Re: [PATCH 2/2] mm/frame-vec: use FOLL_LONGTERM

[Jason Gunthorpe]

On Wed, Oct 07, 2020 at 03:34:01PM +0200, Tomasz Figa wrote:

> I think the userptr zero-copy hack should be able to go away indeed,
> given that we now have CMA that allows having carveouts backed by
> struct pages and having the memory represented as DMA-buf normally.

This also needs to figure out how to get references to CMA pages out
of a VMA. IIRC Daniel said these are not pinnable?

> How about the regular userptr use case, though?

Just call pin_user_pages(), that is the easy case.

> Is your intention to drop get_vaddr_frames() or we could still keep
> using it and if vec->is_pfns is true:

get_vaddr_frames() is dangerous, I would like it to go away.

> a) if CONFIG_VIDEO_LEGACY_PFN_USERPTR is set, taint the kernel
> b) otherwise just undo and fail?

For the CONFIG_VIDEO_LEGACY_PFN_USERPTR case all the follow_pfn
related code in get_vaddr_frames() shold move back into media and be
hidden under this config.

Jason