sh: override DMA interrupt handler to dma-sh driver

While I don't have anything against the idea, I don't care for the approach. It is not acceptable to break the API every time you wish to extend soem basic functionality. The addition of request_dma_bycap() itself should have been a good indicator of this, as we didn't simply hack request_dma() in-place. Part of the reason for this is also because request_dma()/free_dma() are legacy interfaces from the ISA DMA days, and are virtually consistent across the entire tree.

Having said that, wanting to register data with a channel or set a specific handler for it is perfectly reasonable, but this is something you ought to model after the generic hardirq layer and driver model, have special routines that set/get a private data pointer and one for setting the IRQ handler. The other thing to note is that you do not want to blow away the existing IRQ handler in the controller driver either, since that contains logic relative to your channel anyways (ie, TEI handling). You are likely better off introducing a new capability that takes priority over TEI handling and requires a special handler, but note that CHCR handling itself should only be done by the controller driver. If you wish to make use of a DMA channel in a generic driver, that is perfectly fine, but you should not be exposing _any_ dma-sh internals there, especially for the IRQ top half.