embassy/embassy-nrf/src/qspi.rs

//! Quad Serial Peripheral Interface (QSPI) flash driver.

#![macro_use]

use core::future::poll_fn;
use core::marker::PhantomData;
use core::ptr;
use core::task::Poll;

use embassy_hal_common::drop::OnDrop;
use embassy_hal_common::{into_ref, PeripheralRef};
use embedded_storage::nor_flash::{ErrorType, NorFlash, NorFlashError, NorFlashErrorKind, ReadNorFlash};

use crate::gpio::{self, Pin as GpioPin};
use crate::interrupt::{self, Interrupt, InterruptExt};
pub use crate::pac::qspi::ifconfig0::{
    ADDRMODE_A as AddressMode, PPSIZE_A as WritePageSize, READOC_A as ReadOpcode, WRITEOC_A as WriteOpcode,
};
pub use crate::pac::qspi::ifconfig1::SPIMODE_A as SpiMode;
use crate::Peripheral;

/// Deep power-down config.
pub struct DeepPowerDownConfig {
    /// Time required for entering DPM, in units of 16us
    pub enter_time: u16,
    /// Time required for exiting DPM, in units of 16us
    pub exit_time: u16,
}

/// QSPI bus frequency.
pub enum Frequency {
    /// 32 Mhz
    M32 = 0,
    /// 16 Mhz
    M16 = 1,
    /// 10.7 Mhz
    M10_7 = 2,
    /// 8 Mhz
    M8 = 3,
    /// 6.4 Mhz
    M6_4 = 4,
    /// 5.3 Mhz
    M5_3 = 5,
    /// 4.6 Mhz
    M4_6 = 6,
    /// 4 Mhz
    M4 = 7,
    /// 3.6 Mhz
    M3_6 = 8,
    /// 3.2 Mhz
    M3_2 = 9,
    /// 2.9 Mhz
    M2_9 = 10,
    /// 2.7 Mhz
    M2_7 = 11,
    /// 2.5 Mhz
    M2_5 = 12,
    /// 2.3 Mhz
    M2_3 = 13,
    /// 2.1 Mhz
    M2_1 = 14,
    /// 2 Mhz
    M2 = 15,
}

/// QSPI config.
#[non_exhaustive]
pub struct Config {
    /// XIP offset.
    pub xip_offset: u32,
    /// Opcode used for read operations.
    pub read_opcode: ReadOpcode,
    /// Opcode used for write operations.
    pub write_opcode: WriteOpcode,
    /// Page size for write operations.
    pub write_page_size: WritePageSize,
    /// Configuration for deep power down. If None, deep power down is disabled.
    pub deep_power_down: Option<DeepPowerDownConfig>,
    /// QSPI bus frequency.
    pub frequency: Frequency,
    /// Value is specified in number of 16 MHz periods (62.5 ns)
    pub sck_delay: u8,
    /// Whether data is captured on the clock rising edge and data is output on a falling edge (MODE0) or vice-versa (MODE3)
    pub spi_mode: SpiMode,
    /// Addressing mode (24-bit or 32-bit)
    pub address_mode: AddressMode,
    /// Flash memory capacity in bytes. This is the value reported by the `embedded-storage` traits.
    pub capacity: u32,
}

impl Default for Config {
    fn default() -> Self {
        Self {
            read_opcode: ReadOpcode::READ4IO,
            write_opcode: WriteOpcode::PP4IO,
            xip_offset: 0,
            write_page_size: WritePageSize::_256BYTES,
            deep_power_down: None,
            frequency: Frequency::M8,
            sck_delay: 80,
            spi_mode: SpiMode::MODE0,
            address_mode: AddressMode::_24BIT,
            capacity: 0,
        }
    }
}

/// Error
#[derive(Debug, Copy, Clone, Eq, PartialEq)]
#[cfg_attr(feature = "defmt", derive(defmt::Format))]
#[non_exhaustive]
pub enum Error {
    /// Operation address was out of bounds.
    OutOfBounds,
    // TODO add "not in data memory" error and check for it
}

/// Interrupt handler.
pub struct InterruptHandler<T: Instance> {
    _phantom: PhantomData<T>,
}

impl<T: Instance> interrupt::Handler<T::Interrupt> for InterruptHandler<T> {
    unsafe fn on_interrupt() {
        let r = T::regs();
        let s = T::state();

        if r.events_ready.read().bits() != 0 {
            s.waker.wake();
            r.intenclr.write(|w| w.ready().clear());
        }
    }
}

/// QSPI flash driver.
pub struct Qspi<'d, T: Instance> {
    _peri: PeripheralRef<'d, T>,
    dpm_enabled: bool,
    capacity: u32,
}

impl<'d, T: Instance> Qspi<'d, T> {
    /// Create a new QSPI driver.
    pub fn new(
        qspi: impl Peripheral<P = T> + 'd,
        _irq: impl interrupt::Binding<T::Interrupt, InterruptHandler<T>> + 'd,
        sck: impl Peripheral<P = impl GpioPin> + 'd,
        csn: impl Peripheral<P = impl GpioPin> + 'd,
        io0: impl Peripheral<P = impl GpioPin> + 'd,
        io1: impl Peripheral<P = impl GpioPin> + 'd,
        io2: impl Peripheral<P = impl GpioPin> + 'd,
        io3: impl Peripheral<P = impl GpioPin> + 'd,
        config: Config,
    ) -> Self {
        into_ref!(qspi, sck, csn, io0, io1, io2, io3);

        let r = T::regs();

        macro_rules! config_pin {
            ($pin:ident) => {
                $pin.set_high();
                $pin.conf().write(|w| {
                    w.dir().output();
                    w.drive().h0h1();
                    #[cfg(feature = "_nrf5340-s")]
                    w.mcusel().peripheral();
                    w
                });
                r.psel.$pin.write(|w| unsafe { w.bits($pin.psel_bits()) });
            };
        }

        config_pin!(sck);
        config_pin!(csn);
        config_pin!(io0);
        config_pin!(io1);
        config_pin!(io2);
        config_pin!(io3);

        r.ifconfig0.write(|w| {
            w.addrmode().variant(config.address_mode);
            w.dpmenable().bit(config.deep_power_down.is_some());
            w.ppsize().variant(config.write_page_size);
            w.readoc().variant(config.read_opcode);
            w.writeoc().variant(config.write_opcode);
            w
        });

        if let Some(dpd) = &config.deep_power_down {
            r.dpmdur.write(|w| unsafe {
                w.enter().bits(dpd.enter_time);
                w.exit().bits(dpd.exit_time);
                w
            })
        }

        r.ifconfig1.write(|w| unsafe {
            w.sckdelay().bits(config.sck_delay);
            w.dpmen().exit();
            w.spimode().variant(config.spi_mode);
            w.sckfreq().bits(config.frequency as u8);
            w
        });

        r.xipoffset.write(|w| unsafe {
            w.xipoffset().bits(config.xip_offset);
            w
        });

        unsafe { T::Interrupt::steal() }.unpend();
        unsafe { T::Interrupt::steal() }.enable();

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

        let res = Self {
            _peri: qspi,
            dpm_enabled: config.deep_power_down.is_some(),
            capacity: config.capacity,
        };

        r.events_ready.reset();
        r.intenset.write(|w| w.ready().set());

        r.tasks_activate.write(|w| w.tasks_activate().bit(true));

        Self::blocking_wait_ready();

        res
    }

    /// Do a custom QSPI instruction.
    pub async fn custom_instruction(&mut self, opcode: u8, req: &[u8], resp: &mut [u8]) -> Result<(), Error> {
        let ondrop = OnDrop::new(Self::blocking_wait_ready);

        let len = core::cmp::max(req.len(), resp.len()) as u8;
        self.custom_instruction_start(opcode, req, len)?;

        self.wait_ready().await;

        self.custom_instruction_finish(resp)?;

        ondrop.defuse();

        Ok(())
    }

    /// Do a custom QSPI instruction, blocking version.
    pub fn blocking_custom_instruction(&mut self, opcode: u8, req: &[u8], resp: &mut [u8]) -> Result<(), Error> {
        let len = core::cmp::max(req.len(), resp.len()) as u8;
        self.custom_instruction_start(opcode, req, len)?;

        Self::blocking_wait_ready();

        self.custom_instruction_finish(resp)?;

        Ok(())
    }

    fn custom_instruction_start(&mut self, opcode: u8, req: &[u8], len: u8) -> Result<(), Error> {
        assert!(req.len() <= 8);

        let mut dat0: u32 = 0;
        let mut dat1: u32 = 0;

        for i in 0..4 {
            if i < req.len() {
                dat0 |= (req[i] as u32) << (i * 8);
            }
        }
        for i in 0..4 {
            if i + 4 < req.len() {
                dat1 |= (req[i + 4] as u32) << (i * 8);
            }
        }

        let r = T::regs();
        r.cinstrdat0.write(|w| unsafe { w.bits(dat0) });
        r.cinstrdat1.write(|w| unsafe { w.bits(dat1) });

        r.events_ready.reset();
        r.intenset.write(|w| w.ready().set());

        r.cinstrconf.write(|w| {
            let w = unsafe { w.opcode().bits(opcode) };
            let w = unsafe { w.length().bits(len + 1) };
            let w = w.lio2().bit(true);
            let w = w.lio3().bit(true);
            let w = w.wipwait().bit(true);
            let w = w.wren().bit(true);
            let w = w.lfen().bit(false);
            let w = w.lfstop().bit(false);
            w
        });
        Ok(())
    }

    fn custom_instruction_finish(&mut self, resp: &mut [u8]) -> Result<(), Error> {
        let r = T::regs();

        let dat0 = r.cinstrdat0.read().bits();
        let dat1 = r.cinstrdat1.read().bits();
        for i in 0..4 {
            if i < resp.len() {
                resp[i] = (dat0 >> (i * 8)) as u8;
            }
        }
        for i in 0..4 {
            if i + 4 < resp.len() {
                resp[i] = (dat1 >> (i * 8)) as u8;
            }
        }
        Ok(())
    }

    async fn wait_ready(&mut self) {
        poll_fn(move |cx| {
            let r = T::regs();
            let s = T::state();
            s.waker.register(cx.waker());
            if r.events_ready.read().bits() != 0 {
                return Poll::Ready(());
            }
            Poll::Pending
        })
        .await
    }

    fn blocking_wait_ready() {
        loop {
            let r = T::regs();
            if r.events_ready.read().bits() != 0 {
                break;
            }
        }
    }

    fn start_read(&mut self, address: u32, data: &mut [u8]) -> Result<(), Error> {
        // TODO: Return these as errors instead.
        assert_eq!(data.as_ptr() as u32 % 4, 0);
        assert_eq!(data.len() as u32 % 4, 0);
        assert_eq!(address % 4, 0);

        let r = T::regs();

        r.read.src.write(|w| unsafe { w.src().bits(address) });
        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));

        Ok(())
    }

    fn start_write(&mut self, address: u32, data: &[u8]) -> Result<(), Error> {
        // TODO: Return these as errors instead.
        assert_eq!(data.as_ptr() as u32 % 4, 0);
        assert_eq!(data.len() as u32 % 4, 0);
        assert_eq!(address % 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) });
        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));

        Ok(())
    }

    fn start_erase(&mut self, address: u32) -> Result<(), Error> {
        // TODO: Return these as errors instead.
        assert_eq!(address % 4096, 0);

        let r = T::regs();
        r.erase.ptr.write(|w| unsafe { w.ptr().bits(address) });
        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));

        Ok(())
    }

    /// Raw QSPI read.
    ///
    /// The difference with `read` is that this does not do bounds checks
    /// against the flash capacity. It is intended for use when QSPI is used as
    /// a raw bus, not with flash memory.    
    pub async fn read_raw(&mut self, address: u32, data: &mut [u8]) -> Result<(), Error> {
        let ondrop = OnDrop::new(Self::blocking_wait_ready);

        self.start_read(address, data)?;
        self.wait_ready().await;

        ondrop.defuse();

        Ok(())
    }

    /// Raw QSPI write.
    ///
    /// The difference with `write` is that this does not do bounds checks
    /// against the flash capacity. It is intended for use when QSPI is used as
    /// a raw bus, not with flash memory.
    pub async fn write_raw(&mut self, address: u32, data: &[u8]) -> Result<(), Error> {
        let ondrop = OnDrop::new(Self::blocking_wait_ready);

        self.start_write(address, data)?;
        self.wait_ready().await;

        ondrop.defuse();

        Ok(())
    }

    /// Raw QSPI read, blocking version.
    ///
    /// The difference with `blocking_read` is that this does not do bounds checks
    /// against the flash capacity. It is intended for use when QSPI is used as
    /// a raw bus, not with flash memory.
    pub fn blocking_read_raw(&mut self, address: u32, data: &mut [u8]) -> Result<(), Error> {
        self.start_read(address, data)?;
        Self::blocking_wait_ready();
        Ok(())
    }

    /// Raw QSPI write, blocking version.
    ///
    /// The difference with `blocking_write` is that this does not do bounds checks
    /// against the flash capacity. It is intended for use when QSPI is used as
    /// a raw bus, not with flash memory.
    pub fn blocking_write_raw(&mut self, address: u32, data: &[u8]) -> Result<(), Error> {
        self.start_write(address, data)?;
        Self::blocking_wait_ready();
        Ok(())
    }

    /// Read data from the flash memory.
    pub async fn read(&mut self, address: u32, data: &mut [u8]) -> Result<(), Error> {
        self.bounds_check(address, data.len())?;
        self.read_raw(address, data).await
    }

    /// Write data to the flash memory.
    pub async fn write(&mut self, address: u32, data: &[u8]) -> Result<(), Error> {
        self.bounds_check(address, data.len())?;
        self.write_raw(address, data).await
    }

    /// Erase a sector on the flash memory.
    pub async fn erase(&mut self, address: u32) -> Result<(), Error> {
        if address >= self.capacity {
            return Err(Error::OutOfBounds);
        }

        let ondrop = OnDrop::new(Self::blocking_wait_ready);

        self.start_erase(address)?;
        self.wait_ready().await;

        ondrop.defuse();

        Ok(())
    }

    /// Read data from the flash memory, blocking version.
    pub fn blocking_read(&mut self, address: u32, data: &mut [u8]) -> Result<(), Error> {
        self.bounds_check(address, data.len())?;
        self.blocking_read_raw(address, data)
    }

    /// Write data to the flash memory, blocking version.
    pub fn blocking_write(&mut self, address: u32, data: &[u8]) -> Result<(), Error> {
        self.bounds_check(address, data.len())?;
        self.blocking_write_raw(address, data)
    }

    /// Erase a sector on the flash memory, blocking version.
    pub fn blocking_erase(&mut self, address: u32) -> Result<(), Error> {
        if address >= self.capacity {
            return Err(Error::OutOfBounds);
        }

        self.start_erase(address)?;
        Self::blocking_wait_ready();
        Ok(())
    }

    fn bounds_check(&self, address: u32, len: usize) -> Result<(), Error> {
        let len_u32: u32 = len.try_into().map_err(|_| Error::OutOfBounds)?;
        let end_address = address.checked_add(len_u32).ok_or(Error::OutOfBounds)?;
        if end_address > self.capacity {
            return Err(Error::OutOfBounds);
        }
        Ok(())
    }
}

impl<'d, T: Instance> Drop for Qspi<'d, T> {
    fn drop(&mut self) {
        let r = T::regs();

        if self.dpm_enabled {
            trace!("qspi: doing deep powerdown...");

            r.ifconfig1.modify(|_, w| w.dpmen().enter());

            // Wait for DPM enter.
            // Unfortunately we must spin. There's no way to do this interrupt-driven.
            // The READY event does NOT fire on DPM enter (but it does fire on DPM exit :shrug:)
            while r.status.read().dpm().is_disabled() {}

            // Wait MORE for DPM enter.
            // I have absolutely no idea why, but the wait above is not enough :'(
            // Tested with mx25r64 in nrf52840-dk, and with mx25r16 in custom board
            cortex_m::asm::delay(4096);
        }

        // it seems events_ready is not generated in response to deactivate. nrfx doesn't wait for it.
        r.tasks_deactivate.write(|w| w.tasks_deactivate().set_bit());

        // Workaround https://infocenter.nordicsemi.com/topic/errata_nRF52840_Rev1/ERR/nRF52840/Rev1/latest/anomaly_840_122.html?cp=4_0_1_2_1_7
        // Note that the doc has 2 register writes, but the first one is really the write to tasks_deactivate,
        // so we only do the second one here.
        unsafe { ptr::write_volatile(0x40029054 as *mut u32, 1) }

        r.enable.write(|w| w.enable().disabled());

        // Note: we do NOT deconfigure CSN here. If DPM is in use and we disconnect CSN,
        // leaving it floating, the flash chip might read it as zero which would cause it to
        // spuriously exit DPM.
        gpio::deconfigure_pin(r.psel.sck.read().bits());
        gpio::deconfigure_pin(r.psel.io0.read().bits());
        gpio::deconfigure_pin(r.psel.io1.read().bits());
        gpio::deconfigure_pin(r.psel.io2.read().bits());
        gpio::deconfigure_pin(r.psel.io3.read().bits());

        trace!("qspi: dropped");
    }
}

impl<'d, T: Instance> ErrorType for Qspi<'d, T> {
    type Error = Error;
}

impl NorFlashError for Error {
    fn kind(&self) -> NorFlashErrorKind {
        NorFlashErrorKind::Other
    }
}

impl<'d, T: Instance> ReadNorFlash for Qspi<'d, T> {
    const READ_SIZE: usize = 4;

    fn read(&mut self, offset: u32, bytes: &mut [u8]) -> Result<(), Self::Error> {
        self.blocking_read(offset, bytes)?;
        Ok(())
    }

    fn capacity(&self) -> usize {
        self.capacity as usize
    }
}

impl<'d, T: Instance> NorFlash for Qspi<'d, T> {
    const WRITE_SIZE: usize = 4;
    const ERASE_SIZE: usize = 4096;

    fn erase(&mut self, from: u32, to: u32) -> Result<(), Self::Error> {
        for address in (from..to).step_by(<Self as NorFlash>::ERASE_SIZE) {
            self.blocking_erase(address)?;
        }
        Ok(())
    }

    fn write(&mut self, offset: u32, bytes: &[u8]) -> Result<(), Self::Error> {
        self.blocking_write(offset, bytes)?;
        Ok(())
    }
}

#[cfg(feature = "nightly")]
mod _eh1 {
    use embedded_storage_async::nor_flash::{NorFlash as AsyncNorFlash, ReadNorFlash as AsyncReadNorFlash};

    use super::*;

    impl<'d, T: Instance> AsyncNorFlash for Qspi<'d, T> {
        const WRITE_SIZE: usize = <Self as NorFlash>::WRITE_SIZE;
        const ERASE_SIZE: usize = <Self as NorFlash>::ERASE_SIZE;

        async fn write(&mut self, offset: u32, data: &[u8]) -> Result<(), Self::Error> {
            self.write(offset, data).await
        }

        async fn erase(&mut self, from: u32, to: u32) -> Result<(), Self::Error> {
            for address in (from..to).step_by(<Self as AsyncNorFlash>::ERASE_SIZE) {
                self.erase(address).await?
            }
            Ok(())
        }
    }

    impl<'d, T: Instance> AsyncReadNorFlash for Qspi<'d, T> {
        const READ_SIZE: usize = 4;
        async fn read(&mut self, address: u32, data: &mut [u8]) -> Result<(), Self::Error> {
            self.read(address, data).await
        }

        fn capacity(&self) -> usize {
            self.capacity as usize
        }
    }
}

pub(crate) mod sealed {
    use embassy_sync::waitqueue::AtomicWaker;

    /// Peripheral static state
    pub struct State {
        pub waker: AtomicWaker,
    }

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

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

/// QSPI peripheral instance.
pub trait Instance: Peripheral<P = Self> + sealed::Instance + 'static + Send {
    /// Interrupt for this peripheral.
    type Interrupt: Interrupt;
}

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