On Thu, Nov 19, 2009 at 04:48:14PM +0100, Sascha Hauer wrote:
There already is a driver for i.MX in the tree, this one is a nearly complete rewrite. I haven't touched the in Kernel driver with these patches. The in Kernel driver has several problems, it doesn't use ioremap, uses direct pointer derefs instead of proper access functions and the only board supported so far is itself not supported in mainline.
I've CCed in Javier who wrote the currently merged code.
I'll review these patches tomorrow or over the weekend, however this jumps out at me as being something that we'd want to fix before merging since having two drivers for the same bit of hardware always leads to problems going forwards. I don't mind how this is addressed, replacing the existing driver or incrementally fixing it would both work for me (the latter is obviously preferrable but equally well we don't want to get bogged down).
The approach of using FIQ for i.MX3x until SDMA gets merged seems like a reasonable one. It's not ideal but it will work and it shouldn't create any problems transitioning to SDMA when that is submitted.
BTW, might be nice to use --no-chain-reply-to for git send-email - it makes long patch sets look a lot nicer in a threaded mail reader. Not a big deal, though.