Re: [PATCH v2 2/2] dma: add IOMMU static calls with clear default ops

[Christoph Hellwig]

On Thu, Jul 18, 2024 at 10:04:06AM +0300, Leon Romanovsky wrote:
> Whom should I add?

Greg as the device model maintainer.

> > The normal style would be to move the ifdefs outside the helper
> > function.
> 
> I think you are talking about the style of functions declarations in
> header files. This function is inside c-file and it is much easier
> to write it this way. 

But also much harder to read, which will happen much more often.

> > >  #ifdef CONFIG_DMA_OPS_BYPASS
> > > +	WARN_ON_ONCE(dev->dma_ops_bypass && dma_is_default_iommu(dev));
> > >  	if (dev->dma_ops_bypass)
> > 
> > Let's skip this and think about it if we ever use the bypass for
> > something that is not powerpc with it's own dma ops.
> 
> I wanted to catch misconfigurations, but I can remove it. Is this what
> you are suggesting?
> 
>   126 static bool dma_go_direct(struct device *dev, dma_addr_t mask, 
>   127                 const struct dma_map_ops *ops)                
>   128 {                                                            
>   129         if (likely(!ops && !dma_is_default_iommu(dev)))     
>   130                 return true;                               
>   131                                                           
>   132         if (dma_is_default_iommu(dev))    
>   133                 return false;

I'd go for:

		if (use_dma_iommu(dev))
			return false;
		if (likely(!ops))
			return true;
		...