Re: [PATCH] serial: tegra: handle rx race

[Jon Hunter <jonathanh-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>]

On 01/10/15 10:30, Hamza Farooq wrote:
> Hi,
> 
> Sorry if I'm out of context. I saw this thread and found it very
> similar to the problem that we have fixed with the renesas sh-sci
> serial driver. We have made the two handlers (dma packet complete and
> timer expired) mutually exclusive by using spinlocks. Additionally,
> the race condition has been handled by not stopping the dma engine in
> timer expiry handler. If the both happen at the same time, let the dma
> packet complete handler push the data, by simply exiting the timer
> handler. If the interrupts are redirected to CPU in timer expiry
> handler, do it _only_ when you are sure no more data is coming in.

Yes that is the key. I guess if flow-control is used, then you can
prevent more data coming, otherwise I am not sure.

> The development history can be viewed at Geert Uytterhoeven's renesas
> drivers tree:
> https://git.kernel.org/cgit/linux/kernel/git/geert/renesas-drivers.git
> 
> Interesting files are drivers/tty/serial/sh-sci and drivers/dma/sh/rcar-dmac.c

Which branch are you referring to? Is this not in the mainline yet.

Thanks for the info!

Jon