8 files changed, 132 insertions, 58 deletions

diff --git a/embassy-stm32/src/hrtim/mod.rs b/embassy-stm32/src/hrtim/mod.rs
index 6539326b4..1e6626a58 100644
--- a/embassy-stm32/src/hrtim/mod.rs
+++ b/embassy-stm32/src/hrtim/mod.rs
@@ -15,38 +15,42 @@ use crate::rcc::get_freqs;
 use crate::time::Hertz;
 use crate::Peripheral;
 
-pub enum Source {
-    Master,
-    ChA,
-    ChB,
-    ChC,
-    ChD,
-    ChE,
-    #[cfg(hrtim_v2)]
-    ChF,
-}
-
+/// HRTIM burst controller instance.
 pub struct BurstController<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM master instance.
 pub struct Master<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel A instance.
 pub struct ChA<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel B instance.
 pub struct ChB<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel C instance.
 pub struct ChC<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel D instance.
 pub struct ChD<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel E instance.
 pub struct ChE<T: Instance> {
     phantom: PhantomData<T>,
 }
+
+/// HRTIM channel F instance.
 #[cfg(hrtim_v2)]
 pub struct ChF<T: Instance> {
     phantom: PhantomData<T>,
@@ -60,13 +64,16 @@ mod sealed {
     }
 }
 
+/// Advanced channel instance trait.
 pub trait AdvancedChannel<T: Instance>: sealed::AdvancedChannel<T> {}
 
+/// HRTIM PWM pin.
 pub struct PwmPin<'d, Perip, Channel> {
     _pin: PeripheralRef<'d, AnyPin>,
     phantom: PhantomData<(Perip, Channel)>,
 }
 
+/// HRTIM complementary PWM pin.
 pub struct ComplementaryPwmPin<'d, Perip, Channel> {
     _pin: PeripheralRef<'d, AnyPin>,
     phantom: PhantomData<(Perip, Channel)>,
@@ -75,6 +82,7 @@ pub struct ComplementaryPwmPin<'d, Perip, Channel> {
 macro_rules! advanced_channel_impl {
     ($new_chx:ident, $channel:tt, $ch_num:expr, $pin_trait:ident, $complementary_pin_trait:ident) => {
         impl<'d, Perip: Instance> PwmPin<'d, Perip, $channel<Perip>> {
+            #[doc = concat!("Create a new ", stringify!($channel), " PWM pin instance.")]
             pub fn $new_chx(pin: impl Peripheral<P = impl $pin_trait<Perip>> + 'd) -> Self {
                 into_ref!(pin);
                 critical_section::with(|_| {
@@ -91,6 +99,7 @@ macro_rules! advanced_channel_impl {
         }
 
         impl<'d, Perip: Instance> ComplementaryPwmPin<'d, Perip, $channel<Perip>> {
+            #[doc = concat!("Create a new ", stringify!($channel), " complementary PWM pin instance.")]
             pub fn $new_chx(pin: impl Peripheral<P = impl $complementary_pin_trait<Perip>> + 'd) -> Self {
                 into_ref!(pin);
                 critical_section::with(|_| {
@@ -126,18 +135,29 @@ advanced_channel_impl!(new_chf, ChF, 5, ChannelFPin, ChannelFComplementaryPin);
 /// Struct used to divide a high resolution timer into multiple channels
 pub struct AdvancedPwm<'d, T: Instance> {
     _inner: PeripheralRef<'d, T>,
+    /// Master instance.
     pub master: Master<T>,
+    /// Burst controller.
     pub burst_controller: BurstController<T>,
+    /// Channel A.
     pub ch_a: ChA<T>,
+    /// Channel B.
     pub ch_b: ChB<T>,
+    /// Channel C.
     pub ch_c: ChC<T>,
+    /// Channel D.
     pub ch_d: ChD<T>,
+    /// Channel E.
     pub ch_e: ChE<T>,
+    /// Channel F.
     #[cfg(hrtim_v2)]
     pub ch_f: ChF<T>,
 }
 
 impl<'d, T: Instance> AdvancedPwm<'d, T> {
+    /// Create a new HRTIM driver.
+    ///
+    /// This splits the HRTIM into its constituent parts, which you can then use individually.
     pub fn new(
         tim: impl Peripheral<P = T> + 'd,
         _cha: Option<PwmPin<'d, T, ChA<T>>>,
@@ -200,13 +220,7 @@ impl<'d, T: Instance> AdvancedPwm<'d, T> {
     }
 }
 
-impl<T: Instance> BurstController<T> {
-    pub fn set_source(&mut self, _source: Source) {
-        todo!("burst mode control registers not implemented")
-    }
-}
-
-/// Represents a fixed-frequency bridge converter
+/// Fixed-frequency bridge converter driver.
 ///
 /// Our implementation of the bridge converter uses a single channel and three compare registers,
 /// allowing implementation of a synchronous buck or boost converter in continuous or discontinuous
@@ -225,6 +239,7 @@ pub struct BridgeConverter<T: Instance, C: AdvancedChannel<T>> {
 }
 
 impl<T: Instance, C: AdvancedChannel<T>> BridgeConverter<T, C> {
+    /// Create a new HRTIM bridge converter driver.
     pub fn new(_channel: C, frequency: Hertz) -> Self {
         use crate::pac::hrtim::vals::{Activeeffect, Inactiveeffect};
 
@@ -281,14 +296,17 @@ impl<T: Instance, C: AdvancedChannel<T>> BridgeConverter<T, C> {
         }
     }
 
+    /// Start HRTIM.
     pub fn start(&mut self) {
         T::regs().mcr().modify(|w| w.set_tcen(C::raw(), true));
     }
 
+    /// Stop HRTIM.
     pub fn stop(&mut self) {
         T::regs().mcr().modify(|w| w.set_tcen(C::raw(), false));
     }
 
+    /// Enable burst mode.
     pub fn enable_burst_mode(&mut self) {
         T::regs().tim(C::raw()).outr().modify(|w| {
             // Enable Burst Mode
@@ -301,6 +319,7 @@ impl<T: Instance, C: AdvancedChannel<T>> BridgeConverter<T, C> {
         })
     }
 
+    /// Disable burst mode.
     pub fn disable_burst_mode(&mut self) {
         T::regs().tim(C::raw()).outr().modify(|w| {
             // Disable Burst Mode
@@ -357,7 +376,7 @@ impl<T: Instance, C: AdvancedChannel<T>> BridgeConverter<T, C> {
     }
 }
 
-/// Represents a variable-frequency resonant converter
+/// Variable-frequency resonant converter driver.
 ///
 /// This implementation of a resonsant converter is appropriate for a half or full bridge,
 /// but does not include secondary rectification, which is appropriate for applications
@@ -370,6 +389,7 @@ pub struct ResonantConverter<T: Instance, C: AdvancedChannel<T>> {
 }
 
 impl<T: Instance, C: AdvancedChannel<T>> ResonantConverter<T, C> {
+    /// Create a new variable-frequency resonant converter driver.
     pub fn new(_channel: C, min_frequency: Hertz, max_frequency: Hertz) -> Self {
         T::set_channel_frequency(C::raw(), min_frequency);
 
@@ -408,6 +428,7 @@ impl<T: Instance, C: AdvancedChannel<T>> ResonantConverter<T, C> {
         T::set_channel_dead_time(C::raw(), value);
     }
 
+    /// Set the timer period.
     pub fn set_period(&mut self, period: u16) {
         assert!(period < self.max_period);
         assert!(period > self.min_period);
diff --git a/embassy-stm32/src/hrtim/traits.rs b/embassy-stm32/src/hrtim/traits.rs
index 34a363a1f..cfd31c47c 100644
--- a/embassy-stm32/src/hrtim/traits.rs
+++ b/embassy-stm32/src/hrtim/traits.rs
@@ -125,7 +125,6 @@ pub(crate) mod sealed {
         }
 
         /// Set the dead time as a proportion of max_duty
-
         fn set_channel_dead_time(channel: usize, dead_time: u16) {
             let regs = Self::regs();
 
@@ -148,13 +147,10 @@ pub(crate) mod sealed {
                 w.set_dtr(dt_val as u16);
             });
         }
-
-        //        fn enable_outputs(enable: bool);
-        //
-        //        fn enable_channel(&mut self, channel: usize, enable: bool);
     }
 }
 
+/// HRTIM instance trait.
 pub trait Instance: sealed::Instance + 'static {}
 
 foreach_interrupt! {
diff --git a/embassy-stm32/src/lib.rs b/embassy-stm32/src/lib.rs
index fd691a732..5d9b4e6a0 100644
--- a/embassy-stm32/src/lib.rs
+++ b/embassy-stm32/src/lib.rs
@@ -149,33 +149,15 @@ use crate::interrupt::Priority;
 pub use crate::pac::NVIC_PRIO_BITS;
 use crate::rcc::sealed::RccPeripheral;
 
-/// `embassy-stm32` global configuration.
 #[non_exhaustive]
 pub struct Config {
-    /// RCC config.
     pub rcc: rcc::Config,
-
-    /// Enable debug during sleep.
-    ///
-    /// May incrase power consumption. Defaults to true.
     #[cfg(dbgmcu)]
     pub enable_debug_during_sleep: bool,
-
-    /// BDMA interrupt priority.
-    ///
-    /// Defaults to P0 (highest).
     #[cfg(bdma)]
     pub bdma_interrupt_priority: Priority,
-
-    /// DMA interrupt priority.
-    ///
-    /// Defaults to P0 (highest).
     #[cfg(dma)]
     pub dma_interrupt_priority: Priority,
-
-    /// GPDMA interrupt priority.
-    ///
-    /// Defaults to P0 (highest).
     #[cfg(gpdma)]
     pub gpdma_interrupt_priority: Priority,
 }
@@ -196,11 +178,7 @@ impl Default for Config {
     }
 }
 
-/// Initialize the `embassy-stm32` HAL with the provided configuration.
-///
-/// This returns the peripheral singletons that can be used for creating drivers.
-///
-/// This should only be called once at startup, otherwise it panics.
+/// Initialize embassy.
 pub fn init(config: Config) -> Peripherals {
     critical_section::with(|cs| {
         let p = Peripherals::take_with_cs(cs);
diff --git a/embassy-stm32/src/qspi/enums.rs b/embassy-stm32/src/qspi/enums.rs
index e9e7fd482..ecade9b1a 100644
--- a/embassy-stm32/src/qspi/enums.rs
+++ b/embassy-stm32/src/qspi/enums.rs
@@ -1,3 +1,5 @@
+//! Enums used in QSPI configuration.
+
 #[allow(dead_code)]
 #[derive(Copy, Clone)]
 pub(crate) enum QspiMode {
diff --git a/embassy-stm32/src/qspi/mod.rs b/embassy-stm32/src/qspi/mod.rs
index 9ea0a726c..8a709a89e 100644
--- a/embassy-stm32/src/qspi/mod.rs
+++ b/embassy-stm32/src/qspi/mod.rs
@@ -14,6 +14,7 @@ use crate::pac::quadspi::Quadspi as Regs;
 use crate::rcc::RccPeripheral;
 use crate::{peripherals, Peripheral};
 
+/// QSPI transfer configuration.
 pub struct TransferConfig {
     /// Instraction width (IMODE)
     pub iwidth: QspiWidth,
@@ -45,6 +46,7 @@ impl Default for TransferConfig {
     }
 }
 
+/// QSPI driver configuration.
 pub struct Config {
     /// Flash memory size representend as 2^[0-32], as reasonable minimum 1KiB(9) was chosen.
     /// If you need other value the whose predefined use `Other` variant.
@@ -71,6 +73,7 @@ impl Default for Config {
     }
 }
 
+/// QSPI driver.
 #[allow(dead_code)]
 pub struct Qspi<'d, T: Instance, Dma> {
     _peri: PeripheralRef<'d, T>,
@@ -85,6 +88,7 @@ pub struct Qspi<'d, T: Instance, Dma> {
 }
 
 impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
+    /// Create a new QSPI driver for bank 1.
     pub fn new_bk1(
         peri: impl Peripheral<P = T> + 'd,
         d0: impl Peripheral<P = impl BK1D0Pin<T>> + 'd,
@@ -125,6 +129,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         )
     }
 
+    /// Create a new QSPI driver for bank 2.
     pub fn new_bk2(
         peri: impl Peripheral<P = T> + 'd,
         d0: impl Peripheral<P = impl BK2D0Pin<T>> + 'd,
@@ -223,6 +228,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         }
     }
 
+    /// Do a QSPI command.
     pub fn command(&mut self, transaction: TransferConfig) {
         #[cfg(not(stm32h7))]
         T::REGS.cr().modify(|v| v.set_dmaen(false));
@@ -232,6 +238,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         T::REGS.fcr().modify(|v| v.set_ctcf(true));
     }
 
+    /// Blocking read data.
     pub fn blocking_read(&mut self, buf: &mut [u8], transaction: TransferConfig) {
         #[cfg(not(stm32h7))]
         T::REGS.cr().modify(|v| v.set_dmaen(false));
@@ -256,6 +263,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         T::REGS.fcr().modify(|v| v.set_ctcf(true));
     }
 
+    /// Blocking write data.
     pub fn blocking_write(&mut self, buf: &[u8], transaction: TransferConfig) {
         // STM32H7 does not have dmaen
         #[cfg(not(stm32h7))]
@@ -278,6 +286,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         T::REGS.fcr().modify(|v| v.set_ctcf(true));
     }
 
+    /// Blocking read data, using DMA.
     pub fn blocking_read_dma(&mut self, buf: &mut [u8], transaction: TransferConfig)
     where
         Dma: QuadDma<T>,
@@ -310,6 +319,7 @@ impl<'d, T: Instance, Dma> Qspi<'d, T, Dma> {
         transfer.blocking_wait();
     }
 
+    /// Blocking write data, using DMA.
     pub fn blocking_write_dma(&mut self, buf: &[u8], transaction: TransferConfig)
     where
         Dma: QuadDma<T>,
@@ -379,6 +389,7 @@ pub(crate) mod sealed {
     }
 }
 
+/// QSPI instance trait.
 pub trait Instance: Peripheral<P = Self> + sealed::Instance + RccPeripheral {}
 
 pin_trait!(SckPin, Instance);
diff --git a/embassy-stm32/src/sdmmc/mod.rs b/embassy-stm32/src/sdmmc/mod.rs
index 6099b9f43..ab142053a 100644
--- a/embassy-stm32/src/sdmmc/mod.rs
+++ b/embassy-stm32/src/sdmmc/mod.rs
@@ -54,6 +54,7 @@ const SD_INIT_FREQ: Hertz = Hertz(400_000);
 
 /// The signalling scheme used on the SDMMC bus
 #[non_exhaustive]
+#[allow(missing_docs)]
 #[derive(Debug, Copy, Clone, PartialEq, Eq)]
 #[cfg_attr(feature = "defmt", derive(defmt::Format))]
 pub enum Signalling {
@@ -70,6 +71,9 @@ impl Default for Signalling {
     }
 }
 
+/// Aligned data block for SDMMC transfers.
+///
+/// This is a 512-byte array, aligned to 4 bytes to satisfy DMA requirements.
 #[repr(align(4))]
 #[derive(Debug, Clone, PartialEq, Eq)]
 #[cfg_attr(feature = "defmt", derive(defmt::Format))]
@@ -94,17 +98,23 @@ impl DerefMut for DataBlock {
 #[derive(Debug, Copy, Clone, PartialEq, Eq)]
 #[cfg_attr(feature = "defmt", derive(defmt::Format))]
 pub enum Error {
+    /// Timeout reported by the hardware
     Timeout,
+    /// Timeout reported by the software driver.
     SoftwareTimeout,
+    /// Unsupported card version.
     UnsupportedCardVersion,
+    /// Unsupported card type.
     UnsupportedCardType,
+    /// CRC error.
     Crc,
-    DataCrcFail,
-    RxOverFlow,
+    /// No card inserted.
     NoCard,
+    /// Bad clock supplied to the SDMMC peripheral.
     BadClock,
+    /// Signaling switch failed.
     SignalingSwitchFailed,
-    PeripheralBusy,
+    /// ST bit error.
     #[cfg(sdmmc_v1)]
     StBitErr,
 }
@@ -363,6 +373,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T>> Sdmmc<'d, T, Dma> {
 
 #[cfg(sdmmc_v2)]
 impl<'d, T: Instance> Sdmmc<'d, T, NoDma> {
+    /// Create a new SDMMC driver, with 1 data lane.
     pub fn new_1bit(
         sdmmc: impl Peripheral<P = T> + 'd,
         _irq: impl interrupt::typelevel::Binding<T::Interrupt, InterruptHandler<T>> + 'd,
@@ -396,6 +407,7 @@ impl<'d, T: Instance> Sdmmc<'d, T, NoDma> {
         )
     }
 
+    /// Create a new SDMMC driver, with 4 data lanes.
     pub fn new_4bit(
         sdmmc: impl Peripheral<P = T> + 'd,
         _irq: impl interrupt::typelevel::Binding<T::Interrupt, InterruptHandler<T>> + 'd,
@@ -497,7 +509,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
     }
 
     /// Data transfer is in progress
-    #[inline(always)]
+    #[inline]
     fn data_active() -> bool {
         let regs = T::regs();
 
@@ -509,7 +521,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
     }
 
     /// Coammand transfer is in progress
-    #[inline(always)]
+    #[inline]
     fn cmd_active() -> bool {
         let regs = T::regs();
 
@@ -521,7 +533,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
     }
 
     /// Wait idle on CMDACT, RXACT and TXACT (v1) or DOSNACT and CPSMACT (v2)
-    #[inline(always)]
+    #[inline]
     fn wait_idle() {
         while Self::data_active() || Self::cmd_active() {}
     }
@@ -837,7 +849,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
     }
 
     /// Clear flags in interrupt clear register
-    #[inline(always)]
+    #[inline]
     fn clear_interrupt_flags() {
         let regs = T::regs();
         regs.icr().write(|w| {
@@ -1152,7 +1164,8 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
         Ok(())
     }
 
-    #[inline(always)]
+    /// Read a data block.
+    #[inline]
     pub async fn read_block(&mut self, block_idx: u32, buffer: &mut DataBlock) -> Result<(), Error> {
         let card_capacity = self.card()?.card_type;
 
@@ -1204,6 +1217,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
         res
     }
 
+    /// Write a data block.
     pub async fn write_block(&mut self, block_idx: u32, buffer: &DataBlock) -> Result<(), Error> {
         let card = self.card.as_mut().ok_or(Error::NoCard)?;
 
@@ -1283,7 +1297,7 @@ impl<'d, T: Instance, Dma: SdmmcDma<T> + 'd> Sdmmc<'d, T, Dma> {
     ///
     /// Returns Error::NoCard if [`init_card`](#method.init_card)
     /// has not previously succeeded
-    #[inline(always)]
+    #[inline]
     pub fn card(&self) -> Result<&Card, Error> {
         self.card.as_ref().ok_or(Error::NoCard)
     }
@@ -1419,7 +1433,9 @@ pub(crate) mod sealed {
     pub trait Pins<T: Instance> {}
 }
 
+/// SDMMC instance trait.
 pub trait Instance: sealed::Instance + RccPeripheral + 'static {}
+
 pin_trait!(CkPin, Instance);
 pin_trait!(CmdPin, Instance);
 pin_trait!(D0Pin, Instance);
@@ -1434,7 +1450,10 @@ pin_trait!(D7Pin, Instance);
 #[cfg(sdmmc_v1)]
 dma_trait!(SdmmcDma, Instance);
 
-// SDMMCv2 uses internal DMA
+/// DMA instance trait.
+///
+/// This is only implemented for `NoDma`, since SDMMCv2 has DMA built-in, instead of
+/// using ST's system-wide DMA peripheral.
 #[cfg(sdmmc_v2)]
 pub trait SdmmcDma<T: Instance> {}
 #[cfg(sdmmc_v2)]
diff --git a/embassy-stm32/src/spi/mod.rs b/embassy-stm32/src/spi/mod.rs
index 5a1ad3e91..674a5d316 100644
--- a/embassy-stm32/src/spi/mod.rs
+++ b/embassy-stm32/src/spi/mod.rs
@@ -16,27 +16,38 @@ use crate::rcc::RccPeripheral;
 use crate::time::Hertz;
 use crate::{peripherals, Peripheral};
 
+/// SPI error.
 #[derive(Debug, PartialEq, Eq)]
 #[cfg_attr(feature = "defmt", derive(defmt::Format))]
 pub enum Error {
+    /// Invalid framing.
     Framing,
+    /// CRC error (only if hardware CRC checking is enabled).
     Crc,
+    /// Mode fault
     ModeFault,
+    /// Overrun.
     Overrun,
 }
 
-// TODO move upwards in the tree
+/// SPI bit order
 #[derive(Copy, Clone)]
 pub enum BitOrder {
+    /// Least significant bit first.
     LsbFirst,
+    /// Most significant bit first.
     MsbFirst,
 }
 
+/// SPI configuration.
 #[non_exhaustive]
 #[derive(Copy, Clone)]
 pub struct Config {
+    /// SPI mode.
     pub mode: Mode,
+    /// Bit order.
     pub bit_order: BitOrder,
+    /// Clock frequency.
     pub frequency: Hertz,
 }
 
@@ -73,6 +84,7 @@ impl Config {
     }
 }
 
+/// SPI driver.
 pub struct Spi<'d, T: Instance, Tx, Rx> {
     _peri: PeripheralRef<'d, T>,
     sck: Option<PeripheralRef<'d, AnyPin>>,
@@ -84,6 +96,7 @@ pub struct Spi<'d, T: Instance, Tx, Rx> {
 }
 
 impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
+    /// Create a new SPI driver.
     pub fn new(
         peri: impl Peripheral<P = T> + 'd,
         sck: impl Peripheral<P = impl SckPin<T>> + 'd,
@@ -118,6 +131,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         )
     }
 
+    /// Create a new SPI driver, in RX-only mode (only MISO pin, no MOSI).
     pub fn new_rxonly(
         peri: impl Peripheral<P = T> + 'd,
         sck: impl Peripheral<P = impl SckPin<T>> + 'd,
@@ -143,6 +157,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         )
     }
 
+    /// Create a new SPI driver, in TX-only mode (only MOSI pin, no MISO).
     pub fn new_txonly(
         peri: impl Peripheral<P = T> + 'd,
         sck: impl Peripheral<P = impl SckPin<T>> + 'd,
@@ -168,6 +183,9 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         )
     }
 
+    /// Create a new SPI driver, in TX-only mode, without SCK pin.
+    ///
+    /// This can be useful for bit-banging non-SPI protocols.
     pub fn new_txonly_nosck(
         peri: impl Peripheral<P = T> + 'd,
         mosi: impl Peripheral<P = impl MosiPin<T>> + 'd,
@@ -355,6 +373,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// Get current SPI configuration.
     pub fn get_current_config(&self) -> Config {
         #[cfg(any(spi_v1, spi_f1, spi_v2))]
         let cfg = T::REGS.cr1().read();
@@ -444,6 +463,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         self.current_word_size = word_size;
     }
 
+    /// SPI write, using DMA.
     pub async fn write<W: Word>(&mut self, data: &[W]) -> Result<(), Error>
     where
         Tx: TxDma<T>,
@@ -477,6 +497,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// SPI read, using DMA.
     pub async fn read<W: Word>(&mut self, data: &mut [W]) -> Result<(), Error>
     where
         Tx: TxDma<T>,
@@ -580,6 +601,12 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// Bidirectional transfer, using DMA.
+    ///
+    /// This transfers both buffers at the same time, so it is NOT equivalent to `write` followed by `read`.
+    ///
+    /// The transfer runs for `max(read.len(), write.len())` bytes. If `read` is shorter extra bytes are ignored.
+    /// If `write` is shorter it is padded with zero bytes.
     pub async fn transfer<W: Word>(&mut self, read: &mut [W], write: &[W]) -> Result<(), Error>
     where
         Tx: TxDma<T>,
@@ -588,6 +615,9 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         self.transfer_inner(read, write).await
     }
 
+    /// In-place bidirectional transfer, using DMA.
+    ///
+    /// This writes the contents of `data` on MOSI, and puts the received data on MISO in `data`, at the same time.
     pub async fn transfer_in_place<W: Word>(&mut self, data: &mut [W]) -> Result<(), Error>
     where
         Tx: TxDma<T>,
@@ -596,6 +626,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         self.transfer_inner(data, data).await
     }
 
+    /// Blocking write.
     pub fn blocking_write<W: Word>(&mut self, words: &[W]) -> Result<(), Error> {
         T::REGS.cr1().modify(|w| w.set_spe(true));
         flush_rx_fifo(T::REGS);
@@ -606,6 +637,7 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// Blocking read.
     pub fn blocking_read<W: Word>(&mut self, words: &mut [W]) -> Result<(), Error> {
         T::REGS.cr1().modify(|w| w.set_spe(true));
         flush_rx_fifo(T::REGS);
@@ -616,6 +648,9 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// Blocking in-place bidirectional transfer.
+    ///
+    /// This writes the contents of `data` on MOSI, and puts the received data on MISO in `data`, at the same time.
     pub fn blocking_transfer_in_place<W: Word>(&mut self, words: &mut [W]) -> Result<(), Error> {
         T::REGS.cr1().modify(|w| w.set_spe(true));
         flush_rx_fifo(T::REGS);
@@ -626,6 +661,12 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Ok(())
     }
 
+    /// Blocking bidirectional transfer.
+    ///
+    /// This transfers both buffers at the same time, so it is NOT equivalent to `write` followed by `read`.
+    ///
+    /// The transfer runs for `max(read.len(), write.len())` bytes. If `read` is shorter extra bytes are ignored.
+    /// If `write` is shorter it is padded with zero bytes.
     pub fn blocking_transfer<W: Word>(&mut self, read: &mut [W], write: &[W]) -> Result<(), Error> {
         T::REGS.cr1().modify(|w| w.set_spe(true));
         flush_rx_fifo(T::REGS);
@@ -946,6 +987,7 @@ pub(crate) mod sealed {
     }
 }
 
+/// Word sizes usable for SPI.
 pub trait Word: word::Word + sealed::Word {}
 
 macro_rules! impl_word {
@@ -1025,7 +1067,9 @@ mod word_impl {
     impl_word!(u32, 32 - 1);
 }
 
+/// SPI instance trait.
 pub trait Instance: Peripheral<P = Self> + sealed::Instance + RccPeripheral {}
+
 pin_trait!(SckPin, Instance);
 pin_trait!(MosiPin, Instance);
 pin_trait!(MisoPin, Instance);
diff --git a/embassy-stm32/src/time.rs b/embassy-stm32/src/time.rs
index a0bc33944..17690aefc 100644
--- a/embassy-stm32/src/time.rs
+++ b/embassy-stm32/src/time.rs
@@ -8,14 +8,17 @@ use core::ops::{Div, Mul};
 pub struct Hertz(pub u32);
 
 impl Hertz {
+    /// Create a `Hertz` from the given hertz.
     pub const fn hz(hertz: u32) -> Self {
         Self(hertz)
     }
 
+    /// Create a `Hertz` from the given kilohertz.
     pub const fn khz(kilohertz: u32) -> Self {
         Self(kilohertz * 1_000)
     }
 
+    /// Create a `Hertz` from the given megahertz.
     pub const fn mhz(megahertz: u32) -> Self {
         Self(megahertz * 1_000_000)
     }