6 files changed, 403 insertions, 4 deletions

diff --git a/embassy-stm32/build.rs b/embassy-stm32/build.rs
index a00c6c416..dcee535b5 100644
--- a/embassy-stm32/build.rs
+++ b/embassy-stm32/build.rs
@@ -420,6 +420,10 @@ fn main() {
         (("spi", "SCK"), quote!(crate::spi::SckPin)),
         (("spi", "MOSI"), quote!(crate::spi::MosiPin)),
         (("spi", "MISO"), quote!(crate::spi::MisoPin)),
+        (("spi", "NSS"), quote!(crate::spi::CsPin)),
+        (("spi", "I2S_MCK"), quote!(crate::spi::MckPin)),
+        (("spi", "I2S_CK"), quote!(crate::spi::CkPin)),
+        (("spi", "I2S_WS"), quote!(crate::spi::WsPin)),
         (("i2c", "SDA"), quote!(crate::i2c::SdaPin)),
         (("i2c", "SCL"), quote!(crate::i2c::SclPin)),
         (("rcc", "MCO_1"), quote!(crate::rcc::McoPin)),
diff --git a/embassy-stm32/src/i2s.rs b/embassy-stm32/src/i2s.rs
new file mode 100644
index 000000000..2bb199f68
--- /dev/null
+++ b/embassy-stm32/src/i2s.rs
@@ -0,0 +1,310 @@
+use embassy_hal_common::into_ref;
+
+use crate::gpio::sealed::{AFType, Pin as _};
+use crate::gpio::AnyPin;
+use crate::pac::spi::vals;
+use crate::rcc::get_freqs;
+use crate::spi::{Config as SpiConfig, *};
+use crate::time::Hertz;
+use crate::{Peripheral, PeripheralRef};
+
+#[derive(Copy, Clone)]
+pub enum Mode {
+    Master,
+    Slave,
+}
+
+#[derive(Copy, Clone)]
+pub enum Function {
+    Transmit,
+    Receive,
+}
+
+#[derive(Copy, Clone)]
+pub enum Standard {
+    Philips,
+    MsbFirst,
+    LsbFirst,
+    PcmLongSync,
+    PcmShortSync,
+}
+
+impl Standard {
+    #[cfg(any(spi_v1, spi_f1))]
+    pub const fn i2sstd(&self) -> vals::I2sstd {
+        match self {
+            Standard::Philips => vals::I2sstd::PHILIPS,
+            Standard::MsbFirst => vals::I2sstd::MSB,
+            Standard::LsbFirst => vals::I2sstd::LSB,
+            Standard::PcmLongSync => vals::I2sstd::PCM,
+            Standard::PcmShortSync => vals::I2sstd::PCM,
+        }
+    }
+
+    #[cfg(any(spi_v1, spi_f1))]
+    pub const fn pcmsync(&self) -> vals::Pcmsync {
+        match self {
+            Standard::PcmLongSync => vals::Pcmsync::LONG,
+            _ => vals::Pcmsync::SHORT,
+        }
+    }
+}
+
+#[derive(Copy, Clone)]
+pub enum Format {
+    /// 16 bit data length on 16 bit wide channel
+    Data16Channel16,
+    /// 16 bit data length on 32 bit wide channel
+    Data16Channel32,
+    /// 24 bit data length on 32 bit wide channel
+    Data24Channel32,
+    /// 32 bit data length on 32 bit wide channel
+    Data32Channel32,
+}
+
+impl Format {
+    #[cfg(any(spi_v1, spi_f1))]
+    pub const fn datlen(&self) -> vals::Datlen {
+        match self {
+            Format::Data16Channel16 => vals::Datlen::SIXTEENBIT,
+            Format::Data16Channel32 => vals::Datlen::SIXTEENBIT,
+            Format::Data24Channel32 => vals::Datlen::TWENTYFOURBIT,
+            Format::Data32Channel32 => vals::Datlen::THIRTYTWOBIT,
+        }
+    }
+
+    #[cfg(any(spi_v1, spi_f1))]
+    pub const fn chlen(&self) -> vals::Chlen {
+        match self {
+            Format::Data16Channel16 => vals::Chlen::SIXTEENBIT,
+            Format::Data16Channel32 => vals::Chlen::THIRTYTWOBIT,
+            Format::Data24Channel32 => vals::Chlen::THIRTYTWOBIT,
+            Format::Data32Channel32 => vals::Chlen::THIRTYTWOBIT,
+        }
+    }
+}
+
+#[derive(Copy, Clone)]
+pub enum ClockPolarity {
+    IdleLow,
+    IdleHigh,
+}
+
+impl ClockPolarity {
+    #[cfg(any(spi_v1, spi_f1))]
+    pub const fn ckpol(&self) -> vals::Ckpol {
+        match self {
+            ClockPolarity::IdleHigh => vals::Ckpol::IDLEHIGH,
+            ClockPolarity::IdleLow => vals::Ckpol::IDLELOW,
+        }
+    }
+}
+
+/// [`I2S`] configuration.
+///
+///  - `MS`: `Master` or `Slave`
+///  - `TR`: `Transmit` or `Receive`
+///  - `STD`: I2S standard, eg `Philips`
+///  - `FMT`: Frame Format marker, eg `Data16Channel16`
+#[non_exhaustive]
+#[derive(Copy, Clone)]
+pub struct Config {
+    pub mode: Mode,
+    pub function: Function,
+    pub standard: Standard,
+    pub format: Format,
+    pub clock_polarity: ClockPolarity,
+    pub master_clock: bool,
+}
+
+impl Default for Config {
+    fn default() -> Self {
+        Self {
+            mode: Mode::Master,
+            function: Function::Transmit,
+            standard: Standard::Philips,
+            format: Format::Data16Channel16,
+            clock_polarity: ClockPolarity::IdleLow,
+            master_clock: true,
+        }
+    }
+}
+
+pub struct I2S<'d, T: Instance, Tx, Rx> {
+    _peri: Spi<'d, T, Tx, Rx>,
+    sd: Option<PeripheralRef<'d, AnyPin>>,
+    ws: Option<PeripheralRef<'d, AnyPin>>,
+    ck: Option<PeripheralRef<'d, AnyPin>>,
+    mck: Option<PeripheralRef<'d, AnyPin>>,
+}
+
+impl<'d, T: Instance, Tx, Rx> I2S<'d, T, Tx, Rx> {
+    /// Note: Full-Duplex modes are not supported at this time
+    pub fn new(
+        peri: impl Peripheral<P = T> + 'd,
+        sd: impl Peripheral<P = impl MosiPin<T>> + 'd,
+        ws: impl Peripheral<P = impl WsPin<T>> + 'd,
+        ck: impl Peripheral<P = impl CkPin<T>> + 'd,
+        mck: impl Peripheral<P = impl MckPin<T>> + 'd,
+        txdma: impl Peripheral<P = Tx> + 'd,
+        rxdma: impl Peripheral<P = Rx> + 'd,
+        freq: Hertz,
+        config: Config,
+    ) -> Self {
+        into_ref!(sd, ws, ck, mck);
+
+        unsafe {
+            sd.set_as_af(sd.af_num(), AFType::OutputPushPull);
+            sd.set_speed(crate::gpio::Speed::VeryHigh);
+
+            ws.set_as_af(ws.af_num(), AFType::OutputPushPull);
+            ws.set_speed(crate::gpio::Speed::VeryHigh);
+
+            ck.set_as_af(ck.af_num(), AFType::OutputPushPull);
+            ck.set_speed(crate::gpio::Speed::VeryHigh);
+
+            mck.set_as_af(mck.af_num(), AFType::OutputPushPull);
+            mck.set_speed(crate::gpio::Speed::VeryHigh);
+        }
+
+        let spi = Spi::new_internal(peri, txdma, rxdma, freq, SpiConfig::default());
+
+        #[cfg(all(rcc_f4, not(stm32f410)))]
+        let pclk = unsafe { get_freqs() }.plli2s.unwrap();
+
+        #[cfg(stm32f410)]
+        let pclk = T::frequency();
+
+        let (odd, div) = compute_baud_rate(pclk, freq, config.master_clock, config.format);
+
+        #[cfg(any(spi_v1, spi_f1))]
+        unsafe {
+            use stm32_metapac::spi::vals::{I2scfg, Odd};
+
+            // 1. Select the I2SDIV[7:0] bits in the SPI_I2SPR register to define the serial clock baud
+            // rate to reach the proper audio sample frequency. The ODD bit in the SPI_I2SPR
+            // register also has to be defined.
+
+            T::REGS.i2spr().modify(|w| {
+                w.set_i2sdiv(div);
+                w.set_odd(match odd {
+                    true => Odd::ODD,
+                    false => Odd::EVEN,
+                });
+
+                w.set_mckoe(config.master_clock);
+            });
+
+            // 2. Select the CKPOL bit to define the steady level for the communication clock. Set the
+            // MCKOE bit in the SPI_I2SPR register if the master clock MCK needs to be provided to
+            // the external DAC/ADC audio component (the I2SDIV and ODD values should be
+            // computed depending on the state of the MCK output, for more details refer to
+            // Section 28.4.4: Clock generator).
+
+            // 3. Set the I2SMOD bit in SPI_I2SCFGR to activate the I2S functionalities and choose the
+            // I2S standard through the I2SSTD[1:0] and PCMSYNC bits, the data length through the
+            // DATLEN[1:0] bits and the number of bits per channel by configuring the CHLEN bit.
+            // Select also the I2S master mode and direction (Transmitter or Receiver) through the
+            // I2SCFG[1:0] bits in the SPI_I2SCFGR register.
+
+            // 4. If needed, select all the potential interruption sources and the DMA capabilities by
+            // writing the SPI_CR2 register.
+
+            // 5. The I2SE bit in SPI_I2SCFGR register must be set.
+
+            T::REGS.i2scfgr().modify(|w| {
+                w.set_ckpol(config.clock_polarity.ckpol());
+
+                w.set_i2smod(true);
+                w.set_i2sstd(config.standard.i2sstd());
+                w.set_pcmsync(config.standard.pcmsync());
+
+                w.set_datlen(config.format.datlen());
+                w.set_chlen(config.format.chlen());
+
+                w.set_i2scfg(match (config.mode, config.function) {
+                    (Mode::Master, Function::Transmit) => I2scfg::MASTERTX,
+                    (Mode::Master, Function::Receive) => I2scfg::MASTERRX,
+                    (Mode::Slave, Function::Transmit) => I2scfg::SLAVETX,
+                    (Mode::Slave, Function::Receive) => I2scfg::SLAVERX,
+                });
+
+                w.set_i2se(true)
+            });
+        }
+        #[cfg(spi_v2)]
+        unsafe {}
+        #[cfg(any(spi_v3, spi_v4))]
+        unsafe {}
+
+        Self {
+            _peri: spi,
+            sd: Some(sd.map_into()),
+            ws: Some(ws.map_into()),
+            ck: Some(ck.map_into()),
+            mck: Some(mck.map_into()),
+        }
+    }
+
+    pub async fn write<W: Word>(&mut self, data: &[W]) -> Result<(), Error>
+    where
+        Tx: TxDma<T>,
+    {
+        self._peri.write(data).await
+    }
+
+    pub async fn read<W: Word>(&mut self, data: &mut [W]) -> Result<(), Error>
+    where
+        Tx: TxDma<T>,
+        Rx: RxDma<T>,
+    {
+        self._peri.read(data).await
+    }
+}
+
+impl<'d, T: Instance, Tx, Rx> Drop for I2S<'d, T, Tx, Rx> {
+    fn drop(&mut self) {
+        unsafe {
+            self.sd.as_ref().map(|x| x.set_as_disconnected());
+            self.ws.as_ref().map(|x| x.set_as_disconnected());
+            self.ck.as_ref().map(|x| x.set_as_disconnected());
+            self.mck.as_ref().map(|x| x.set_as_disconnected());
+        }
+    }
+}
+
+// Note, calculation details:
+// Fs = i2s_clock / [256 * ((2 * div) + odd)] when master clock is enabled
+// Fs = i2s_clock / [(channel_length * 2) * ((2 * div) + odd)]` when master clock is disabled
+// channel_length is 16 or 32
+//
+// can be rewritten as
+// Fs = i2s_clock / (coef * division)
+// where coef is a constant equal to 256, 64 or 32 depending channel length and master clock
+// and where division = (2 * div) + odd
+//
+// Equation can be rewritten as
+// division = i2s_clock/ (coef * Fs)
+//
+// note: division = (2 * div) + odd = (div << 1) + odd
+// in other word, from bits point of view, division[8:1] = div[7:0] and division[0] = odd
+fn compute_baud_rate(i2s_clock: Hertz, request_freq: Hertz, mclk: bool, data_format: Format) -> (bool, u8) {
+    let coef = if mclk {
+        256
+    } else if let Format::Data16Channel16 = data_format {
+        32
+    } else {
+        64
+    };
+
+    let (n, d) = (i2s_clock.0, coef * request_freq.0);
+    let division = (n + (d >> 1)) / d;
+
+    if division < 4 {
+        (false, 2)
+    } else if division > 511 {
+        (true, 255)
+    } else {
+        ((division & 1) == 1, (division >> 1) as u8)
+    }
+}
diff --git a/embassy-stm32/src/lib.rs b/embassy-stm32/src/lib.rs
index 6722658c9..c9df5c1b2 100644
--- a/embassy-stm32/src/lib.rs
+++ b/embassy-stm32/src/lib.rs
@@ -39,6 +39,8 @@ pub mod i2c;
 #[cfg(crc)]
 pub mod crc;
 pub mod flash;
+#[cfg(all(spi_v1, rcc_f4))]
+pub mod i2s;
 #[cfg(stm32wb)]
 pub mod ipcc;
 pub mod pwm;
diff --git a/embassy-stm32/src/spi/mod.rs b/embassy-stm32/src/spi/mod.rs
index 7390830dc..580971e45 100644
--- a/embassy-stm32/src/spi/mod.rs
+++ b/embassy-stm32/src/spi/mod.rs
@@ -212,6 +212,17 @@ impl<'d, T: Instance, Tx, Rx> Spi<'d, T, Tx, Rx> {
         Self::new_inner(peri, None, None, None, txdma, rxdma, freq, config)
     }
 
+    #[allow(dead_code)]
+    pub(crate) fn new_internal(
+        peri: impl Peripheral<P = T> + 'd,
+        txdma: impl Peripheral<P = Tx> + 'd,
+        rxdma: impl Peripheral<P = Rx> + 'd,
+        freq: Hertz,
+        config: Config,
+    ) -> Self {
+        Self::new_inner(peri, None, None, None, txdma, rxdma, freq, config)
+    }
+
     fn new_inner(
         peri: impl Peripheral<P = T> + 'd,
         sck: Option<PeripheralRef<'d, AnyPin>>,
@@ -1044,6 +1055,10 @@ pub trait Instance: Peripheral<P = Self> + sealed::Instance + RccPeripheral {}
 pin_trait!(SckPin, Instance);
 pin_trait!(MosiPin, Instance);
 pin_trait!(MisoPin, Instance);
+pin_trait!(CsPin, Instance);
+pin_trait!(MckPin, Instance);
+pin_trait!(CkPin, Instance);
+pin_trait!(WsPin, Instance);
 dma_trait!(RxDma, Instance);
 dma_trait!(TxDma, Instance);
 
diff --git a/embassy-stm32/src/wdg/mod.rs b/embassy-stm32/src/wdg/mod.rs
index 92b9a5ca8..18ebf97d8 100644
--- a/embassy-stm32/src/wdg/mod.rs
+++ b/embassy-stm32/src/wdg/mod.rs
@@ -13,13 +13,13 @@ pub struct IndependentWatchdog<'d, T: Instance> {
 const MAX_RL: u16 = 0xFFF;
 
 /// Calculates maximum watchdog timeout in us (RL = 0xFFF) for a given prescaler
-const fn max_timeout(prescaler: u16) -> u32 {
-    1_000_000 * MAX_RL as u32 / (LSI_FREQ.0 / prescaler as u32)
+const fn get_timeout_us(prescaler: u16, reload_value: u16) -> u32 {
+    1_000_000 * (reload_value + 1) as u32 / (LSI_FREQ.0 / prescaler as u32)
 }
 
 /// Calculates watchdog reload value for the given prescaler and desired timeout
 const fn reload_value(prescaler: u16, timeout_us: u32) -> u16 {
-    (timeout_us / prescaler as u32 * LSI_FREQ.0 / 1_000_000) as u16
+    (timeout_us / prescaler as u32 * LSI_FREQ.0 / 1_000_000) as u16 - 1
 }
 
 impl<'d, T: Instance> IndependentWatchdog<'d, T> {
@@ -34,7 +34,7 @@ impl<'d, T: Instance> IndependentWatchdog<'d, T> {
         // This iterates from 4 (2^2) to 256 (2^8).
         let psc_power = unwrap!((2..=8).find(|psc_power| {
             let psc = 2u16.pow(*psc_power);
-            timeout_us <= max_timeout(psc)
+            timeout_us <= get_timeout_us(psc, MAX_RL)
         }));
 
         // Prescaler value
@@ -54,6 +54,14 @@ impl<'d, T: Instance> IndependentWatchdog<'d, T> {
             wdg.rlr().write(|w| w.set_rl(rl));
         }
 
+        trace!(
+            "Watchdog configured with {}us timeout, desired was {}us (PR={}, RL={})",
+            get_timeout_us(psc, rl),
+            timeout_us,
+            pr,
+            rl
+        );
+
         IndependentWatchdog {
             wdg: PhantomData::default(),
         }
@@ -87,3 +95,27 @@ foreach_peripheral!(
         impl Instance for crate::peripherals::$inst {}
     };
 );
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn can_compute_timeout_us() {
+        assert_eq!(125, get_timeout_us(4, 0));
+        assert_eq!(512_000, get_timeout_us(4, MAX_RL));
+
+        assert_eq!(8_000, get_timeout_us(256, 0));
+        assert_eq!(32768_000, get_timeout_us(256, MAX_RL));
+
+        assert_eq!(8000_000, get_timeout_us(64, 3999));
+    }
+
+    #[test]
+    fn can_compute_reload_value() {
+        assert_eq!(0xFFF, reload_value(4, 512_000));
+        assert_eq!(0xFFF, reload_value(256, 32768_000));
+
+        assert_eq!(3999, reload_value(64, 8000_000));
+    }
+}
diff --git a/examples/stm32f4/src/bin/i2s_dma.rs b/examples/stm32f4/src/bin/i2s_dma.rs
new file mode 100644
index 000000000..e8d7b5f77
--- /dev/null
+++ b/examples/stm32f4/src/bin/i2s_dma.rs
@@ -0,0 +1,36 @@
+#![no_std]
+#![no_main]
+#![feature(type_alias_impl_trait)]
+
+use core::fmt::Write;
+
+use defmt::*;
+use embassy_executor::Spawner;
+use embassy_stm32::i2s::{Config, I2S};
+use embassy_stm32::time::Hertz;
+use heapless::String;
+use {defmt_rtt as _, panic_probe as _};
+
+#[embassy_executor::main]
+async fn main(_spawner: Spawner) {
+    let p = embassy_stm32::init(Default::default());
+    info!("Hello World!");
+
+    let mut i2s = I2S::new(
+        p.SPI2,
+        p.PC3,  // sd
+        p.PB12, // ws
+        p.PB10, // ck
+        p.PC6,  // mck
+        p.DMA1_CH4,
+        p.DMA1_CH3,
+        Hertz(1_000_000),
+        Config::default(),
+    );
+
+    for n in 0u32.. {
+        let mut write: String<128> = String::new();
+        core::write!(&mut write, "Hello DMA World {}!\r\n", n).unwrap();
+        i2s.write(&mut write.as_bytes()).await.ok();
+    }
+}