#![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::{AtomicWaker, Unborrow};
use embassy_extras::unborrow;
use futures::future::poll_fn;
use traits::spi::FullDuplex;

use crate::gpio;
use crate::gpio::sealed::Pin as _;
use crate::gpio::{OptionalPin, Pin as GpioPin};
use crate::interrupt::Interrupt;
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 pac::spim0::frequency::FREQUENCY_A as Frequency;

#[derive(Debug, Clone, Copy, PartialEq, Eq)]
#[cfg_attr(feature = "defmt", derive(defmt::Format))]
#[non_exhaustive]
pub enum Error {
    TxBufferTooLong,
    RxBufferTooLong,
    /// EasyDMA can only read from data memory, read only buffers in flash will fail.
    DMABufferNotInDataMemory,
}

pub struct Spim<'d, T: Instance> {
    phantom: PhantomData<&'d mut T>,
}

#[non_exhaustive]
pub struct Config {
    pub frequency: Frequency,
    pub mode: Mode,
    pub orc: u8,
}

impl Default for Config {
    fn default() -> Self {
        Self {
            frequency: Frequency::M1,
            mode: MODE_0,
            orc: 0x00,
        }
    }
}

impl<'d, T: Instance> Spim<'d, T> {
    pub fn new(
        _spim: impl Unborrow<Target = T> + 'd,
        irq: impl Unborrow<Target = T::Interrupt> + 'd,
        sck: impl Unborrow<Target = impl GpioPin> + 'd,
        miso: impl Unborrow<Target = impl OptionalPin> + 'd,
        mosi: impl Unborrow<Target = impl OptionalPin> + 'd,
        config: Config,
    ) -> Self {
        unborrow!(irq, sck, miso, mosi);

        let r = T::regs();

        // Configure pins
        sck.conf().write(|w| w.dir().output().drive().h0h1());
        if let Some(mosi) = mosi.pin_mut() {
            mosi.conf().write(|w| w.dir().output().drive().h0h1());
        }
        if let Some(miso) = miso.pin_mut() {
            miso.conf().write(|w| w.input().connect().drive().h0h1());
        }

        match config.mode.polarity {
            Polarity::IdleHigh => {
                sck.set_high();
                if let Some(mosi) = mosi.pin_mut() {
                    mosi.set_high();
                }
            }
            Polarity::IdleLow => {
                sck.set_low();
                if let Some(mosi) = mosi.pin_mut() {
                    mosi.set_low();
                }
            }
        }

        // Select pins.
        // Note: OptionalPin reports 'disabled' for psel_bits when no pin was selected.
        r.psel.sck.write(|w| unsafe { w.bits(sck.psel_bits()) });
        r.psel.mosi.write(|w| unsafe { w.bits(mosi.psel_bits()) });
        r.psel.miso.write(|w| unsafe { w.bits(miso.psel_bits()) });

        // Enable SPIM instance.
        r.enable.write(|w| w.enable().enabled());

        // Configure mode.
        let mode = config.mode;
        r.config.write(|w| {
            match mode {
                MODE_0 => {
                    w.order().msb_first();
                    w.cpol().active_high();
                    w.cpha().leading();
                }
                MODE_1 => {
                    w.order().msb_first();
                    w.cpol().active_high();
                    w.cpha().trailing();
                }
                MODE_2 => {
                    w.order().msb_first();
                    w.cpol().active_low();
                    w.cpha().leading();
                }
                MODE_3 => {
                    w.order().msb_first();
                    w.cpol().active_low();
                    w.cpha().trailing();
                }
            }

            w
        });

        // Configure frequency.
        let frequency = config.frequency;
        r.frequency.write(|w| w.frequency().variant(frequency));

        // 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) });

        // Disable all events interrupts
        r.intenclr.write(|w| unsafe { w.bits(0xFFFF_FFFF) });

        irq.set_handler(Self::on_interrupt);
        irq.unpend();
        irq.enable();

        Self {
            phantom: PhantomData,
        }
    }

    fn on_interrupt(_: *mut ()) {
        let r = T::regs();
        let s = T::state();

        if r.events_end.read().bits() != 0 {
            s.end_waker.wake();
            r.intenclr.write(|w| w.end().clear());
        }
    }
}

impl<'d, T: Instance> Drop for Spim<'d, T> {
    fn drop(&mut self) {
        info!("spim drop");

        // TODO check for abort, wait for xxxstopped

        // 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());

        info!("spim drop: done");
    }
}

impl<'d, T: Instance> FullDuplex<u8> for Spim<'d, T> {
    type Error = Error;

    #[rustfmt::skip]
    type WriteFuture<'a> where Self: 'a = impl Future<Output = Result<(), Self::Error>> + 'a;
    #[rustfmt::skip]
    type ReadFuture<'a> where Self: 'a = impl Future<Output = Result<(), Self::Error>> + 'a;
    #[rustfmt::skip]
    type WriteReadFuture<'a> where Self: 'a = impl Future<Output = Result<(), Self::Error>> + 'a;

    fn read<'a>(&'a mut self, data: &'a mut [u8]) -> Self::ReadFuture<'a> {
        self.read_write(data, &[])
    }
    fn write<'a>(&'a mut self, data: &'a [u8]) -> Self::WriteFuture<'a> {
        self.read_write(&mut [], data)
    }

    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(());
                }

                Poll::Pending
            })
            .await;

            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);

        let r = T::regs();

        // 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 _) });

        // 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();

        // Start SPI transaction.
        r.tasks_start.write(|w| unsafe { w.bits(1) });

        // Wait for 'end' event.
        while r.events_end.read().bits() == 0 {}

        // 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);

        Ok(words)
    }
}

impl<'d, T: Instance> embedded_hal::blocking::spi::Write<u8> for Spim<'d, T> {
    type Error = Error;

    fn write(&mut self, words: &[u8]) -> Result<(), Self::Error> {
        slice_in_ram_or(words, Error::DMABufferNotInDataMemory)?;
        let recv: &mut [u8] = &mut [];

        // 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();

        // 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 _) });

        // 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();

        // Start SPI transaction.
        r.tasks_start.write(|w| unsafe { w.bits(1) });

        // Wait for 'end' event.
        while r.events_end.read().bits() == 0 {}

        // 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);

        Ok(())
    }
}

pub(crate) mod sealed {
    use super::*;

    pub struct State {
        pub end_waker: AtomicWaker,
    }

    impl State {
        pub const fn new() -> Self {
            Self {
                end_waker: AtomicWaker::new(),
            }
        }
    }

    pub trait Instance {
        fn regs() -> &'static pac::spim0::RegisterBlock;
        fn state() -> &'static State;
    }
}

pub trait Instance: Unborrow<Target = Self> + sealed::Instance + 'static {
    type Interrupt: Interrupt;
}

macro_rules! impl_spim {
    ($type:ident, $pac_type:ident, $irq:ident) => {
        impl crate::spim::sealed::Instance for peripherals::$type {
            fn regs() -> &'static pac::spim0::RegisterBlock {
                unsafe { &*pac::$pac_type::ptr() }
            }
            fn state() -> &'static crate::spim::sealed::State {
                static STATE: crate::spim::sealed::State = crate::spim::sealed::State::new();
                &STATE
            }
        }
        impl crate::spim::Instance for peripherals::$type {
            type Interrupt = crate::interrupt::$irq;
        }
    };
}