Re: [PATCH for-next] RDMA/rxe: fix regression caused by recent patch

[Bob Pearson <rpearsonhpe@gmail.com>]

On 10/29/20 4:25 PM, Bob Pearson wrote:
> The commit referenced below performs additional checking on
> devices used for DMA. Specifically it checks that
> 
> device->dma_mask != NULL
> 
> Rdma_rxe uses this device when pinning MR memory but did not
> set the value of dma_mask. In fact rdma_rxe does not perform
> any DMA operations so the value is never used but is checked.
> 
> This patch gives dma_mask a valid value. Without this patch
> rdma_rxe does not function at all.
> 
> Fixes: f959dcd6ddfd2 ("dma-direct: Fix potential NULL pointer dereference")
> Signed-off-by: Bob Pearson <rpearson@hpe.com>
> ---
>  drivers/infiniband/sw/rxe/rxe_verbs.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
> index 7652d53af2c1..116a234e92db 100644
> --- a/drivers/infiniband/sw/rxe/rxe_verbs.c
> +++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
> @@ -1134,8 +1134,15 @@ int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name)
>  	dev->node_type = RDMA_NODE_IB_CA;
>  	dev->phys_port_cnt = 1;
>  	dev->num_comp_vectors = num_possible_cpus();
> +
> +	/* rdma_rxe never does real DMA but does rely on
> +	 * pinning user memory in MRs to avoid page faults
> +	 * in responder and completer tasklets
> +	 */
>  	dev->dev.parent = rxe_dma_device(rxe);
> +	dev->dev.dma_mask = DMA_BIT_MASK(64);
>  	dev->local_dma_lkey = 0;
> +
>  	addrconf_addr_eui48((unsigned char *)&dev->node_guid,
>  			    rxe->ndev->dev_addr);
>  	dev->dev.dma_parms = &rxe->dma_parms;
>

Ignore this patch. It turns out it works because any nonzero number in dma_mask will stop the check that is failing and since rxe never uses DMA it won't affect anything. But, it doesn't compile cleanly because the dma_mask is a pointer to the actual dma_mask and not the mask. Somehow I missed the warning. I have a newer version that uses the function dma_coerce_mask_and_coherent() and also works. (Works means it gets to the next problem as mentioned in the prvious note.)

Bob