Re: [PATCH v5 13/15] iommu/dma: Force bouncing if the size is not cacheline-aligned

[Robin Murphy <robin.murphy@arm.com>]

On 26/05/2023 8:22 pm, Catalin Marinas wrote:
> On Sat, May 27, 2023 at 12:36:30AM +0800, Jisheng Zhang wrote:
>> On Wed, May 24, 2023 at 06:19:02PM +0100, Catalin Marinas wrote:
>>> diff --git a/include/linux/scatterlist.h b/include/linux/scatterlist.h
>>> index 87aaf8b5cdb4..330a157c5501 100644
>>> --- a/include/linux/scatterlist.h
>>> +++ b/include/linux/scatterlist.h
>>> @@ -248,6 +248,29 @@ static inline void sg_unmark_end(struct scatterlist *sg)
>>>   	sg->page_link &= ~SG_END;
>>>   }
>>>   
>>> +#define SG_DMA_BUS_ADDRESS	(1 << 0)
>>> +#define SG_DMA_USE_SWIOTLB	(1 << 1)
>>> +
>>> +#ifdef CONFIG_SWIOTLB
>>
>> s/CONFIG_SWIOTLB/CONFIG_NEED_SG_DMA_FLAGS ?
>> Otherwise, there's compiler error if SWIOTLB=y but IOMMU=n
> 
> Yes, I pushed a fixup to the kmalloc-minalign branch.

I'd agree that CONFIG_NEED_SG_DMA_FLAGS is the better option.

>>> +static inline bool sg_is_dma_use_swiotlb(struct scatterlist *sg)
>>> +{
>>> +	return sg->dma_flags & SG_DMA_USE_SWIOTLB;
>>> +}
>>> +
>>> +static inline void sg_dma_mark_use_swiotlb(struct scatterlist *sg)
>>> +{
>>> +	sg->dma_flags |= SG_DMA_USE_SWIOTLB;
>>> +}
>>> +#else
>>> +static inline bool sg_is_dma_use_swiotlb(struct scatterlist *sg)
>>> +{
>>> +	return false;
>>> +}
>>> +static inline void sg_dma_mark_use_swiotlb(struct scatterlist *sg)
>>> +{
>>> +}
>>> +#endif
> 
> And I wonder whether we should keep these accessors in the scatterlist.h
> file. They are only used by the dma-iommu.c code.

Nah, logically they belong with the definition of the flag itself, which 
certainly should be here. Also once this does land, dma-direct and 
possibly others will be able to take advantage if it too (as a small win 
over repeating the is_swiotlb_buffer() check a lot).

Thanks,
Robin.