[PATCH 6/7] fix COH 901 318 memcpy and slave issues

[Dan Williams]

On Thu, Feb 18, 2010 at 5:06 PM, Linus Walleij
<linus.walleij at stericsson.com> wrote:
> This patch fixes several issues with the COH 901 318 DMA engine:
>
> - A questionable mechanism for counting the number of interrupts
>  was removed. There is never more than one interrupt fired for
>  a transaction in this hardware.
> - A debug print was broken to the point that it caused a bug
>  when swithed on.
> - Dynamically configure the slave controller for mem->dev
>  or dev->mem transfers, this was previously hard-coded in
>  the channel configurations, but we have a channel which
>  is actually bidirectionals (MMC/SD) so this doesn't work
>  out!
> - Refactored a bit to get a clearer program flow especially in
>  the dma_tasklet() and also remove some duplicated code, minor
>  nitwits, spelling etc.
> - Also removed the redundant BUG_ON() located by Julia Lawall
>  using coccinelle.
>

Sorry for the delay.  This looks ok to me as far as I can tell, but it
makes me a bit nervous to make all the unrelated changes in one
commit.  For bisection purposes I think it would be better to split
the pure cleanups from the fixes, agree?

I may proceed with the initial dmaengine pull request as early as
tomorrow, but I think this is something that we can justify later in
the merge window or even shortly after -rc1.

--
Dan