Re: [PATCH 4/6] MIPS: refactor the maybe coherent DMA indicators

[Christoph Hellwig]

On Tue, Feb 09, 2021 at 02:12:37PM +0100, Thomas Bogendoerfer wrote:
> > +#ifdef CONFIG_DMA_MAYBE_COHERENT
> > +extern bool dma_default_coherent;
> >  static inline bool dev_is_dma_coherent(struct device *dev)
> >  {
> > -	return coherentio == IO_COHERENCE_ENABLED ||
> > -		(coherentio == IO_COHERENCE_DEFAULT && hw_coherentio);
> > +	return dma_default_coherent;
> 
> this breaks overriding of coherentio via command line. plat_mem_setup/
> plat_setup_iocoherency is called before earlyparams are handled. So
> changing coherentio after that doesn't have any effect.

Hmm.  In that case a manual override does actually work for alchemy,
as that initializes coherentio from plat_mem_setup().  But the
elaborate sanity checking that malta performs in plat_setup_iocoherency
is rather pointless then, as coherentio will always be set to
IO_COHERENCE_DISABLED at this point.