22 files changed, 1594 insertions, 1085 deletions

diff --git a/embassy-nrf/Cargo.toml b/embassy-nrf/Cargo.toml
index 3d0c171f3..c9bd2bc14 100644
--- a/embassy-nrf/Cargo.toml
+++ b/embassy-nrf/Cargo.toml
@@ -12,6 +12,9 @@ edition = "2018"
 # There are no plans to make this stable.
 unstable-pac = []
 
+# Implement embedded-hal 1.0 alpha and embedded-hal-async traits.
+unstable-traits = ["embedded-hal-1", "embedded-hal-async"]
+
 nrf52805 = ["nrf52805-pac", "_ppi"]
 nrf52810 = ["nrf52810-pac", "_ppi"]
 nrf52811 = ["nrf52811-pac", "_ppi"]
@@ -47,11 +50,14 @@ embassy = { version = "0.1.0", path = "../embassy" }
 embassy-macros = { version = "0.1.0", path = "../embassy-macros", features = ["nrf"]}
 embassy-hal-common = {version = "0.1.0", path = "../embassy-hal-common" }
 
+embedded-hal-02 = { package = "embedded-hal", version = "0.2.6" }
+embedded-hal-1 = { package = "embedded-hal", version = "1.0.0-alpha.6", git = "https://github.com/embassy-rs/embedded-hal", branch = "embassy", optional = true}
+embedded-hal-async = { version = "0.0.1", git = "https://github.com/embassy-rs/embedded-hal", branch = "embassy", optional = true}
+
 defmt = { version = "0.3", optional = true }
 log = { version = "0.4.14", optional = true }
 cortex-m-rt = ">=0.6.15,<0.8"
 cortex-m = "0.7.3"
-embedded-hal    = "0.2.6"
 embedded-dma    = "0.1.2"
 futures     = { version = "0.3.17", default-features = false }
 critical-section = "0.2.5"
diff --git a/embassy-nrf/src/buffered_uarte.rs b/embassy-nrf/src/buffered_uarte.rs
index 45e8afc4b..2880c84f6 100644
--- a/embassy-nrf/src/buffered_uarte.rs
+++ b/embassy-nrf/src/buffered_uarte.rs
@@ -213,9 +213,6 @@ impl<'d, U: UarteInstance, T: TimerInstance> AsyncBufRead for BufferedUarte<'d,
         cx: &mut Context<'_>,
     ) -> Poll<embassy::io::Result<&[u8]>> {
         self.inner.with(|state| {
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started
             compiler_fence(Ordering::SeqCst);
             trace!("poll_read");
 
@@ -265,9 +262,6 @@ impl<'d, U: UarteInstance, T: TimerInstance> AsyncWrite for BufferedUarte<'d, U,
 
             trace!("poll_write: queued {:?}", n);
 
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started
             compiler_fence(Ordering::SeqCst);
 
             Poll::Ready(Ok(n))
@@ -347,9 +341,7 @@ impl<'a, U: UarteInstance, T: TimerInstance> PeripheralState for StateInner<'a,
                         trace!("  irq_rx: buf {:?} {:?}", buf.as_ptr() as u32, buf.len());
 
                         // Start UARTE Receive transaction
-                        r.tasks_startrx.write(|w|
-                            // `1` is a valid value to write to task registers.
-                            unsafe { w.bits(1) });
+                        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
                     }
                     break;
                 }
@@ -397,9 +389,7 @@ impl<'a, U: UarteInstance, T: TimerInstance> PeripheralState for StateInner<'a,
                             unsafe { w.maxcnt().bits(buf.len() as _) });
 
                         // Start UARTE Transmit transaction
-                        r.tasks_starttx.write(|w|
-                            // `1` is a valid value to write to task registers.
-                            unsafe { w.bits(1) });
+                        r.tasks_starttx.write(|w| unsafe { w.bits(1) });
                     }
                     break;
                 }
diff --git a/embassy-nrf/src/gpio.rs b/embassy-nrf/src/gpio.rs
index ddc84cf5d..3f204d564 100644
--- a/embassy-nrf/src/gpio.rs
+++ b/embassy-nrf/src/gpio.rs
@@ -7,7 +7,6 @@ use core::marker::PhantomData;
 use cfg_if::cfg_if;
 use embassy::util::Unborrow;
 use embassy_hal_common::{unborrow, unsafe_impl_unborrow};
-use embedded_hal::digital::v2::{InputPin, OutputPin, StatefulOutputPin};
 use gpio::pin_cnf::DRIVE_A;
 
 use crate::pac;
@@ -37,12 +36,12 @@ pub enum Pull {
 
 /// GPIO input driver.
 pub struct Input<'d, T: Pin> {
-    pub(crate) pin: FlexPin<'d, T>,
+    pub(crate) pin: Flex<'d, T>,
 }
 
 impl<'d, T: Pin> Input<'d, T> {
     pub fn new(pin: impl Unborrow<Target = T> + 'd, pull: Pull) -> Self {
-        let mut pin = FlexPin::new(pin);
+        let mut pin = Flex::new(pin);
         pin.set_as_input(pull);
 
         Self { pin }
@@ -57,18 +56,6 @@ impl<'d, T: Pin> Input<'d, T> {
     }
 }
 
-impl<'d, T: Pin> InputPin for Input<'d, T> {
-    type Error = Infallible;
-
-    fn is_high(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_high())
-    }
-
-    fn is_low(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_low())
-    }
-}
-
 /// Digital input or output level.
 #[derive(Debug, Eq, PartialEq)]
 #[cfg_attr(feature = "defmt", derive(defmt::Format))]
@@ -102,7 +89,7 @@ pub enum OutputDrive {
 
 /// GPIO output driver.
 pub struct Output<'d, T: Pin> {
-    pub(crate) pin: FlexPin<'d, T>,
+    pub(crate) pin: Flex<'d, T>,
 }
 
 impl<'d, T: Pin> Output<'d, T> {
@@ -111,7 +98,7 @@ impl<'d, T: Pin> Output<'d, T> {
         initial_output: Level,
         drive: OutputDrive,
     ) -> Self {
-        let mut pin = FlexPin::new(pin);
+        let mut pin = Flex::new(pin);
         match initial_output {
             Level::High => pin.set_high(),
             Level::Low => pin.set_low(),
@@ -142,40 +129,18 @@ impl<'d, T: Pin> Output<'d, T> {
     }
 }
 
-impl<'d, T: Pin> OutputPin for Output<'d, T> {
-    type Error = Infallible;
-
-    fn set_high(&mut self) -> Result<(), Self::Error> {
-        Ok(self.set_high())
-    }
-
-    fn set_low(&mut self) -> Result<(), Self::Error> {
-        Ok(self.set_low())
-    }
-}
-
-impl<'d, T: Pin> StatefulOutputPin for Output<'d, T> {
-    fn is_set_high(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_set_high())
-    }
-
-    fn is_set_low(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_set_low())
-    }
-}
-
 /// GPIO flexible pin.
 ///
 /// This pin can either be a disconnected, input, or output pin. The level register bit will remain
 /// set while not in output mode, so the pin's level will be 'remembered' when it is not in output
 /// mode.
-pub struct FlexPin<'d, T: Pin> {
+pub struct Flex<'d, T: Pin> {
     pub(crate) pin: T,
     phantom: PhantomData<&'d mut T>,
 }
 
-impl<'d, T: Pin> FlexPin<'d, T> {
-    /// Wrap the pin in a `FlexPin`.
+impl<'d, T: Pin> Flex<'d, T> {
+    /// Wrap the pin in a `Flex`.
     ///
     /// The pin remains disconnected. The initial output level is unspecified, but can be changed
     /// before the pin is put into output mode.
@@ -270,49 +235,12 @@ impl<'d, T: Pin> FlexPin<'d, T> {
     }
 }
 
-impl<'d, T: Pin> Drop for FlexPin<'d, T> {
+impl<'d, T: Pin> Drop for Flex<'d, T> {
     fn drop(&mut self) {
         self.pin.conf().reset();
     }
 }
 
-/// Implement [`InputPin`] for [`FlexPin`];
-///
-/// If the pin is not in input mode the result is unspecified.
-impl<'d, T: Pin> InputPin for FlexPin<'d, T> {
-    type Error = Infallible;
-
-    fn is_high(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_high())
-    }
-
-    fn is_low(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_low())
-    }
-}
-
-impl<'d, T: Pin> OutputPin for FlexPin<'d, T> {
-    type Error = Infallible;
-
-    fn set_high(&mut self) -> Result<(), Self::Error> {
-        Ok(self.set_high())
-    }
-
-    fn set_low(&mut self) -> Result<(), Self::Error> {
-        Ok(self.set_low())
-    }
-}
-
-impl<'d, T: Pin> StatefulOutputPin for FlexPin<'d, T> {
-    fn is_set_high(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_set_high())
-    }
-
-    fn is_set_low(&self) -> Result<bool, Self::Error> {
-        Ok(self.is_set_low())
-    }
-}
-
 pub(crate) mod sealed {
     use super::*;
 
@@ -350,17 +278,13 @@ pub(crate) mod sealed {
         /// Set the output as high.
         #[inline]
         fn set_high(&self) {
-            unsafe {
-                self.block().outset.write(|w| w.bits(1u32 << self._pin()));
-            }
+            unsafe { self.block().outset.write(|w| w.bits(1u32 << self._pin())) }
         }
 
         /// Set the output as low.
         #[inline]
         fn set_low(&self) {
-            unsafe {
-                self.block().outclr.write(|w| w.bits(1u32 << self._pin()));
-            }
+            unsafe { self.block().outclr.write(|w| w.bits(1u32 << self._pin())) }
         }
     }
 
@@ -495,3 +419,160 @@ macro_rules! impl_pin {
         }
     };
 }
+
+// ====================
+
+mod eh02 {
+    use super::*;
+
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::InputPin for Input<'d, T> {
+        type Error = Infallible;
+
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_high())
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::OutputPin for Output<'d, T> {
+        type Error = Infallible;
+
+        fn set_high(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_high())
+        }
+
+        fn set_low(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::StatefulOutputPin for Output<'d, T> {
+        fn is_set_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_high())
+        }
+
+        fn is_set_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_low())
+        }
+    }
+
+    /// Implement [`InputPin`] for [`Flex`];
+    ///
+    /// If the pin is not in input mode the result is unspecified.
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::InputPin for Flex<'d, T> {
+        type Error = Infallible;
+
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_high())
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::OutputPin for Flex<'d, T> {
+        type Error = Infallible;
+
+        fn set_high(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_high())
+        }
+
+        fn set_low(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_02::digital::v2::StatefulOutputPin for Flex<'d, T> {
+        fn is_set_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_high())
+        }
+
+        fn is_set_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_low())
+        }
+    }
+}
+
+#[cfg(feature = "unstable-traits")]
+mod eh1 {
+    use super::*;
+
+    impl<'d, T: Pin> embedded_hal_1::digital::ErrorType for Input<'d, T> {
+        type Error = Infallible;
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::InputPin for Input<'d, T> {
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_high())
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::ErrorType for Output<'d, T> {
+        type Error = Infallible;
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::OutputPin for Output<'d, T> {
+        fn set_high(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_high())
+        }
+
+        fn set_low(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::StatefulOutputPin for Output<'d, T> {
+        fn is_set_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_high())
+        }
+
+        fn is_set_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::ErrorType for Flex<'d, T> {
+        type Error = Infallible;
+    }
+
+    /// Implement [`InputPin`] for [`Flex`];
+    ///
+    /// If the pin is not in input mode the result is unspecified.
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::InputPin for Flex<'d, T> {
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_high())
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::OutputPin for Flex<'d, T> {
+        fn set_high(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_high())
+        }
+
+        fn set_low(&mut self) -> Result<(), Self::Error> {
+            Ok(self.set_low())
+        }
+    }
+
+    impl<'d, T: Pin> embedded_hal_1::digital::blocking::StatefulOutputPin for Flex<'d, T> {
+        fn is_set_high(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_high())
+        }
+
+        fn is_set_low(&self) -> Result<bool, Self::Error> {
+            Ok(self.is_set_low())
+        }
+    }
+}
diff --git a/embassy-nrf/src/gpiote.rs b/embassy-nrf/src/gpiote.rs
index 99dfd1ebd..a4c24058c 100644
--- a/embassy-nrf/src/gpiote.rs
+++ b/embassy-nrf/src/gpiote.rs
@@ -5,11 +5,10 @@ use core::task::{Context, Poll};
 use embassy::interrupt::{Interrupt, InterruptExt};
 use embassy::waitqueue::AtomicWaker;
 use embassy_hal_common::unsafe_impl_unborrow;
-use embedded_hal::digital::v2::InputPin;
 use futures::future::poll_fn;
 
 use crate::gpio::sealed::Pin as _;
-use crate::gpio::{AnyPin, FlexPin, Input, Output, Pin as GpioPin};
+use crate::gpio::{AnyPin, Flex, Input, Output, Pin as GpioPin};
 use crate::pac;
 use crate::ppi::{Event, Task};
 use crate::{interrupt, peripherals};
@@ -216,18 +215,6 @@ impl<'d, C: Channel, T: GpioPin> InputChannel<'d, C, T> {
     }
 }
 
-impl<'d, C: Channel, T: GpioPin> InputPin for InputChannel<'d, C, T> {
-    type Error = Infallible;
-
-    fn is_high(&self) -> Result<bool, Self::Error> {
-        self.pin.is_high()
-    }
-
-    fn is_low(&self) -> Result<bool, Self::Error> {
-        self.pin.is_low()
-    }
-}
-
 /// GPIOTE channel driver in output mode
 pub struct OutputChannel<'d, C: Channel, T: GpioPin> {
     ch: C,
@@ -342,78 +329,60 @@ impl<'a> Future for PortInputFuture<'a> {
     }
 }
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForHigh for Input<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
-
-    fn wait_for_high<'a>(&'a mut self) -> Self::Future<'a> {
-        self.pin.wait_for_high()
+impl<'d, T: GpioPin> Input<'d, T> {
+    pub async fn wait_for_high(&mut self) {
+        self.pin.wait_for_high().await
     }
-}
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForLow for Input<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
+    pub async fn wait_for_low(&mut self) {
+        self.pin.wait_for_low().await
+    }
 
-    fn wait_for_low<'a>(&'a mut self) -> Self::Future<'a> {
-        self.pin.wait_for_low()
+    pub async fn wait_for_rising_edge(&mut self) {
+        self.pin.wait_for_rising_edge().await
     }
-}
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForAnyEdge for Input<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
+    pub async fn wait_for_falling_edge(&mut self) {
+        self.pin.wait_for_falling_edge().await
+    }
 
-    fn wait_for_any_edge<'a>(&'a mut self) -> Self::Future<'a> {
-        self.pin.wait_for_any_edge()
+    pub async fn wait_for_any_edge(&mut self) {
+        self.pin.wait_for_any_edge().await
     }
 }
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForHigh for FlexPin<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
-
-    fn wait_for_high<'a>(&'a mut self) -> Self::Future<'a> {
+impl<'d, T: GpioPin> Flex<'d, T> {
+    pub async fn wait_for_high(&mut self) {
         self.pin.conf().modify(|_, w| w.sense().high());
 
         PortInputFuture {
             pin_port: self.pin.pin_port(),
             phantom: PhantomData,
         }
+        .await
     }
-}
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForLow for FlexPin<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
-
-    fn wait_for_low<'a>(&'a mut self) -> Self::Future<'a> {
+    pub async fn wait_for_low(&mut self) {
         self.pin.conf().modify(|_, w| w.sense().low());
 
         PortInputFuture {
             pin_port: self.pin.pin_port(),
             phantom: PhantomData,
         }
+        .await
+    }
+
+    pub async fn wait_for_rising_edge(&mut self) {
+        self.wait_for_low().await;
+        self.wait_for_high().await;
     }
-}
 
-impl<'d, T: GpioPin> embassy::traits::gpio::WaitForAnyEdge for FlexPin<'d, T> {
-    type Future<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = ()> + Unpin + 'a;
+    pub async fn wait_for_falling_edge(&mut self) {
+        self.wait_for_high().await;
+        self.wait_for_low().await;
+    }
 
-    fn wait_for_any_edge<'a>(&'a mut self) -> Self::Future<'a> {
+    pub async fn wait_for_any_edge(&mut self) {
         if self.is_high() {
             self.pin.conf().modify(|_, w| w.sense().low());
         } else {
@@ -423,6 +392,7 @@ impl<'d, T: GpioPin> embassy::traits::gpio::WaitForAnyEdge for FlexPin<'d, T> {
             pin_port: self.pin.pin_port(),
             phantom: PhantomData,
         }
+        .await
     }
 }
 
@@ -471,3 +441,137 @@ impl_channel!(GPIOTE_CH4, 4);
 impl_channel!(GPIOTE_CH5, 5);
 impl_channel!(GPIOTE_CH6, 6);
 impl_channel!(GPIOTE_CH7, 7);
+
+// ====================
+
+mod eh02 {
+    use super::*;
+
+    impl<'d, C: Channel, T: GpioPin> embedded_hal_02::digital::v2::InputPin for InputChannel<'d, C, T> {
+        type Error = Infallible;
+
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            self.pin.is_high()
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            self.pin.is_low()
+        }
+    }
+}
+
+#[cfg(feature = "unstable-traits")]
+mod eh1 {
+    use super::*;
+    use futures::FutureExt;
+
+    impl<'d, C: Channel, T: GpioPin> embedded_hal_1::digital::ErrorType for InputChannel<'d, C, T> {
+        type Error = Infallible;
+    }
+
+    impl<'d, C: Channel, T: GpioPin> embedded_hal_1::digital::blocking::InputPin
+        for InputChannel<'d, C, T>
+    {
+        fn is_high(&self) -> Result<bool, Self::Error> {
+            self.pin.is_high()
+        }
+
+        fn is_low(&self) -> Result<bool, Self::Error> {
+            self.pin.is_low()
+        }
+    }
+
+    impl<'d, T: GpioPin> embedded_hal_async::digital::Wait for Input<'d, T> {
+        type WaitForHighFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_high<'a>(&'a mut self) -> Self::WaitForHighFuture<'a> {
+            self.wait_for_high().map(Ok)
+        }
+
+        type WaitForLowFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_low<'a>(&'a mut self) -> Self::WaitForLowFuture<'a> {
+            self.wait_for_low().map(Ok)
+        }
+
+        type WaitForRisingEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_rising_edge<'a>(&'a mut self) -> Self::WaitForRisingEdgeFuture<'a> {
+            self.wait_for_rising_edge().map(Ok)
+        }
+
+        type WaitForFallingEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_falling_edge<'a>(&'a mut self) -> Self::WaitForFallingEdgeFuture<'a> {
+            self.wait_for_falling_edge().map(Ok)
+        }
+
+        type WaitForAnyEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_any_edge<'a>(&'a mut self) -> Self::WaitForAnyEdgeFuture<'a> {
+            self.wait_for_any_edge().map(Ok)
+        }
+    }
+
+    impl<'d, T: GpioPin> embedded_hal_async::digital::Wait for Flex<'d, T> {
+        type WaitForHighFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_high<'a>(&'a mut self) -> Self::WaitForHighFuture<'a> {
+            self.wait_for_high().map(Ok)
+        }
+
+        type WaitForLowFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_low<'a>(&'a mut self) -> Self::WaitForLowFuture<'a> {
+            self.wait_for_low().map(Ok)
+        }
+
+        type WaitForRisingEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_rising_edge<'a>(&'a mut self) -> Self::WaitForRisingEdgeFuture<'a> {
+            self.wait_for_rising_edge().map(Ok)
+        }
+
+        type WaitForFallingEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_falling_edge<'a>(&'a mut self) -> Self::WaitForFallingEdgeFuture<'a> {
+            self.wait_for_falling_edge().map(Ok)
+        }
+
+        type WaitForAnyEdgeFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn wait_for_any_edge<'a>(&'a mut self) -> Self::WaitForAnyEdgeFuture<'a> {
+            self.wait_for_any_edge().map(Ok)
+        }
+    }
+}
diff --git a/embassy-nrf/src/qspi.rs b/embassy-nrf/src/qspi.rs
index 6afba4931..89262ac05 100644
--- a/embassy-nrf/src/qspi.rs
+++ b/embassy-nrf/src/qspi.rs
@@ -1,11 +1,9 @@
 #![macro_use]
 
-use core::future::Future;
 use core::marker::PhantomData;
 use core::ptr;
 use core::task::Poll;
 use embassy::interrupt::{Interrupt, InterruptExt};
-use embassy::traits::flash::{Error, Flash};
 use embassy::util::Unborrow;
 use embassy_hal_common::drop::DropBomb;
 use embassy_hal_common::unborrow;
@@ -58,6 +56,13 @@ impl Default for Config {
     }
 }
 
+#[derive(Debug, Copy, Clone, Eq, PartialEq)]
+#[cfg_attr(feature = "defmt", derive(defmt::Format))]
+#[non_exhaustive]
+pub enum Error {
+    // TODO add "not in data memory" error and check for it
+}
+
 pub struct Qspi<'d, T: Instance> {
     dpm_enabled: bool,
     phantom: PhantomData<&'d mut T>,
@@ -240,6 +245,87 @@ impl<'d, T: Instance> Qspi<'d, T> {
         })
         .await
     }
+
+    pub async fn read(&mut self, address: usize, data: &mut [u8]) -> Result<(), Error> {
+        let bomb = DropBomb::new();
+
+        assert_eq!(data.as_ptr() as u32 % 4, 0);
+        assert_eq!(data.len() as u32 % 4, 0);
+        assert_eq!(address as u32 % 4, 0);
+
+        let r = T::regs();
+
+        r.read
+            .src
+            .write(|w| unsafe { w.src().bits(address as u32) });
+        r.read
+            .dst
+            .write(|w| unsafe { w.dst().bits(data.as_ptr() as u32) });
+        r.read
+            .cnt
+            .write(|w| unsafe { w.cnt().bits(data.len() as u32) });
+
+        r.events_ready.reset();
+        r.intenset.write(|w| w.ready().set());
+        r.tasks_readstart.write(|w| w.tasks_readstart().bit(true));
+
+        self.wait_ready().await;
+
+        bomb.defuse();
+
+        Ok(())
+    }
+
+    pub async fn write(&mut self, address: usize, data: &[u8]) -> Result<(), Error> {
+        let bomb = DropBomb::new();
+
+        assert_eq!(data.as_ptr() as u32 % 4, 0);
+        assert_eq!(data.len() as u32 % 4, 0);
+        assert_eq!(address as u32 % 4, 0);
+
+        let r = T::regs();
+        r.write
+            .src
+            .write(|w| unsafe { w.src().bits(data.as_ptr() as u32) });
+        r.write
+            .dst
+            .write(|w| unsafe { w.dst().bits(address as u32) });
+        r.write
+            .cnt
+            .write(|w| unsafe { w.cnt().bits(data.len() as u32) });
+
+        r.events_ready.reset();
+        r.intenset.write(|w| w.ready().set());
+        r.tasks_writestart.write(|w| w.tasks_writestart().bit(true));
+
+        self.wait_ready().await;
+
+        bomb.defuse();
+
+        Ok(())
+    }
+
+    pub async fn erase(&mut self, address: usize) -> Result<(), Error> {
+        let bomb = DropBomb::new();
+
+        assert_eq!(address as u32 % 4096, 0);
+
+        let r = T::regs();
+        r.erase
+            .ptr
+            .write(|w| unsafe { w.ptr().bits(address as u32) });
+        r.erase.len.write(|w| w.len()._4kb());
+
+        r.events_ready.reset();
+        r.intenset.write(|w| w.ready().set());
+        r.tasks_erasestart.write(|w| w.tasks_erasestart().bit(true));
+
+        self.wait_ready().await;
+
+        bomb.defuse();
+
+        Ok(())
+    }
 }
 
 impl<'d, T: Instance> Drop for Qspi<'d, T> {
@@ -285,124 +371,6 @@ impl<'d, T: Instance> Drop for Qspi<'d, T> {
     }
 }
 
-impl<'d, T: Instance> Flash for Qspi<'d, T> {
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Error>> + 'a;
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Error>> + 'a;
-    type ErasePageFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Error>> + 'a;
-
-    fn read<'a>(&'a mut self, address: usize, data: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        async move {
-            let bomb = DropBomb::new();
-
-            assert_eq!(data.as_ptr() as u32 % 4, 0);
-            assert_eq!(data.len() as u32 % 4, 0);
-            assert_eq!(address as u32 % 4, 0);
-
-            let r = T::regs();
-
-            r.read
-                .src
-                .write(|w| unsafe { w.src().bits(address as u32) });
-            r.read
-                .dst
-                .write(|w| unsafe { w.dst().bits(data.as_ptr() as u32) });
-            r.read
-                .cnt
-                .write(|w| unsafe { w.cnt().bits(data.len() as u32) });
-
-            r.events_ready.reset();
-            r.intenset.write(|w| w.ready().set());
-            r.tasks_readstart.write(|w| w.tasks_readstart().bit(true));
-
-            self.wait_ready().await;
-
-            bomb.defuse();
-
-            Ok(())
-        }
-    }
-
-    fn write<'a>(&'a mut self, address: usize, data: &'a [u8]) -> Self::WriteFuture<'a> {
-        async move {
-            let bomb = DropBomb::new();
-
-            assert_eq!(data.as_ptr() as u32 % 4, 0);
-            assert_eq!(data.len() as u32 % 4, 0);
-            assert_eq!(address as u32 % 4, 0);
-
-            let r = T::regs();
-            r.write
-                .src
-                .write(|w| unsafe { w.src().bits(data.as_ptr() as u32) });
-            r.write
-                .dst
-                .write(|w| unsafe { w.dst().bits(address as u32) });
-            r.write
-                .cnt
-                .write(|w| unsafe { w.cnt().bits(data.len() as u32) });
-
-            r.events_ready.reset();
-            r.intenset.write(|w| w.ready().set());
-            r.tasks_writestart.write(|w| w.tasks_writestart().bit(true));
-
-            self.wait_ready().await;
-
-            bomb.defuse();
-
-            Ok(())
-        }
-    }
-
-    fn erase<'a>(&'a mut self, address: usize) -> Self::ErasePageFuture<'a> {
-        async move {
-            let bomb = DropBomb::new();
-
-            assert_eq!(address as u32 % 4096, 0);
-
-            let r = T::regs();
-            r.erase
-                .ptr
-                .write(|w| unsafe { w.ptr().bits(address as u32) });
-            r.erase.len.write(|w| w.len()._4kb());
-
-            r.events_ready.reset();
-            r.intenset.write(|w| w.ready().set());
-            r.tasks_erasestart.write(|w| w.tasks_erasestart().bit(true));
-
-            self.wait_ready().await;
-
-            bomb.defuse();
-
-            Ok(())
-        }
-    }
-
-    fn size(&self) -> usize {
-        256 * 4096 // TODO
-    }
-
-    fn read_size(&self) -> usize {
-        4 // TODO
-    }
-
-    fn write_size(&self) -> usize {
-        4 // TODO
-    }
-
-    fn erase_size(&self) -> usize {
-        4096 // TODO
-    }
-}
-
 pub(crate) mod sealed {
     use embassy::waitqueue::AtomicWaker;
 
diff --git a/embassy-nrf/src/rng.rs b/embassy-nrf/src/rng.rs
index 6b36cbb88..98833c52b 100644
--- a/embassy-nrf/src/rng.rs
+++ b/embassy-nrf/src/rng.rs
@@ -1,5 +1,3 @@
-use core::convert::Infallible;
-use core::future::Future;
 use core::marker::PhantomData;
 use core::ptr;
 use core::sync::atomic::AtomicPtr;
@@ -7,13 +5,11 @@ use core::sync::atomic::Ordering;
 use core::task::Poll;
 
 use embassy::interrupt::InterruptExt;
-use embassy::traits;
 use embassy::util::Unborrow;
 use embassy::waitqueue::AtomicWaker;
 use embassy_hal_common::drop::OnDrop;
 use embassy_hal_common::unborrow;
 use futures::future::poll_fn;
-use rand_core::RngCore;
 
 use crate::interrupt;
 use crate::pac;
@@ -39,7 +35,7 @@ struct State {
 
 /// A wrapper around an nRF RNG peripheral.
 ///
-/// It has a non-blocking API, through `embassy::traits::Rng`, and a blocking api through `rand`.
+/// It has a non-blocking API, and a blocking api through `rand`.
 pub struct Rng<'d> {
     irq: interrupt::RNG,
     phantom: PhantomData<(&'d mut RNG, &'d mut interrupt::RNG)>,
@@ -146,72 +142,51 @@ impl<'d> Rng<'d> {
     pub fn bias_correction(&self, enable: bool) {
         RNG::regs().config.write(|w| w.dercen().bit(enable))
     }
-}
-
-impl<'d> Drop for Rng<'d> {
-    fn drop(&mut self) {
-        self.irq.disable()
-    }
-}
-
-impl<'d> traits::rng::Rng for Rng<'d> {
-    type Error = Infallible;
 
-    type RngFuture<'a>
-    where
-        'd: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-
-    fn fill_bytes<'a>(&'a mut self, dest: &'a mut [u8]) -> Self::RngFuture<'a> {
-        async move {
-            if dest.len() == 0 {
-                return Ok(()); // Nothing to fill
-            }
-
-            let range = dest.as_mut_ptr_range();
-            // Even if we've preempted the interrupt, it can't preempt us again,
-            // so we don't need to worry about the order we write these in.
-            STATE.ptr.store(range.start, Ordering::Relaxed);
-            STATE.end.store(range.end, Ordering::Relaxed);
+    pub async fn fill_bytes(&mut self, dest: &mut [u8]) {
+        if dest.len() == 0 {
+            return; // Nothing to fill
+        }
 
-            self.enable_irq();
-            self.start();
+        let range = dest.as_mut_ptr_range();
+        // Even if we've preempted the interrupt, it can't preempt us again,
+        // so we don't need to worry about the order we write these in.
+        STATE.ptr.store(range.start, Ordering::Relaxed);
+        STATE.end.store(range.end, Ordering::Relaxed);
 
-            let on_drop = OnDrop::new(|| {
-                self.stop();
-                self.disable_irq();
+        self.enable_irq();
+        self.start();
 
-                // The interrupt is now disabled and can't preempt us anymore, so the order doesn't matter here.
-                STATE.ptr.store(ptr::null_mut(), Ordering::Relaxed);
-                STATE.end.store(ptr::null_mut(), Ordering::Relaxed);
-            });
+        let on_drop = OnDrop::new(|| {
+            self.stop();
+            self.disable_irq();
 
-            poll_fn(|cx| {
-                STATE.waker.register(cx.waker());
+            // The interrupt is now disabled and can't preempt us anymore, so the order doesn't matter here.
+            STATE.ptr.store(ptr::null_mut(), Ordering::Relaxed);
+            STATE.end.store(ptr::null_mut(), Ordering::Relaxed);
+        });
 
-                // The interrupt will never modify `end`, so load it first and then get the most up-to-date `ptr`.
-                let end = STATE.end.load(Ordering::Relaxed);
-                let ptr = STATE.ptr.load(Ordering::Relaxed);
+        poll_fn(|cx| {
+            STATE.waker.register(cx.waker());
 
-                if ptr == end {
-                    // We're done.
-                    Poll::Ready(())
-                } else {
-                    Poll::Pending
-                }
-            })
-            .await;
+            // The interrupt will never modify `end`, so load it first and then get the most up-to-date `ptr`.
+            let end = STATE.end.load(Ordering::Relaxed);
+            let ptr = STATE.ptr.load(Ordering::Relaxed);
 
-            // Trigger the teardown
-            drop(on_drop);
+            if ptr == end {
+                // We're done.
+                Poll::Ready(())
+            } else {
+                Poll::Pending
+            }
+        })
+        .await;
 
-            Ok(())
-        }
+        // Trigger the teardown
+        drop(on_drop);
     }
-}
 
-impl<'d> RngCore for Rng<'d> {
-    fn fill_bytes(&mut self, dest: &mut [u8]) {
+    pub fn blocking_fill_bytes(&mut self, dest: &mut [u8]) {
         self.start();
 
         for byte in dest.iter_mut() {
@@ -223,24 +198,36 @@ impl<'d> RngCore for Rng<'d> {
 
         self.stop();
     }
+}
+
+impl<'d> Drop for Rng<'d> {
+    fn drop(&mut self) {
+        self.irq.disable()
+    }
+}
+
+impl<'d> rand_core::RngCore for Rng<'d> {
+    fn fill_bytes(&mut self, dest: &mut [u8]) {
+        self.blocking_fill_bytes(dest);
+    }
 
     fn next_u32(&mut self) -> u32 {
         let mut bytes = [0; 4];
-        self.fill_bytes(&mut bytes);
+        self.blocking_fill_bytes(&mut bytes);
         // We don't care about the endianness, so just use the native one.
         u32::from_ne_bytes(bytes)
     }
 
     fn next_u64(&mut self) -> u64 {
         let mut bytes = [0; 8];
-        self.fill_bytes(&mut bytes);
+        self.blocking_fill_bytes(&mut bytes);
         u64::from_ne_bytes(bytes)
     }
 
     fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), rand_core::Error> {
-        self.fill_bytes(dest);
+        self.blocking_fill_bytes(dest);
         Ok(())
     }
 }
 
-// TODO: Should `Rng` implement `CryptoRng`? It's 'suitable for cryptographic purposes' according to the specification.
+impl<'d> rand_core::CryptoRng for Rng<'d> {}
diff --git a/embassy-nrf/src/spim.rs b/embassy-nrf/src/spim.rs
index bc5823f6a..cd43b26e6 100644
--- a/embassy-nrf/src/spim.rs
+++ b/embassy-nrf/src/spim.rs
@@ -1,23 +1,21 @@
 #![macro_use]
 
-use core::future::Future;
 use core::marker::PhantomData;
 use core::sync::atomic::{compiler_fence, Ordering};
 use core::task::Poll;
 use embassy::interrupt::InterruptExt;
-use embassy::traits;
 use embassy::util::Unborrow;
 use embassy_hal_common::unborrow;
 use futures::future::poll_fn;
-use traits::spi::{FullDuplex, Read, Spi, Write};
 
 use crate::gpio;
 use crate::gpio::sealed::Pin as _;
 use crate::gpio::{OptionalPin, Pin as GpioPin};
 use crate::interrupt::Interrupt;
+use crate::util::{slice_ptr_parts, slice_ptr_parts_mut};
 use crate::{pac, util::slice_in_ram_or};
 
-pub use embedded_hal::spi::{Mode, Phase, Polarity, MODE_0, MODE_1, MODE_2, MODE_3};
+pub use embedded_hal_02::spi::{Mode, Phase, Polarity, MODE_0, MODE_1, MODE_2, MODE_3};
 pub use pac::spim0::frequency::FREQUENCY_A as Frequency;
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
@@ -132,9 +130,7 @@ impl<'d, T: Instance> Spim<'d, T> {
 
         // Set over-read character
         let orc = config.orc;
-        r.orc.write(|w|
-            // The ORC field is 8 bits long, so any u8 is a valid value to write.
-            unsafe { w.orc().bits(orc) });
+        r.orc.write(|w| unsafe { w.orc().bits(orc) });
 
         // Disable all events interrupts
         r.intenclr.write(|w| unsafe { w.bits(0xFFFF_FFFF) });
@@ -157,210 +153,114 @@ impl<'d, T: Instance> Spim<'d, T> {
             r.intenclr.write(|w| w.end().clear());
         }
     }
-}
 
-impl<'d, T: Instance> Drop for Spim<'d, T> {
-    fn drop(&mut self) {
-        trace!("spim drop");
+    fn prepare(&mut self, rx: *mut [u8], tx: *const [u8]) -> Result<(), Error> {
+        slice_in_ram_or(tx, Error::DMABufferNotInDataMemory)?;
+        // NOTE: RAM slice check for rx is not necessary, as a mutable
+        // slice can only be built from data located in RAM.
 
-        // TODO check for abort, wait for xxxstopped
+        compiler_fence(Ordering::SeqCst);
 
-        // disable!
         let r = T::regs();
-        r.enable.write(|w| w.enable().disabled());
 
-        gpio::deconfigure_pin(r.psel.sck.read().bits());
-        gpio::deconfigure_pin(r.psel.miso.read().bits());
-        gpio::deconfigure_pin(r.psel.mosi.read().bits());
+        // Set up the DMA write.
+        let (ptr, len) = slice_ptr_parts(tx);
+        r.txd.ptr.write(|w| unsafe { w.ptr().bits(ptr as _) });
+        r.txd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
 
-        trace!("spim drop: done");
-    }
-}
+        // Set up the DMA read.
+        let (ptr, len) = slice_ptr_parts_mut(rx);
+        r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as _) });
+        r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
 
-impl<'d, T: Instance> Spi<u8> for Spim<'d, T> {
-    type Error = Error;
-}
+        // Reset and enable the event
+        r.events_end.reset();
+        r.intenset.write(|w| w.end().set());
 
-impl<'d, T: Instance> Read<u8> for Spim<'d, T> {
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
+        // Start SPI transaction.
+        r.tasks_start.write(|w| unsafe { w.bits(1) });
 
-    fn read<'a>(&'a mut self, data: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        self.read_write(data, &[])
+        Ok(())
     }
-}
 
-impl<'d, T: Instance> Write<u8> for Spim<'d, T> {
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-
-    fn write<'a>(&'a mut self, data: &'a [u8]) -> Self::WriteFuture<'a> {
-        self.read_write(&mut [], data)
-    }
-}
+    fn blocking_inner(&mut self, rx: *mut [u8], tx: *const [u8]) -> Result<(), Error> {
+        self.prepare(rx, tx)?;
 
-impl<'d, T: Instance> FullDuplex<u8> for Spim<'d, T> {
-    type WriteReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-
-    fn read_write<'a>(&'a mut self, rx: &'a mut [u8], tx: &'a [u8]) -> Self::WriteReadFuture<'a> {
-        async move {
-            slice_in_ram_or(tx, Error::DMABufferNotInDataMemory)?;
-            // NOTE: RAM slice check for rx is not necessary, as a mutable
-            // slice can only be built from data located in RAM.
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started.
-            compiler_fence(Ordering::SeqCst);
-
-            let r = T::regs();
-            let s = T::state();
-
-            // Set up the DMA write.
-            r.txd
-                .ptr
-                .write(|w| unsafe { w.ptr().bits(tx.as_ptr() as u32) });
-            r.txd
-                .maxcnt
-                .write(|w| unsafe { w.maxcnt().bits(tx.len() as _) });
-
-            // Set up the DMA read.
-            r.rxd
-                .ptr
-                .write(|w| unsafe { w.ptr().bits(rx.as_mut_ptr() as u32) });
-            r.rxd
-                .maxcnt
-                .write(|w| unsafe { w.maxcnt().bits(rx.len() as _) });
-
-            // Reset and enable the event
-            r.events_end.reset();
-            r.intenset.write(|w| w.end().set());
-
-            // Start SPI transaction.
-            r.tasks_start.write(|w| unsafe { w.bits(1) });
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // after all possible DMA actions have completed.
-            compiler_fence(Ordering::SeqCst);
-
-            // Wait for 'end' event.
-            poll_fn(|cx| {
-                s.end_waker.register(cx.waker());
-                if r.events_end.read().bits() != 0 {
-                    return Poll::Ready(());
-                }
+        // Wait for 'end' event.
+        while T::regs().events_end.read().bits() == 0 {}
 
-                Poll::Pending
-            })
-            .await;
+        compiler_fence(Ordering::SeqCst);
 
-            Ok(())
-        }
+        Ok(())
     }
-}
 
-// Blocking functions are provided by implementing `embedded_hal` traits.
-//
-// Code could be shared between traits to reduce code size.
-impl<'d, T: Instance> embedded_hal::blocking::spi::Transfer<u8> for Spim<'d, T> {
-    type Error = Error;
-    fn transfer<'w>(&mut self, words: &'w mut [u8]) -> Result<&'w [u8], Self::Error> {
-        slice_in_ram_or(words, Error::DMABufferNotInDataMemory)?;
-
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // before any DMA action has started.
-        compiler_fence(Ordering::SeqCst);
+    async fn async_inner(&mut self, rx: *mut [u8], tx: *const [u8]) -> Result<(), Error> {
+        self.prepare(rx, tx)?;
 
-        let r = T::regs();
+        // Wait for 'end' event.
+        poll_fn(|cx| {
+            T::state().end_waker.register(cx.waker());
+            if T::regs().events_end.read().bits() != 0 {
+                return Poll::Ready(());
+            }
 
-        // Set up the DMA write.
-        r.txd
-            .ptr
-            .write(|w| unsafe { w.ptr().bits(words.as_ptr() as u32) });
-        r.txd
-            .maxcnt
-            .write(|w| unsafe { w.maxcnt().bits(words.len() as _) });
+            Poll::Pending
+        })
+        .await;
 
-        // Set up the DMA read.
-        r.rxd
-            .ptr
-            .write(|w| unsafe { w.ptr().bits(words.as_mut_ptr() as u32) });
-        r.rxd
-            .maxcnt
-            .write(|w| unsafe { w.maxcnt().bits(words.len() as _) });
-
-        // Disable the end event since we are busy-polling.
-        r.events_end.reset();
+        compiler_fence(Ordering::SeqCst);
 
-        // Start SPI transaction.
-        r.tasks_start.write(|w| unsafe { w.bits(1) });
+        Ok(())
+    }
 
-        // Wait for 'end' event.
-        while r.events_end.read().bits() == 0 {}
+    pub fn blocking_read(&mut self, data: &mut [u8]) -> Result<(), Error> {
+        self.blocking_inner(data, &[])
+    }
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // after all possible DMA actions have completed.
-        compiler_fence(Ordering::SeqCst);
+    pub fn blocking_transfer(&mut self, read: &mut [u8], write: &[u8]) -> Result<(), Error> {
+        self.blocking_inner(read, write)
+    }
 
-        Ok(words)
+    pub fn blocking_transfer_in_place(&mut self, data: &mut [u8]) -> Result<(), Error> {
+        self.blocking_inner(data, data)
     }
-}
 
-impl<'d, T: Instance> embedded_hal::blocking::spi::Write<u8> for Spim<'d, T> {
-    type Error = Error;
+    pub fn blocking_write(&mut self, data: &[u8]) -> Result<(), Error> {
+        self.blocking_inner(&mut [], data)
+    }
 
-    fn write(&mut self, words: &[u8]) -> Result<(), Self::Error> {
-        slice_in_ram_or(words, Error::DMABufferNotInDataMemory)?;
-        let recv: &mut [u8] = &mut [];
+    pub async fn read(&mut self, data: &mut [u8]) -> Result<(), Error> {
+        self.async_inner(data, &[]).await
+    }
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // before any DMA action has started.
-        compiler_fence(Ordering::SeqCst);
+    pub async fn transfer(&mut self, read: &mut [u8], write: &[u8]) -> Result<(), Error> {
+        self.async_inner(read, write).await
+    }
 
-        let r = T::regs();
+    pub async fn transfer_in_place(&mut self, data: &mut [u8]) -> Result<(), Error> {
+        self.async_inner(data, data).await
+    }
 
-        // Set up the DMA write.
-        r.txd
-            .ptr
-            .write(|w| unsafe { w.ptr().bits(words.as_ptr() as u32) });
-        r.txd
-            .maxcnt
-            .write(|w| unsafe { w.maxcnt().bits(words.len() as _) });
+    pub async fn write(&mut self, data: &[u8]) -> Result<(), Error> {
+        self.async_inner(&mut [], data).await
+    }
+}
 
-        // Set up the DMA read.
-        r.rxd
-            .ptr
-            .write(|w| unsafe { w.ptr().bits(recv.as_mut_ptr() as u32) });
-        r.rxd
-            .maxcnt
-            .write(|w| unsafe { w.maxcnt().bits(recv.len() as _) });
-
-        // Disable the end event since we are busy-polling.
-        r.events_end.reset();
+impl<'d, T: Instance> Drop for Spim<'d, T> {
+    fn drop(&mut self) {
+        trace!("spim drop");
 
-        // Start SPI transaction.
-        r.tasks_start.write(|w| unsafe { w.bits(1) });
+        // TODO check for abort, wait for xxxstopped
 
-        // Wait for 'end' event.
-        while r.events_end.read().bits() == 0 {}
+        // disable!
+        let r = T::regs();
+        r.enable.write(|w| w.enable().disabled());
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // after all possible DMA actions have completed.
-        compiler_fence(Ordering::SeqCst);
+        gpio::deconfigure_pin(r.psel.sck.read().bits());
+        gpio::deconfigure_pin(r.psel.miso.read().bits());
+        gpio::deconfigure_pin(r.psel.mosi.read().bits());
 
-        Ok(())
+        trace!("spim drop: done");
     }
 }
 
@@ -407,3 +307,209 @@ macro_rules! impl_spim {
         }
     };
 }
+
+// ====================
+
+mod eh02 {
+    use super::*;
+
+    impl<'d, T: Instance> embedded_hal_02::blocking::spi::Transfer<u8> for Spim<'d, T> {
+        type Error = Error;
+        fn transfer<'w>(&mut self, words: &'w mut [u8]) -> Result<&'w [u8], Self::Error> {
+            self.blocking_transfer_in_place(words)?;
+            Ok(words)
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_02::blocking::spi::Write<u8> for Spim<'d, T> {
+        type Error = Error;
+
+        fn write(&mut self, words: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(words)
+        }
+    }
+}
+
+#[cfg(feature = "unstable-traits")]
+mod eh1 {
+    use super::*;
+    use core::future::Future;
+
+    impl embedded_hal_1::spi::Error for Error {
+        fn kind(&self) -> embedded_hal_1::spi::ErrorKind {
+            match *self {
+                Self::TxBufferTooLong => embedded_hal_1::spi::ErrorKind::Other,
+                Self::RxBufferTooLong => embedded_hal_1::spi::ErrorKind::Other,
+                Self::DMABufferNotInDataMemory => embedded_hal_1::spi::ErrorKind::Other,
+            }
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::spi::ErrorType for Spim<'d, T> {
+        type Error = Error;
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::spi::blocking::Read<u8> for Spim<'d, T> {
+        fn read(&mut self, words: &mut [u8]) -> Result<(), Self::Error> {
+            self.blocking_transfer(words, &[])
+        }
+
+        fn read_transaction(&mut self, words: &mut [&mut [u8]]) -> Result<(), Self::Error> {
+            for buf in words {
+                self.blocking_read(buf)?
+            }
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::spi::blocking::Write<u8> for Spim<'d, T> {
+        fn write(&mut self, words: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(words)
+        }
+
+        fn write_transaction(&mut self, words: &[&[u8]]) -> Result<(), Self::Error> {
+            for buf in words {
+                self.blocking_write(buf)?
+            }
+            Ok(())
+        }
+
+        fn write_iter<WI>(&mut self, words: WI) -> Result<(), Self::Error>
+        where
+            WI: IntoIterator<Item = u8>,
+        {
+            for w in words {
+                self.blocking_write(&[w])?;
+            }
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::spi::blocking::ReadWrite<u8> for Spim<'d, T> {
+        fn transfer(&mut self, read: &mut [u8], write: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_transfer(read, write)
+        }
+
+        fn transfer_in_place(&mut self, words: &mut [u8]) -> Result<(), Self::Error> {
+            self.blocking_transfer_in_place(words)
+        }
+
+        fn transaction<'a>(
+            &mut self,
+            operations: &mut [embedded_hal_async::spi::Operation<'a, u8>],
+        ) -> Result<(), Self::Error> {
+            use embedded_hal_1::spi::blocking::Operation;
+            for o in operations {
+                match o {
+                    Operation::Read(b) => self.blocking_read(b)?,
+                    Operation::Write(b) => self.blocking_write(b)?,
+                    Operation::Transfer(r, w) => self.blocking_transfer(r, w)?,
+                    Operation::TransferInPlace(b) => self.blocking_transfer_in_place(b)?,
+                }
+            }
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::spi::Read<u8> for Spim<'d, T> {
+        type ReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn read<'a>(&'a mut self, words: &'a mut [u8]) -> Self::ReadFuture<'a> {
+            self.read(words)
+        }
+
+        type ReadTransactionFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn read_transaction<'a>(
+            &'a mut self,
+            words: &'a mut [&'a mut [u8]],
+        ) -> Self::ReadTransactionFuture<'a> {
+            async move {
+                for buf in words {
+                    self.read(buf).await?
+                }
+                Ok(())
+            }
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::spi::Write<u8> for Spim<'d, T> {
+        type WriteFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write<'a>(&'a mut self, data: &'a [u8]) -> Self::WriteFuture<'a> {
+            self.write(data)
+        }
+
+        type WriteTransactionFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write_transaction<'a>(
+            &'a mut self,
+            words: &'a [&'a [u8]],
+        ) -> Self::WriteTransactionFuture<'a> {
+            async move {
+                for buf in words {
+                    self.write(buf).await?
+                }
+                Ok(())
+            }
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::spi::ReadWrite<u8> for Spim<'d, T> {
+        type TransferFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn transfer<'a>(&'a mut self, rx: &'a mut [u8], tx: &'a [u8]) -> Self::TransferFuture<'a> {
+            self.transfer(rx, tx)
+        }
+
+        type TransferInPlaceFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn transfer_in_place<'a>(
+            &'a mut self,
+            words: &'a mut [u8],
+        ) -> Self::TransferInPlaceFuture<'a> {
+            self.transfer_in_place(words)
+        }
+
+        type TransactionFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn transaction<'a>(
+            &'a mut self,
+            operations: &'a mut [embedded_hal_async::spi::Operation<'a, u8>],
+        ) -> Self::TransactionFuture<'a> {
+            use embedded_hal_1::spi::blocking::Operation;
+            async move {
+                for o in operations {
+                    match o {
+                        Operation::Read(b) => self.read(b).await?,
+                        Operation::Write(b) => self.write(b).await?,
+                        Operation::Transfer(r, w) => self.transfer(r, w).await?,
+                        Operation::TransferInPlace(b) => self.transfer_in_place(b).await?,
+                    }
+                }
+                Ok(())
+            }
+        }
+    }
+}
diff --git a/embassy-nrf/src/twim.rs b/embassy-nrf/src/twim.rs
index 3cc79227e..4cd47c897 100644
--- a/embassy-nrf/src/twim.rs
+++ b/embassy-nrf/src/twim.rs
@@ -11,12 +11,10 @@ use core::marker::PhantomData;
 use core::sync::atomic::{compiler_fence, Ordering::SeqCst};
 use core::task::Poll;
 use embassy::interrupt::{Interrupt, InterruptExt};
-use embassy::traits;
 use embassy::util::Unborrow;
 use embassy::waitqueue::AtomicWaker;
 use embassy_hal_common::unborrow;
 use futures::future::poll_fn;
-use traits::i2c::I2c;
 
 use crate::chip::{EASY_DMA_SIZE, FORCE_COPY_BUFFER_SIZE};
 use crate::gpio;
@@ -50,6 +48,22 @@ impl Default for Config {
     }
 }
 
+#[derive(Debug, Copy, Clone, Eq, PartialEq)]
+#[cfg_attr(feature = "defmt", derive(defmt::Format))]
+#[non_exhaustive]
+pub enum Error {
+    TxBufferTooLong,
+    RxBufferTooLong,
+    TxBufferZeroLength,
+    RxBufferZeroLength,
+    Transmit,
+    Receive,
+    DMABufferNotInDataMemory,
+    AddressNack,
+    DataNack,
+    Overrun,
+}
+
 /// Interface to a TWIM instance.
 pub struct Twim<'d, T: Instance> {
     phantom: PhantomData<&'d mut T>,
@@ -201,7 +215,7 @@ impl<'d, T: Instance> Twim<'d, T> {
     }
 
     /// Get Error instance, if any occurred.
-    fn read_errorsrc(&self) -> Result<(), Error> {
+    fn check_errorsrc(&self) -> Result<(), Error> {
         let r = T::regs();
 
         let err = r.errorsrc.read();
@@ -217,8 +231,26 @@ impl<'d, T: Instance> Twim<'d, T> {
         Ok(())
     }
 
+    fn check_rx(&self, len: usize) -> Result<(), Error> {
+        let r = T::regs();
+        if r.rxd.amount.read().bits() != len as u32 {
+            Err(Error::Receive)
+        } else {
+            Ok(())
+        }
+    }
+
+    fn check_tx(&self, len: usize) -> Result<(), Error> {
+        let r = T::regs();
+        if r.txd.amount.read().bits() != len as u32 {
+            Err(Error::Transmit)
+        } else {
+            Ok(())
+        }
+    }
+
     /// Wait for stop or error
-    fn wait(&mut self) {
+    fn blocking_wait(&mut self) {
         let r = T::regs();
         loop {
             if r.events_stopped.read().bits() != 0 {
@@ -232,16 +264,32 @@ impl<'d, T: Instance> Twim<'d, T> {
         }
     }
 
-    /// Write to an I2C slave.
-    ///
-    /// The buffer must have a length of at most 255 bytes on the nRF52832
-    /// and at most 65535 bytes on the nRF52840.
-    pub fn write(&mut self, address: u8, buffer: &[u8]) -> Result<(), Error> {
+    /// Wait for stop or error
+    fn async_wait(&mut self) -> impl Future<Output = ()> {
+        poll_fn(move |cx| {
+            let r = T::regs();
+            let s = T::state();
+
+            s.end_waker.register(cx.waker());
+            if r.events_stopped.read().bits() != 0 {
+                r.events_stopped.reset();
+
+                return Poll::Ready(());
+            }
+
+            // stop if an error occured
+            if r.events_error.read().bits() != 0 {
+                r.events_error.reset();
+                r.tasks_stop.write(|w| unsafe { w.bits(1) });
+            }
+
+            Poll::Pending
+        })
+    }
+
+    fn setup_write(&mut self, address: u8, buffer: &[u8], inten: bool) -> Result<(), Error> {
         let r = T::regs();
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // before any DMA action has started.
         compiler_fence(SeqCst);
 
         r.address.write(|w| unsafe { w.address().bits(address) });
@@ -255,38 +303,21 @@ impl<'d, T: Instance> Twim<'d, T> {
         r.events_lasttx.reset();
         self.clear_errorsrc();
 
-        // Start write operation.
-        r.shorts.write(|w| w.lasttx_stop().enabled());
-        r.tasks_starttx.write(|w|
-            // `1` is a valid value to write to task registers.
-            unsafe { w.bits(1) });
-
-        self.wait();
-
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // after all possible DMA actions have completed.
-        compiler_fence(SeqCst);
-
-        self.read_errorsrc()?;
-
-        if r.txd.amount.read().bits() != buffer.len() as u32 {
-            return Err(Error::Transmit);
+        if inten {
+            r.intenset.write(|w| w.stopped().set().error().set());
+        } else {
+            r.intenclr.write(|w| w.stopped().clear().error().clear());
         }
 
+        // Start write operation.
+        r.shorts.write(|w| w.lasttx_stop().enabled());
+        r.tasks_starttx.write(|w| unsafe { w.bits(1) });
         Ok(())
     }
 
-    /// Read from an I2C slave.
-    ///
-    /// The buffer must have a length of at most 255 bytes on the nRF52832
-    /// and at most 65535 bytes on the nRF52840.
-    pub fn read(&mut self, address: u8, buffer: &mut [u8]) -> Result<(), Error> {
+    fn setup_read(&mut self, address: u8, buffer: &mut [u8], inten: bool) -> Result<(), Error> {
         let r = T::regs();
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // before any DMA action has started.
         compiler_fence(SeqCst);
 
         r.address.write(|w| unsafe { w.address().bits(address) });
@@ -299,44 +330,27 @@ impl<'d, T: Instance> Twim<'d, T> {
         r.events_error.reset();
         self.clear_errorsrc();
 
-        // Start read operation.
-        r.shorts.write(|w| w.lastrx_stop().enabled());
-        r.tasks_startrx.write(|w|
-            // `1` is a valid value to write to task registers.
-            unsafe { w.bits(1) });
-
-        self.wait();
-
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // after all possible DMA actions have completed.
-        compiler_fence(SeqCst);
-
-        self.read_errorsrc()?;
-
-        if r.rxd.amount.read().bits() != buffer.len() as u32 {
-            return Err(Error::Receive);
+        if inten {
+            r.intenset.write(|w| w.stopped().set().error().set());
+        } else {
+            r.intenclr.write(|w| w.stopped().clear().error().clear());
         }
 
+        // Start read operation.
+        r.shorts.write(|w| w.lastrx_stop().enabled());
+        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
         Ok(())
     }
 
-    /// Write data to an I2C slave, then read data from the slave without
-    /// triggering a stop condition between the two.
-    ///
-    /// The buffers must have a length of at most 255 bytes on the nRF52832
-    /// and at most 65535 bytes on the nRF52840.
-    pub fn write_then_read(
+    fn setup_write_read(
         &mut self,
         address: u8,
         wr_buffer: &[u8],
         rd_buffer: &mut [u8],
+        inten: bool,
     ) -> Result<(), Error> {
         let r = T::regs();
 
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // before any DMA action has started.
         compiler_fence(SeqCst);
 
         r.address.write(|w| unsafe { w.address().bits(address) });
@@ -352,35 +366,65 @@ impl<'d, T: Instance> Twim<'d, T> {
         r.events_error.reset();
         self.clear_errorsrc();
 
+        if inten {
+            r.intenset.write(|w| w.stopped().set().error().set());
+        } else {
+            r.intenclr.write(|w| w.stopped().clear().error().clear());
+        }
+
         // Start write+read operation.
         r.shorts.write(|w| {
             w.lasttx_startrx().enabled();
             w.lastrx_stop().enabled();
             w
         });
-        // `1` is a valid value to write to task registers.
         r.tasks_starttx.write(|w| unsafe { w.bits(1) });
+        Ok(())
+    }
 
-        self.wait();
-
-        // Conservative compiler fence to prevent optimizations that do not
-        // take in to account actions by DMA. The fence has been placed here,
-        // after all possible DMA actions have completed.
+    /// Write to an I2C slave.
+    ///
+    /// The buffer must have a length of at most 255 bytes on the nRF52832
+    /// and at most 65535 bytes on the nRF52840.
+    pub fn blocking_write(&mut self, address: u8, buffer: &[u8]) -> Result<(), Error> {
+        self.setup_write(address, buffer, false)?;
+        self.blocking_wait();
         compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_tx(buffer.len())?;
+        Ok(())
+    }
 
-        self.read_errorsrc()?;
-
-        let bad_write = r.txd.amount.read().bits() != wr_buffer.len() as u32;
-        let bad_read = r.rxd.amount.read().bits() != rd_buffer.len() as u32;
-
-        if bad_write {
-            return Err(Error::Transmit);
-        }
-
-        if bad_read {
-            return Err(Error::Receive);
-        }
+    /// Read from an I2C slave.
+    ///
+    /// The buffer must have a length of at most 255 bytes on the nRF52832
+    /// and at most 65535 bytes on the nRF52840.
+    pub fn blocking_read(&mut self, address: u8, buffer: &mut [u8]) -> Result<(), Error> {
+        self.setup_read(address, buffer, false)?;
+        self.blocking_wait();
+        compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_rx(buffer.len())?;
+        Ok(())
+    }
 
+    /// Write data to an I2C slave, then read data from the slave without
+    /// triggering a stop condition between the two.
+    ///
+    /// The buffers must have a length of at most 255 bytes on the nRF52832
+    /// and at most 65535 bytes on the nRF52840.
+    pub fn blocking_write_read(
+        &mut self,
+        address: u8,
+        wr_buffer: &[u8],
+        rd_buffer: &mut [u8],
+    ) -> Result<(), Error> {
+        self.setup_write_read(address, wr_buffer, rd_buffer, false)?;
+        self.blocking_wait();
+        compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_tx(wr_buffer.len())?;
+        self.check_rx(rd_buffer.len())?;
         Ok(())
     }
 
@@ -388,7 +432,7 @@ impl<'d, T: Instance> Twim<'d, T> {
     ///
     /// The write buffer must have a length of at most 255 bytes on the nRF52832
     /// and at most 1024 bytes on the nRF52840.
-    pub fn copy_write(&mut self, address: u8, wr_buffer: &[u8]) -> Result<(), Error> {
+    pub fn blocking_copy_write(&mut self, address: u8, wr_buffer: &[u8]) -> Result<(), Error> {
         if wr_buffer.len() > FORCE_COPY_BUFFER_SIZE {
             return Err(Error::TxBufferTooLong);
         }
@@ -397,7 +441,7 @@ impl<'d, T: Instance> Twim<'d, T> {
         let wr_ram_buffer = &mut [0; FORCE_COPY_BUFFER_SIZE][..wr_buffer.len()];
         wr_ram_buffer.copy_from_slice(wr_buffer);
 
-        self.write(address, wr_ram_buffer)
+        self.blocking_write(address, wr_ram_buffer)
     }
 
     /// Copy data into RAM and write to an I2C slave, then read data from the slave without
@@ -408,7 +452,7 @@ impl<'d, T: Instance> Twim<'d, T> {
     ///
     /// The read buffer must have a length of at most 255 bytes on the nRF52832
     /// and at most 65535 bytes on the nRF52840.
-    pub fn copy_write_then_read(
+    pub fn blocking_copy_write_read(
         &mut self,
         address: u8,
         wr_buffer: &[u8],
@@ -422,27 +466,40 @@ impl<'d, T: Instance> Twim<'d, T> {
         let wr_ram_buffer = &mut [0; FORCE_COPY_BUFFER_SIZE][..wr_buffer.len()];
         wr_ram_buffer.copy_from_slice(wr_buffer);
 
-        self.write_then_read(address, wr_ram_buffer, rd_buffer)
+        self.blocking_write_read(address, wr_ram_buffer, rd_buffer)
     }
 
-    fn wait_for_stopped_event(cx: &mut core::task::Context) -> Poll<()> {
-        let r = T::regs();
-        let s = T::state();
-
-        s.end_waker.register(cx.waker());
-        if r.events_stopped.read().bits() != 0 {
-            r.events_stopped.reset();
-
-            return Poll::Ready(());
-        }
+    pub async fn read(&mut self, address: u8, buffer: &mut [u8]) -> Result<(), Error> {
+        self.setup_read(address, buffer, true)?;
+        self.async_wait().await;
+        compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_rx(buffer.len())?;
+        Ok(())
+    }
 
-        // stop if an error occured
-        if r.events_error.read().bits() != 0 {
-            r.events_error.reset();
-            r.tasks_stop.write(|w| unsafe { w.bits(1) });
-        }
+    pub async fn write(&mut self, address: u8, buffer: &[u8]) -> Result<(), Error> {
+        self.setup_write(address, buffer, true)?;
+        self.async_wait().await;
+        compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_tx(buffer.len())?;
+        Ok(())
+    }
 
-        Poll::Pending
+    pub async fn write_read(
+        &mut self,
+        address: u8,
+        wr_buffer: &[u8],
+        rd_buffer: &mut [u8],
+    ) -> Result<(), Error> {
+        self.setup_write_read(address, wr_buffer, rd_buffer, true)?;
+        self.async_wait().await;
+        compiler_fence(SeqCst);
+        self.check_errorsrc()?;
+        self.check_tx(wr_buffer.len())?;
+        self.check_rx(rd_buffer.len())?;
+        Ok(())
     }
 }
 
@@ -450,7 +507,7 @@ impl<'a, T: Instance> Drop for Twim<'a, T> {
     fn drop(&mut self) {
         trace!("twim drop");
 
-        // TODO when implementing async here, check for abort
+        // TODO: check for abort
 
         // disable!
         let r = T::regs();
@@ -463,292 +520,228 @@ impl<'a, T: Instance> Drop for Twim<'a, T> {
     }
 }
 
-impl<'d, T> I2c for Twim<'d, T>
-where
-    T: Instance,
-{
-    type Error = Error;
-
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-    type WriteReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), Self::Error>> + 'a;
-
-    fn read<'a>(&'a mut self, address: u8, buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        async move {
-            // NOTE: RAM slice check for buffer is not necessary, as a mutable
-            // slice can only be built from data located in RAM.
-
-            let r = T::regs();
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started.
-            compiler_fence(SeqCst);
-
-            r.address.write(|w| unsafe { w.address().bits(address) });
-
-            // Set up the DMA read.
-            unsafe { self.set_rx_buffer(buffer)? };
-
-            // Reset events
-            r.events_stopped.reset();
-            r.events_error.reset();
-            self.clear_errorsrc();
-
-            // Enable events
-            r.intenset.write(|w| w.stopped().set().error().set());
-
-            // Start read operation.
-            r.shorts.write(|w| w.lastrx_stop().enabled());
-            r.tasks_startrx.write(|w|
-            // `1` is a valid value to write to task registers.
-            unsafe { w.bits(1) });
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // after all possible DMA actions have completed.
-            compiler_fence(SeqCst);
-
-            // Wait for 'stopped' event.
-            poll_fn(Self::wait_for_stopped_event).await;
+pub(crate) mod sealed {
+    use super::*;
 
-            self.read_errorsrc()?;
+    pub struct State {
+        pub end_waker: AtomicWaker,
+    }
 
-            if r.rxd.amount.read().bits() != buffer.len() as u32 {
-                return Err(Error::Receive);
+    impl State {
+        pub const fn new() -> Self {
+            Self {
+                end_waker: AtomicWaker::new(),
             }
-
-            Ok(())
         }
     }
 
-    fn write<'a>(&'a mut self, address: u8, bytes: &'a [u8]) -> Self::WriteFuture<'a> {
-        async move {
-            slice_in_ram_or(bytes, Error::DMABufferNotInDataMemory)?;
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started.
-            compiler_fence(SeqCst);
-
-            let r = T::regs();
+    pub trait Instance {
+        fn regs() -> &'static pac::twim0::RegisterBlock;
+        fn state() -> &'static State;
+    }
+}
 
-            // Set up current address we're trying to talk to
-            r.address.write(|w| unsafe { w.address().bits(address) });
+pub trait Instance: Unborrow<Target = Self> + sealed::Instance + 'static {
+    type Interrupt: Interrupt;
+}
 
-            // Set up DMA write.
-            unsafe {
-                self.set_tx_buffer(bytes)?;
+macro_rules! impl_twim {
+    ($type:ident, $pac_type:ident, $irq:ident) => {
+        impl crate::twim::sealed::Instance for peripherals::$type {
+            fn regs() -> &'static pac::twim0::RegisterBlock {
+                unsafe { &*pac::$pac_type::ptr() }
             }
-
-            // Reset events
-            r.events_stopped.reset();
-            r.events_error.reset();
-            r.events_lasttx.reset();
-            self.clear_errorsrc();
-
-            // Enable events
-            r.intenset.write(|w| w.stopped().set().error().set());
-
-            // Start write operation.
-            r.shorts.write(|w| w.lasttx_stop().enabled());
-            r.tasks_starttx.write(|w|
-            // `1` is a valid value to write to task registers.
-            unsafe { w.bits(1) });
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // after all possible DMA actions have completed.
-            compiler_fence(SeqCst);
-
-            // Wait for 'stopped' event.
-            poll_fn(Self::wait_for_stopped_event).await;
-
-            self.read_errorsrc()?;
-
-            if r.txd.amount.read().bits() != bytes.len() as u32 {
-                return Err(Error::Transmit);
+            fn state() -> &'static crate::twim::sealed::State {
+                static STATE: crate::twim::sealed::State = crate::twim::sealed::State::new();
+                &STATE
             }
-
-            Ok(())
         }
-    }
-
-    fn write_read<'a>(
-        &'a mut self,
-        address: u8,
-        bytes: &'a [u8],
-        buffer: &'a mut [u8],
-    ) -> Self::WriteReadFuture<'a> {
-        async move {
-            slice_in_ram_or(bytes, Error::DMABufferNotInDataMemory)?;
-            // NOTE: RAM slice check for buffer is not necessary, as a mutable
-            // slice can only be built from data located in RAM.
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // before any DMA action has started.
-            compiler_fence(SeqCst);
+        impl crate::twim::Instance for peripherals::$type {
+            type Interrupt = crate::interrupt::$irq;
+        }
+    };
+}
 
-            let r = T::regs();
+// ====================
 
-            // Set up current address we're trying to talk to
-            r.address.write(|w| unsafe { w.address().bits(address) });
+mod eh02 {
+    use super::*;
 
-            // Set up DMA buffers.
-            unsafe {
-                self.set_tx_buffer(bytes)?;
-                self.set_rx_buffer(buffer)?;
+    impl<'a, T: Instance> embedded_hal_02::blocking::i2c::Write for Twim<'a, T> {
+        type Error = Error;
+
+        fn write<'w>(&mut self, addr: u8, bytes: &'w [u8]) -> Result<(), Error> {
+            if slice_in_ram(bytes) {
+                self.blocking_write(addr, bytes)
+            } else {
+                let buf = &mut [0; FORCE_COPY_BUFFER_SIZE][..];
+                for chunk in bytes.chunks(FORCE_COPY_BUFFER_SIZE) {
+                    buf[..chunk.len()].copy_from_slice(chunk);
+                    self.blocking_write(addr, &buf[..chunk.len()])?;
+                }
+                Ok(())
             }
+        }
+    }
 
-            // Reset events
-            r.events_stopped.reset();
-            r.events_error.reset();
-            r.events_lasttx.reset();
-            self.clear_errorsrc();
-
-            // Enable events
-            r.intenset.write(|w| w.stopped().set().error().set());
-
-            // Start write+read operation.
-            r.shorts.write(|w| {
-                w.lasttx_startrx().enabled();
-                w.lastrx_stop().enabled();
-                w
-            });
-            // `1` is a valid value to write to task registers.
-            r.tasks_starttx.write(|w| unsafe { w.bits(1) });
-
-            // Conservative compiler fence to prevent optimizations that do not
-            // take in to account actions by DMA. The fence has been placed here,
-            // after all possible DMA actions have completed.
-            compiler_fence(SeqCst);
-
-            // Wait for 'stopped' event.
-            poll_fn(Self::wait_for_stopped_event).await;
-
-            self.read_errorsrc()?;
-
-            let bad_write = r.txd.amount.read().bits() != bytes.len() as u32;
-            let bad_read = r.rxd.amount.read().bits() != buffer.len() as u32;
+    impl<'a, T: Instance> embedded_hal_02::blocking::i2c::Read for Twim<'a, T> {
+        type Error = Error;
 
-            if bad_write {
-                return Err(Error::Transmit);
-            }
+        fn read<'w>(&mut self, addr: u8, bytes: &'w mut [u8]) -> Result<(), Error> {
+            self.blocking_read(addr, bytes)
+        }
+    }
 
-            if bad_read {
-                return Err(Error::Receive);
+    impl<'a, T: Instance> embedded_hal_02::blocking::i2c::WriteRead for Twim<'a, T> {
+        type Error = Error;
+
+        fn write_read<'w>(
+            &mut self,
+            addr: u8,
+            bytes: &'w [u8],
+            buffer: &'w mut [u8],
+        ) -> Result<(), Error> {
+            if slice_in_ram(bytes) {
+                self.blocking_write_read(addr, bytes, buffer)
+            } else {
+                self.blocking_copy_write_read(addr, bytes, buffer)
             }
-
-            Ok(())
         }
     }
 }
 
-impl<'a, T: Instance> embedded_hal::blocking::i2c::Write for Twim<'a, T> {
-    type Error = Error;
+#[cfg(feature = "unstable-traits")]
+mod eh1 {
+    use super::*;
 
-    fn write<'w>(&mut self, addr: u8, bytes: &'w [u8]) -> Result<(), Error> {
-        if slice_in_ram(bytes) {
-            self.write(addr, bytes)
-        } else {
-            let buf = &mut [0; FORCE_COPY_BUFFER_SIZE][..];
-            for chunk in bytes.chunks(FORCE_COPY_BUFFER_SIZE) {
-                buf[..chunk.len()].copy_from_slice(chunk);
-                self.write(addr, &buf[..chunk.len()])?;
+    impl embedded_hal_1::i2c::Error for Error {
+        fn kind(&self) -> embedded_hal_1::i2c::ErrorKind {
+            match *self {
+                Self::TxBufferTooLong => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::RxBufferTooLong => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::TxBufferZeroLength => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::RxBufferZeroLength => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::Transmit => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::Receive => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::DMABufferNotInDataMemory => embedded_hal_1::i2c::ErrorKind::Other,
+                Self::AddressNack => embedded_hal_1::i2c::ErrorKind::NoAcknowledge(
+                    embedded_hal_1::i2c::NoAcknowledgeSource::Address,
+                ),
+                Self::DataNack => embedded_hal_1::i2c::ErrorKind::NoAcknowledge(
+                    embedded_hal_1::i2c::NoAcknowledgeSource::Data,
+                ),
+                Self::Overrun => embedded_hal_1::i2c::ErrorKind::Overrun,
             }
-            Ok(())
         }
     }
-}
 
-impl<'a, T: Instance> embedded_hal::blocking::i2c::Read for Twim<'a, T> {
-    type Error = Error;
-
-    fn read<'w>(&mut self, addr: u8, bytes: &'w mut [u8]) -> Result<(), Error> {
-        self.read(addr, bytes)
+    impl<'d, T: Instance> embedded_hal_1::i2c::ErrorType for Twim<'d, T> {
+        type Error = Error;
     }
-}
 
-impl<'a, T: Instance> embedded_hal::blocking::i2c::WriteRead for Twim<'a, T> {
-    type Error = Error;
+    impl<'d, T: Instance> embedded_hal_1::i2c::blocking::I2c for Twim<'d, T> {
+        fn read(&mut self, address: u8, buffer: &mut [u8]) -> Result<(), Self::Error> {
+            self.blocking_read(address, buffer)
+        }
 
-    fn write_read<'w>(
-        &mut self,
-        addr: u8,
-        bytes: &'w [u8],
-        buffer: &'w mut [u8],
-    ) -> Result<(), Error> {
-        if slice_in_ram(bytes) {
-            self.write_then_read(addr, bytes, buffer)
-        } else {
-            self.copy_write_then_read(addr, bytes, buffer)
+        fn write(&mut self, address: u8, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(address, buffer)
         }
-    }
-}
 
-#[derive(Debug, Copy, Clone, Eq, PartialEq)]
-#[cfg_attr(feature = "defmt", derive(defmt::Format))]
-pub enum Error {
-    TxBufferTooLong,
-    RxBufferTooLong,
-    TxBufferZeroLength,
-    RxBufferZeroLength,
-    Transmit,
-    Receive,
-    DMABufferNotInDataMemory,
-    AddressNack,
-    DataNack,
-    Overrun,
-}
+        fn write_iter<B>(&mut self, _address: u8, _bytes: B) -> Result<(), Self::Error>
+        where
+            B: IntoIterator<Item = u8>,
+        {
+            todo!();
+        }
 
-pub(crate) mod sealed {
-    use super::*;
+        fn write_iter_read<B>(
+            &mut self,
+            _address: u8,
+            _bytes: B,
+            _buffer: &mut [u8],
+        ) -> Result<(), Self::Error>
+        where
+            B: IntoIterator<Item = u8>,
+        {
+            todo!();
+        }
 
-    pub struct State {
-        pub end_waker: AtomicWaker,
-    }
+        fn write_read(
+            &mut self,
+            address: u8,
+            wr_buffer: &[u8],
+            rd_buffer: &mut [u8],
+        ) -> Result<(), Self::Error> {
+            self.blocking_write_read(address, wr_buffer, rd_buffer)
+        }
 
-    impl State {
-        pub const fn new() -> Self {
-            Self {
-                end_waker: AtomicWaker::new(),
-            }
+        fn transaction<'a>(
+            &mut self,
+            _address: u8,
+            _operations: &mut [embedded_hal_async::i2c::Operation<'a>],
+        ) -> Result<(), Self::Error> {
+            todo!();
         }
-    }
 
-    pub trait Instance {
-        fn regs() -> &'static pac::twim0::RegisterBlock;
-        fn state() -> &'static State;
+        fn transaction_iter<'a, O>(
+            &mut self,
+            _address: u8,
+            _operations: O,
+        ) -> Result<(), Self::Error>
+        where
+            O: IntoIterator<Item = embedded_hal_async::i2c::Operation<'a>>,
+        {
+            todo!();
+        }
     }
-}
 
-pub trait Instance: Unborrow<Target = Self> + sealed::Instance + 'static {
-    type Interrupt: Interrupt;
-}
+    impl<'d, T: Instance> embedded_hal_async::i2c::I2c for Twim<'d, T> {
+        type ReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
 
-macro_rules! impl_twim {
-    ($type:ident, $pac_type:ident, $irq:ident) => {
-        impl crate::twim::sealed::Instance for peripherals::$type {
-            fn regs() -> &'static pac::twim0::RegisterBlock {
-                unsafe { &*pac::$pac_type::ptr() }
-            }
-            fn state() -> &'static crate::twim::sealed::State {
-                static STATE: crate::twim::sealed::State = crate::twim::sealed::State::new();
-                &STATE
-            }
+        fn read<'a>(&'a mut self, address: u8, buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
+            self.read(address, buffer)
         }
-        impl crate::twim::Instance for peripherals::$type {
-            type Interrupt = crate::interrupt::$irq;
+
+        type WriteFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write<'a>(&'a mut self, address: u8, bytes: &'a [u8]) -> Self::WriteFuture<'a> {
+            self.write(address, bytes)
         }
-    };
+
+        type WriteReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write_read<'a>(
+            &'a mut self,
+            address: u8,
+            wr_buffer: &'a [u8],
+            rd_buffer: &'a mut [u8],
+        ) -> Self::WriteReadFuture<'a> {
+            self.write_read(address, wr_buffer, rd_buffer)
+        }
+
+        type TransactionFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn transaction<'a>(
+            &'a mut self,
+            address: u8,
+            operations: &mut [embedded_hal_async::i2c::Operation<'a>],
+        ) -> Self::TransactionFuture<'a> {
+            let _ = address;
+            let _ = operations;
+            async move { todo!() }
+        }
+    }
 }
diff --git a/embassy-nrf/src/uarte.rs b/embassy-nrf/src/uarte.rs
index 38480ecca..b10e55a05 100644
--- a/embassy-nrf/src/uarte.rs
+++ b/embassy-nrf/src/uarte.rs
@@ -13,12 +13,10 @@
 //! memory may be used given that buffers are passed in directly to its read and write
 //! methods.
 
-use core::future::Future;
 use core::marker::PhantomData;
 use core::sync::atomic::{compiler_fence, Ordering};
 use core::task::Poll;
 use embassy::interrupt::InterruptExt;
-use embassy::traits::uart::{Error as TraitError, Read, ReadUntilIdle, Write};
 use embassy::util::Unborrow;
 use embassy_hal_common::drop::OnDrop;
 use embassy_hal_common::unborrow;
@@ -32,6 +30,7 @@ use crate::pac;
 use crate::ppi::{AnyConfigurableChannel, ConfigurableChannel, Event, Ppi, Task};
 use crate::timer::Instance as TimerInstance;
 use crate::timer::{Frequency, Timer};
+use crate::util::slice_in_ram_or;
 
 // Re-export SVD variants to allow user to directly set values.
 pub use pac::uarte0::{baudrate::BAUDRATE_A as Baudrate, config::PARITY_A as Parity};
@@ -51,6 +50,16 @@ impl Default for Config {
     }
 }
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+#[cfg_attr(feature = "defmt", derive(defmt::Format))]
+#[non_exhaustive]
+pub enum Error {
+    BufferTooLong,
+    BufferZeroLength,
+    DMABufferNotInDataMemory,
+    // TODO: add other error variants.
+}
+
 /// Interface to the UARTE peripheral
 pub struct Uarte<'d, T: Instance> {
     phantom: PhantomData<&'d mut T>,
@@ -139,8 +148,12 @@ impl<'d, T: Instance> Uarte<'d, T> {
 
         Self {
             phantom: PhantomData,
-            tx: UarteTx::new(),
-            rx: UarteRx::new(),
+            tx: UarteTx {
+                phantom: PhantomData,
+            },
+            rx: UarteRx {
+                phantom: PhantomData,
+            },
         }
     }
 
@@ -170,92 +183,110 @@ impl<'d, T: Instance> Uarte<'d, T> {
             r.intenclr.write(|w| w.endtx().clear());
         }
     }
-}
 
-impl<'d, T: Instance> Read for Uarte<'d, T> {
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
+    pub async fn read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        self.rx.read(buffer).await
+    }
 
-    fn read<'a>(&'a mut self, rx_buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        self.rx.read(rx_buffer)
+    pub async fn write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        self.tx.write(buffer).await
     }
-}
 
-impl<'d, T: Instance> Write for Uarte<'d, T> {
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
+    pub fn blocking_read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        self.rx.blocking_read(buffer)
+    }
 
-    fn write<'a>(&'a mut self, tx_buffer: &'a [u8]) -> Self::WriteFuture<'a> {
-        self.tx.write(tx_buffer)
+    pub fn blocking_write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        self.tx.blocking_write(buffer)
     }
 }
 
 impl<'d, T: Instance> UarteTx<'d, T> {
-    pub fn new() -> Self {
-        Self {
-            phantom: PhantomData,
+    pub async fn write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        slice_in_ram_or(buffer, Error::DMABufferNotInDataMemory)?;
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
+        }
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
         }
-    }
-}
 
-impl<'d, T: Instance> Write for UarteTx<'d, T> {
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
 
-    fn write<'a>(&'a mut self, tx_buffer: &'a [u8]) -> Self::WriteFuture<'a> {
-        async move {
-            let ptr = tx_buffer.as_ptr();
-            let len = tx_buffer.len();
-            assert!(len <= EASY_DMA_SIZE);
-            // TODO: panic if buffer is not in SRAM
+        let r = T::regs();
+        let s = T::state();
 
-            let r = T::regs();
-            let s = T::state();
+        let drop = OnDrop::new(move || {
+            trace!("write drop: stopping");
 
-            let drop = OnDrop::new(move || {
-                trace!("write drop: stopping");
+            r.intenclr.write(|w| w.endtx().clear());
+            r.events_txstopped.reset();
+            r.tasks_stoptx.write(|w| unsafe { w.bits(1) });
 
-                r.intenclr.write(|w| w.endtx().clear());
-                r.events_txstopped.reset();
-                r.tasks_stoptx.write(|w| unsafe { w.bits(1) });
+            // TX is stopped almost instantly, spinning is fine.
+            while r.events_endtx.read().bits() == 0 {}
+            trace!("write drop: stopped");
+        });
 
-                // TX is stopped almost instantly, spinning is fine.
-                while r.events_endtx.read().bits() == 0 {}
-                trace!("write drop: stopped");
-            });
+        r.txd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.txd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
 
-            r.txd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
-            r.txd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+        r.events_endtx.reset();
+        r.intenset.write(|w| w.endtx().set());
 
-            r.events_endtx.reset();
-            r.intenset.write(|w| w.endtx().set());
+        compiler_fence(Ordering::SeqCst);
 
-            compiler_fence(Ordering::SeqCst);
+        trace!("starttx");
+        r.tasks_starttx.write(|w| unsafe { w.bits(1) });
 
-            trace!("starttx");
-            r.tasks_starttx.write(|w| unsafe { w.bits(1) });
+        poll_fn(|cx| {
+            s.endtx_waker.register(cx.waker());
+            if r.events_endtx.read().bits() != 0 {
+                return Poll::Ready(());
+            }
+            Poll::Pending
+        })
+        .await;
 
-            poll_fn(|cx| {
-                s.endtx_waker.register(cx.waker());
-                if r.events_endtx.read().bits() != 0 {
-                    return Poll::Ready(());
-                }
-                Poll::Pending
-            })
-            .await;
+        compiler_fence(Ordering::SeqCst);
+        r.events_txstarted.reset();
+        drop.defuse();
 
-            compiler_fence(Ordering::SeqCst);
-            r.events_txstarted.reset();
-            drop.defuse();
+        Ok(())
+    }
 
-            Ok(())
+    pub fn blocking_write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        slice_in_ram_or(buffer, Error::DMABufferNotInDataMemory)?;
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
+        }
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
         }
+
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
+
+        let r = T::regs();
+
+        r.txd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.txd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+
+        r.events_endtx.reset();
+        r.intenclr.write(|w| w.endtx().clear());
+
+        compiler_fence(Ordering::SeqCst);
+
+        trace!("starttx");
+        r.tasks_starttx.write(|w| unsafe { w.bits(1) });
+
+        while r.events_endtx.read().bits() == 0 {}
+
+        compiler_fence(Ordering::SeqCst);
+        r.events_txstarted.reset();
+
+        Ok(())
     }
 }
 
@@ -278,66 +309,89 @@ impl<'a, T: Instance> Drop for UarteTx<'a, T> {
 }
 
 impl<'d, T: Instance> UarteRx<'d, T> {
-    pub fn new() -> Self {
-        Self {
-            phantom: PhantomData,
+    pub async fn read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
+        }
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
         }
-    }
-}
 
-impl<'d, T: Instance> Read for UarteRx<'d, T> {
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
 
-    fn read<'a>(&'a mut self, rx_buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        async move {
-            let ptr = rx_buffer.as_ptr();
-            let len = rx_buffer.len();
-            assert!(len <= EASY_DMA_SIZE);
+        let r = T::regs();
+        let s = T::state();
 
-            let r = T::regs();
-            let s = T::state();
+        let drop = OnDrop::new(move || {
+            trace!("read drop: stopping");
 
-            let drop = OnDrop::new(move || {
-                trace!("read drop: stopping");
+            r.intenclr.write(|w| w.endrx().clear());
+            r.events_rxto.reset();
+            r.tasks_stoprx.write(|w| unsafe { w.bits(1) });
 
-                r.intenclr.write(|w| w.endrx().clear());
-                r.events_rxto.reset();
-                r.tasks_stoprx.write(|w| unsafe { w.bits(1) });
+            while r.events_endrx.read().bits() == 0 {}
 
-                while r.events_endrx.read().bits() == 0 {}
+            trace!("read drop: stopped");
+        });
 
-                trace!("read drop: stopped");
-            });
+        r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
 
-            r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
-            r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+        r.events_endrx.reset();
+        r.intenset.write(|w| w.endrx().set());
 
-            r.events_endrx.reset();
-            r.intenset.write(|w| w.endrx().set());
+        compiler_fence(Ordering::SeqCst);
 
-            compiler_fence(Ordering::SeqCst);
+        trace!("startrx");
+        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
 
-            trace!("startrx");
-            r.tasks_startrx.write(|w| unsafe { w.bits(1) });
+        poll_fn(|cx| {
+            s.endrx_waker.register(cx.waker());
+            if r.events_endrx.read().bits() != 0 {
+                return Poll::Ready(());
+            }
+            Poll::Pending
+        })
+        .await;
 
-            poll_fn(|cx| {
-                s.endrx_waker.register(cx.waker());
-                if r.events_endrx.read().bits() != 0 {
-                    return Poll::Ready(());
-                }
-                Poll::Pending
-            })
-            .await;
+        compiler_fence(Ordering::SeqCst);
+        r.events_rxstarted.reset();
+        drop.defuse();
 
-            compiler_fence(Ordering::SeqCst);
-            r.events_rxstarted.reset();
-            drop.defuse();
+        Ok(())
+    }
 
-            Ok(())
+    pub fn blocking_read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
+        }
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
         }
+
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
+
+        let r = T::regs();
+
+        r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+
+        r.events_endrx.reset();
+        r.intenclr.write(|w| w.endrx().clear());
+
+        compiler_fence(Ordering::SeqCst);
+
+        trace!("startrx");
+        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
+
+        while r.events_endrx.read().bits() == 0 {}
+
+        compiler_fence(Ordering::SeqCst);
+        r.events_rxstarted.reset();
+
+        Ok(())
     }
 }
 
@@ -439,14 +493,7 @@ pub struct UarteWithIdle<'d, U: Instance, T: TimerInstance> {
 impl<'d, U: Instance, T: TimerInstance> UarteWithIdle<'d, U, T> {
     /// Creates the interface to a UARTE instance.
     /// Sets the baud rate, parity and assigns the pins to the UARTE peripheral.
-    ///
-    /// # Safety
-    ///
-    /// The returned API is safe unless you use `mem::forget` (or similar safe mechanisms)
-    /// on stack allocated buffers which which have been passed to [`send()`](Uarte::send)
-    /// or [`receive`](Uarte::receive).
-    #[allow(unused_unsafe)]
-    pub unsafe fn new(
+    pub fn new(
         uarte: impl Unborrow<Target = U> + 'd,
         timer: impl Unborrow<Target = T> + 'd,
         ppi_ch1: impl Unborrow<Target = impl ConfigurableChannel + 'd> + 'd,
@@ -501,93 +548,119 @@ impl<'d, U: Instance, T: TimerInstance> UarteWithIdle<'d, U, T> {
             _ppi_ch2: ppi_ch2,
         }
     }
-}
 
-impl<'d, U: Instance, T: TimerInstance> ReadUntilIdle for UarteWithIdle<'d, U, T> {
-    type ReadUntilIdleFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<usize, TraitError>> + 'a;
-    fn read_until_idle<'a>(&'a mut self, rx_buffer: &'a mut [u8]) -> Self::ReadUntilIdleFuture<'a> {
-        async move {
-            let ptr = rx_buffer.as_ptr();
-            let len = rx_buffer.len();
-            assert!(len <= EASY_DMA_SIZE);
+    pub async fn read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        self.ppi_ch1.disable();
+        self.uarte.read(buffer).await
+    }
 
-            let r = U::regs();
-            let s = U::state();
+    pub async fn write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        self.uarte.write(buffer).await
+    }
 
-            let drop = OnDrop::new(|| {
-                trace!("read drop: stopping");
+    pub fn blocking_read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
+        self.ppi_ch1.disable();
+        self.uarte.blocking_read(buffer)
+    }
 
-                self.timer.stop();
+    pub fn blocking_write(&mut self, buffer: &[u8]) -> Result<(), Error> {
+        self.uarte.blocking_write(buffer)
+    }
 
-                r.intenclr.write(|w| w.endrx().clear());
-                r.events_rxto.reset();
-                r.tasks_stoprx.write(|w| unsafe { w.bits(1) });
+    pub async fn read_until_idle(&mut self, buffer: &mut [u8]) -> Result<usize, Error> {
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
+        }
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
+        }
 
-                while r.events_endrx.read().bits() == 0 {}
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
 
-                trace!("read drop: stopped");
-            });
+        let r = U::regs();
+        let s = U::state();
 
-            r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
-            r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+        self.ppi_ch1.enable();
 
-            r.events_endrx.reset();
-            r.intenset.write(|w| w.endrx().set());
+        let drop = OnDrop::new(|| {
+            trace!("read drop: stopping");
 
-            compiler_fence(Ordering::SeqCst);
+            self.timer.stop();
 
-            trace!("startrx");
-            r.tasks_startrx.write(|w| unsafe { w.bits(1) });
+            r.intenclr.write(|w| w.endrx().clear());
+            r.events_rxto.reset();
+            r.tasks_stoprx.write(|w| unsafe { w.bits(1) });
 
-            poll_fn(|cx| {
-                s.endrx_waker.register(cx.waker());
-                if r.events_endrx.read().bits() != 0 {
-                    return Poll::Ready(());
-                }
-                Poll::Pending
-            })
-            .await;
+            while r.events_endrx.read().bits() == 0 {}
 
-            compiler_fence(Ordering::SeqCst);
-            let n = r.rxd.amount.read().amount().bits() as usize;
+            trace!("read drop: stopped");
+        });
 
-            // Stop timer
-            self.timer.stop();
-            r.events_rxstarted.reset();
+        r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
 
-            drop.defuse();
+        r.events_endrx.reset();
+        r.intenset.write(|w| w.endrx().set());
 
-            Ok(n)
-        }
+        compiler_fence(Ordering::SeqCst);
+
+        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
+
+        poll_fn(|cx| {
+            s.endrx_waker.register(cx.waker());
+            if r.events_endrx.read().bits() != 0 {
+                return Poll::Ready(());
+            }
+            Poll::Pending
+        })
+        .await;
+
+        compiler_fence(Ordering::SeqCst);
+        let n = r.rxd.amount.read().amount().bits() as usize;
+
+        self.timer.stop();
+        r.events_rxstarted.reset();
+
+        drop.defuse();
+
+        Ok(n)
     }
-}
 
-impl<'d, U: Instance, T: TimerInstance> Read for UarteWithIdle<'d, U, T> {
-    type ReadFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
-    fn read<'a>(&'a mut self, rx_buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
-        async move {
-            self.ppi_ch1.disable();
-            let result = self.uarte.read(rx_buffer).await;
-            self.ppi_ch1.enable();
-            result
+    pub fn blocking_read_until_idle(&mut self, buffer: &mut [u8]) -> Result<usize, Error> {
+        if buffer.len() == 0 {
+            return Err(Error::BufferZeroLength);
         }
-    }
-}
+        if buffer.len() > EASY_DMA_SIZE {
+            return Err(Error::BufferTooLong);
+        }
+
+        let ptr = buffer.as_ptr();
+        let len = buffer.len();
+
+        let r = U::regs();
+
+        self.ppi_ch1.enable();
+
+        r.rxd.ptr.write(|w| unsafe { w.ptr().bits(ptr as u32) });
+        r.rxd.maxcnt.write(|w| unsafe { w.maxcnt().bits(len as _) });
+
+        r.events_endrx.reset();
+        r.intenclr.write(|w| w.endrx().clear());
+
+        compiler_fence(Ordering::SeqCst);
+
+        r.tasks_startrx.write(|w| unsafe { w.bits(1) });
 
-impl<'d, U: Instance, T: TimerInstance> Write for UarteWithIdle<'d, U, T> {
-    type WriteFuture<'a>
-    where
-        Self: 'a,
-    = impl Future<Output = Result<(), TraitError>> + 'a;
+        while r.events_endrx.read().bits() == 0 {}
 
-    fn write<'a>(&'a mut self, tx_buffer: &'a [u8]) -> Self::WriteFuture<'a> {
-        self.uarte.write(tx_buffer)
+        compiler_fence(Ordering::SeqCst);
+        let n = r.rxd.amount.read().amount().bits() as usize;
+
+        self.timer.stop();
+        r.events_rxstarted.reset();
+
+        Ok(n)
     }
 }
 
@@ -639,3 +712,206 @@ macro_rules! impl_uarte {
         }
     };
 }
+
+// ====================
+
+mod eh02 {
+    use super::*;
+
+    impl<'d, T: Instance> embedded_hal_02::blocking::serial::Write<u8> for Uarte<'d, T> {
+        type Error = Error;
+
+        fn bwrite_all(&mut self, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(buffer)
+        }
+
+        fn bflush(&mut self) -> Result<(), Self::Error> {
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_02::blocking::serial::Write<u8> for UarteTx<'d, T> {
+        type Error = Error;
+
+        fn bwrite_all(&mut self, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(buffer)
+        }
+
+        fn bflush(&mut self) -> Result<(), Self::Error> {
+            Ok(())
+        }
+    }
+
+    impl<'d, U: Instance, T: TimerInstance> embedded_hal_02::blocking::serial::Write<u8>
+        for UarteWithIdle<'d, U, T>
+    {
+        type Error = Error;
+
+        fn bwrite_all(&mut self, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(buffer)
+        }
+
+        fn bflush(&mut self) -> Result<(), Self::Error> {
+            Ok(())
+        }
+    }
+}
+
+#[cfg(feature = "unstable-traits")]
+mod eh1 {
+    use super::*;
+    use core::future::Future;
+
+    impl embedded_hal_1::serial::Error for Error {
+        fn kind(&self) -> embedded_hal_1::serial::ErrorKind {
+            match *self {
+                Self::BufferTooLong => embedded_hal_1::serial::ErrorKind::Other,
+                Self::BufferZeroLength => embedded_hal_1::serial::ErrorKind::Other,
+                Self::DMABufferNotInDataMemory => embedded_hal_1::serial::ErrorKind::Other,
+            }
+        }
+    }
+
+    // =====================
+
+    impl<'d, T: Instance> embedded_hal_1::serial::ErrorType for Uarte<'d, T> {
+        type Error = Error;
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::serial::blocking::Write for Uarte<'d, T> {
+        fn write(&mut self, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(buffer)
+        }
+
+        fn flush(&mut self) -> Result<(), Self::Error> {
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::serial::Read for Uarte<'d, T> {
+        type ReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn read<'a>(&'a mut self, buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
+            self.read(buffer)
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::serial::Write for Uarte<'d, T> {
+        type WriteFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write<'a>(&'a mut self, buffer: &'a [u8]) -> Self::WriteFuture<'a> {
+            self.write(buffer)
+        }
+
+        type FlushFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn flush<'a>(&'a mut self) -> Self::FlushFuture<'a> {
+            async move { Ok(()) }
+        }
+    }
+
+    // =====================
+
+    impl<'d, T: Instance> embedded_hal_1::serial::ErrorType for UarteTx<'d, T> {
+        type Error = Error;
+    }
+
+    impl<'d, T: Instance> embedded_hal_1::serial::blocking::Write for UarteTx<'d, T> {
+        fn write(&mut self, buffer: &[u8]) -> Result<(), Self::Error> {
+            self.blocking_write(buffer)
+        }
+
+        fn flush(&mut self) -> Result<(), Self::Error> {
+            Ok(())
+        }
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::serial::Write for UarteTx<'d, T> {
+        type WriteFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write<'a>(&'a mut self, buffer: &'a [u8]) -> Self::WriteFuture<'a> {
+            self.write(buffer)
+        }
+
+        type FlushFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn flush<'a>(&'a mut self) -> Self::FlushFuture<'a> {
+            async move { Ok(()) }
+        }
+    }
+
+    // =====================
+
+    impl<'d, T: Instance> embedded_hal_1::serial::ErrorType for UarteRx<'d, T> {
+        type Error = Error;
+    }
+
+    impl<'d, T: Instance> embedded_hal_async::serial::Read for UarteRx<'d, T> {
+        type ReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn read<'a>(&'a mut self, buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
+            self.read(buffer)
+        }
+    }
+
+    // =====================
+
+    impl<'d, U: Instance, T: TimerInstance> embedded_hal_1::serial::ErrorType
+        for UarteWithIdle<'d, U, T>
+    {
+        type Error = Error;
+    }
+
+    impl<'d, U: Instance, T: TimerInstance> embedded_hal_async::serial::Read
+        for UarteWithIdle<'d, U, T>
+    {
+        type ReadFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn read<'a>(&'a mut self, buffer: &'a mut [u8]) -> Self::ReadFuture<'a> {
+            self.read(buffer)
+        }
+    }
+
+    impl<'d, U: Instance, T: TimerInstance> embedded_hal_async::serial::Write
+        for UarteWithIdle<'d, U, T>
+    {
+        type WriteFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn write<'a>(&'a mut self, buffer: &'a [u8]) -> Self::WriteFuture<'a> {
+            self.write(buffer)
+        }
+
+        type FlushFuture<'a>
+        where
+            Self: 'a,
+        = impl Future<Output = Result<(), Self::Error>> + 'a;
+
+        fn flush<'a>(&'a mut self) -> Self::FlushFuture<'a> {
+            async move { Ok(()) }
+        }
+    }
+}
diff --git a/embassy-nrf/src/util.rs b/embassy-nrf/src/util.rs
index 2fd0bc5a8..76162b701 100644
--- a/embassy-nrf/src/util.rs
+++ b/embassy-nrf/src/util.rs
@@ -1,16 +1,28 @@
+use core::mem;
+
 const SRAM_LOWER: usize = 0x2000_0000;
 const SRAM_UPPER: usize = 0x3000_0000;
 
+// TODO: replace transmutes with core::ptr::metadata once it's stable
+
+pub(crate) fn slice_ptr_parts<T>(slice: *const [T]) -> (usize, usize) {
+    unsafe { mem::transmute(slice) }
+}
+
+pub(crate) fn slice_ptr_parts_mut<T>(slice: *mut [T]) -> (usize, usize) {
+    unsafe { mem::transmute(slice) }
+}
+
 /// Does this slice reside entirely within RAM?
-pub(crate) fn slice_in_ram<T>(slice: &[T]) -> bool {
-    let ptr = slice.as_ptr() as usize;
-    ptr >= SRAM_LOWER && (ptr + slice.len() * core::mem::size_of::<T>()) < SRAM_UPPER
+pub(crate) fn slice_in_ram<T>(slice: *const [T]) -> bool {
+    let (ptr, len) = slice_ptr_parts(slice);
+    ptr >= SRAM_LOWER && (ptr + len * core::mem::size_of::<T>()) < SRAM_UPPER
 }
 
 /// Return an error if slice is not in RAM.
 #[cfg(not(feature = "nrf51"))]
-pub(crate) fn slice_in_ram_or<T, E>(slice: &[T], err: E) -> Result<(), E> {
-    if slice.is_empty() || slice_in_ram(slice) {
+pub(crate) fn slice_in_ram_or<T, E>(slice: *const [T], err: E) -> Result<(), E> {
+    if slice_in_ram(slice) {
         Ok(())
     } else {
         Err(err)
diff --git a/examples/nrf/Cargo.toml b/examples/nrf/Cargo.toml
index ca013f8b7..fa7286923 100644
--- a/examples/nrf/Cargo.toml
+++ b/examples/nrf/Cargo.toml
@@ -15,7 +15,6 @@ defmt-rtt = "0.3"
 
 cortex-m = "0.7.3"
 cortex-m-rt = "0.7.0"
-embedded-hal = "0.2.6"
 panic-probe = { version = "0.3", features = ["print-defmt"] }
 futures = { version = "0.3.17", default-features = false, features = ["async-await"] }
 rand = { version = "0.8.4", default-features = false }
diff --git a/examples/nrf/src/bin/gpiote_port.rs b/examples/nrf/src/bin/gpiote_port.rs
index 76c861d95..3ea9a6e67 100644
--- a/examples/nrf/src/bin/gpiote_port.rs
+++ b/examples/nrf/src/bin/gpiote_port.rs
@@ -6,7 +6,6 @@
 mod example_common;
 
 use embassy::executor::Spawner;
-use embassy::traits::gpio::{WaitForHigh, WaitForLow};
 use embassy_nrf::gpio::{AnyPin, Input, Pin as _, Pull};
 use embassy_nrf::Peripherals;
 use example_common::*;
diff --git a/examples/nrf/src/bin/qspi.rs b/examples/nrf/src/bin/qspi.rs
index c7f9503bb..b2e6bfc15 100644
--- a/examples/nrf/src/bin/qspi.rs
+++ b/examples/nrf/src/bin/qspi.rs
@@ -7,7 +7,6 @@ mod example_common;
 
 use defmt::assert_eq;
 use embassy::executor::Spawner;
-use embassy::traits::flash::Flash;
 use embassy_nrf::Peripherals;
 use embassy_nrf::{interrupt, qspi};
 use example_common::*;
diff --git a/examples/nrf/src/bin/qspi_lowpower.rs b/examples/nrf/src/bin/qspi_lowpower.rs
index e9e1bac42..4e264ef2f 100644
--- a/examples/nrf/src/bin/qspi_lowpower.rs
+++ b/examples/nrf/src/bin/qspi_lowpower.rs
@@ -8,7 +8,6 @@ mod example_common;
 use core::mem;
 use embassy::executor::Spawner;
 use embassy::time::{Duration, Timer};
-use embassy::traits::flash::Flash;
 use embassy_nrf::Peripherals;
 use embassy_nrf::{interrupt, qspi};
 use example_common::*;
diff --git a/examples/nrf/src/bin/rng.rs b/examples/nrf/src/bin/rng.rs
index 2f1a26993..a35a9fa85 100644
--- a/examples/nrf/src/bin/rng.rs
+++ b/examples/nrf/src/bin/rng.rs
@@ -5,9 +5,7 @@
 #[path = "../example_common.rs"]
 mod example_common;
 
-use defmt::unwrap;
 use embassy::executor::Spawner;
-use embassy::traits::rng::Rng as _;
 use embassy_nrf::interrupt;
 use embassy_nrf::rng::Rng;
 use embassy_nrf::Peripherals;
@@ -19,14 +17,14 @@ async fn main(_spawner: Spawner, p: Peripherals) {
 
     // Async API
     let mut bytes = [0; 4];
-    unwrap!(rng.fill_bytes(&mut bytes).await); // nRF RNG is infallible
+    rng.fill_bytes(&mut bytes).await;
     defmt::info!("Some random bytes: {:?}", bytes);
 
     // Sync API with `rand`
     defmt::info!("A random number from 1 to 10: {:?}", rng.gen_range(1..=10));
 
     let mut bytes = [0; 1024];
-    unwrap!(rng.fill_bytes(&mut bytes).await);
+    rng.fill_bytes(&mut bytes).await;
     let zero_count: u32 = bytes.iter().fold(0, |acc, val| acc + val.count_zeros());
     let one_count: u32 = bytes.iter().fold(0, |acc, val| acc + val.count_ones());
     defmt::info!(
diff --git a/examples/nrf/src/bin/spim.rs b/examples/nrf/src/bin/spim.rs
index fc31d140a..cda3baa26 100644
--- a/examples/nrf/src/bin/spim.rs
+++ b/examples/nrf/src/bin/spim.rs
@@ -9,7 +9,6 @@ use embassy::executor::Spawner;
 use embassy_nrf::gpio::{Level, Output, OutputDrive};
 use embassy_nrf::Peripherals;
 use embassy_nrf::{interrupt, spim};
-use embassy_traits::spi::FullDuplex;
 use example_common::*;
 
 #[embassy::main]
@@ -31,7 +30,7 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     ncs.set_low();
     cortex_m::asm::delay(5);
     let tx = [0xFF];
-    unwrap!(spim.read_write(&mut [], &tx).await);
+    unwrap!(spim.transfer(&mut [], &tx).await);
     cortex_m::asm::delay(10);
     ncs.set_high();
 
@@ -44,7 +43,7 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     ncs.set_low();
     cortex_m::asm::delay(5000);
     let tx = [0b000_11101, 0];
-    unwrap!(spim.read_write(&mut rx, &tx).await);
+    unwrap!(spim.transfer(&mut rx, &tx).await);
     cortex_m::asm::delay(5000);
     ncs.set_high();
     info!("estat: {=[?]}", rx);
@@ -54,7 +53,7 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     ncs.set_low();
     cortex_m::asm::delay(5);
     let tx = [0b100_11111, 0b11];
-    unwrap!(spim.read_write(&mut rx, &tx).await);
+    unwrap!(spim.transfer(&mut rx, &tx).await);
     cortex_m::asm::delay(10);
     ncs.set_high();
 
@@ -63,7 +62,7 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     ncs.set_low();
     cortex_m::asm::delay(5);
     let tx = [0b000_10010, 0];
-    unwrap!(spim.read_write(&mut rx, &tx).await);
+    unwrap!(spim.transfer(&mut rx, &tx).await);
     cortex_m::asm::delay(10);
     ncs.set_high();
 
diff --git a/examples/nrf/src/bin/twim.rs b/examples/nrf/src/bin/twim.rs
index 4c2f2bf2e..1ac3a3945 100644
--- a/examples/nrf/src/bin/twim.rs
+++ b/examples/nrf/src/bin/twim.rs
@@ -26,7 +26,7 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     info!("Reading...");
 
     let mut buf = [0u8; 16];
-    unwrap!(twi.write_then_read(ADDRESS, &mut [0x00], &mut buf));
+    unwrap!(twi.blocking_write_read(ADDRESS, &mut [0x00], &mut buf));
 
     info!("Read: {=[u8]:x}", buf);
 }
diff --git a/examples/nrf/src/bin/twim_lowpower.rs b/examples/nrf/src/bin/twim_lowpower.rs
index 62a5f0c94..4a0596437 100644
--- a/examples/nrf/src/bin/twim_lowpower.rs
+++ b/examples/nrf/src/bin/twim_lowpower.rs
@@ -36,7 +36,7 @@ async fn main(_spawner: Spawner, mut p: Peripherals) {
         info!("Reading...");
 
         let mut buf = [0u8; 16];
-        unwrap!(twi.write_then_read(ADDRESS, &mut [0x00], &mut buf));
+        unwrap!(twi.blocking_write_read(ADDRESS, &mut [0x00], &mut buf));
 
         info!("Read: {=[u8]:x}", buf);
 
diff --git a/examples/nrf/src/bin/uart.rs b/examples/nrf/src/bin/uart.rs
index 208961c8b..68ee3978a 100644
--- a/examples/nrf/src/bin/uart.rs
+++ b/examples/nrf/src/bin/uart.rs
@@ -7,7 +7,6 @@ mod example_common;
 use example_common::*;
 
 use embassy::executor::Spawner;
-use embassy::traits::uart::{Read, Write};
 use embassy_nrf::gpio::NoPin;
 use embassy_nrf::{interrupt, uarte, Peripherals};
 
diff --git a/examples/nrf/src/bin/uart_idle.rs b/examples/nrf/src/bin/uart_idle.rs
index ec9a36026..76449c0e3 100644
--- a/examples/nrf/src/bin/uart_idle.rs
+++ b/examples/nrf/src/bin/uart_idle.rs
@@ -4,11 +4,9 @@
 
 #[path = "../example_common.rs"]
 mod example_common;
-use embassy_traits::uart::ReadUntilIdle;
 use example_common::*;
 
 use embassy::executor::Spawner;
-use embassy::traits::uart::Write;
 use embassy_nrf::gpio::NoPin;
 use embassy_nrf::{interrupt, uarte, Peripherals};
 
@@ -19,11 +17,9 @@ async fn main(_spawner: Spawner, p: Peripherals) {
     config.baudrate = uarte::Baudrate::BAUD115200;
 
     let irq = interrupt::take!(UARTE0_UART0);
-    let mut uart = unsafe {
-        uarte::UarteWithIdle::new(
-            p.UARTE0, p.TIMER0, p.PPI_CH0, p.PPI_CH1, irq, p.P0_08, p.P0_06, NoPin, NoPin, config,
-        )
-    };
+    let mut uart = uarte::UarteWithIdle::new(
+        p.UARTE0, p.TIMER0, p.PPI_CH0, p.PPI_CH1, irq, p.P0_08, p.P0_06, NoPin, NoPin, config,
+    );
 
     info!("uarte initialized!");
 
diff --git a/examples/nrf/src/bin/uart_split.rs b/examples/nrf/src/bin/uart_split.rs
index 4b5dbb21f..a9c02e796 100644
--- a/examples/nrf/src/bin/uart_split.rs
+++ b/examples/nrf/src/bin/uart_split.rs
@@ -4,16 +4,15 @@
 
 #[path = "../example_common.rs"]
 mod example_common;
+use example_common::*;
+
 use embassy::blocking_mutex::kind::Noop;
 use embassy::channel::mpsc::{self, Channel, Sender};
+use embassy::executor::Spawner;
 use embassy::util::Forever;
+use embassy_nrf::gpio::NoPin;
 use embassy_nrf::peripherals::UARTE0;
 use embassy_nrf::uarte::UarteRx;
-use example_common::*;
-
-use embassy::executor::Spawner;
-use embassy::traits::uart::{Read, Write};
-use embassy_nrf::gpio::NoPin;
 use embassy_nrf::{interrupt, uarte, Peripherals};
 
 static CHANNEL: Forever<Channel<Noop, [u8; 8], 1>> = Forever::new();
diff --git a/examples/nrf/src/bin/wdt.rs b/examples/nrf/src/bin/wdt.rs
index 78c2205d9..cc199be9d 100644
--- a/examples/nrf/src/bin/wdt.rs
+++ b/examples/nrf/src/bin/wdt.rs
@@ -10,7 +10,6 @@ use embassy::executor::Spawner;
 use embassy_nrf::gpio::{Input, Pull};
 use embassy_nrf::wdt::{Config, Watchdog};
 use embassy_nrf::Peripherals;
-use embassy_traits::gpio::{WaitForHigh, WaitForLow};
 
 #[embassy::main]
 async fn main(_spawner: Spawner, p: Peripherals) {