Re: [PATCH V10 6/8] PCI/P2PDMA: Add a 10-Bit Tag check in P2PDMA

[Bjorn Helgaas <helgaas@kernel.org>]

On Sat, Oct 09, 2021 at 06:49:36PM +0800, Dongdong Liu wrote:
> Add a 10-Bit Tag check in the P2PDMA code to ensure that a device with
> 10-Bit Tag Requester doesn't interact with a device that does not
> support 10-Bit Tag Completer. Before that happens, the kernel should
> emit a warning.

> Signed-off-by: Dongdong Liu <liudongdong3@huawei.com>
> Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

> @@ -532,6 +577,9 @@ calc_map_type_and_dist(struct pci_dev *provider, struct pci_dev *client,
>  		map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
>  	}
>  done:
> +	if (pci_10bit_tags_unsupported(client, provider, verbose))
> +		map_type = PCI_P2PDMA_MAP_NOT_SUPPORTED;

I need to be convinced that this check is in the right spot to catch
all potential P2PDMA situations.  The pci_p2pmem_find() and
pci_p2pdma_distance() interfaces eventually call
calc_map_type_and_dist().  But those interfaces don't actually produce
DMA bus addresses, and I'm not convinced that all P2PDMA users use
them.

nvme *does* use them, but infiniband (rdma_rw_map_sg()) does not, and
it calls pci_p2pdma_map_sg().

amdgpu_dma_buf_attach() calls pci_p2pdma_distance_many() but I don't
know where it sets up P2PDMA transactions.

cxgb4 and qed mention "peer2peer", but I don't know whether they are
related; they don't seem to use any pci_p2p.* interfaces.

>  	rcu_read_lock();
>  	p2pdma = rcu_dereference(provider->p2pdma);
>  	if (p2pdma)
> -- 
> 2.22.0
>