On Mon, Dec 10, 2018 at 07:19:30PM +0000, Robin Murphy wrote:
> On 08/12/2018 17:36, Christoph Hellwig wrote:
>> There is no need to have an additional kernel mapping for a contiguous
>> allocation if the device already is DMA coherent, so skip it.
>
> FWIW, the "need" was that it kept the code in this path simple and the
> mapping behaviour consistent with the regular iommu_dma_alloc() case. One
> could quite easily retort that there is no need for the extra complexity of
> this patch, since vmalloc is cheap on a 64-bit architecture ;)
Heh. Well, without the remap we do less work, we prepare for a simple
implementation of DMA_ATTR_NON_CONSISTENT, and also prepapre the code
to be better reusable for architectures that don't do remapping of
DMA allocations at all.
>> if (addr) {
>> memset(addr, 0, size);
>> - if (!coherent)
>> - __dma_flush_area(page_to_virt(page), iosize);
>> + __dma_flush_area(page_to_virt(page), iosize);
>
> Oh poo - seems I missed it at the time but the existing logic here is
> wrong. Let me send a separate fix to flip those statements into the correct
> order...
Yes, flushing the remapped alias only after zeroing it looks odd.
_______________________________________________
linux-snps-arc mailing list
linux-snps-arc@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/linux-snps-arc
