[U-Boot] [PATCH v4] nios2: convert dma_alloc_coherent to use malloc_cache_aligned

[Marek Vasut]

On Tuesday, October 13, 2015 at 02:17:53 AM, Thomas Chou wrote:
> Hi  Marek,

Hi!

> On 10/12/2015 09:46 PM, Marek Vasut wrote:
> > On Monday, October 12, 2015 at 02:55:03 PM, Thomas Chou wrote:
> >> Hi Marek,
> > 
> > Hi Thomas,
> > 
> >> On 10/12/2015 06:32 PM, Marek Vasut wrote:
> >>> Wouldn't invalidate_dcache_range() be enough here ? You don't care
> >>> about the data in the newly allocated area at this point I guess --
> >>> either you fill them in and then flush, for DMA from CPU to device OR
> >>> you receive data from device to CPU and then you invalidate this
> >>> buffer again.
> >> 
> >> No. We cannot use invalidate cache here. This is related to cache design
> >> of nios2, kind of direct mapped cache.
> > 
> > Can you please expand on this ?
> 
> This is rather hardware depended.
> 
> Please refer to page 7 about cache implementation on the "Nios II Core
> Implementation Detail" manual,
> 
> https://www.altera.com/content/dam/altera-www/global/en_US/pdfs/literature/
> hb/nios2/n2cpu_nii51015.pdf
> 
> And page 227, 8-44 about initd and initda instructions on the "Nios II
> Classic Processor Reference Guide",
> 
> https://www.altera.com/content/dam/altera-www/global/en_US/pdfs/literature/
> hb/nios2/n2cpu_nii5v1.pdf
> 
> The initda instruction might do the job.

I'd say that's the instruction we need, yep.

> But nios2 cpu with 4 bytes
> dcache line size does not support this instruction. So we don't
> implement the invalidate_dcache_range/all() in u-boot yet.

Where did you find this information ?

> Of course, we can add the invalidate cache ops if we can drop the
> support of nios2 cpu with 4 bytes dcache line size.
> 
> > btw. I was thinking about this whole cache situation. Please don't get me
> > wrong, my intention is not to put way more work unto you and/or grind
> > you about minor details. I hope it doesn't look that way to you.
> 
> No worries. It is just that I had very little free time last night to
> give more details. You are welcome.

Thanks!

Best regards,
Marek Vasut