Ssd1309 display with async DMA only fills the first block of the screen, then hangs

[Patrickfen]

I'm working on a small project combining an SSD1309 display driven by a SAMD21G18A mcu. I'm hoping to use embassy in order to display graphics until a user presses a button, but I'm having some strange issues with async:

bind_interrupts!(struct Irqs {
    EIC => eic::InterruptHandler;
    SERCOM0 => sercom::i2c::InterruptHandler<sercom::Sercom0>;
    DMAC => dmac::InterruptHandler;
});

#[embassy_executor::main]
async fn main(_spawn: Spawner) {
    // Setup
    let mut peripherals = pac::Peripherals::take().unwrap();
    let _core = pac::CorePeripherals::take().unwrap();
    let mut clocks = clock::GenericClockController::with_internal_8mhz(
        peripherals.gclk,
        &mut peripherals.pm,
        &mut peripherals.sysctrl,
        &mut peripherals.nvmctrl,
    );
    let pins = gpio::Pins::new(peripherals.port);

    // Initialize DMA Controller
    let mut dmac =
        dmac::DmaController::init(peripherals.dmac, &mut peripherals.pm).into_future(Irqs);
    let channels: dmac::FutureChannels = dmac.split();
    let channel: dmac::Channel<dmac::Ch0, dmac::ReadyFuture> =
        channels.0.init(dmac::PriorityLevel::Lvl0);

    // I2C
    let gclk0 = clocks.gclk0();
    let sercom0_clock = &clocks.sercom0_core(&gclk0).unwrap();
    let pads = sercom::i2c::Pads::new(
        pins.pa08.into_alternate::<gpio::C>(), // A4
        pins.pa09.into_alternate::<gpio::C>(), // A5
    );
    let i2c = sercom::i2c::Config::new(
        &peripherals.pm,
        peripherals.sercom0,
        pads,
        sercom0_clock.freq(),
    )
    .baud(400.kHz())
    .enable()
    .into_future(Irqs)
    .with_dma_channel(channel); // -- works if removed --

    // Display
    let di: I2CInterface<_> = display_interface_i2c::I2CInterface::new(i2c, 0x3C, 0x40);
    let mut display: GraphicsMode<_, _> =
        oled_async::Builder::new(oled_async::displays::ssd1309::Ssd1309_128_64 {})
            .connect(di)
            .into();
    display.init().await.unwrap();
    display.clear();
    display.fill_solid(&display.bounding_box(), On).unwrap();
    display.flush().await.unwrap();

    loop {}
}

I can see the display spring to life and display a block of white, however it doesn't fill the entire screen, and adding panics to the code shows me that it never reaches beyond display.flush().await.unwrap();. Interestingly, as soon as I remove .with_dma_channel(channel); from the i2c config, everything does work, though slow enough to see blocks of the screen update in sequence (exactly what I hoped to prevent by using DMA). Could this be an issue with the DMA implementation in atsamd hal not waking up the executor? From what I saw in the existing examples there doesn't seem to be any additional setup or interrupts that need to be configured to get DMA working.