| // Copyright 2016 `multipart` Crate Developers | |
| // | |
| // Licensed under the Apache License, Version 2.0, <LICENSE-APACHE or | |
| // http://apache.org/licenses/LICENSE-2.0> or the MIT license <LICENSE-MIT or | |
| // http://opensource.org/licenses/MIT>, at your option. This file may not be | |
| // copied, modified, or distributed except according to those terms. | |
| //! Boundary parsing for `multipart` requests. | |
| use ::safemem; | |
| use super::buf_redux::BufReader; | |
| use super::buf_redux::policy::MinBuffered; | |
| use super::twoway; | |
| use std::cmp; | |
| use std::borrow::Borrow; | |
| use std::io; | |
| use std::io::prelude::*; | |
| use self::State::*; | |
| pub const MIN_BUF_SIZE: usize = 1024; | |
| #[derive(Debug, PartialEq, Eq)] | |
| enum State { | |
| Searching, | |
| BoundaryRead, | |
| AtEnd | |
| } | |
| /// A struct implementing `Read` and `BufRead` that will yield bytes until it sees a given sequence. | |
| #[derive(Debug)] | |
| pub struct BoundaryReader<R> { | |
| source: BufReader<R, MinBuffered>, | |
| boundary: Vec<u8>, | |
| search_idx: usize, | |
| state: State, | |
| } | |
| impl<R> BoundaryReader<R> where R: Read { | |
| /// Internal API | |
| pub fn from_reader<B: Into<Vec<u8>>>(reader: R, boundary: B) -> BoundaryReader<R> { | |
| let mut boundary = boundary.into(); | |
| safemem::prepend(b"--", &mut boundary); | |
| let source = BufReader::new(reader).set_policy(MinBuffered(MIN_BUF_SIZE)); | |
| BoundaryReader { | |
| source, | |
| boundary, | |
| search_idx: 0, | |
| state: Searching, | |
| } | |
| } | |
| fn read_to_boundary(&mut self) -> io::Result<&[u8]> { | |
| let buf = self.source.fill_buf()?; | |
| trace!("Buf: {:?}", String::from_utf8_lossy(buf)); | |
| debug!("Before search Buf len: {} Search idx: {} State: {:?}", | |
| buf.len(), self.search_idx, self.state); | |
| if self.state == BoundaryRead || self.state == AtEnd { | |
| return Ok(&buf[..self.search_idx]) | |
| } | |
| if self.state == Searching && self.search_idx < buf.len() { | |
| let lookahead = &buf[self.search_idx..]; | |
| // Look for the boundary, or if it isn't found, stop near the end. | |
| match find_boundary(lookahead, &self.boundary) { | |
| Ok(found_idx) => { | |
| self.search_idx += found_idx; | |
| self.state = BoundaryRead; | |
| }, | |
| Err(yield_len) => { | |
| self.search_idx += yield_len; | |
| } | |
| } | |
| } | |
| debug!("After search Buf len: {} Search idx: {} State: {:?}", | |
| buf.len(), self.search_idx, self.state); | |
| // back up the cursor to before the boundary's preceding CRLF if we haven't already | |
| if self.search_idx >= 2 && !buf[self.search_idx..].starts_with(b"\r\n") { | |
| let two_bytes_before = &buf[self.search_idx - 2 .. self.search_idx]; | |
| trace!("Two bytes before: {:?} ({:?}) (\"\\r\\n\": {:?})", | |
| String::from_utf8_lossy(two_bytes_before), two_bytes_before, b"\r\n"); | |
| if two_bytes_before == *b"\r\n" { | |
| debug!("Subtract two!"); | |
| self.search_idx -= 2; | |
| } | |
| } | |
| let ret_buf = &buf[..self.search_idx]; | |
| trace!("Returning buf: {:?}", String::from_utf8_lossy(ret_buf)); | |
| Ok(ret_buf) | |
| } | |
| pub fn set_min_buf_size(&mut self, min_buf_size: usize) { | |
| // ensure the minimum buf size is at least enough to find a boundary with some extra | |
| let min_buf_size = cmp::max(self.boundary.len() * 2, min_buf_size); | |
| self.source.policy_mut().0 = min_buf_size; | |
| } | |
| pub fn consume_boundary(&mut self) -> io::Result<bool> { | |
| if self.state == AtEnd { | |
| return Ok(false); | |
| } | |
| while self.state == Searching { | |
| debug!("Boundary not found yet"); | |
| let buf_len = self.read_to_boundary()?.len(); | |
| if buf_len == 0 && self.state == Searching { | |
| return Err(io::Error::new(io::ErrorKind::UnexpectedEof, | |
| "unexpected end of request body")); | |
| } | |
| debug!("Discarding {} bytes", buf_len); | |
| self.consume(buf_len); | |
| } | |
| let consume_amt = { | |
| let buf = self.source.fill_buf()?; | |
| // if the boundary is found we should have at least this much in-buffer | |
| let mut consume_amt = self.search_idx + self.boundary.len(); | |
| // we don't care about data before the cursor | |
| let bnd_segment = &buf[self.search_idx..]; | |
| if bnd_segment.starts_with(b"\r\n") { | |
| // preceding CRLF needs to be consumed as well | |
| consume_amt += 2; | |
| // assert that we've found the boundary after the CRLF | |
| debug_assert_eq!(*self.boundary, bnd_segment[2 .. self.boundary.len() + 2]); | |
| } else { | |
| // assert that we've found the boundary | |
| debug_assert_eq!(*self.boundary, bnd_segment[..self.boundary.len()]); | |
| } | |
| // include the trailing CRLF or -- | |
| consume_amt += 2; | |
| if buf.len() < consume_amt { | |
| return Err(io::Error::new(io::ErrorKind::UnexpectedEof, | |
| "not enough bytes to verify boundary")); | |
| } | |
| // we have enough bytes to verify | |
| self.state = Searching; | |
| let last_two = &buf[consume_amt - 2 .. consume_amt]; | |
| match last_two { | |
| b"\r\n" => self.state = Searching, | |
| b"--" => self.state = AtEnd, | |
| _ => return Err(io::Error::new( | |
| io::ErrorKind::InvalidData, | |
| format!("unexpected bytes following multipart boundary: {:X} {:X}", | |
| last_two[0], last_two[1]) | |
| )), | |
| } | |
| consume_amt | |
| }; | |
| trace!("Consuming {} bytes, remaining buf: {:?}", | |
| consume_amt, | |
| String::from_utf8_lossy(self.source.buffer())); | |
| self.source.consume(consume_amt); | |
| if cfg!(debug_assertions) { | |
| } | |
| self.search_idx = 0; | |
| trace!("Consumed boundary (state: {:?}), remaining buf: {:?}", self.state, | |
| String::from_utf8_lossy(self.source.buffer())); | |
| Ok(self.state != AtEnd) | |
| } | |
| } | |
| /// Find the boundary occurrence or the highest length to safely yield | |
| fn find_boundary(buf: &[u8], boundary: &[u8]) -> Result<usize, usize> { | |
| if let Some(idx) = twoway::find_bytes(buf, boundary) { | |
| return Ok(idx); | |
| } | |
| let search_start = buf.len().saturating_sub(boundary.len()); | |
| // search for just the boundary fragment | |
| for i in search_start .. buf.len() { | |
| if boundary.starts_with(&buf[i..]) { | |
| return Err(i); | |
| } | |
| } | |
| Err(buf.len()) | |
| } | |
| #[cfg(feature = "bench")] | |
| impl<'a> BoundaryReader<io::Cursor<&'a [u8]>> { | |
| fn new_with_bytes(bytes: &'a [u8], boundary: &str) -> Self { | |
| Self::from_reader(io::Cursor::new(bytes), boundary) | |
| } | |
| fn reset(&mut self) { | |
| // Dump buffer and reset cursor | |
| self.source.seek(io::SeekFrom::Start(0)); | |
| self.state = Searching; | |
| self.search_idx = 0; | |
| } | |
| } | |
| impl<R> Borrow<R> for BoundaryReader<R> { | |
| fn borrow(&self) -> &R { | |
| self.source.get_ref() | |
| } | |
| } | |
| impl<R> Read for BoundaryReader<R> where R: Read { | |
| fn read(&mut self, out: &mut [u8]) -> io::Result<usize> { | |
| let read = { | |
| let mut buf = self.read_to_boundary()?; | |
| // This shouldn't ever be an error so unwrapping is fine. | |
| buf.read(out).unwrap() | |
| }; | |
| self.consume(read); | |
| Ok(read) | |
| } | |
| } | |
| impl<R> BufRead for BoundaryReader<R> where R: Read { | |
| fn fill_buf(&mut self) -> io::Result<&[u8]> { | |
| self.read_to_boundary() | |
| } | |
| fn consume(&mut self, amt: usize) { | |
| let true_amt = cmp::min(amt, self.search_idx); | |
| debug!("Consume! amt: {} true amt: {}", amt, true_amt); | |
| self.source.consume(true_amt); | |
| self.search_idx -= true_amt; | |
| } | |
| } | |
| #[cfg(test)] | |
| mod test { | |
| use super::BoundaryReader; | |
| use std::io; | |
| use std::io::prelude::*; | |
| const BOUNDARY: &'static str = "boundary"; | |
| const TEST_VAL: &'static str = "--boundary\r\n\ | |
| dashed-value-1\r\n\ | |
| --boundary\r\n\ | |
| dashed-value-2\r\n\ | |
| --boundary--"; | |
| #[test] | |
| fn test_boundary() { | |
| ::init_log(); | |
| debug!("Testing boundary (no split)"); | |
| let src = &mut TEST_VAL.as_bytes(); | |
| let mut reader = BoundaryReader::from_reader(src, BOUNDARY); | |
| let mut buf = String::new(); | |
| test_boundary_reader(&mut reader, &mut buf); | |
| } | |
| struct SplitReader<'a> { | |
| left: &'a [u8], | |
| right: &'a [u8], | |
| } | |
| impl<'a> SplitReader<'a> { | |
| fn split(data: &'a [u8], at: usize) -> SplitReader<'a> { | |
| let (left, right) = data.split_at(at); | |
| SplitReader { | |
| left: left, | |
| right: right, | |
| } | |
| } | |
| } | |
| impl<'a> Read for SplitReader<'a> { | |
| fn read(&mut self, dst: &mut [u8]) -> io::Result<usize> { | |
| fn copy_bytes_partial(src: &mut &[u8], dst: &mut [u8]) -> usize { | |
| src.read(dst).unwrap() | |
| } | |
| let mut copy_amt = copy_bytes_partial(&mut self.left, dst); | |
| if copy_amt == 0 { | |
| copy_amt = copy_bytes_partial(&mut self.right, dst) | |
| }; | |
| Ok(copy_amt) | |
| } | |
| } | |
| #[test] | |
| fn test_split_boundary() { | |
| ::init_log(); | |
| debug!("Testing boundary (split)"); | |
| let mut buf = String::new(); | |
| // Substitute for `.step_by()` being unstable. | |
| for split_at in 0 .. TEST_VAL.len(){ | |
| debug!("Testing split at: {}", split_at); | |
| let src = SplitReader::split(TEST_VAL.as_bytes(), split_at); | |
| let mut reader = BoundaryReader::from_reader(src, BOUNDARY); | |
| test_boundary_reader(&mut reader, &mut buf); | |
| } | |
| } | |
| fn test_boundary_reader<R: Read>(reader: &mut BoundaryReader<R>, buf: &mut String) { | |
| buf.clear(); | |
| debug!("Read 1"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert!(buf.is_empty(), "Buffer not empty: {:?}", buf); | |
| buf.clear(); | |
| debug!("Consume 1"); | |
| reader.consume_boundary().unwrap(); | |
| debug!("Read 2"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "dashed-value-1"); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| reader.consume_boundary().unwrap(); | |
| debug!("Read 3"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "dashed-value-2"); | |
| buf.clear(); | |
| debug!("Consume 3"); | |
| reader.consume_boundary().unwrap(); | |
| debug!("Read 4"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, ""); | |
| } | |
| #[test] | |
| fn test_empty_body() { | |
| ::init_log(); | |
| // empty body contains closing boundary only | |
| let mut body: &[u8] = b"--boundary--"; | |
| let ref mut buf = String::new(); | |
| let mut reader = BoundaryReader::from_reader(&mut body, BOUNDARY); | |
| debug!("Consume 1"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| debug!("Read 1"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, ""); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| } | |
| #[test] | |
| fn test_leading_crlf() { | |
| ::init_log(); | |
| let mut body: &[u8] = b"\r\n\r\n--boundary\r\n\ | |
| asdf1234\ | |
| \r\n\r\n--boundary--"; | |
| let ref mut buf = String::new(); | |
| let mut reader = BoundaryReader::from_reader(&mut body, BOUNDARY); | |
| debug!("Consume 1"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 1"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "asdf1234\r\n"); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| debug!("Read 2 (empty)"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, ""); | |
| } | |
| #[test] | |
| fn test_trailing_crlf() { | |
| ::init_log(); | |
| let mut body: &[u8] = b"--boundary\r\n\ | |
| asdf1234\ | |
| \r\n\r\n--boundary\r\n\ | |
| hjkl5678\r\n--boundary--"; | |
| let ref mut buf = String::new(); | |
| let mut reader = BoundaryReader::from_reader(&mut body, BOUNDARY); | |
| debug!("Consume 1"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 1"); | |
| // Repro for https://github.com/abonander/multipart/issues/93 | |
| // These two reads should produce the same buffer | |
| let buf1 = reader.read_to_boundary().unwrap().to_owned(); | |
| let buf2 = reader.read_to_boundary().unwrap().to_owned(); | |
| assert_eq!(buf1, buf2); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "asdf1234\r\n"); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 2"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "hjkl5678"); | |
| buf.clear(); | |
| debug!("Consume 3"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| debug!("Read 3 (empty)"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, ""); | |
| } | |
| // https://github.com/abonander/multipart/issues/93#issuecomment-343610587 | |
| #[test] | |
| fn test_trailing_lflf() { | |
| ::init_log(); | |
| let mut body: &[u8] = b"--boundary\r\n\ | |
| asdf1234\ | |
| \n\n\r\n--boundary\r\n\ | |
| hjkl5678\r\n--boundary--"; | |
| let ref mut buf = String::new(); | |
| let mut reader = BoundaryReader::from_reader(&mut body, BOUNDARY); | |
| debug!("Consume 1"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 1"); | |
| // same as above | |
| let buf1 = reader.read_to_boundary().unwrap().to_owned(); | |
| let buf2 = reader.read_to_boundary().unwrap().to_owned(); | |
| assert_eq!(buf1, buf2); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "asdf1234\n\n"); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 2"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "hjkl5678"); | |
| buf.clear(); | |
| debug!("Consume 3"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| debug!("Read 3 (empty)"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, ""); | |
| } | |
| // https://github.com/abonander/multipart/issues/104 | |
| #[test] | |
| fn test_unterminated_body() { | |
| ::init_log(); | |
| let mut body: &[u8] = b"--boundary\r\n\ | |
| asdf1234\ | |
| \n\n\r\n--boundary\r\n\ | |
| hjkl5678 "; | |
| let ref mut buf = String::new(); | |
| let mut reader = BoundaryReader::from_reader(&mut body, BOUNDARY); | |
| debug!("Consume 1"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 1"); | |
| // same as above | |
| let buf1 = reader.read_to_boundary().unwrap().to_owned(); | |
| let buf2 = reader.read_to_boundary().unwrap().to_owned(); | |
| assert_eq!(buf1, buf2); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "asdf1234\n\n"); | |
| buf.clear(); | |
| debug!("Consume 2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| debug!("Read 2"); | |
| let _ = reader.read_to_string(buf).unwrap(); | |
| assert_eq!(buf, "hjkl5678 "); | |
| buf.clear(); | |
| debug!("Consume 3 - expecting error"); | |
| reader.consume_boundary().unwrap_err(); | |
| } | |
| #[test] | |
| fn test_lone_boundary() { | |
| let mut body: &[u8] = b"--boundary"; | |
| let mut reader = BoundaryReader::from_reader(&mut body, "boundary"); | |
| reader.consume_boundary().unwrap_err(); | |
| } | |
| #[test] | |
| fn test_invalid_boundary() { | |
| let mut body: &[u8] = b"--boundary\x00\x00"; | |
| let mut reader = BoundaryReader::from_reader(&mut body, "boundary"); | |
| reader.consume_boundary().unwrap_err(); | |
| } | |
| #[test] | |
| fn test_skip_field() { | |
| let mut body: &[u8] = b"--boundary\r\nfield1\r\n--boundary\r\nfield2\r\n--boundary--"; | |
| let mut reader = BoundaryReader::from_reader(&mut body, "boundary"); | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| // skip `field1` | |
| assert_eq!(reader.consume_boundary().unwrap(), true); | |
| let mut buf = String::new(); | |
| reader.read_to_string(&mut buf).unwrap(); | |
| assert_eq!(buf, "field2"); | |
| assert_eq!(reader.consume_boundary().unwrap(), false); | |
| } | |
| #[cfg(feature = "bench")] | |
| mod bench { | |
| extern crate test; | |
| use self::test::Bencher; | |
| use super::*; | |
| #[bench] | |
| fn bench_boundary_reader(b: &mut Bencher) { | |
| let mut reader = BoundaryReader::new_with_bytes(TEST_VAL.as_bytes(), BOUNDARY); | |
| let mut buf = String::with_capacity(256); | |
| b.iter(|| { | |
| reader.reset(); | |
| test_boundary_reader(&mut reader, &mut buf); | |
| }); | |
| } | |
| } | |
| } |