Current high-level transfer API is not designed to support ping-pong buffering using CHAIN_TO.

It is crucial when dealing for example with camera data, where it is needed to fill a DMA buffer with a few lines and processing it very quickly while the second buffer is filled meanwhile.

I've tried double buffer approach without chaining:

let transfer_a = rx_fifo.dma_pull(dma_channel.reborrow(), buffer_a, false);
transfer_a.await;
  
loop {        
      {
          let transfer_b = rx_fifo.dma_pull(dma_channel.reborrow(), buffer_b, false);
          process_data(buffer_a);
          transfer_b.await;
      }
      {
          let transfer_a = rx_fifo.dma_pull(dma_channel.reborrow(), buffer_a, false);
          process_data(buffer_b);
          transfer_a.await;
      }
}

However it is not reliable and I observe missing data quite quickly (it probably depends on other running tasks).