Re: [PATCH 7/8] mci: sdhci: Force DMA update to the next block boundary

[Jules Maselbas <jmaselbas@kalray.eu>]

On Mon, Jul 10, 2023 at 07:23:34PM +0200, Jules Maselbas wrote:
> When working with 64-bit DMA the DMA addresse is stored in two 32-bit
> registers, different from the usual 32-bit one. Copy what is done by
> Linux which simple and is also a workaround for some broken controller.
> 
> Note: in case the 64-bit DMA address needs to be read, it must be done
> by reading the 32-bit LSB first (SDHCI_ADMA_ADDRESS) and then the MSB
> (SDHCI_ADMA_ADDRESS_HI).
> 
> Adapted from Linux commit:
> 8<----------------------------------------------------------------------
> commit f6a03cbf43e586211f8ea088148c8ecd3fc4b5be
> Author: Mikko Vinni <mmvinni@yahoo.com>
> Date:   Tue Apr 12 09:36:18 2011 -0400
> 
>     mmc: sdhci: work around broken dma boundary behavior
> 
>     Some SD host controllers (noticed on an integrated JMicron SD reader on an
>     HP Pavilion dv5-1250eo laptop) don't update the dma address register before
>     signaling a dma interrupt due to a dma boundary. Update the register
>     manually to the next boundary (by default 512KiB), at which the transfer
>     stopped.
> 
>     As long as each transfer is at most 512KiB in size (guaranteed by a BUG_ON
>     in sdhci_prepare_data()) and the boundary is kept at the default value,
>     this fix is needed at most once per transfer. Smaller boundaries are taken
>     care of by counting the transferred bytes.
> 
>     Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=28462
> 
>     Signed-off-by: Mikko Vinni <mmvinni@yahoo.com>
>     Signed-off-by: Chris Ball <cjb@laptop.org>
> 8<----------------------------------------------------------------------
> 
> Signed-off-by: Jules Maselbas <jmaselbas@kalray.eu>
> ---
>  drivers/mci/sdhci.c | 12 +++++++++++-
>  1 file changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/mci/sdhci.c b/drivers/mci/sdhci.c
> index 4aca3af5aa..4c8e73d94d 100644
> --- a/drivers/mci/sdhci.c
> +++ b/drivers/mci/sdhci.c
> @@ -279,8 +279,18 @@ int sdhci_transfer_data_dma(struct sdhci *sdhci, struct mci_data *data,
>  			goto out;
>  		}
>  
> +		/*
> +		 * We currently don't do anything fancy with DMA
> +		 * boundaries, but as we can't disable the feature
> +		 * we need to at least restart the transfer.
> +		 *
> +		 * According to the spec sdhci_readl(host, SDHCI_DMA_ADDRESS)
> +		 * should return a valid address to continue from, but as
> +		 * some controllers are faulty, don't trust them.
> +		 */
>  		if (irqstat & SDHCI_INT_DMA) {
> -			u32 addr = sdhci_read32(sdhci, SDHCI_DMA_ADDRESS);
> +			/* Force update to the next DMA block boundary. */
> +			dma += sdhci->sdma_boundary;
I've modified this to increase the dma addresse instead of aligning the addresse
to the next boundary...
I don't know why this works in my case, but by reading the SDHCI spec this should
really be aligned on a boundary, maybe the behavior is different when v4_mode is
enabled.

>  
>  			/*
>  			 * DMA engine has stopped on buffer boundary. Acknowledge
> -- 
> 2.17.1
>