[RFC PATCH 3/5] mm/vma: add support for peer to peer to device vma

[Logan Gunthorpe]

On 2019-01-30 12:59 p.m., Jason Gunthorpe wrote:
> On Wed, Jan 30, 2019 at 12:45:46PM -0700, Logan Gunthorpe wrote:
>>
>>
>> On 2019-01-30 12:06 p.m., Jason Gunthorpe wrote:
>>>> Way less problems than not having struct page for doing anything
>>>> non-trivial.  If you map the BAR to userspace with remap_pfn_range
>>>> and friends the mapping is indeed very simple.  But any operation
>>>> that expects a page structure, which is at least everything using
>>>> get_user_pages won't work.
>>>
>>> GUP doesn't work anyhow today, and won't work with BAR struct pages in
>>> the forseeable future (Logan has sent attempts on this before).
>>
>> I don't recall ever attempting that... But patching GUP for special
>> pages or VMAS; or working around by not calling it in some cases seems
>> like the thing that's going to need to be done one way or another.
> 
> Remember, the long discussion we had about how to get the IOMEM
> annotation into SGL? That is a necessary pre-condition to doing
> anything with GUP in DMA using drivers as GUP -> SGL -> DMA map is
> pretty much the standard flow.

Yes, but that was unrelated to GUP even if that might have been the
eventual direction.

And I feel the GUP->SGL->DMA flow should still be what we are aiming
for. Even if we need a special GUP for special pages, and a special DMA
map; and the SGL still has to be homogenous....

> So, I see Jerome solving the GUP problem by replacing GUP entirely
> using an API that is more suited to what these sorts of drivers
> actually need.

Yes, this is what I'm expecting and what I want. Not bypassing the whole
thing by doing special things with VMAs.

Logan