On Fri, 18 Oct 2013, Russell King - ARM Linux wrote:

On Fri, Oct 18, 2013 at 04:46:00PM +0100, Stefano Stabellini wrote:

quoted

On Fri, 18 Oct 2013, Russell King - ARM Linux wrote:

quoted

On Fri, Oct 18, 2013 at 01:22:18PM +0100, Stefano Stabellini wrote:

quoted

Russell,
are you OK with this patch?

Only concern is why the change is soo large.

quoted

quoted

-static inline struct dma_map_ops *get_dma_ops(struct device *dev)
+static inline struct dma_map_ops *__generic_dma_ops(struct device *dev)
 {
 	if (dev && dev->archdata.dma_ops)
 		return dev->archdata.dma_ops;
 	return &arm_dma_ops;
 }
 
+static inline struct dma_map_ops *get_dma_ops(struct device *dev)
+{
+	if (xen_initial_domain())
+		return xen_dma_ops;
+	else
+		return __generic_dma_ops(dev);
+}
+

What's wrong with:

 static inline struct dma_map_ops *get_dma_ops(struct device *dev)
 {
+	if (xen_initial_domain())
+		return xen_dma_ops;
 	if (dev && dev->archdata.dma_ops)
 		return dev->archdata.dma_ops;
 	return &arm_dma_ops;
 }

xen_dma_ops functions might have to call back the native implementation.
With the above there is no way for a xen_dma_ops function to get to
archdata.dma_ops or arm_dma_ops.
This is the reason why this patch introduces __generic_dma_ops.

Ah, that's fine then, but having that described in the commit message
would be useful.

Acked-by: Russell King <redacted>

Thanks.

Thank you. I'll add a note to the commit message.