[+Robin]
On Mon, Apr 20, 2020 at 10:11:31AM +0200, Ard Biesheuvel wrote:
> When calculating the DMA mask from the address limit provided by the
> firmware, we add one to the ilog2() of the end address, and pass the
> result to DMA_BIT_MASK().
>
> For an end address that is not a power-of-2 minus 1, this will result
> in the mask to be wider than the limit, and cover memory that is not
> addressable by the device. Instead, we should add 1 to 'end' before
> taking the log, so that a limit of, say, 0x3fffffff gets translated
> to a mask of 30, but any value below it gets translated to 29.
>
> Signed-off-by: Ard Biesheuvel <ardb@xxxxxxxxxx>
> ---
> drivers/acpi/arm64/iort.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
Need Robin's feedback on this - I was looking at:
a7ba70f1787f ("dma-mapping: treat dev->bus_dma_mask as a DMA limit")
I assume current code is *intended* but I shall let Robin comment
on this.
Thanks,
Lorenzo
> diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
> index 7d04424189df..aab2f51eff14 100644
> --- a/drivers/acpi/arm64/iort.c
> +++ b/drivers/acpi/arm64/iort.c
> @@ -1162,7 +1162,7 @@ void iort_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
> * firmware.
> */
> end = dmaaddr + size - 1;
> - mask = DMA_BIT_MASK(ilog2(end) + 1);
> + mask = DMA_BIT_MASK(ilog2(end + 1));
> dev->bus_dma_limit = end;
> dev->coherent_dma_mask = mask;
> *dev->dma_mask = mask;
> --
> 2.17.1
>
