Re: i915 ttm_tt shmem backend

From: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>
To: "Matthew Auld" <matthew.william.auld@gmail.com>,
	"Christian König" <ckoenig.leichtzumerken@gmail.com>,
	"Christian König" <christian.koenig@amd.com>
Cc: ML dri-devel <dri-devel@lists.freedesktop.org>
Subject: Re: i915 ttm_tt shmem backend
Date: Fri, 10 Sep 2021 09:46:38 +0200	[thread overview]
Message-ID: <c5edb73d-72fc-6cef-2735-9e6af55a8bd9@linux.intel.com> (raw)
In-Reply-To: <CAM0jSHO+vBwJmB=bsYcdjzHFfnaLYSDrPYWNUmEe1BqmrVOBxA@mail.gmail.com>

Hi,

On 9/9/21 4:56 PM, Matthew Auld wrote:
> Hi Christian,
>
> We are looking into using shmem as a ttm_tt backend in i915 for cached
> system memory objects. We would also like to make such objects visible
> to the i915-gem shrinker, so that they may be swapped out or discarded
> when under memory pressure.
>
> One idea for handling this is roughly something like:
> - Add a new TTM_PAGE_FLAG_SHMEM flag, or similar.
> - Skip the ttm_pages_allocated accounting on such objects, similar to
> how FLAG_SG is already handled.
> - Skip all the page->mapping and page->index related bits, like in
> tt_add_mapping, since it looks like these are set and used by shmem.
> Not sure what functionally this might break, but looks like it's maybe
> only driver specific?

IIrc the page->mapping and index is needed when doing dirty-tracking 
using mkwrite and by vmwgfx at some point when doing fb_defio on top of 
TTM buffers. I don't think vmwgfx does that anymore, but it still does 
dirty-tracking.

/Thomas