Re: [PATCH] dmaengine: sun6i: bound burst selection before shifting

[Frank Li]

On Wed, Oct 21, 2026 at 03:06:33PM -0600, James Hilliard wrote:
> find_burst_size() tests BIT(maxburst) before reducing the client's maximum
> to a supported burst length. DesignWare UARTs request a maximum burst of
> one quarter of their FIFO size, so a 256-byte FIFO supplies 64. Shifting
> an unsigned long by 64 is undefined even on a 64-bit CPU. The fallback
> loop can repeat the same invalid shift.
>
> Limit the search to the highest burst represented in the controller's
> nonzero capability mask before either shift. Larger client limits still
> select the largest supported burst, while smaller limits retain the
> existing power-of-two selection and the zero/default case is unchanged.
>
> Fixes: 7178c3586ab4 ("dmaengine: sun6i: Choose appropriate burst length under maxburst")
> Signed-off-by: James Hilliard <james.hilliard1@xxxxxxxxx>
> ---

Reviewed-by: Frank Li <Frank.Li@xxxxxxx>

>  drivers/dma/sun6i-dma.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/dma/sun6i-dma.c b/drivers/dma/sun6i-dma.c
> index 7704b016aed8..a8b12262b3c3 100644
> --- a/drivers/dma/sun6i-dma.c
> +++ b/drivers/dma/sun6i-dma.c
> @@ -590,6 +590,9 @@ static u32 find_burst_size(const u32 burst_lengths, u32 maxburst)
>  	if (!maxburst)
>  		return 1;
>
> +	/* maxburst is a transfer limit, not necessarily a valid bit index. */
> +	maxburst = min_t(u32, maxburst, fls(burst_lengths) - 1);
> +
>  	if (BIT(maxburst) & burst_lengths)
>  		return maxburst;
>
>
> ---
> base-commit: 93f51579e7df248780214094418f205253383cc5
> change-id: 20261021-submit-sun6i-dma-burst-limit-v1-47e32971471e
>
> Best regards,
> --
> James Hilliard <james.hilliard1@xxxxxxxxx>
>