On 09/03/18 18:29, Nipun Gupta wrote:

quoted

-----Original Message-----
From: Christoph Hellwig [mailto:hch at lst.de]
Sent: Thursday, March 08, 2018 13:11

On Tue, Mar 06, 2018 at 04:41:56AM +0000, Nipun Gupta wrote:

quoted

Sorry for asking a trivial question - looking into dma_configure() I see that
PCI is used in the start and the end of the API.
In the end part pci_put_host_bridge_device() is called.
So are two bus callbacks something like 'dma_config_start' &

'dma_config_end'

quoted

will be required where the former one will return "dma_dev"?

I'd just use dma_configure as the callback.

This would be a decent stuff.

Yes, the PCI version should end up looking a lot like the old 
pci_dma_configure() used to, before everything got pulled together.

quoted

Currently the of_dma_configure and acpi_dma_configure are only used
for PCI anyway, as no one else sets a non-NULL dma dev.

My understanding is that even the platform bus uses the of_dma_configure
and probably acpi_dma_configure too. So platform bus may also need the
callback implemented. Please correct me if my understanding is wrong.

Indeed, platform and AMBA will want an implementation to dispatch 
between {of,acpi}_dma_configre() per the current logic. Host1x doesn't 
seem to care about ACPI so probably wants its own just calling 
of_dma_configure().

I will submit the patch with 'dma_configure' callback implemented for
the busses shortly.

Cheers - I've been trying to find some time to take a look myself, but 
having something to review instead would certainly be easier :)

Robin.