What exactly are you guys doing?
The API Anton introduced: dma_mapping_error() takes only a virtual
address as the argument (no struct device or anyting), so the additional
API's pci_dma_mapping_error() and vio_dma_mapping_error have absolutely
no choice but to do the same thing as dma_mapping_error() (because the
error return cannot be bus or device specific).
So, why bother introducing all these superfluous APIs in the first
place? Just stick to the single dma_mapping_error(); it will be much
easier.
James
> What exactly are you guys doing?
>
> The API Anton introduced: dma_mapping_error() takes only a virtual
> address as the argument (no struct device or anyting), so the additional
> API's pci_dma_mapping_error() and vio_dma_mapping_error have absolutely
> no choice but to do the same thing as dma_mapping_error() (because the
> error return cannot be bus or device specific).
Umm yeah, we should probably just use dma_mapping_error. I want to see
vio_* disappear anyway.
Anton