Re: [PATCH] rpmsg_ns: Work around TI non-standard message

[Mathieu Poirier <mathieu.poirier@linaro.org>]

On Tue, Oct 15, 2024 at 06:58:33PM +0200, Richard Weinberger wrote:
> Mathieu,
> 
> Am Dienstag, 15. Oktober 2024, 18:48:08 CEST schrieb Mathieu Poirier:
> > Good morning Richard,
> > 
> > On Fri, Oct 11, 2024 at 02:39:22PM +0200, Richard Weinberger wrote:
> > > Texas Instruments ships a patch in their vendor kernels,
> > > which adds a new NS message that includes a description field.
> > > While TI is free to do whatever they want in their copy of the kernel,
> > > it becomes a mess when people switch to a mainline kernel and want
> > > to use their existing DSP programs with it.
> > 
> > I suspect there is a lot more things to change when going from downstream to a
> > mainline kernel.
> 
> Not really.
> I had to revert c6aed238b7a9b ("remoteproc: modify vring allocation to rely on centralized carveout allocator")
> because the DSP has a sub-optimal resource table, and this workaround.
> With that the DSP program worked as-is on kernel 6.6.
> Downstream was 4.19 TI.
> 
> > > 
> > > To make it easier to migrate to a mainline kernel,
> > > let's make the kernel aware of their non-standard extension but
> > > briefly ignore the description field.
> > 
> > In my opinion the real fix here is to get TI to use the standard message
> > announcement structure.  The ->desc field doesn't seem to be that useful since
> > it gets discarted.
> 
> This is for the future, the goal of my patch is helping people to
> get existing DSP programs work with mainline.
> Not everyone can or want to rebuild theirs DSP programs when moving to a mainline
> kernel.

That's an even better argument to adopt the standard structure as soon as
possible.  Modifying the mainline kernel to adapt to vendors' quirks doesn't
scale.  

> 
> Thanks,
> //richard
> 
> -- 
> ​​​​​sigma star gmbh | Eduard-Bodem-Gasse 6, 6020 Innsbruck, AUT
> UID/VAT Nr: ATU 66964118 | FN: 374287y
> 
>