mbox series

[00/45] TTM move refactoring

Message ID 20200924051845.397177-1-airlied@gmail.com
Headers show
Series TTM move refactoring | expand

Message

Dave Airlie Sept. 24, 2020, 5:18 a.m. UTC
This refactors how TTM moves get called between core and drivers.

1) puts the driver in charge of all moves, and enforces
drivers have a move callback.
2) moves move_notify actions around moves to the driver side
3) moves binding/unbinding completely into move and driver side
4) add a new invalidate callback to replace the last use of move_notify
5) add some helpers to cleanup the resulting move code

There's a bit of internal churn to try and make each patch logical, so
don't get too caught up in each patches changes unless the end result
is a problem.

Dave.

Comments

Dave Airlie Sept. 24, 2020, 6:40 a.m. UTC | #1
On Thu, 24 Sep 2020 at 15:19, Dave Airlie <airlied@gmail.com> wrote:
>
> This refactors how TTM moves get called between core and drivers.
>

Full series is also:

https://github.com/airlied/linux/tree/ttm-no-more-bind

(it's based on top of Christian's last series).

Dave.
Ben Skeggs Sept. 30, 2020, 5:51 a.m. UTC | #2
On Thu, 24 Sep 2020 at 15:19, Dave Airlie <airlied@gmail.com> wrote:
>
> This refactors how TTM moves get called between core and drivers.
>
> 1) puts the driver in charge of all moves, and enforces
> drivers have a move callback.
> 2) moves move_notify actions around moves to the driver side
> 3) moves binding/unbinding completely into move and driver side
> 4) add a new invalidate callback to replace the last use of move_notify
> 5) add some helpers to cleanup the resulting move code
>
> There's a bit of internal churn to try and make each patch logical, so
> don't get too caught up in each patches changes unless the end result
> is a problem.
I've looked over the nouveau-specific patches, and the ttm ones
(mostly ignoring the changes to other drivers beyond a quick glance
over).  For all but 37/45 and the patches that depend on it:

Reviewed-by: Ben Skeggs <bskeggs@redhat.com>

I'll add some more specific comments to one of the patches, but it's
also fine as-is.

Ben.

>
> Dave.
>
>
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel