diff --git a/embassy-rp/src/dma.rs b/embassy-rp/src/dma.rs
index acf33822..b256cc2f 100644
--- a/embassy-rp/src/dma.rs
+++ b/embassy-rp/src/dma.rs
@@ -56,6 +56,25 @@ pub unsafe fn read<'a, C: Channel, W: Word>(
     )
 }
 
+pub unsafe fn read_repeated<'a, C: Channel, W: Word>(
+    ch: impl Peripheral<P = C> + 'a,
+    from: *const W,
+    len: usize,
+    dreq: u8,
+) -> Transfer<'a, C> {
+    let mut dummy: u32 = 0;
+    copy_inner(
+        ch,
+        from as *const u32,
+        &mut dummy as *mut u32,
+        len,
+        W::size(),
+        false,
+        false,
+        dreq,
+    )
+}
+
 pub unsafe fn write<'a, C: Channel, W: Word>(
     ch: impl Peripheral<P = C> + 'a,
     from: *const [W],
@@ -75,6 +94,25 @@ pub unsafe fn write<'a, C: Channel, W: Word>(
     )
 }
 
+pub unsafe fn write_repeated<'a, C: Channel, W: Word>(
+    ch: impl Peripheral<P = C> + 'a,
+    to: *mut W,
+    len: usize,
+    dreq: u8,
+) -> Transfer<'a, C> {
+    let dummy: u32 = 0;
+    copy_inner(
+        ch,
+        &dummy as *const u32,
+        to as *mut u32,
+        len,
+        W::size(),
+        false,
+        false,
+        dreq,
+    )
+}
+
 pub unsafe fn copy<'a, C: Channel, W: Word>(
     ch: impl Peripheral<P = C> + 'a,
     from: &[W],
diff --git a/embassy-rp/src/spi.rs b/embassy-rp/src/spi.rs
index 74f0b04d..e7cd9992 100644
--- a/embassy-rp/src/spi.rs
+++ b/embassy-rp/src/spi.rs
@@ -325,30 +325,53 @@ impl<'d, T: Instance> Spi<'d, T, Async> {
     }
 
     pub async fn write(&mut self, buffer: &[u8]) -> Result<(), Error> {
-        let ch = self.tx_dma.as_mut().unwrap();
-        let transfer = unsafe {
-            self.inner.regs().dmacr().modify(|reg| {
+        unsafe {
+            self.inner.regs().dmacr().write(|reg| {
+                reg.set_rxdmae(true);
                 reg.set_txdmae(true);
-            });
+            })
+        };
+        let tx_ch = self.tx_dma.as_mut().unwrap();
+        let tx_transfer = unsafe {
             // If we don't assign future to a variable, the data register pointer
             // is held across an await and makes the future non-Send.
-            crate::dma::write(ch, buffer, self.inner.regs().dr().ptr() as *mut _, T::TX_DREQ)
+            crate::dma::write(tx_ch, buffer, self.inner.regs().dr().ptr() as *mut _, T::TX_DREQ)
         };
-        transfer.await;
+        let rx_ch = self.rx_dma.as_mut().unwrap();
+        let rx_transfer = unsafe {
+            // If we don't assign future to a variable, the data register pointer
+            // is held across an await and makes the future non-Send.
+            crate::dma::read_repeated(
+                rx_ch,
+                self.inner.regs().dr().ptr() as *const u8,
+                buffer.len(),
+                T::RX_DREQ,
+            )
+        };
+        join(tx_transfer, rx_transfer).await;
         Ok(())
     }
 
     pub async fn read(&mut self, buffer: &mut [u8]) -> Result<(), Error> {
-        let ch = self.rx_dma.as_mut().unwrap();
-        let transfer = unsafe {
-            self.inner.regs().dmacr().modify(|reg| {
+        unsafe {
+            self.inner.regs().dmacr().write(|reg| {
                 reg.set_rxdmae(true);
-            });
+                reg.set_txdmae(true);
+            })
+        };
+        let tx_ch = self.tx_dma.as_mut().unwrap();
+        let tx_transfer = unsafe {
             // If we don't assign future to a variable, the data register pointer
             // is held across an await and makes the future non-Send.
-            crate::dma::read(ch, self.inner.regs().dr().ptr() as *const _, buffer, T::RX_DREQ)
+            crate::dma::write_repeated(tx_ch, self.inner.regs().dr().ptr() as *mut u8, buffer.len(), T::TX_DREQ)
         };
-        transfer.await;
+        let rx_ch = self.rx_dma.as_mut().unwrap();
+        let rx_transfer = unsafe {
+            // If we don't assign future to a variable, the data register pointer
+            // is held across an await and makes the future non-Send.
+            crate::dma::read(rx_ch, self.inner.regs().dr().ptr() as *const _, buffer, T::RX_DREQ)
+        };
+        join(tx_transfer, rx_transfer).await;
         Ok(())
     }
 
@@ -364,20 +387,20 @@ impl<'d, T: Instance> Spi<'d, T, Async> {
         let (_, from_len) = crate::dma::slice_ptr_parts(tx_ptr);
         let (_, to_len) = crate::dma::slice_ptr_parts_mut(rx_ptr);
         assert_eq!(from_len, to_len);
+        unsafe {
+            self.inner.regs().dmacr().write(|reg| {
+                reg.set_rxdmae(true);
+                reg.set_txdmae(true);
+            })
+        };
         let tx_ch = self.tx_dma.as_mut().unwrap();
         let tx_transfer = unsafe {
-            self.inner.regs().dmacr().modify(|reg| {
-                reg.set_txdmae(true);
-            });
             // If we don't assign future to a variable, the data register pointer
             // is held across an await and makes the future non-Send.
             crate::dma::write(tx_ch, tx_ptr, self.inner.regs().dr().ptr() as *mut _, T::TX_DREQ)
         };
         let rx_ch = self.rx_dma.as_mut().unwrap();
         let rx_transfer = unsafe {
-            self.inner.regs().dmacr().modify(|reg| {
-                reg.set_rxdmae(true);
-            });
             // If we don't assign future to a variable, the data register pointer
             // is held across an await and makes the future non-Send.
             crate::dma::read(rx_ch, self.inner.regs().dr().ptr() as *const _, rx_ptr, T::RX_DREQ)