
)),
2018-04-29 05:20:32 +00:00
}
}
fn read_body_lf(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\n' => Ok(Async::Ready(ChunkedState::Size)),
2018-04-29 16:09:08 +00:00
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk body LF",
)),
2018-04-29 05:20:32 +00:00
}
}
fn read_end_cr(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\r' => Ok(Async::Ready(ChunkedState::EndLf)),
2018-04-29 16:09:08 +00:00
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk end CR",
)),
2018-04-29 05:20:32 +00:00
}
}
fn read_end_lf(rdr: &mut BytesMut) -> Poll<ChunkedState, io::Error> {
match byte!(rdr) {
b'\n' => Ok(Async::Ready(ChunkedState::End)),
2018-04-29 16:09:08 +00:00
_ => Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Invalid chunk end LF",
)),
2018-04-29 05:20:32 +00:00
}
}
}