CONFIG_ARM

[igalkr]

Hi,

Great job on the driver.
Could you elaborate on the CONFIG_ARM, where is it set and when?

This is actually brings me to the next questions:

1. Why the dma_mmap_coherent is utilized when CONFIG_ARM is not defined?
2. There is no mapping applied in the udmabuf_driver_vma_fault - as I saw in other samples which have utilized remap_pfn_range - https://github.com/claudioscordino/mmap_alloc/blob/master/mmap_alloc.c
3. https://www.embeddedrelated.com/showthread/comp.arch.embedded/217578-1.php states that the vma->vm_pgoff has to be set to zero prior to invoking dma_mmap_coherent, why you do not reset this value?

Thanks,
Igal

[KAWAZOME Ichiro]

Thank you for the issue.

A.1

Originally I made udmabuf assuming ARM architecture only.
At that time, cache coherency control between buffer and CPU was necessary.
For that reason I did not use dma_mmap_coherent (), because I did not know how to specify cache coherency control with dma_mmap_coherent ().

At one time I try to build on the x86 architecture as well as the ARM architecture.
At that time, the following two problems occurred.
 1. pgprot_dmacoherent () was undefined on x86, pgprot_dmacoherent () seems to be for ARM only.
 2. udmabuf used on-demand paging mechanism for page mapping, but it did not work well on x86.

So I changed it so that I do not use on-demand paging outside the ARM architecture.
However, as a side effect, caching coherency control can not be controlled outside the ARM architecture.
That's why we switched the code with the definition of SYNC_ENABLE.

A.2

udmabuf_driver_vma_fault() referred to [Linux Device Drivers, Third Edition] (https://lwn.net/Kernel/LDD 3/) , "Chapter 15: Memory Mapping and DMA", "Mapping Memory with nopage".

A.3

I did not know this information.
Thank you very much.
I am going to use it as an example.

[igalkr]

Thank you very much for your time and explanations. Igal Igal Kroyter

…

On Mon, Jun 26, 2017 at 7:28 AM, KAWAZOME Ichiro ***@***.***> wrote: Thank you for the issue. A.1 Originally I made udmabuf assuming ARM architecture only. At that time, cache coherency control between buffer and CPU was necessary. For that reason I did not use dma_mmap_coherent (), because I did not know how to specify cache coherency control with dma_mmap_coherent (). At one time I try to build on the x86 architecture as well as the ARM architecture. At that time, the following two problems occurred. 1. pgprot_dmacoherent () was undefined on x86, pgprot_dmacoherent () seems to be for ARM only. 2. udmabuf used on-demand paging mechanism for page mapping, but it did not work well on x86. So I changed it so that I do not use on-demand paging outside the ARM architecture. However, as a side effect, caching coherency control can not be controlled outside the ARM architecture. That's why we switched the code with the definition of SYNC_ENABLE. A.2 udmabuf_driver_vma_fault() referred to [Linux Device Drivers, Third Edition] (https://lwn.net/Kernel/LDD 3/) , "Chapter 15: Memory Mapping and DMA", "Mapping Memory with nopage". A.3 I did not know this information. Thank you very much. I am going to use it as an example. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#2 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AcTDZtrz_mEo9XE-1AV7D-bWeEGGhKhFks5sHzN-gaJpZM4OEQp2> .