Re: [PATCH 34/49] gma500: the GEM and GTT code is device independant

[Rob Clark]

On Mon, Oct 17, 2011 at 5:34 PM, Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> On Mon, 17 Oct 2011, Alan Cox wrote:
>> > It feels to me like GEM is pulling shmem in an ever more alien direction:
>> > these device constraints are so foreign to the nature of tmpfs; and
>> > beyond my expertise, so that I'd be ever more likely to make the wrong
>> > decisions (mixing swap and uncached pages? hmmm).
>>
>> For the most part we fixed that. You can now have a GEM object that is
>> backed by a private memory object rather than having to be tmpfs.
>> GMA500 uses it to attach 'stolen' memory to GEM handles, and at least
>> one other pending submission uses it with a private CMA style allocator.
>
> Sounds good.

Well, that means we could implement our own shmem/tmpfs.. although I'm
not entirely sure that I would want to..  a couple drm drivers are
using private memory objects, but I think primarily for scanout
buffers (which can't really be swapped out anyways).  I don't think
we'd want to use this for everything.

What troubles could you see for swap+uncached (or more likely,
writecombine) pages?

BR,
-R

>>
>> The gma500 report seems an odd one - no GMA500 box has >4GB memory so how
>> did the test code get a page that was unsuitable - is the test buggy ?
>
> This is the first I heard of some gma500 test seeing a problem with >4GB.
>
> In this thread we have Patrik hitting the oops in read_cache_page_gfp()
> because of some changes he is preparing, but I thought he and Rob were
> just thinking ahead when they raise the 4GB issue.
>
> Can you point us to another thread on a another list?
>
> Hugh
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/