Snap for 7105071 from 002e162b898f5e57fe5ec8d61457c3bfb8e07864 to sc-d1-release

Change-Id: I8f3a992c07246a0d8ca4adaad2a44d43102278f2

31 files changed, 709 insertions, 571 deletions

diff --git a/.cargo_vcs_info.json b/.cargo_vcs_info.json
index dd25c50..8a97d76 100644
--- a/.cargo_vcs_info.json
+++ b/.cargo_vcs_info.json
@@ -1,5 +1,5 @@
 {
   "git": {
-    "sha1": "6fdb7391ce83dc71ccaeda6a54211ce723e5d9a5"
+    "sha1": "7b18c1c076adf9dff4dca28ffc1bbd8759860798"
   }
 }
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 8b99832..fc03588 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -11,6 +11,7 @@ on:
 env:
   RUSTFLAGS: -Dwarnings
   RUST_BACKTRACE: 1
+  nightly: nightly-2020-12-17
 
 defaults:
   run:
@@ -74,14 +75,11 @@ jobs:
   # Nightly
   nightly:
     name: nightly
-    env:
-      # Pin nightly to avoid being impacted by breakage
-      RUST_VERSION: nightly-2019-09-25
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
       - name: Install Rust
-        run: rustup update $RUST_VERSION && rustup default $RUST_VERSION
+        run: rustup update $nightly && rustup default $nightly
       - name: Test
         run: . ci/test-stable.sh test
 
@@ -120,23 +118,27 @@ jobs:
     steps:
       - uses: actions/checkout@v2
       - name: Install Rust
-        run: rustup update nightly && rustup default nightly
+        run: rustup update $nightly && rustup default $nightly
       - name: Install rust-src
         run: rustup component add rust-src
       - name: ASAN / TSAN
         run: . ci/tsan.sh
+  miri:
+    name: miri
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - name: Miri
+        run: ci/miri.sh
 
   # Loom
   loom:
     name: loom
-    env:
-      # Pin nightly to avoid being impacted by breakage
-      RUST_VERSION: nightly-2020-05-19
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
       - name: Install Rust
-        run: rustup update $RUST_VERSION && rustup default $RUST_VERSION
+        run: rustup update $nightly && rustup default $nightly
       - name: Loom tests
         run: RUSTFLAGS="--cfg loom -Dwarnings" cargo test --lib
 
@@ -149,6 +151,8 @@ jobs:
       - nightly
       - minrust
       - cross
+      - tsan
+      - loom
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1b821da..a53a165 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,36 @@
+# 1.0.1 (January 11, 2021)
+
+### Changed
+- mark `Vec::put_slice` with `#[inline]` (#459)
+
+### Fixed
+- Fix deprecation warning (#457)
+- use `Box::into_raw` instead of `mem::forget`-in-disguise (#458)
+
+# 1.0.0 (December 22, 2020)
+
+### Changed
+- Rename Buf/BufMut, methods to chunk/chunk_mut (#450)
+
+### Removed
+- remove unused Buf implementation. (#449)
+
+# 0.6.0 (October 21, 2020)
+
+API polish in preparation for a 1.0 release.
+
+### Changed
+- `BufMut` is now an `unsafe` trait (#432).
+- `BufMut::bytes_mut()` returns `&mut UninitSlice`, a type owned by `bytes` to
+  avoid undefined behavior (#433).
+- `Buf::copy_to_bytes(len)` replaces `Buf::into_bytes()` (#439).
+- `Buf`/`BufMut` utility methods are moved onto the trait and `*Ext` traits are
+  removed (#431).
+
+### Removed
+- `BufMut::bytes_vectored_mut()` (#430).
+- `new` methods on combinator types (#434).
+
 # 0.5.6 (July 13, 2020)
 
 - Improve `BytesMut` to reuse buffer when fully `advance`d.
diff --git a/Cargo.toml b/Cargo.toml
index 81a7224..d102cf0 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -13,10 +13,10 @@
 [package]
 edition = "2018"
 name = "bytes"
-version = "0.5.6"
+version = "1.0.1"
 authors = ["Carl Lerche <me@carllerche.com>", "Sean McArthur <sean@seanmonstar.com>"]
 description = "Types and traits for working with bytes"
-documentation = "https://docs.rs/bytes"
+documentation = "https://docs.rs/bytes/1.0.1/bytes/"
 readme = "README.md"
 keywords = ["buffers", "zero-copy", "io"]
 categories = ["network-programming", "data-structures"]
@@ -34,4 +34,4 @@ version = "1.0"
 default = ["std"]
 std = []
 [target."cfg(loom)".dev-dependencies.loom]
-version = "0.3"
+version = "0.4"
diff --git a/Cargo.toml.orig b/Cargo.toml.orig
index bcb8f17..8f4e224 100644
--- a/Cargo.toml.orig
+++ b/Cargo.toml.orig
@@ -5,15 +5,15 @@ name = "bytes"
 # - Update html_root_url.
 # - Update CHANGELOG.md.
 # - Update doc URL.
-# - Create "v0.5.x" git tag.
-version = "0.5.6"
+# - Create "v1.0.x" git tag.
+version = "1.0.1"
 license = "MIT"
 authors = [
     "Carl Lerche <me@carllerche.com>",
     "Sean McArthur <sean@seanmonstar.com>",
 ]
 description = "Types and traits for working with bytes"
-documentation = "https://docs.rs/bytes"
+documentation = "https://docs.rs/bytes/1.0.1/bytes/"
 repository = "https://github.com/tokio-rs/bytes"
 readme = "README.md"
 keywords = ["buffers", "zero-copy", "io"]
@@ -31,4 +31,4 @@ serde = { version = "1.0.60", optional = true, default-features = false, feature
 serde_test = "1.0"
 
 [target.'cfg(loom)'.dev-dependencies]
-loom = "0.3"
+loom = "0.4"
diff --git a/METADATA b/METADATA
index 47e5bdd..9c2bf5a 100644
--- a/METADATA
+++ b/METADATA
@@ -7,13 +7,13 @@ third_party {
   }
   url {
     type: ARCHIVE
-    value: "https://static.crates.io/crates/bytes/bytes-0.5.6.crate"
+    value: "https://static.crates.io/crates/bytes/bytes-1.0.1.crate"
   }
-  version: "0.5.6"
+  version: "1.0.1"
   license_type: NOTICE
   last_upgrade_date {
-    year: 2020
-    month: 7
-    day: 17
+    year: 2021
+    month: 1
+    day: 11
   }
 }
diff --git a/README.md b/README.md
index 73c43ab..468485d 100644
--- a/README.md
+++ b/README.md
@@ -18,7 +18,7 @@ To use `bytes`, first add this to your `Cargo.toml`:
 
 ```toml
 [dependencies]
-bytes = "0.5"
+bytes = "1"
 ```
 
 Next, add this to your crate:
@@ -33,7 +33,7 @@ Serde support is optional and disabled by default. To enable use the feature `se
 
 ```toml
 [dependencies]
-bytes = { version = "0.5", features = ["serde"] }
+bytes = { version = "1", features = ["serde"] }
 ```
 
 ## License
diff --git a/benches/buf.rs b/benches/buf.rs
index 77b0633..6dc8516 100644
--- a/benches/buf.rs
+++ b/benches/buf.rs
@@ -53,7 +53,7 @@ impl Buf for TestBuf {
         assert!(self.pos <= self.buf.len());
         self.next_readlen();
     }
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         if self.readlen == 0 {
             Default::default()
         } else {
@@ -87,8 +87,8 @@ impl Buf for TestBufC {
         self.inner.advance(cnt)
     }
     #[inline(never)]
-    fn bytes(&self) -> &[u8] {
-        self.inner.bytes()
+    fn chunk(&self) -> &[u8] {
+        self.inner.chunk()
     }
 }
 
@@ -159,7 +159,6 @@ macro_rules! bench_group {
 mod get_u8 {
     use super::*;
     bench_group!(get_u8);
-    bench!(option, option);
 }
 mod get_u16 {
     use super::*;
diff --git a/ci/miri.sh b/ci/miri.sh
new file mode 100755
index 0000000..88d2b6a
--- /dev/null
+++ b/ci/miri.sh
@@ -0,0 +1,11 @@
+#!/bin/bash
+set -e
+
+MIRI_NIGHTLY=nightly-$(curl -s https://rust-lang.github.io/rustup-components-history/x86_64-unknown-linux-gnu/miri)
+echo "Installing latest nightly with Miri: $MIRI_NIGHTLY"
+rustup set profile minimal
+rustup default "$MIRI_NIGHTLY"
+rustup component add miri
+
+cargo miri test
+cargo miri test --target mips64-unknown-linux-gnuabi64
diff --git a/src/buf/buf_impl.rs b/src/buf/buf_impl.rs
index 5cd7c68..16ad8a7 100644
--- a/src/buf/buf_impl.rs
+++ b/src/buf/buf_impl.rs
@@ -1,3 +1,7 @@
+#[cfg(feature = "std")]
+use crate::buf::{reader, Reader};
+use crate::buf::{take, Chain, Take};
+
 use core::{cmp, mem, ptr};
 
 #[cfg(feature = "std")]
@@ -12,7 +16,7 @@ macro_rules! buf_get_impl {
         // this Option<ret> trick is to avoid keeping a borrow on self
         // when advance() is called (mut borrow) and to call bytes() only once
         let ret = $this
-            .bytes()
+            .chunk()
             .get(..SIZE)
             .map(|src| unsafe { $typ::$conv(*(src as *const _ as *const [_; SIZE])) });
 
@@ -74,7 +78,7 @@ pub trait Buf {
     /// the buffer.
     ///
     /// This value is greater than or equal to the length of the slice returned
-    /// by `bytes`.
+    /// by `chunk()`.
     ///
     /// # Examples
     ///
@@ -111,31 +115,31 @@ pub trait Buf {
     ///
     /// let mut buf = &b"hello world"[..];
     ///
-    /// assert_eq!(buf.bytes(), &b"hello world"[..]);
+    /// assert_eq!(buf.chunk(), &b"hello world"[..]);
     ///
     /// buf.advance(6);
     ///
-    /// assert_eq!(buf.bytes(), &b"world"[..]);
+    /// assert_eq!(buf.chunk(), &b"world"[..]);
     /// ```
     ///
     /// # Implementer notes
     ///
     /// This function should never panic. Once the end of the buffer is reached,
-    /// i.e., `Buf::remaining` returns 0, calls to `bytes` should return an
+    /// i.e., `Buf::remaining` returns 0, calls to `chunk()` should return an
     /// empty slice.
-    fn bytes(&self) -> &[u8];
+    fn chunk(&self) -> &[u8];
 
     /// Fills `dst` with potentially multiple slices starting at `self`'s
     /// current position.
     ///
-    /// If the `Buf` is backed by disjoint slices of bytes, `bytes_vectored` enables
+    /// If the `Buf` is backed by disjoint slices of bytes, `chunk_vectored` enables
     /// fetching more than one slice at once. `dst` is a slice of `IoSlice`
     /// references, enabling the slice to be directly used with [`writev`]
     /// without any further conversion. The sum of the lengths of all the
     /// buffers in `dst` will be less than or equal to `Buf::remaining()`.
     ///
     /// The entries in `dst` will be overwritten, but the data **contained** by
-    /// the slices **will not** be modified. If `bytes_vectored` does not fill every
+    /// the slices **will not** be modified. If `chunk_vectored` does not fill every
     /// entry in `dst`, then `dst` is guaranteed to contain all remaining slices
     /// in `self.
     ///
@@ -145,7 +149,7 @@ pub trait Buf {
     /// # Implementer notes
     ///
     /// This function should never panic. Once the end of the buffer is reached,
-    /// i.e., `Buf::remaining` returns 0, calls to `bytes_vectored` must return 0
+    /// i.e., `Buf::remaining` returns 0, calls to `chunk_vectored` must return 0
     /// without mutating `dst`.
     ///
     /// Implementations should also take care to properly handle being called
@@ -153,13 +157,13 @@ pub trait Buf {
     ///
     /// [`writev`]: http://man7.org/linux/man-pages/man2/readv.2.html
     #[cfg(feature = "std")]
-    fn bytes_vectored<'a>(&'a self, dst: &mut [IoSlice<'a>]) -> usize {
+    fn chunks_vectored<'a>(&'a self, dst: &mut [IoSlice<'a>]) -> usize {
         if dst.is_empty() {
             return 0;
         }
 
         if self.has_remaining() {
-            dst[0] = IoSlice::new(self.bytes());
+            dst[0] = IoSlice::new(self.chunk());
             1
         } else {
             0
@@ -168,7 +172,7 @@ pub trait Buf {
 
     /// Advance the internal cursor of the Buf
     ///
-    /// The next call to `bytes` will return a slice starting `cnt` bytes
+    /// The next call to `chunk()` will return a slice starting `cnt` bytes
     /// further into the underlying buffer.
     ///
     /// # Examples
@@ -178,11 +182,11 @@ pub trait Buf {
     ///
     /// let mut buf = &b"hello world"[..];
     ///
-    /// assert_eq!(buf.bytes(), &b"hello world"[..]);
+    /// assert_eq!(buf.chunk(), &b"hello world"[..]);
     ///
     /// buf.advance(6);
     ///
-    /// assert_eq!(buf.bytes(), &b"world"[..]);
+    /// assert_eq!(buf.chunk(), &b"world"[..]);
     /// ```
     ///
     /// # Panics
@@ -249,7 +253,7 @@ pub trait Buf {
             let cnt;
 
             unsafe {
-                let src = self.bytes();
+                let src = self.chunk();
                 cnt = cmp::min(src.len(), dst.len() - off);
 
                 ptr::copy_nonoverlapping(src.as_ptr(), dst[off..].as_mut_ptr(), cnt);
@@ -279,7 +283,7 @@ pub trait Buf {
     /// This function panics if there is no more remaining data in `self`.
     fn get_u8(&mut self) -> u8 {
         assert!(self.remaining() >= 1);
-        let ret = self.bytes()[0];
+        let ret = self.chunk()[0];
         self.advance(1);
         ret
     }
@@ -302,7 +306,7 @@ pub trait Buf {
     /// This function panics if there is no more remaining data in `self`.
     fn get_i8(&mut self) -> i8 {
         assert!(self.remaining() >= 1);
-        let ret = self.bytes()[0] as i8;
+        let ret = self.chunk()[0] as i8;
         self.advance(1);
         ret
     }
@@ -791,22 +795,111 @@ pub trait Buf {
         f64::from_bits(Self::get_u64_le(self))
     }
 
-    /// Consumes remaining bytes inside self and returns new instance of `Bytes`
+    /// Consumes `len` bytes inside self and returns new instance of `Bytes`
+    /// with this data.
+    ///
+    /// This function may be optimized by the underlying type to avoid actual
+    /// copies. For example, `Bytes` implementation will do a shallow copy
+    /// (ref-count increment).
     ///
     /// # Examples
     ///
     /// ```
     /// use bytes::Buf;
     ///
-    /// let bytes = (&b"hello world"[..]).to_bytes();
-    /// assert_eq!(&bytes[..], &b"hello world"[..]);
+    /// let bytes = (&b"hello world"[..]).copy_to_bytes(5);
+    /// assert_eq!(&bytes[..], &b"hello"[..]);
     /// ```
-    fn to_bytes(&mut self) -> crate::Bytes {
+    fn copy_to_bytes(&mut self, len: usize) -> crate::Bytes {
         use super::BufMut;
-        let mut ret = crate::BytesMut::with_capacity(self.remaining());
-        ret.put(self);
+
+        assert!(len <= self.remaining(), "`len` greater than remaining");
+
+        let mut ret = crate::BytesMut::with_capacity(len);
+        ret.put(self.take(len));
         ret.freeze()
     }
+
+    /// Creates an adaptor which will read at most `limit` bytes from `self`.
+    ///
+    /// This function returns a new instance of `Buf` which will read at most
+    /// `limit` bytes.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::{Buf, BufMut};
+    ///
+    /// let mut buf = b"hello world"[..].take(5);
+    /// let mut dst = vec![];
+    ///
+    /// dst.put(&mut buf);
+    /// assert_eq!(dst, b"hello");
+    ///
+    /// let mut buf = buf.into_inner();
+    /// dst.clear();
+    /// dst.put(&mut buf);
+    /// assert_eq!(dst, b" world");
+    /// ```
+    fn take(self, limit: usize) -> Take<Self>
+    where
+        Self: Sized,
+    {
+        take::new(self, limit)
+    }
+
+    /// Creates an adaptor which will chain this buffer with another.
+    ///
+    /// The returned `Buf` instance will first consume all bytes from `self`.
+    /// Afterwards the output is equivalent to the output of next.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::Buf;
+    ///
+    /// let mut chain = b"hello "[..].chain(&b"world"[..]);
+    ///
+    /// let full = chain.copy_to_bytes(11);
+    /// assert_eq!(full.chunk(), b"hello world");
+    /// ```
+    fn chain<U: Buf>(self, next: U) -> Chain<Self, U>
+    where
+        Self: Sized,
+    {
+        Chain::new(self, next)
+    }
+
+    /// Creates an adaptor which implements the `Read` trait for `self`.
+    ///
+    /// This function returns a new value which implements `Read` by adapting
+    /// the `Read` trait functions to the `Buf` trait functions. Given that
+    /// `Buf` operations are infallible, none of the `Read` functions will
+    /// return with `Err`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::{Bytes, Buf};
+    /// use std::io::Read;
+    ///
+    /// let buf = Bytes::from("hello world");
+    ///
+    /// let mut reader = buf.reader();
+    /// let mut dst = [0; 1024];
+    ///
+    /// let num = reader.read(&mut dst).unwrap();
+    ///
+    /// assert_eq!(11, num);
+    /// assert_eq!(&dst[..11], &b"hello world"[..]);
+    /// ```
+    #[cfg(feature = "std")]
+    fn reader(self) -> Reader<Self>
+    where
+        Self: Sized,
+    {
+        reader::new(self)
+    }
 }
 
 macro_rules! deref_forward_buf {
@@ -815,13 +908,13 @@ macro_rules! deref_forward_buf {
             (**self).remaining()
         }
 
-        fn bytes(&self) -> &[u8] {
-            (**self).bytes()
+        fn chunk(&self) -> &[u8] {
+            (**self).chunk()
         }
 
         #[cfg(feature = "std")]
-        fn bytes_vectored<'b>(&'b self, dst: &mut [IoSlice<'b>]) -> usize {
-            (**self).bytes_vectored(dst)
+        fn chunks_vectored<'b>(&'b self, dst: &mut [IoSlice<'b>]) -> usize {
+            (**self).chunks_vectored(dst)
         }
 
         fn advance(&mut self, cnt: usize) {
@@ -908,8 +1001,8 @@ macro_rules! deref_forward_buf {
             (**self).get_int_le(nbytes)
         }
 
-        fn to_bytes(&mut self) -> crate::Bytes {
-            (**self).to_bytes()
+        fn copy_to_bytes(&mut self, len: usize) -> crate::Bytes {
+            (**self).copy_to_bytes(len)
         }
     };
 }
@@ -929,7 +1022,7 @@ impl Buf for &[u8] {
     }
 
     #[inline]
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         self
     }
 
@@ -939,35 +1032,6 @@ impl Buf for &[u8] {
     }
 }
 
-impl Buf for Option<[u8; 1]> {
-    fn remaining(&self) -> usize {
-        if self.is_some() {
-            1
-        } else {
-            0
-        }
-    }
-
-    fn bytes(&self) -> &[u8] {
-        self.as_ref()
-            .map(AsRef::as_ref)
-            .unwrap_or(Default::default())
-    }
-
-    fn advance(&mut self, cnt: usize) {
-        if cnt == 0 {
-            return;
-        }
-
-        if self.is_none() {
-            panic!("overflow");
-        } else {
-            assert_eq!(1, cnt);
-            *self = None;
-        }
-    }
-}
-
 #[cfg(feature = "std")]
 impl<T: AsRef<[u8]>> Buf for std::io::Cursor<T> {
     fn remaining(&self) -> usize {
@@ -981,7 +1045,7 @@ impl<T: AsRef<[u8]>> Buf for std::io::Cursor<T> {
         len - pos as usize
     }
 
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         let len = self.get_ref().as_ref().len();
         let pos = self.position();
 
diff --git a/src/buf/buf_mut.rs b/src/buf/buf_mut.rs
index 628b240..f736727 100644
--- a/src/buf/buf_mut.rs
+++ b/src/buf/buf_mut.rs
@@ -1,11 +1,8 @@
-use core::{
-    cmp,
-    mem::{self, MaybeUninit},
-    ptr, usize,
-};
-
+use crate::buf::{limit, Chain, Limit, UninitSlice};
 #[cfg(feature = "std")]
-use std::fmt;
+use crate::buf::{writer, Writer};
+
+use core::{cmp, mem, ptr, usize};
 
 use alloc::{boxed::Box, vec::Vec};
 
@@ -29,12 +26,12 @@ use alloc::{boxed::Box, vec::Vec};
 ///
 /// assert_eq!(buf, b"hello world");
 /// ```
-pub trait BufMut {
+pub unsafe trait BufMut {
     /// Returns the number of bytes that can be written from the current
     /// position until the end of the buffer is reached.
     ///
     /// This value is greater than or equal to the length of the slice returned
-    /// by `bytes_mut`.
+    /// by `chunk_mut()`.
     ///
     /// # Examples
     ///
@@ -59,7 +56,7 @@ pub trait BufMut {
 
     /// Advance the internal cursor of the BufMut
     ///
-    /// The next call to `bytes_mut` will return a slice starting `cnt` bytes
+    /// The next call to `chunk_mut` will return a slice starting `cnt` bytes
     /// further into the underlying buffer.
     ///
     /// This function is unsafe because there is no guarantee that the bytes
@@ -72,19 +69,14 @@ pub trait BufMut {
     ///
     /// let mut buf = Vec::with_capacity(16);
     ///
-    /// unsafe {
-    ///     // MaybeUninit::as_mut_ptr
-    ///     buf.bytes_mut()[0].as_mut_ptr().write(b'h');
-    ///     buf.bytes_mut()[1].as_mut_ptr().write(b'e');
-    ///
-    ///     buf.advance_mut(2);
+    /// // Write some data
+    /// buf.chunk_mut()[0..2].copy_from_slice(b"he");
+    /// unsafe { buf.advance_mut(2) };
     ///
-    ///     buf.bytes_mut()[0].as_mut_ptr().write(b'l');
-    ///     buf.bytes_mut()[1].as_mut_ptr().write(b'l');
-    ///     buf.bytes_mut()[2].as_mut_ptr().write(b'o');
+    /// // write more bytes
+    /// buf.chunk_mut()[0..3].copy_from_slice(b"llo");
     ///
-    ///     buf.advance_mut(3);
-    /// }
+    /// unsafe { buf.advance_mut(3); }
     ///
     /// assert_eq!(5, buf.len());
     /// assert_eq!(buf, b"hello");
@@ -143,14 +135,14 @@ pub trait BufMut {
     ///
     /// unsafe {
     ///     // MaybeUninit::as_mut_ptr
-    ///     buf.bytes_mut()[0].as_mut_ptr().write(b'h');
-    ///     buf.bytes_mut()[1].as_mut_ptr().write(b'e');
+    ///     buf.chunk_mut()[0..].as_mut_ptr().write(b'h');
+    ///     buf.chunk_mut()[1..].as_mut_ptr().write(b'e');
     ///
     ///     buf.advance_mut(2);
     ///
-    ///     buf.bytes_mut()[0].as_mut_ptr().write(b'l');
-    ///     buf.bytes_mut()[1].as_mut_ptr().write(b'l');
-    ///     buf.bytes_mut()[2].as_mut_ptr().write(b'o');
+    ///     buf.chunk_mut()[0..].as_mut_ptr().write(b'l');
+    ///     buf.chunk_mut()[1..].as_mut_ptr().write(b'l');
+    ///     buf.chunk_mut()[2..].as_mut_ptr().write(b'o');
     ///
     ///     buf.advance_mut(3);
     /// }
@@ -161,54 +153,12 @@ pub trait BufMut {
     ///
     /// # Implementer notes
     ///
-    /// This function should never panic. `bytes_mut` should return an empty
-    /// slice **if and only if** `remaining_mut` returns 0. In other words,
-    /// `bytes_mut` returning an empty slice implies that `remaining_mut` will
-    /// return 0 and `remaining_mut` returning 0 implies that `bytes_mut` will
+    /// This function should never panic. `chunk_mut` should return an empty
+    /// slice **if and only if** `remaining_mut()` returns 0. In other words,
+    /// `chunk_mut()` returning an empty slice implies that `remaining_mut()` will
+    /// return 0 and `remaining_mut()` returning 0 implies that `chunk_mut()` will
     /// return an empty slice.
-    fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>];
-
-    /// Fills `dst` with potentially multiple mutable slices starting at `self`'s
-    /// current position.
-    ///
-    /// If the `BufMut` is backed by disjoint slices of bytes, `bytes_vectored_mut`
-    /// enables fetching more than one slice at once. `dst` is a slice of
-    /// mutable `IoSliceMut` references, enabling the slice to be directly used with
-    /// [`readv`] without any further conversion. The sum of the lengths of all
-    /// the buffers in `dst` will be less than or equal to
-    /// `Buf::remaining_mut()`.
-    ///
-    /// The entries in `dst` will be overwritten, but the data **contained** by
-    /// the slices **will not** be modified. If `bytes_vectored_mut` does not fill every
-    /// entry in `dst`, then `dst` is guaranteed to contain all remaining slices
-    /// in `self.
-    ///
-    /// This is a lower level function. Most operations are done with other
-    /// functions.
-    ///
-    /// # Implementer notes
-    ///
-    /// This function should never panic. Once the end of the buffer is reached,
-    /// i.e., `BufMut::remaining_mut` returns 0, calls to `bytes_vectored_mut` must
-    /// return 0 without mutating `dst`.
-    ///
-    /// Implementations should also take care to properly handle being called
-    /// with `dst` being a zero length slice.
-    ///
-    /// [`readv`]: http://man7.org/linux/man-pages/man2/readv.2.html
-    #[cfg(feature = "std")]
-    fn bytes_vectored_mut<'a>(&'a mut self, dst: &mut [IoSliceMut<'a>]) -> usize {
-        if dst.is_empty() {
-            return 0;
-        }
-
-        if self.has_remaining_mut() {
-            dst[0] = IoSliceMut::from(self.bytes_mut());
-            1
-        } else {
-            0
-        }
-    }
+    fn chunk_mut(&mut self) -> &mut UninitSlice;
 
     /// Transfer bytes into `self` from `src` and advance the cursor by the
     /// number of bytes written.
@@ -240,8 +190,8 @@ pub trait BufMut {
             let l;
 
             unsafe {
-                let s = src.bytes();
-                let d = self.bytes_mut();
+                let s = src.chunk();
+                let d = self.chunk_mut();
                 l = cmp::min(s.len(), d.len());
 
                 ptr::copy_nonoverlapping(s.as_ptr(), d.as_mut_ptr() as *mut u8, l);
@@ -287,7 +237,7 @@ pub trait BufMut {
             let cnt;
 
             unsafe {
-                let dst = self.bytes_mut();
+                let dst = self.chunk_mut();
                 cnt = cmp::min(dst.len(), src.len() - off);
 
                 ptr::copy_nonoverlapping(src[off..].as_ptr(), dst.as_mut_ptr() as *mut u8, cnt);
@@ -878,6 +828,83 @@ pub trait BufMut {
     fn put_f64_le(&mut self, n: f64) {
         self.put_u64_le(n.to_bits());
     }
+
+    /// Creates an adaptor which can write at most `limit` bytes to `self`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::BufMut;
+    ///
+    /// let arr = &mut [0u8; 128][..];
+    /// assert_eq!(arr.remaining_mut(), 128);
+    ///
+    /// let dst = arr.limit(10);
+    /// assert_eq!(dst.remaining_mut(), 10);
+    /// ```
+    fn limit(self, limit: usize) -> Limit<Self>
+    where
+        Self: Sized,
+    {
+        limit::new(self, limit)
+    }
+
+    /// Creates an adaptor which implements the `Write` trait for `self`.
+    ///
+    /// This function returns a new value which implements `Write` by adapting
+    /// the `Write` trait functions to the `BufMut` trait functions. Given that
+    /// `BufMut` operations are infallible, none of the `Write` functions will
+    /// return with `Err`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::BufMut;
+    /// use std::io::Write;
+    ///
+    /// let mut buf = vec![].writer();
+    ///
+    /// let num = buf.write(&b"hello world"[..]).unwrap();
+    /// assert_eq!(11, num);
+    ///
+    /// let buf = buf.into_inner();
+    ///
+    /// assert_eq!(*buf, b"hello world"[..]);
+    /// ```
+    #[cfg(feature = "std")]
+    fn writer(self) -> Writer<Self>
+    where
+        Self: Sized,
+    {
+        writer::new(self)
+    }
+
+    /// Creates an adapter which will chain this buffer with another.
+    ///
+    /// The returned `BufMut` instance will first write to all bytes from
+    /// `self`. Afterwards, it will write to `next`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::BufMut;
+    ///
+    /// let mut a = [0u8; 5];
+    /// let mut b = [0u8; 6];
+    ///
+    /// let mut chain = (&mut a[..]).chain_mut(&mut b[..]);
+    ///
+    /// chain.put_slice(b"hello world");
+    ///
+    /// assert_eq!(&a[..], b"hello");
+    /// assert_eq!(&b[..], b" world");
+    /// ```
+    fn chain_mut<U: BufMut>(self, next: U) -> Chain<Self, U>
+    where
+        Self: Sized,
+    {
+        Chain::new(self, next)
+    }
 }
 
 macro_rules! deref_forward_bufmut {
@@ -886,13 +913,8 @@ macro_rules! deref_forward_bufmut {
             (**self).remaining_mut()
         }
 
-        fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>] {
-            (**self).bytes_mut()
-        }
-
-        #[cfg(feature = "std")]
-        fn bytes_vectored_mut<'b>(&'b mut self, dst: &mut [IoSliceMut<'b>]) -> usize {
-            (**self).bytes_vectored_mut(dst)
+        fn chunk_mut(&mut self) -> &mut UninitSlice {
+            (**self).chunk_mut()
         }
 
         unsafe fn advance_mut(&mut self, cnt: usize) {
@@ -961,24 +983,24 @@ macro_rules! deref_forward_bufmut {
     };
 }
 
-impl<T: BufMut + ?Sized> BufMut for &mut T {
+unsafe impl<T: BufMut + ?Sized> BufMut for &mut T {
     deref_forward_bufmut!();
 }
 
-impl<T: BufMut + ?Sized> BufMut for Box<T> {
+unsafe impl<T: BufMut + ?Sized> BufMut for Box<T> {
     deref_forward_bufmut!();
 }
 
-impl BufMut for &mut [u8] {
+unsafe impl BufMut for &mut [u8] {
     #[inline]
     fn remaining_mut(&self) -> usize {
         self.len()
     }
 
     #[inline]
-    fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>] {
-        // MaybeUninit is repr(transparent), so safe to transmute
-        unsafe { mem::transmute(&mut **self) }
+    fn chunk_mut(&mut self) -> &mut UninitSlice {
+        // UninitSlice is repr(transparent), so safe to transmute
+        unsafe { &mut *(*self as *mut [u8] as *mut _) }
     }
 
     #[inline]
@@ -989,7 +1011,7 @@ impl BufMut for &mut [u8] {
     }
 }
 
-impl BufMut for Vec<u8> {
+unsafe impl BufMut for Vec<u8> {
     #[inline]
     fn remaining_mut(&self) -> usize {
         usize::MAX - self.len()
@@ -1011,9 +1033,7 @@ impl BufMut for Vec<u8> {
     }
 
     #[inline]
-    fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>] {
-        use core::slice;
-
+    fn chunk_mut(&mut self) -> &mut UninitSlice {
         if self.capacity() == self.len() {
             self.reserve(64); // Grow the vec
         }
@@ -1021,13 +1041,12 @@ impl BufMut for Vec<u8> {
         let cap = self.capacity();
         let len = self.len();
 
-        let ptr = self.as_mut_ptr() as *mut MaybeUninit<u8>;
-        unsafe { &mut slice::from_raw_parts_mut(ptr, cap)[len..] }
+        let ptr = self.as_mut_ptr();
+        unsafe { &mut UninitSlice::from_raw_parts_mut(ptr, cap)[len..] }
     }
 
     // Specialize these methods so they can skip checking `remaining_mut`
     // and `advance_mut`.
-
     fn put<T: super::Buf>(&mut self, mut src: T)
     where
         Self: Sized,
@@ -1040,7 +1059,7 @@ impl BufMut for Vec<u8> {
 
             // a block to contain the src.bytes() borrow
             {
-                let s = src.bytes();
+                let s = src.chunk();
                 l = s.len();
                 self.extend_from_slice(s);
             }
@@ -1049,6 +1068,7 @@ impl BufMut for Vec<u8> {
         }
     }
 
+    #[inline]
     fn put_slice(&mut self, src: &[u8]) {
         self.extend_from_slice(src);
     }
@@ -1057,44 +1077,3 @@ impl BufMut for Vec<u8> {
 // The existence of this function makes the compiler catch if the BufMut
 // trait is "object-safe" or not.
 fn _assert_trait_object(_b: &dyn BufMut) {}
-
-// ===== impl IoSliceMut =====
-
-/// A buffer type used for `readv`.
-///
-/// This is a wrapper around an `std::io::IoSliceMut`, but does not expose
-/// the inner bytes in a safe API, as they may point at uninitialized memory.
-///
-/// This is `repr(transparent)` of the `std::io::IoSliceMut`, so it is valid to
-/// transmute them. However, as the memory might be uninitialized, care must be
-/// taken to not *read* the internal bytes, only *write* to them.
-#[repr(transparent)]
-#[cfg(feature = "std")]
-pub struct IoSliceMut<'a>(std::io::IoSliceMut<'a>);
-
-#[cfg(feature = "std")]
-impl fmt::Debug for IoSliceMut<'_> {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        f.debug_struct("IoSliceMut")
-            .field("len", &self.0.len())
-            .finish()
-    }
-}
-
-#[cfg(feature = "std")]
-impl<'a> From<&'a mut [u8]> for IoSliceMut<'a> {
-    fn from(buf: &'a mut [u8]) -> IoSliceMut<'a> {
-        IoSliceMut(std::io::IoSliceMut::new(buf))
-    }
-}
-
-#[cfg(feature = "std")]
-impl<'a> From<&'a mut [MaybeUninit<u8>]> for IoSliceMut<'a> {
-    fn from(buf: &'a mut [MaybeUninit<u8>]) -> IoSliceMut<'a> {
-        IoSliceMut(std::io::IoSliceMut::new(unsafe {
-            // We don't look at the contents, and `std::io::IoSliceMut`
-            // doesn't either.
-            mem::transmute::<&'a mut [MaybeUninit<u8>], &'a mut [u8]>(buf)
-        }))
-    }
-}
diff --git a/src/buf/ext/chain.rs b/src/buf/chain.rs
index e62e2f1..d68bc2d 100644
--- a/src/buf/ext/chain.rs
+++ b/src/buf/chain.rs
@@ -1,10 +1,6 @@
-use crate::buf::IntoIter;
+use crate::buf::{IntoIter, UninitSlice};
 use crate::{Buf, BufMut};
 
-use core::mem::MaybeUninit;
-
-#[cfg(feature = "std")]
-use crate::buf::IoSliceMut;
 #[cfg(feature = "std")]
 use std::io::IoSlice;
 
@@ -20,12 +16,12 @@ use std::io::IoSlice;
 /// # Examples
 ///
 /// ```
-/// use bytes::{Bytes, Buf, buf::BufExt};
+/// use bytes::{Bytes, Buf};
 ///
 /// let mut buf = (&b"hello "[..])
 ///     .chain(&b"world"[..]);
 ///
-/// let full: Bytes = buf.to_bytes();
+/// let full: Bytes = buf.copy_to_bytes(11);
 /// assert_eq!(full[..], b"hello world"[..]);
 /// ```
 ///
@@ -40,7 +36,7 @@ pub struct Chain<T, U> {
 
 impl<T, U> Chain<T, U> {
     /// Creates a new `Chain` sequencing the provided values.
-    pub fn new(a: T, b: U) -> Chain<T, U> {
+    pub(crate) fn new(a: T, b: U) -> Chain<T, U> {
         Chain { a, b }
     }
 
@@ -49,7 +45,7 @@ impl<T, U> Chain<T, U> {
     /// # Examples
     ///
     /// ```
-    /// use bytes::buf::BufExt;
+    /// use bytes::Buf;
     ///
     /// let buf = (&b"hello"[..])
     ///     .chain(&b"world"[..]);
@@ -65,14 +61,14 @@ impl<T, U> Chain<T, U> {
     /// # Examples
     ///
     /// ```
-    /// use bytes::{Buf, buf::BufExt};
+    /// use bytes::Buf;
     ///
     /// let mut buf = (&b"hello"[..])
     ///     .chain(&b"world"[..]);
     ///
     /// buf.first_mut().advance(1);
     ///
-    /// let full = buf.to_bytes();
+    /// let full = buf.copy_to_bytes(9);
     /// assert_eq!(full, b"elloworld"[..]);
     /// ```
     pub fn first_mut(&mut self) -> &mut T {
@@ -84,7 +80,7 @@ impl<T, U> Chain<T, U> {
     /// # Examples
     ///
     /// ```
-    /// use bytes::buf::BufExt;
+    /// use bytes::Buf;
     ///
     /// let buf = (&b"hello"[..])
     ///     .chain(&b"world"[..]);
@@ -100,14 +96,14 @@ impl<T, U> Chain<T, U> {
     /// # Examples
     ///
     /// ```
-    /// use bytes::{Buf, buf::BufExt};
+    /// use bytes::Buf;
     ///
     /// let mut buf = (&b"hello "[..])
     ///     .chain(&b"world"[..]);
     ///
     /// buf.last_mut().advance(1);
     ///
-    /// let full = buf.to_bytes();
+    /// let full = buf.copy_to_bytes(10);
     /// assert_eq!(full, b"hello orld"[..]);
     /// ```
     pub fn last_mut(&mut self) -> &mut U {
@@ -119,7 +115,7 @@ impl<T, U> Chain<T, U> {
     /// # Examples
     ///
     /// ```
-    /// use bytes::buf::BufExt;
+    /// use bytes::Buf;
     ///
     /// let chain = (&b"hello"[..])
     ///     .chain(&b"world"[..]);
@@ -142,11 +138,11 @@ where
         self.a.remaining() + self.b.remaining()
     }
 
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         if self.a.has_remaining() {
-            self.a.bytes()
+            self.a.chunk()
         } else {
-            self.b.bytes()
+            self.b.chunk()
         }
     }
 
@@ -169,14 +165,14 @@ where
     }
 
     #[cfg(feature = "std")]
-    fn bytes_vectored<'a>(&'a self, dst: &mut [IoSlice<'a>]) -> usize {
-        let mut n = self.a.bytes_vectored(dst);
-        n += self.b.bytes_vectored(&mut dst[n..]);
+    fn chunks_vectored<'a>(&'a self, dst: &mut [IoSlice<'a>]) -> usize {
+        let mut n = self.a.chunks_vectored(dst);
+        n += self.b.chunks_vectored(&mut dst[n..]);
         n
     }
 }
 
-impl<T, U> BufMut for Chain<T, U>
+unsafe impl<T, U> BufMut for Chain<T, U>
 where
     T: BufMut,
     U: BufMut,
@@ -185,11 +181,11 @@ where
         self.a.remaining_mut() + self.b.remaining_mut()
     }
 
-    fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>] {
+    fn chunk_mut(&mut self) -> &mut UninitSlice {
         if self.a.has_remaining_mut() {
-            self.a.bytes_mut()
+            self.a.chunk_mut()
         } else {
-            self.b.bytes_mut()
+            self.b.chunk_mut()
         }
     }
 
@@ -210,13 +206,6 @@ where
 
         self.b.advance_mut(cnt);
     }
-
-    #[cfg(feature = "std")]
-    fn bytes_vectored_mut<'a>(&'a mut self, dst: &mut [IoSliceMut<'a>]) -> usize {
-        let mut n = self.a.bytes_vectored_mut(dst);
-        n += self.b.bytes_vectored_mut(&mut dst[n..]);
-        n
-    }
 }
 
 impl<T, U> IntoIterator for Chain<T, U>
diff --git a/src/buf/ext/mod.rs b/src/buf/ext/mod.rs
deleted file mode 100644
index 4a29267..0000000
--- a/src/buf/ext/mod.rs
+++ /dev/null
@@ -1,186 +0,0 @@
-//! Extra utilities for `Buf` and `BufMut` types.
-
-use super::{Buf, BufMut};
-
-mod chain;
-mod limit;
-#[cfg(feature = "std")]
-mod reader;
-mod take;
-#[cfg(feature = "std")]
-mod writer;
-
-pub use self::chain::Chain;
-pub use self::limit::Limit;
-pub use self::take::Take;
-
-#[cfg(feature = "std")]
-pub use self::{reader::Reader, writer::Writer};
-
-/// Extra methods for implementations of `Buf`.
-pub trait BufExt: Buf {
-    /// Creates an adaptor which will read at most `limit` bytes from `self`.
-    ///
-    /// This function returns a new instance of `Buf` which will read at most
-    /// `limit` bytes.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::{BufMut, buf::BufExt};
-    ///
-    /// let mut buf = b"hello world"[..].take(5);
-    /// let mut dst = vec![];
-    ///
-    /// dst.put(&mut buf);
-    /// assert_eq!(dst, b"hello");
-    ///
-    /// let mut buf = buf.into_inner();
-    /// dst.clear();
-    /// dst.put(&mut buf);
-    /// assert_eq!(dst, b" world");
-    /// ```
-    fn take(self, limit: usize) -> Take<Self>
-    where
-        Self: Sized,
-    {
-        take::new(self, limit)
-    }
-
-    /// Creates an adaptor which will chain this buffer with another.
-    ///
-    /// The returned `Buf` instance will first consume all bytes from `self`.
-    /// Afterwards the output is equivalent to the output of next.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::{Buf, buf::BufExt};
-    ///
-    /// let mut chain = b"hello "[..].chain(&b"world"[..]);
-    ///
-    /// let full = chain.to_bytes();
-    /// assert_eq!(full.bytes(), b"hello world");
-    /// ```
-    fn chain<U: Buf>(self, next: U) -> Chain<Self, U>
-    where
-        Self: Sized,
-    {
-        Chain::new(self, next)
-    }
-
-    /// Creates an adaptor which implements the `Read` trait for `self`.
-    ///
-    /// This function returns a new value which implements `Read` by adapting
-    /// the `Read` trait functions to the `Buf` trait functions. Given that
-    /// `Buf` operations are infallible, none of the `Read` functions will
-    /// return with `Err`.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::{Bytes, buf::BufExt};
-    /// use std::io::Read;
-    ///
-    /// let buf = Bytes::from("hello world");
-    ///
-    /// let mut reader = buf.reader();
-    /// let mut dst = [0; 1024];
-    ///
-    /// let num = reader.read(&mut dst).unwrap();
-    ///
-    /// assert_eq!(11, num);
-    /// assert_eq!(&dst[..11], &b"hello world"[..]);
-    /// ```
-    #[cfg(feature = "std")]
-    fn reader(self) -> Reader<Self>
-    where
-        Self: Sized,
-    {
-        reader::new(self)
-    }
-}
-
-impl<B: Buf + ?Sized> BufExt for B {}
-
-/// Extra methods for implementations of `BufMut`.
-pub trait BufMutExt: BufMut {
-    /// Creates an adaptor which can write at most `limit` bytes to `self`.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::{BufMut, buf::BufMutExt};
-    ///
-    /// let arr = &mut [0u8; 128][..];
-    /// assert_eq!(arr.remaining_mut(), 128);
-    ///
-    /// let dst = arr.limit(10);
-    /// assert_eq!(dst.remaining_mut(), 10);
-    /// ```
-    fn limit(self, limit: usize) -> Limit<Self>
-    where
-        Self: Sized,
-    {
-        limit::new(self, limit)
-    }
-
-    /// Creates an adaptor which implements the `Write` trait for `self`.
-    ///
-    /// This function returns a new value which implements `Write` by adapting
-    /// the `Write` trait functions to the `BufMut` trait functions. Given that
-    /// `BufMut` operations are infallible, none of the `Write` functions will
-    /// return with `Err`.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::buf::BufMutExt;
-    /// use std::io::Write;
-    ///
-    /// let mut buf = vec![].writer();
-    ///
-    /// let num = buf.write(&b"hello world"[..]).unwrap();
-    /// assert_eq!(11, num);
-    ///
-    /// let buf = buf.into_inner();
-    ///
-    /// assert_eq!(*buf, b"hello world"[..]);
-    /// ```
-    #[cfg(feature = "std")]
-    fn writer(self) -> Writer<Self>
-    where
-        Self: Sized,
-    {
-        writer::new(self)
-    }
-
-    /// Creates an adapter which will chain this buffer with another.
-    ///
-    /// The returned `BufMut` instance will first write to all bytes from
-    /// `self`. Afterwards, it will write to `next`.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use bytes::{BufMut, buf::BufMutExt};
-    ///
-    /// let mut a = [0u8; 5];
-    /// let mut b = [0u8; 6];
-    ///
-    /// let mut chain = (&mut a[..]).chain_mut(&mut b[..]);
-    ///
-    /// chain.put_slice(b"hello world");
-    ///
-    /// assert_eq!(&a[..], b"hello");
-    /// assert_eq!(&b[..], b" world");
-    /// ```
-    fn chain_mut<U: BufMut>(self, next: U) -> Chain<Self, U>
-    where
-        Self: Sized,
-    {
-        Chain::new(self, next)
-    }
-}
-
-impl<B: BufMut + ?Sized> BufMutExt for B {}
diff --git a/src/buf/iter.rs b/src/buf/iter.rs
index 0f9bdc0..8914a40 100644
--- a/src/buf/iter.rs
+++ b/src/buf/iter.rs
@@ -34,17 +34,16 @@ impl<T> IntoIter<T> {
     ///
     /// ```
     /// use bytes::Bytes;
-    /// use bytes::buf::IntoIter;
     ///
     /// let buf = Bytes::from_static(b"abc");
-    /// let mut iter = IntoIter::new(buf);
+    /// let mut iter = buf.into_iter();
     ///
     /// assert_eq!(iter.next(), Some(b'a'));
     /// assert_eq!(iter.next(), Some(b'b'));
     /// assert_eq!(iter.next(), Some(b'c'));
     /// assert_eq!(iter.next(), None);
     /// ```
-    pub fn new(inner: T) -> IntoIter<T> {
+    pub(crate) fn new(inner: T) -> IntoIter<T> {
         IntoIter { inner }
     }
 
@@ -118,7 +117,7 @@ impl<T: Buf> Iterator for IntoIter<T> {
             return None;
         }
 
-        let b = self.inner.bytes()[0];
+        let b = self.inner.chunk()[0];
         self.inner.advance(1);
 
         Some(b)
diff --git a/src/buf/ext/limit.rs b/src/buf/limit.rs
index a36ecee..b422be5 100644
--- a/src/buf/ext/limit.rs
+++ b/src/buf/limit.rs
@@ -1,6 +1,7 @@
+use crate::buf::UninitSlice;
 use crate::BufMut;
 
-use core::{cmp, mem::MaybeUninit};
+use core::cmp;
 
 /// A `BufMut` adapter which limits the amount of bytes that can be written
 /// to an underlying buffer.
@@ -55,13 +56,13 @@ impl<T> Limit<T> {
     }
 }
 
-impl<T: BufMut> BufMut for Limit<T> {
+unsafe impl<T: BufMut> BufMut for Limit<T> {
     fn remaining_mut(&self) -> usize {
         cmp::min(self.inner.remaining_mut(), self.limit)
     }
 
-    fn bytes_mut(&mut self) -> &mut [MaybeUninit<u8>] {
-        let bytes = self.inner.bytes_mut();
+    fn chunk_mut(&mut self) -> &mut UninitSlice {
+        let bytes = self.inner.chunk_mut();
         let end = cmp::min(bytes.len(), self.limit);
         &mut bytes[..end]
     }
diff --git a/src/buf/mod.rs b/src/buf/mod.rs
index 1d7292c..c4c0a57 100644
--- a/src/buf/mod.rs
+++ b/src/buf/mod.rs
@@ -18,13 +18,24 @@
 
 mod buf_impl;
 mod buf_mut;
-pub mod ext;
+mod chain;
 mod iter;
+mod limit;
+#[cfg(feature = "std")]
+mod reader;
+mod take;
+mod uninit_slice;
 mod vec_deque;
+#[cfg(feature = "std")]
+mod writer;
 
 pub use self::buf_impl::Buf;
 pub use self::buf_mut::BufMut;
-#[cfg(feature = "std")]
-pub use self::buf_mut::IoSliceMut;
-pub use self::ext::{BufExt, BufMutExt};
+pub use self::chain::Chain;
 pub use self::iter::IntoIter;
+pub use self::limit::Limit;
+pub use self::take::Take;
+pub use self::uninit_slice::UninitSlice;
+
+#[cfg(feature = "std")]
+pub use self::{reader::Reader, writer::Writer};
diff --git a/src/buf/ext/reader.rs b/src/buf/reader.rs
index dde3548..f2b4d98 100644
--- a/src/buf/ext/reader.rs
+++ b/src/buf/reader.rs
@@ -24,7 +24,7 @@ impl<B: Buf> Reader<B> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::buf::BufExt;
+    /// use bytes::Buf;
     ///
     /// let buf = b"hello world".reader();
     ///
@@ -46,7 +46,7 @@ impl<B: Buf> Reader<B> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::{Buf, buf::BufExt};
+    /// use bytes::Buf;
     /// use std::io;
     ///
     /// let mut buf = b"hello world".reader();
@@ -73,7 +73,7 @@ impl<B: Buf + Sized> io::Read for Reader<B> {
 
 impl<B: Buf + Sized> io::BufRead for Reader<B> {
     fn fill_buf(&mut self) -> io::Result<&[u8]> {
-        Ok(self.buf.bytes())
+        Ok(self.buf.chunk())
     }
     fn consume(&mut self, amt: usize) {
         self.buf.advance(amt)
diff --git a/src/buf/ext/take.rs b/src/buf/take.rs
index 1d84868..1747f6e 100644
--- a/src/buf/ext/take.rs
+++ b/src/buf/take.rs
@@ -22,7 +22,7 @@ impl<T> Take<T> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::buf::{BufMut, BufExt};
+    /// use bytes::{Buf, BufMut};
     ///
     /// let mut buf = b"hello world".take(2);
     /// let mut dst = vec![];
@@ -47,7 +47,7 @@ impl<T> Take<T> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::{Buf, buf::BufExt};
+    /// use bytes::Buf;
     ///
     /// let buf = b"hello world".take(2);
     ///
@@ -64,7 +64,7 @@ impl<T> Take<T> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::{Buf, BufMut, buf::BufExt};
+    /// use bytes::{Buf, BufMut};
     ///
     /// let mut buf = b"hello world".take(2);
     /// let mut dst = vec![];
@@ -88,7 +88,7 @@ impl<T> Take<T> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::{Buf, buf::BufExt};
+    /// use bytes::Buf;
     ///
     /// let mut buf = b"hello world".take(2);
     ///
@@ -110,7 +110,7 @@ impl<T> Take<T> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::{BufMut, buf::BufExt};
+    /// use bytes::{Buf, BufMut};
     ///
     /// let mut buf = b"hello world".take(2);
     /// let mut dst = vec![];
@@ -134,8 +134,8 @@ impl<T: Buf> Buf for Take<T> {
         cmp::min(self.inner.remaining(), self.limit)
     }
 
-    fn bytes(&self) -> &[u8] {
-        let bytes = self.inner.bytes();
+    fn chunk(&self) -> &[u8] {
+        let bytes = self.inner.chunk();
         &bytes[..cmp::min(bytes.len(), self.limit)]
     }
 
diff --git a/src/buf/uninit_slice.rs b/src/buf/uninit_slice.rs
new file mode 100644
index 0000000..73f4e89
--- /dev/null
+++ b/src/buf/uninit_slice.rs
@@ -0,0 +1,176 @@
+use core::fmt;
+use core::mem::MaybeUninit;
+use core::ops::{
+    Index, IndexMut, Range, RangeFrom, RangeFull, RangeInclusive, RangeTo, RangeToInclusive,
+};
+
+/// Uninitialized byte slice.
+///
+/// Returned by `BufMut::chunk_mut()`, the referenced byte slice may be
+/// uninitialized. The wrapper provides safe access without introducing
+/// undefined behavior.
+///
+/// The safety invariants of this wrapper are:
+///
+///  1. Reading from an `UninitSlice` is undefined behavior.
+///  2. Writing uninitialized bytes to an `UninitSlice` is undefined behavior.
+///
+/// The difference between `&mut UninitSlice` and `&mut [MaybeUninit<u8>]` is
+/// that it is possible in safe code to write uninitialized bytes to an
+/// `&mut [MaybeUninit<u8>]`, which this type prohibits.
+#[repr(transparent)]
+pub struct UninitSlice([MaybeUninit<u8>]);
+
+impl UninitSlice {
+    /// Create a `&mut UninitSlice` from a pointer and a length.
+    ///
+    /// # Safety
+    ///
+    /// The caller must ensure that `ptr` references a valid memory region owned
+    /// by the caller representing a byte slice for the duration of `'a`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::buf::UninitSlice;
+    ///
+    /// let bytes = b"hello world".to_vec();
+    /// let ptr = bytes.as_ptr() as *mut _;
+    /// let len = bytes.len();
+    ///
+    /// let slice = unsafe { UninitSlice::from_raw_parts_mut(ptr, len) };
+    /// ```
+    pub unsafe fn from_raw_parts_mut<'a>(ptr: *mut u8, len: usize) -> &'a mut UninitSlice {
+        let maybe_init: &mut [MaybeUninit<u8>] =
+            core::slice::from_raw_parts_mut(ptr as *mut _, len);
+        &mut *(maybe_init as *mut [MaybeUninit<u8>] as *mut UninitSlice)
+    }
+
+    /// Write a single byte at the specified offset.
+    ///
+    /// # Panics
+    ///
+    /// The function panics if `index` is out of bounds.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::buf::UninitSlice;
+    ///
+    /// let mut data = [b'f', b'o', b'o'];
+    /// let slice = unsafe { UninitSlice::from_raw_parts_mut(data.as_mut_ptr(), 3) };
+    ///
+    /// slice.write_byte(0, b'b');
+    ///
+    /// assert_eq!(b"boo", &data[..]);
+    /// ```
+    pub fn write_byte(&mut self, index: usize, byte: u8) {
+        assert!(index < self.len());
+
+        unsafe { self[index..].as_mut_ptr().write(byte) }
+    }
+
+    /// Copies bytes  from `src` into `self`.
+    ///
+    /// The length of `src` must be the same as `self`.
+    ///
+    /// # Panics
+    ///
+    /// The function panics if `src` has a different length than `self`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::buf::UninitSlice;
+    ///
+    /// let mut data = [b'f', b'o', b'o'];
+    /// let slice = unsafe { UninitSlice::from_raw_parts_mut(data.as_mut_ptr(), 3) };
+    ///
+    /// slice.copy_from_slice(b"bar");
+    ///
+    /// assert_eq!(b"bar", &data[..]);
+    /// ```
+    pub fn copy_from_slice(&mut self, src: &[u8]) {
+        use core::ptr;
+
+        assert_eq!(self.len(), src.len());
+
+        unsafe {
+            ptr::copy_nonoverlapping(src.as_ptr(), self.as_mut_ptr(), self.len());
+        }
+    }
+
+    /// Return a raw pointer to the slice's buffer.
+    ///
+    /// # Safety
+    ///
+    /// The caller **must not** read from the referenced memory and **must not**
+    /// write **uninitialized** bytes to the slice either.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::BufMut;
+    ///
+    /// let mut data = [0, 1, 2];
+    /// let mut slice = &mut data[..];
+    /// let ptr = BufMut::chunk_mut(&mut slice).as_mut_ptr();
+    /// ```
+    pub fn as_mut_ptr(&mut self) -> *mut u8 {
+        self.0.as_mut_ptr() as *mut _
+    }
+
+    /// Returns the number of bytes in the slice.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use bytes::BufMut;
+    ///
+    /// let mut data = [0, 1, 2];
+    /// let mut slice = &mut data[..];
+    /// let len = BufMut::chunk_mut(&mut slice).len();
+    ///
+    /// assert_eq!(len, 3);
+    /// ```
+    pub fn len(&self) -> usize {
+        self.0.len()
+    }
+}
+
+impl fmt::Debug for UninitSlice {
+    fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
+        fmt.debug_struct("UninitSlice[...]").finish()
+    }
+}
+
+macro_rules! impl_index {
+    ($($t:ty),*) => {
+        $(
+            impl Index<$t> for UninitSlice {
+                type Output = UninitSlice;
+
+                fn index(&self, index: $t) -> &UninitSlice {
+                    let maybe_uninit: &[MaybeUninit<u8>] = &self.0[index];
+                    unsafe { &*(maybe_uninit as *const [MaybeUninit<u8>] as *const UninitSlice) }
+                }
+            }
+
+            impl IndexMut<$t> for UninitSlice {
+                fn index_mut(&mut self, index: $t) -> &mut UninitSlice {
+                    let maybe_uninit: &mut [MaybeUninit<u8>] = &mut self.0[index];
+                    unsafe { &mut *(maybe_uninit as *mut [MaybeUninit<u8>] as *mut UninitSlice) }
+                }
+            }
+        )*
+    };
+}
+
+impl_index!(
+    Range<usize>,
+    RangeFrom<usize>,
+    RangeFull,
+    RangeInclusive<usize>,
+    RangeTo<usize>,
+    RangeToInclusive<usize>
+);
diff --git a/src/buf/vec_deque.rs b/src/buf/vec_deque.rs
index 195e689..263167e 100644
--- a/src/buf/vec_deque.rs
+++ b/src/buf/vec_deque.rs
@@ -7,7 +7,7 @@ impl Buf for VecDeque<u8> {
         self.len()
     }
 
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         let (s1, s2) = self.as_slices();
         if s1.is_empty() {
             s2
diff --git a/src/buf/ext/writer.rs b/src/buf/writer.rs
index a14197c..261d7cd 100644
--- a/src/buf/ext/writer.rs
+++ b/src/buf/writer.rs
@@ -24,7 +24,7 @@ impl<B: BufMut> Writer<B> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::buf::BufMutExt;
+    /// use bytes::BufMut;
     ///
     /// let buf = Vec::with_capacity(1024).writer();
     ///
@@ -41,7 +41,7 @@ impl<B: BufMut> Writer<B> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::buf::BufMutExt;
+    /// use bytes::BufMut;
     ///
     /// let mut buf = vec![].writer();
     ///
@@ -58,7 +58,7 @@ impl<B: BufMut> Writer<B> {
     /// # Examples
     ///
     /// ```rust
-    /// use bytes::buf::BufMutExt;
+    /// use bytes::BufMut;
     /// use std::io;
     ///
     /// let mut buf = vec![].writer();
diff --git a/src/bytes.rs b/src/bytes.rs
index 79a09f3..b1b35ea 100644
--- a/src/bytes.rs
+++ b/src/bytes.rs
@@ -10,16 +10,23 @@ use crate::loom::sync::atomic::AtomicMut;
 use crate::loom::sync::atomic::{self, AtomicPtr, AtomicUsize, Ordering};
 use crate::Buf;
 
-/// A reference counted contiguous slice of memory.
+/// A cheaply cloneable and sliceable chunk of contiguous memory.
 ///
 /// `Bytes` is an efficient container for storing and operating on contiguous
 /// slices of memory. It is intended for use primarily in networking code, but
 /// could have applications elsewhere as well.
 ///
 /// `Bytes` values facilitate zero-copy network programming by allowing multiple
-/// `Bytes` objects to point to the same underlying memory. This is managed by
-/// using a reference count to track when the memory is no longer needed and can
-/// be freed.
+/// `Bytes` objects to point to the same underlying memory.
+///
+/// `Bytes` does not have a single implementation. It is an interface, whose
+/// exact behavior is implemented through dynamic dispatch in several underlying
+/// implementations of `Bytes`.
+///
+/// All `Bytes` implementations must fulfill the following requirements:
+/// - They are cheaply cloneable and thereby shareable between an unlimited amount
+///   of components, for example by modifying a reference count.
+/// - Instances can be sliced to refer to a subset of the the original buffer.
 ///
 /// ```
 /// use bytes::Bytes;
@@ -41,17 +48,33 @@ use crate::Buf;
 /// to track information about which segment of the underlying memory the
 /// `Bytes` handle has access to.
 ///
-/// `Bytes` keeps both a pointer to the shared `Arc` containing the full memory
+/// `Bytes` keeps both a pointer to the shared state containing the full memory
 /// slice and a pointer to the start of the region visible by the handle.
 /// `Bytes` also tracks the length of its view into the memory.
 ///
 /// # Sharing
 ///
-/// The memory itself is reference counted, and multiple `Bytes` objects may
-/// point to the same region. Each `Bytes` handle point to different sections within
-/// the memory region, and `Bytes` handle may or may not have overlapping views
+/// `Bytes` contains a vtable, which allows implementations of `Bytes` to define
+/// how sharing/cloneing is implemented in detail.
+/// When `Bytes::clone()` is called, `Bytes` will call the vtable function for
+/// cloning the backing storage in order to share it behind between multiple
+/// `Bytes` instances.
+///
+/// For `Bytes` implementations which refer to constant memory (e.g. created
+/// via `Bytes::from_static()`) the cloning implementation will be a no-op.
+///
+/// For `Bytes` implementations which point to a reference counted shared storage
+/// (e.g. an `Arc<[u8]>`), sharing will be implemented by increasing the
+/// the reference count.
+///
+/// Due to this mechanism, multiple `Bytes` instances may point to the same
+/// shared memory region.
+/// Each `Bytes` instance can point to different sections within that
+/// memory region, and `Bytes` instances may or may not have overlapping views
 /// into the memory.
 ///
+/// The following diagram visualizes a scenario where 2 `Bytes` instances make
+/// use of an `Arc`-based backing storage, and provide access to different views:
 ///
 /// ```text
 ///
@@ -175,7 +198,7 @@ impl Bytes {
         self.len == 0
     }
 
-    ///Creates `Bytes` instance from slice, by copying it.
+    /// Creates `Bytes` instance from slice, by copying it.
     pub fn copy_from_slice(data: &[u8]) -> Self {
         data.to_vec().into()
     }
@@ -214,7 +237,7 @@ impl Bytes {
         };
 
         let end = match range.end_bound() {
-            Bound::Included(&n) => n + 1,
+            Bound::Included(&n) => n.checked_add(1).expect("out of range"),
             Bound::Excluded(&n) => n,
             Bound::Unbounded => len,
         };
@@ -507,7 +530,7 @@ impl Buf for Bytes {
     }
 
     #[inline]
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         self.as_slice()
     }
 
@@ -525,8 +548,14 @@ impl Buf for Bytes {
         }
     }
 
-    fn to_bytes(&mut self) -> crate::Bytes {
-        core::mem::replace(self, Bytes::new())
+    fn copy_to_bytes(&mut self, len: usize) -> crate::Bytes {
+        if len == self.remaining() {
+            core::mem::replace(self, Bytes::new())
+        } else {
+            let ret = self.slice(..len);
+            self.advance(len);
+            ret
+        }
     }
 }
 
@@ -777,8 +806,7 @@ impl From<Vec<u8>> for Bytes {
 
         let slice = vec.into_boxed_slice();
         let len = slice.len();
-        let ptr = slice.as_ptr();
-        drop(Box::into_raw(slice));
+        let ptr = Box::into_raw(slice) as *mut u8;
 
         if ptr as usize & 0x1 == 0 {
             let data = ptr as usize | KIND_VEC;
@@ -994,33 +1022,35 @@ unsafe fn shallow_clone_vec(
     // `Release` is used synchronize with other threads that
     // will load the `arc` field.
     //
-    // If the `compare_and_swap` fails, then the thread lost the
+    // If the `compare_exchange` fails, then the thread lost the
     // race to promote the buffer to shared. The `Acquire`
-    // ordering will synchronize with the `compare_and_swap`
+    // ordering will synchronize with the `compare_exchange`
     // that happened in the other thread and the `Shared`
     // pointed to by `actual` will be visible.
-    let actual = atom.compare_and_swap(ptr as _, shared as _, Ordering::AcqRel);
-
-    if actual as usize == ptr as usize {
-        // The upgrade was successful, the new handle can be
-        // returned.
-        return Bytes {
-            ptr: offset,
-            len,
-            data: AtomicPtr::new(shared as _),
-            vtable: &SHARED_VTABLE,
-        };
+    match atom.compare_exchange(ptr as _, shared as _, Ordering::AcqRel, Ordering::Acquire) {
+        Ok(actual) => {
+            debug_assert!(actual as usize == ptr as usize);
+            // The upgrade was successful, the new handle can be
+            // returned.
+            Bytes {
+                ptr: offset,
+                len,
+                data: AtomicPtr::new(shared as _),
+                vtable: &SHARED_VTABLE,
+            }
+        }
+        Err(actual) => {
+            // The upgrade failed, a concurrent clone happened. Release
+            // the allocation that was made in this thread, it will not
+            // be needed.
+            let shared = Box::from_raw(shared);
+            mem::forget(*shared);
+
+            // Buffer already promoted to shared storage, so increment ref
+            // count.
+            shallow_clone_arc(actual as _, offset, len)
+        }
     }
-
-    // The upgrade failed, a concurrent clone happened. Release
-    // the allocation that was made in this thread, it will not
-    // be needed.
-    let shared = Box::from_raw(shared);
-    mem::forget(*shared);
-
-    // Buffer already promoted to shared storage, so increment ref
-    // count.
-    shallow_clone_arc(actual as _, offset, len)
 }
 
 unsafe fn release_shared(ptr: *mut Shared) {
diff --git a/src/bytes_mut.rs b/src/bytes_mut.rs
index a7a8e57..61c0460 100644
--- a/src/bytes_mut.rs
+++ b/src/bytes_mut.rs
@@ -11,7 +11,7 @@ use alloc::{
     vec::Vec,
 };
 
-use crate::buf::IntoIter;
+use crate::buf::{IntoIter, UninitSlice};
 use crate::bytes::Vtable;
 #[allow(unused)]
 use crate::loom::sync::atomic::AtomicMut;
@@ -445,7 +445,7 @@ impl BytesMut {
             let additional = new_len - len;
             self.reserve(additional);
             unsafe {
-                let dst = self.bytes_mut().as_mut_ptr();
+                let dst = self.chunk_mut().as_mut_ptr();
                 ptr::write_bytes(dst, value, additional);
                 self.set_len(new_len);
             }
@@ -684,7 +684,7 @@ impl BytesMut {
         self.reserve(cnt);
 
         unsafe {
-            let dst = self.maybe_uninit_bytes();
+            let dst = self.uninit_slice();
             // Reserved above
             debug_assert!(dst.len() >= cnt);
 
@@ -910,12 +910,12 @@ impl BytesMut {
     }
 
     #[inline]
-    fn maybe_uninit_bytes(&mut self) -> &mut [mem::MaybeUninit<u8>] {
+    fn uninit_slice(&mut self) -> &mut UninitSlice {
         unsafe {
             let ptr = self.ptr.as_ptr().offset(self.len as isize);
             let len = self.cap - self.len;
 
-            slice::from_raw_parts_mut(ptr as *mut mem::MaybeUninit<u8>, len)
+            UninitSlice::from_raw_parts_mut(ptr, len)
         }
     }
 }
@@ -944,7 +944,7 @@ impl Buf for BytesMut {
     }
 
     #[inline]
-    fn bytes(&self) -> &[u8] {
+    fn chunk(&self) -> &[u8] {
         self.as_slice()
     }
 
@@ -961,12 +961,12 @@ impl Buf for BytesMut {
         }
     }
 
-    fn to_bytes(&mut self) -> crate::Bytes {
-        self.split().freeze()
+    fn copy_to_bytes(&mut self, len: usize) -> crate::Bytes {
+        self.split_to(len).freeze()
     }
 }
 
-impl BufMut for BytesMut {
+unsafe impl BufMut for BytesMut {
     #[inline]
     fn remaining_mut(&self) -> usize {
         usize::MAX - self.len()
@@ -985,11 +985,11 @@ impl BufMut for BytesMut {
     }
 
     #[inline]
-    fn bytes_mut(&mut self) -> &mut [mem::MaybeUninit<u8>] {
+    fn chunk_mut(&mut self) -> &mut UninitSlice {
         if self.capacity() == self.len() {
             self.reserve(64);
         }
-        self.maybe_uninit_bytes()
+        self.uninit_slice()
     }
 
     // Specialize these methods so they can skip checking `remaining_mut`
@@ -1000,7 +1000,7 @@ impl BufMut for BytesMut {
         Self: Sized,
     {
         while src.has_remaining() {
-            let s = src.bytes();
+            let s = src.chunk();
             let l = s.len();
             self.extend_from_slice(s);
             src.advance(l);
diff --git a/src/lib.rs b/src/lib.rs
index e375c01..dd8cc96 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -3,7 +3,7 @@
     no_crate_inject,
     attr(deny(warnings, rust_2018_idioms), allow(dead_code, unused_variables))
 ))]
-#![doc(html_root_url = "https://docs.rs/bytes/0.5.6")]
+#![doc(html_root_url = "https://docs.rs/bytes/1.0.1")]
 #![no_std]
 
 //! Provides abstractions for working with bytes.
diff --git a/tests/test_buf.rs b/tests/test_buf.rs
index 17bdd54..fbad003 100644
--- a/tests/test_buf.rs
+++ b/tests/test_buf.rs
@@ -9,17 +9,17 @@ fn test_fresh_cursor_vec() {
     let mut buf = &b"hello"[..];
 
     assert_eq!(buf.remaining(), 5);
-    assert_eq!(buf.bytes(), b"hello");
+    assert_eq!(buf.chunk(), b"hello");
 
     buf.advance(2);
 
     assert_eq!(buf.remaining(), 3);
-    assert_eq!(buf.bytes(), b"llo");
+    assert_eq!(buf.chunk(), b"llo");
 
     buf.advance(3);
 
     assert_eq!(buf.remaining(), 0);
-    assert_eq!(buf.bytes(), b"");
+    assert_eq!(buf.chunk(), b"");
 }
 
 #[test]
@@ -53,7 +53,7 @@ fn test_bufs_vec() {
 
     let mut dst = [IoSlice::new(b1), IoSlice::new(b2)];
 
-    assert_eq!(1, buf.bytes_vectored(&mut dst[..]));
+    assert_eq!(1, buf.chunks_vectored(&mut dst[..]));
 }
 
 #[test]
@@ -63,9 +63,9 @@ fn test_vec_deque() {
     let mut buffer: VecDeque<u8> = VecDeque::new();
     buffer.extend(b"hello world");
     assert_eq!(11, buffer.remaining());
-    assert_eq!(b"hello world", buffer.bytes());
+    assert_eq!(b"hello world", buffer.chunk());
     buffer.advance(6);
-    assert_eq!(b"world", buffer.bytes());
+    assert_eq!(b"world", buffer.chunk());
     buffer.extend(b" piece");
     let mut out = [0; 11];
     buffer.copy_to_slice(&mut out);
@@ -81,8 +81,8 @@ fn test_deref_buf_forwards() {
             unreachable!("remaining");
         }
 
-        fn bytes(&self) -> &[u8] {
-            unreachable!("bytes");
+        fn chunk(&self) -> &[u8] {
+            unreachable!("chunk");
         }
 
         fn advance(&mut self, _: usize) {
@@ -101,3 +101,20 @@ fn test_deref_buf_forwards() {
     assert_eq!((Box::new(Special) as Box<dyn Buf>).get_u8(), b'x');
     assert_eq!(Box::new(Special).get_u8(), b'x');
 }
+
+#[test]
+fn copy_to_bytes_less() {
+    let mut buf = &b"hello world"[..];
+
+    let bytes = buf.copy_to_bytes(5);
+    assert_eq!(bytes, &b"hello"[..]);
+    assert_eq!(buf, &b" world"[..])
+}
+
+#[test]
+#[should_panic]
+fn copy_to_bytes_overflow() {
+    let mut buf = &b"hello world"[..];
+
+    let _bytes = buf.copy_to_bytes(12);
+}
diff --git a/tests/test_buf_mut.rs b/tests/test_buf_mut.rs
index b91e2e5..8d270e3 100644
--- a/tests/test_buf_mut.rs
+++ b/tests/test_buf_mut.rs
@@ -1,7 +1,6 @@
 #![warn(rust_2018_idioms)]
 
-#[cfg(feature = "std")]
-use bytes::buf::IoSliceMut;
+use bytes::buf::UninitSlice;
 use bytes::{BufMut, BytesMut};
 use core::fmt::Write;
 use core::usize;
@@ -12,7 +11,7 @@ fn test_vec_as_mut_buf() {
 
     assert_eq!(buf.remaining_mut(), usize::MAX);
 
-    assert!(buf.bytes_mut().len() >= 64);
+    assert!(buf.chunk_mut().len() >= 64);
 
     buf.put(&b"zomg"[..]);
 
@@ -66,23 +65,6 @@ fn test_clone() {
     assert!(buf != buf2);
 }
 
-#[cfg(feature = "std")]
-#[test]
-fn test_bufs_vec_mut() {
-    let b1: &mut [u8] = &mut [];
-    let b2: &mut [u8] = &mut [];
-    let mut dst = [IoSliceMut::from(b1), IoSliceMut::from(b2)];
-
-    // with no capacity
-    let mut buf = BytesMut::new();
-    assert_eq!(buf.capacity(), 0);
-    assert_eq!(1, buf.bytes_vectored_mut(&mut dst[..]));
-
-    // with capacity
-    let mut buf = BytesMut::with_capacity(64);
-    assert_eq!(1, buf.bytes_vectored_mut(&mut dst[..]));
-}
-
 #[test]
 fn test_mut_slice() {
     let mut v = vec![0, 0, 0, 0];
@@ -94,13 +76,13 @@ fn test_mut_slice() {
 fn test_deref_bufmut_forwards() {
     struct Special;
 
-    impl BufMut for Special {
+    unsafe impl BufMut for Special {
         fn remaining_mut(&self) -> usize {
             unreachable!("remaining_mut");
         }
 
-        fn bytes_mut(&mut self) -> &mut [std::mem::MaybeUninit<u8>] {
-            unreachable!("bytes_mut");
+        fn chunk_mut(&mut self) -> &mut UninitSlice {
+            unreachable!("chunk_mut");
         }
 
         unsafe fn advance_mut(&mut self, _: usize) {
@@ -118,3 +100,30 @@ fn test_deref_bufmut_forwards() {
     (Box::new(Special) as Box<dyn BufMut>).put_u8(b'x');
     Box::new(Special).put_u8(b'x');
 }
+
+#[test]
+#[should_panic]
+fn write_byte_panics_if_out_of_bounds() {
+    let mut data = [b'b', b'a', b'r'];
+
+    let slice = unsafe { UninitSlice::from_raw_parts_mut(data.as_mut_ptr(), 3) };
+    slice.write_byte(4, b'f');
+}
+
+#[test]
+#[should_panic]
+fn copy_from_slice_panics_if_different_length_1() {
+    let mut data = [b'b', b'a', b'r'];
+
+    let slice = unsafe { UninitSlice::from_raw_parts_mut(data.as_mut_ptr(), 3) };
+    slice.copy_from_slice(b"a");
+}
+
+#[test]
+#[should_panic]
+fn copy_from_slice_panics_if_different_length_2() {
+    let mut data = [b'b', b'a', b'r'];
+
+    let slice = unsafe { UninitSlice::from_raw_parts_mut(data.as_mut_ptr(), 3) };
+    slice.copy_from_slice(b"abcd");
+}
diff --git a/tests/test_bytes.rs b/tests/test_bytes.rs
index 6b106a6..b9e6ce4 100644
--- a/tests/test_bytes.rs
+++ b/tests/test_bytes.rs
@@ -461,6 +461,7 @@ fn reserve_allocates_at_least_original_capacity() {
 }
 
 #[test]
+#[cfg_attr(miri, ignore)] // Miri is too slow
 fn reserve_max_original_capacity_value() {
     const SIZE: usize = 128 * 1024;
 
@@ -608,15 +609,15 @@ fn advance_past_len() {
 
 #[test]
 // Only run these tests on little endian systems. CI uses qemu for testing
-// little endian... and qemu doesn't really support threading all that well.
-#[cfg(target_endian = "little")]
+// big endian... and qemu doesn't really support threading all that well.
+#[cfg(any(miri, target_endian = "little"))]
 fn stress() {
     // Tests promoting a buffer from a vec -> shared in a concurrent situation
     use std::sync::{Arc, Barrier};
     use std::thread;
 
     const THREADS: usize = 8;
-    const ITERS: usize = 1_000;
+    const ITERS: usize = if cfg!(miri) { 100 } else { 1_000 };
 
     for i in 0..ITERS {
         let data = [i as u8; 256];
@@ -912,20 +913,20 @@ fn bytes_buf_mut_advance() {
     let mut bytes = BytesMut::with_capacity(1024);
 
     unsafe {
-        let ptr = bytes.bytes_mut().as_ptr();
-        assert_eq!(1024, bytes.bytes_mut().len());
+        let ptr = bytes.chunk_mut().as_mut_ptr();
+        assert_eq!(1024, bytes.chunk_mut().len());
 
         bytes.advance_mut(10);
 
-        let next = bytes.bytes_mut().as_ptr();
-        assert_eq!(1024 - 10, bytes.bytes_mut().len());
+        let next = bytes.chunk_mut().as_mut_ptr();
+        assert_eq!(1024 - 10, bytes.chunk_mut().len());
         assert_eq!(ptr.offset(10), next);
 
         // advance to the end
         bytes.advance_mut(1024 - 10);
 
         // The buffer size is doubled
-        assert_eq!(1024, bytes.bytes_mut().len());
+        assert_eq!(1024, bytes.chunk_mut().len());
     }
 }
 
diff --git a/tests/test_bytes_odd_alloc.rs b/tests/test_bytes_odd_alloc.rs
index 4ce424b..04ba7c2 100644
--- a/tests/test_bytes_odd_alloc.rs
+++ b/tests/test_bytes_odd_alloc.rs
@@ -1,6 +1,8 @@
 //! Test using `Bytes` with an allocator that hands out "odd" pointers for
 //! vectors (pointers where the LSB is set).
 
+#![cfg(not(miri))] // Miri does not support custom allocators (also, Miri is "odd" by default with 50% chance)
+
 use std::alloc::{GlobalAlloc, Layout, System};
 use std::ptr;
 
diff --git a/tests/test_chain.rs b/tests/test_chain.rs
index 6dbc45d..500ccd4 100644
--- a/tests/test_chain.rs
+++ b/tests/test_chain.rs
@@ -1,6 +1,5 @@
 #![warn(rust_2018_idioms)]
 
-use bytes::buf::{BufExt, BufMutExt};
 use bytes::{Buf, BufMut, Bytes};
 #[cfg(feature = "std")]
 use std::io::IoSlice;
@@ -10,7 +9,7 @@ fn collect_two_bufs() {
     let a = Bytes::from(&b"hello"[..]);
     let b = Bytes::from(&b"world"[..]);
 
-    let res = a.chain(b).to_bytes();
+    let res = a.chain(b).copy_to_bytes(10);
     assert_eq!(res, &b"helloworld"[..]);
 }
 
@@ -63,7 +62,7 @@ fn vectored_read() {
             IoSlice::new(b4),
         ];
 
-        assert_eq!(2, buf.bytes_vectored(&mut iovecs));
+        assert_eq!(2, buf.chunks_vectored(&mut iovecs));
         assert_eq!(iovecs[0][..], b"hello"[..]);
         assert_eq!(iovecs[1][..], b"world"[..]);
         assert_eq!(iovecs[2][..], b""[..]);
@@ -84,7 +83,7 @@ fn vectored_read() {
             IoSlice::new(b4),
         ];
 
-        assert_eq!(2, buf.bytes_vectored(&mut iovecs));
+        assert_eq!(2, buf.chunks_vectored(&mut iovecs));
         assert_eq!(iovecs[0][..], b"llo"[..]);
         assert_eq!(iovecs[1][..], b"world"[..]);
         assert_eq!(iovecs[2][..], b""[..]);
@@ -105,7 +104,7 @@ fn vectored_read() {
             IoSlice::new(b4),
         ];
 
-        assert_eq!(1, buf.bytes_vectored(&mut iovecs));
+        assert_eq!(1, buf.chunks_vectored(&mut iovecs));
         assert_eq!(iovecs[0][..], b"world"[..]);
         assert_eq!(iovecs[1][..], b""[..]);
         assert_eq!(iovecs[2][..], b""[..]);
@@ -126,7 +125,7 @@ fn vectored_read() {
             IoSlice::new(b4),
         ];
 
-        assert_eq!(1, buf.bytes_vectored(&mut iovecs));
+        assert_eq!(1, buf.chunks_vectored(&mut iovecs));
         assert_eq!(iovecs[0][..], b"ld"[..]);
         assert_eq!(iovecs[1][..], b""[..]);
         assert_eq!(iovecs[2][..], b""[..]);
diff --git a/tests/test_reader.rs b/tests/test_reader.rs
index 10b480f..897aff6 100644
--- a/tests/test_reader.rs
+++ b/tests/test_reader.rs
@@ -3,13 +3,13 @@
 
 use std::io::{BufRead, Read};
 
-use bytes::buf::BufExt;
+use bytes::Buf;
 
 #[test]
 fn read() {
     let buf1 = &b"hello "[..];
     let buf2 = &b"world"[..];
-    let buf = BufExt::chain(buf1, buf2); // Disambiguate with Read::chain
+    let buf = Buf::chain(buf1, buf2); // Disambiguate with Read::chain
     let mut buffer = Vec::new();
     buf.reader().read_to_end(&mut buffer).unwrap();
     assert_eq!(b"hello world", &buffer[..]);
@@ -19,7 +19,7 @@ fn read() {
 fn buf_read() {
     let buf1 = &b"hell"[..];
     let buf2 = &b"o\nworld"[..];
-    let mut reader = BufExt::chain(buf1, buf2).reader();
+    let mut reader = Buf::chain(buf1, buf2).reader();
     let mut line = String::new();
     reader.read_line(&mut line).unwrap();
     assert_eq!("hello\n", &line);
diff --git a/tests/test_take.rs b/tests/test_take.rs
index 0afb28b..a23a29e 100644
--- a/tests/test_take.rs
+++ b/tests/test_take.rs
@@ -1,6 +1,6 @@
 #![warn(rust_2018_idioms)]
 
-use bytes::buf::{Buf, BufExt};
+use bytes::buf::Buf;
 
 #[test]
 fn long_take() {
@@ -8,5 +8,5 @@ fn long_take() {
     // overrun the buffer. Regression test for #138.
     let buf = b"hello world".take(100);
     assert_eq!(11, buf.remaining());
-    assert_eq!(b"hello world", buf.bytes());
+    assert_eq!(b"hello world", buf.chunk());
 }