feat(dma): add DMA driver with 10 verified examples

Initial DMA driver implementation for MCXA276 with:

Core DMA Features:
- DmaChannel type with ownership tracking via Channel trait
- Transfer, RingBuffer, and ScatterGatherBuilder abstractions
- Support for mem-to-mem, mem-to-peripheral, peripheral-to-mem transfers
- Interrupt-driven completion with embassy async/await integration
- Word size abstraction (u8, u16, u32) via Word trait

LPUART DMA Integration:
- LpuartTxDma and LpuartRxDma drivers for async UART with DMA
- LpuartDma combined TX/RX driver
- Automatic chunking for buffers > 0x7FFF bytes
- DMA guards with Drop impl for safe cancellation

10 Verified Examples:
- dma_mem2mem: Basic memory-to-memory copy
- dma_memset: Memory fill with pattern
- dma_uart_tx: UART transmit via DMA
- dma_uart_rx: UART receive via DMA
- dma_uart_loopback: Combined TX/RX loopback test
- dma_scatter_gather: Linked descriptor chains
- dma_channel_linking: Major/minor loop channel linking
- dma_ring_buffer: Circular buffer for continuous streaming
- dma_ping_pong: Double-buffering pattern
- dma_software_trigger: Manual transfer triggering

PR Feedback Addressed:
- Use PAC accessor for LPUART DATA register instead of manual offset
- Add EnableInterrupt enum to replace boolean parameter for readability
- Add DMA guards with Drop impl for safe async cancellation
- Automatic chunking for large buffers instead of returning error
- Use NonNull<[W]> + PhantomData for RingBuffer (DMA acts like separate thread)
- Remove edma parameter from all methods (single eDMA instance steals ptr internally)
- Make edma_tcd() non-public (HAL should not expose PAC items)

1 files changed, 231 insertions, 0 deletions

diff --git a/examples/src/bin/dma_wrap_transfer.rs b/examples/src/bin/dma_wrap_transfer.rs
new file mode 100644
index 000000000..b115a2c19
--- /dev/null
+++ b/examples/src/bin/dma_wrap_transfer.rs
@@ -0,0 +1,231 @@
+//! DMA wrap transfer example for MCXA276.
+//!
+//! This example demonstrates using DMA with modulo addressing to wrap around
+//! a source buffer, effectively repeating the source data in the destination.
+//!
+//! # Embassy-style features demonstrated:
+//! - `dma::edma_tcd()` accessor for simplified register access
+//! - `DmaChannel::is_done()` and `clear_done()` helper methods
+//! - No need to pass register block around
+
+#![no_std]
+#![no_main]
+
+use embassy_executor::Spawner;
+use embassy_mcxa::clocks::config::Div8;
+use embassy_mcxa::clocks::Gate;
+use embassy_mcxa::dma::{edma_tcd, DmaChannel, DmaCh0InterruptHandler};
+use embassy_mcxa::{bind_interrupts, dma};
+use embassy_mcxa::lpuart::{Blocking, Config, Lpuart, LpuartTx};
+use embassy_mcxa::pac;
+use {defmt_rtt as _, embassy_mcxa as hal, panic_probe as _};
+
+// Bind DMA channel 0 interrupt using Embassy-style macro
+bind_interrupts!(struct Irqs {
+    DMA_CH0 => DmaCh0InterruptHandler;
+});
+
+// Source buffer: 4 words (16 bytes), aligned to 16 bytes for modulo
+#[repr(align(16))]
+struct AlignedSrc([u32; 4]);
+
+static mut SRC: AlignedSrc = AlignedSrc([0; 4]);
+static mut DST: [u32; 8] = [0; 8];
+
+/// Helper to write a u32 as decimal ASCII to UART
+fn write_u32(tx: &mut LpuartTx<'_, Blocking>, val: u32) {
+    let mut buf = [0u8; 10];
+    let mut n = val;
+    let mut i = buf.len();
+
+    if n == 0 {
+        tx.blocking_write(b"0").ok();
+        return;
+    }
+
+    while n > 0 {
+        i -= 1;
+        buf[i] = b'0' + (n % 10) as u8;
+        n /= 10;
+    }
+
+    tx.blocking_write(&buf[i..]).ok();
+}
+
+/// Helper to print a buffer to UART
+fn print_buffer(tx: &mut LpuartTx<'_, Blocking>, buf_ptr: *const u32, len: usize) {
+    tx.blocking_write(b"[").ok();
+    unsafe {
+        for i in 0..len {
+            write_u32(tx, *buf_ptr.add(i));
+            if i < len - 1 {
+                tx.blocking_write(b", ").ok();
+            }
+        }
+    }
+    tx.blocking_write(b"]").ok();
+}
+
+#[embassy_executor::main]
+async fn main(_spawner: Spawner) {
+    // Small delay to allow probe-rs to attach after reset
+    for _ in 0..100_000 {
+        cortex_m::asm::nop();
+    }
+
+    let mut cfg = hal::config::Config::default();
+    cfg.clock_cfg.sirc.fro_12m_enabled = true;
+    cfg.clock_cfg.sirc.fro_lf_div = Some(Div8::no_div());
+    let p = hal::init(cfg);
+
+    defmt::info!("DMA wrap transfer example starting...");
+
+    // Enable DMA0 clock and release reset
+    unsafe {
+        hal::peripherals::DMA0::enable_clock();
+        hal::peripherals::DMA0::release_reset();
+    }
+
+    let pac_periphs = unsafe { pac::Peripherals::steal() };
+
+    unsafe {
+        dma::init(&pac_periphs);
+    }
+
+    // Enable DMA interrupt
+    unsafe {
+        cortex_m::peripheral::NVIC::unmask(pac::Interrupt::DMA_CH0);
+    }
+
+    let config = Config {
+        baudrate_bps: 115_200,
+        enable_tx: true,
+        enable_rx: false,
+        ..Default::default()
+    };
+
+    let lpuart = Lpuart::new_blocking(p.LPUART2, p.P2_2, p.P2_3, config).unwrap();
+    let (mut tx, _rx) = lpuart.split();
+
+    tx.blocking_write(b"EDMA wrap transfer example begin.\r\n\r\n")
+        .unwrap();
+
+    // Initialize buffers
+    unsafe {
+        SRC.0 = [1, 2, 3, 4];
+        DST = [0; 8];
+    }
+
+    tx.blocking_write(b"Source Buffer:              ").unwrap();
+    print_buffer(&mut tx, unsafe { core::ptr::addr_of!(SRC.0) } as *const u32, 4);
+    tx.blocking_write(b"\r\n").unwrap();
+
+    tx.blocking_write(b"Destination Buffer (before): ").unwrap();
+    print_buffer(&mut tx, core::ptr::addr_of!(DST) as *const u32, 8);
+    tx.blocking_write(b"\r\n").unwrap();
+
+    tx.blocking_write(b"Configuring DMA with Embassy-style API...\r\n")
+        .unwrap();
+
+    // Create DMA channel using Embassy-style API
+    let dma_ch0 = DmaChannel::new(p.DMA_CH0);
+
+    // Use edma_tcd() accessor instead of passing register block around
+    let edma = edma_tcd();
+
+    // Configure wrap transfer using direct TCD access:
+    // SRC is 16 bytes (4 * u32). We want to transfer 32 bytes (8 * u32).
+    // SRC modulo is 16 bytes (2^4 = 16) - wraps source address.
+    // DST modulo is 0 (disabled).
+    // This causes the source address to wrap around after 16 bytes,
+    // effectively repeating the source data.
+    unsafe {
+        let t = edma.tcd(0);
+
+        // Reset channel state
+        t.ch_csr().write(|w| {
+            w.erq().disable()
+                .earq().disable()
+                .eei().no_error()
+                .ebw().disable()
+                .done().clear_bit_by_one()
+        });
+        t.ch_es().write(|w| w.bits(0));
+        t.ch_int().write(|w| w.int().clear_bit_by_one());
+
+        // Source/destination addresses
+        t.tcd_saddr().write(|w| w.saddr().bits(core::ptr::addr_of!(SRC.0) as u32));
+        t.tcd_daddr().write(|w| w.daddr().bits(core::ptr::addr_of_mut!(DST) as u32));
+
+        // Offsets: both increment by 4 bytes
+        t.tcd_soff().write(|w| w.soff().bits(4));
+        t.tcd_doff().write(|w| w.doff().bits(4));
+
+        // Attributes: 32-bit transfers (size = 2)
+        // SMOD = 4 (2^4 = 16 byte modulo for source), DMOD = 0 (disabled)
+        t.tcd_attr().write(|w| {
+            w.ssize().bits(2)
+                .dsize().bits(2)
+                .smod().bits(4)  // Source modulo: 2^4 = 16 bytes
+                .dmod().bits(0)  // Dest modulo: disabled
+        });
+
+        // Transfer 32 bytes total in one minor loop
+        let nbytes = 32u32;
+        t.tcd_nbytes_mloffno().write(|w| w.nbytes().bits(nbytes));
+
+        // Source wraps via modulo, no adjustment needed
+        t.tcd_slast_sda().write(|w| w.slast_sda().bits(0));
+        // Reset dest address after major loop
+        t.tcd_dlast_sga().write(|w| w.dlast_sga().bits(-(nbytes as i32) as u32));
+
+        // Major loop count = 1
+        t.tcd_biter_elinkno().write(|w| w.biter().bits(1));
+        t.tcd_citer_elinkno().write(|w| w.citer().bits(1));
+
+        // Enable interrupt on major loop completion
+        t.tcd_csr().write(|w| w.intmajor().set_bit());
+
+        cortex_m::asm::dsb();
+
+        tx.blocking_write(b"Triggering transfer...\r\n").unwrap();
+        dma_ch0.trigger_start(edma);
+    }
+
+    // Wait for completion using channel helper method
+    while !dma_ch0.is_done(edma) {
+        cortex_m::asm::nop();
+    }
+    unsafe { dma_ch0.clear_done(edma); }
+
+    tx.blocking_write(b"\r\nEDMA wrap transfer example finish.\r\n\r\n")
+        .unwrap();
+    tx.blocking_write(b"Destination Buffer (after):  ").unwrap();
+    print_buffer(&mut tx, core::ptr::addr_of!(DST) as *const u32, 8);
+    tx.blocking_write(b"\r\n\r\n").unwrap();
+
+    // Verify: DST should be [1, 2, 3, 4, 1, 2, 3, 4]
+    let expected = [1u32, 2, 3, 4, 1, 2, 3, 4];
+    let mut mismatch = false;
+    unsafe {
+        for i in 0..8 {
+            if DST[i] != expected[i] {
+                mismatch = true;
+                break;
+            }
+        }
+    }
+
+    if mismatch {
+        tx.blocking_write(b"FAIL: Mismatch detected!\r\n").unwrap();
+        defmt::error!("FAIL: Mismatch detected!");
+    } else {
+        tx.blocking_write(b"PASS: Data verified.\r\n").unwrap();
+        defmt::info!("PASS: Data verified.");
+    }
+
+    loop {
+        cortex_m::asm::wfe();
+    }
+}
+