Re: pcm|dmaengine|imx-sdma race condition on i.MX6

[Richard Leitner]

On Fri, Aug 21, 2020 at 09:21:37AM +0000, Robin Gong wrote:
> On 2020/08/21 12:34 Richard Leitner <richard.leitner@xxxxxxxxxxx> wrote: 
> > On Thu, Aug 20, 2020 at 03:01:44PM +0000, Robin Gong wrote:
> > > On 2020/08/19 22:26 Benjamin Bara - SKIDATA <Benjamin.Bara@xxxxxxxxxxx>
> > wrote:
> > > >
> > > > @Robin:
> > > > Is it possible to tag the commits for the stable-tree
> > > > Cc: stable@xxxxxxxxxxxxxxx?
> > > Could my patch work in your side? If yes, I will add
> > > Cc: stable@xxxxxxxxxxxxxxx
> > 
> > I've tested the patches 3 & 4 (removing sdmac->context_loaded) of the series
> > you mentioned and sent Tested-by tags for them [1,2], as they fix the EIO
> > problems for our use case.
> > 
> > So from our side they are fine for stable.
> > 
> Okay, I thought that's just decrease the issue in your side not totally fix, and the patch

As Benjamin mentioned the issue isn't "fixed" for us from the logical/
technical side.
Nonetheless the EIO error won't occur anymore with the patches applied.
Therefore the issue is for me "fixed from a userspace point of view", as
they don't get/see the error anymore.

> I post in https://www.spinics.net/lists/arm-kernel/msg829972.html
> could resolve the potential next descriptor wrongly freed by vchan_get_all_descriptors
> in sdma_channel_terminate_work. Anyway, I'll add ' Cc: stable@xxxxxxxxxxxxxxx' and
> your Tested-by tags in 3&4, then resend it again, thanks.

Great. Thank you!

regards;rl