On Thu, May 31, 2012 at 09:14:15AM +0800, Linus Walleij wrote:

On Tue, May 29, 2012 at 5:43 PM, Russell King
[off-list ref] wrote:

quoted

Rather than code the de-queue of the txd several times, move that into
the start_txd function. ?Rename this to better illustrate what it's
now doing, and call this function when starting a delayed memcpy().

Signed-off-by: Russell King <redacted>

quoted

+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? /*
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?* Eww. ?We know this isn't going to deadlock
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?* but lockdep probably doens't.

Speling...

quoted

+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?*/
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? spin_lock(&waiting->lock);
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? pl08x_start_next_txd(waiting);
+ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? spin_unlock(&waiting->lock);

I've seen this construct in the past and have no better suggestions...

I have no way to test whether my statement is true as I can't get enough
DMA engine usage through the PL011 transmit support to check what happens.
So the statement is purely theoretical.

Acked-by: Linus Walleij <redacted>

Yours,
Linus Walleij