Re: [PATCH] Revert "of/device: Really only set bus DMA mask when appropriate"

[Robin Murphy]

On 2019-02-28 10:33 pm, Alistair Francis wrote:
> This reverts commit 6778be4e520959659b27a441c06a84c9cb009085.
>
> Reverting the commit fixes these error messages and an non-functioning
> USB bus when attaching a USB to PCIe card to a RISC-V board:
>      xhci_hcd 0000:03:00.0: Host took too long to start, waited 16000 microseconds.
>      xhci_hcd 0000:03:00.0: startup error -19
>      xhci_hcd 0000:03:00.0: USB bus 2 deregistered
>      xhci_hcd 0000:03:00.0: WARNING: Host System Error
>      xhci_hcd 0000:03:00.0: remove, state 1
>
> Signed-off-by: Alistair Francis <alistair.francis@xxxxxxx>
> ---
> I'm not sure if this is the correct fix, it's possible something else is
> wrong, but I wanted to get some feedback.

Given that the bug which this revert would reintroduce cripples DMA on 
just about every 64-bit DT system, probably not. Can you clarify what 
your DT looks like, and turn on the debug output for of_dma_get_range() 
to see what it says?

Robin.

> drivers/of/device.c | 4 +---
>   1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/of/device.c b/drivers/of/device.c
> index 3717f2a20d0d..8299f8055da7 100644
> --- a/drivers/of/device.c
> +++ b/drivers/of/device.c
> @@ -149,11 +149,9 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
>   	 * set by the driver.
>   	 */
>   	mask = DMA_BIT_MASK(ilog2(dma_addr + size - 1) + 1);
> +	dev->bus_dma_mask = mask;
>   	dev->coherent_dma_mask &= mask;
>   	*dev->dma_mask &= mask;
> -	/* ...but only set bus mask if we found valid dma-ranges earlier */
> -	if (!ret)
> -		dev->bus_dma_mask = mask;
>   
>   	coherent = of_dma_is_coherent(np);
>   	dev_dbg(dev, "device is%sdma coherent\n",