[PATCH] PCI/P2PDMA: Fix missing check for dma_virt_ops

[PATCH] PCI/P2PDMA: Fix missing check for dma_virt_ops

[Logan Gunthorpe]

Drivers that use dma_virt_ops were meant to be rejected when testing
compatibility for P2PDMA.

This check got inadvertantly dropped in one of the later versions of the
original patchset, so add it back.

Fixes: 52916982af48 ("PCI/P2PDMA: Support peer-to-peer memory")
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 drivers/pci/p2pdma.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index a4994aa3acc0..ab48babdf214 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -487,6 +487,14 @@ int pci_p2pdma_distance_many(struct pci_dev *provider, struct device **clients,
 		return -1;
 
 	for (i = 0; i < num_clients; i++) {
+		if (IS_ENABLED(CONFIG_DMA_VIRT_OPS) &&
+		    clients[i]->dma_ops == &dma_virt_ops) {
+			if (verbose)
+				dev_warn(clients[i],
+					 "cannot be used for peer-to-peer DMA because the driver makes use of dma_virt_ops\n");
+			return -1;
+		}
+
 		pci_client = find_parent_pci_dev(clients[i]);
 		if (!pci_client) {
 			if (verbose)
@@ -765,7 +773,7 @@ int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg, int nents,
 	 * p2pdma mappings are not compatible with devices that use
 	 * dma_virt_ops. If the upper layers do the right thing
 	 * this should never happen because it will be prevented
-	 * by the check in pci_p2pdma_add_client()
+	 * by the check in pci_p2pdma_distance_many()
 	 */
 	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_DMA_VIRT_OPS) &&
 			 dev->dma_ops == &dma_virt_ops))
-- 
2.20.1

Re: [PATCH] PCI/P2PDMA: Fix missing check for dma_virt_ops

[Bjorn Helgaas]

On Tue, Jul 02, 2019 at 11:35:44AM -0600, Logan Gunthorpe wrote:
> Drivers that use dma_virt_ops were meant to be rejected when testing
> compatibility for P2PDMA.
> 
> This check got inadvertantly dropped in one of the later versions of the
> original patchset, so add it back.
> 
> Fixes: 52916982af48 ("PCI/P2PDMA: Support peer-to-peer memory")
> Signed-off-by: Logan Gunthorpe <logang@deltatee.com>

Applied to pci/peer-to-peer for v5.3, thanks!

> ---
>  drivers/pci/p2pdma.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
> index a4994aa3acc0..ab48babdf214 100644
> --- a/drivers/pci/p2pdma.c
> +++ b/drivers/pci/p2pdma.c
> @@ -487,6 +487,14 @@ int pci_p2pdma_distance_many(struct pci_dev *provider, struct device **clients,
>  		return -1;
>  
>  	for (i = 0; i < num_clients; i++) {
> +		if (IS_ENABLED(CONFIG_DMA_VIRT_OPS) &&
> +		    clients[i]->dma_ops == &dma_virt_ops) {
> +			if (verbose)
> +				dev_warn(clients[i],
> +					 "cannot be used for peer-to-peer DMA because the driver makes use of dma_virt_ops\n");
> +			return -1;
> +		}
> +
>  		pci_client = find_parent_pci_dev(clients[i]);
>  		if (!pci_client) {
>  			if (verbose)
> @@ -765,7 +773,7 @@ int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg, int nents,
>  	 * p2pdma mappings are not compatible with devices that use
>  	 * dma_virt_ops. If the upper layers do the right thing
>  	 * this should never happen because it will be prevented
> -	 * by the check in pci_p2pdma_add_client()
> +	 * by the check in pci_p2pdma_distance_many()
>  	 */
>  	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_DMA_VIRT_OPS) &&
>  			 dev->dma_ops == &dma_virt_ops))
> -- 
> 2.20.1
>