On Fri, May 30, 2014 at 02:16:30PM +0100, Chris Wilson wrote:
> Fallout from
>
> commit 46470fc932ac8a0e8317a220b3f4ea4ed903338e
> Author: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx>
> Date: Wed May 21 19:01:06 2014 +0300
>
> drm/i915: Add null state batch to active list
>
> undid the earlier fix of only marking the ctx as initialised after it is
> saved by the hardware during a SET_CONTEXT operation.
>
> Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> Cc: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> Cc: Damien Lespiau <damien.lespiau@xxxxxxxxx>
> Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
> Cc: Ben Widawsky <ben@xxxxxxxxxxxx>
> ---
> drivers/gpu/drm/i915/i915_gem_context.c | 8 +++++---
> 1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 5a71ef1975b3..34a0b49e6add 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -641,6 +641,7 @@ static int do_switch(struct intel_engine_cs *ring,
> struct intel_context *from = ring->last_context;
> struct i915_hw_ppgtt *ppgtt = ctx_to_ppgtt(to);
> u32 hw_flags = 0;
> + bool uninitialized = false;
> int ret, i;
>
> if (from != NULL && ring == &dev_priv->ring[RCS]) {
> @@ -739,18 +740,19 @@ static int do_switch(struct intel_engine_cs *ring,
> i915_gem_context_unreference(from);
> }
>
> + uninitialized = !to->is_initialized && from == NULL;
> + to->is_initialized = true;
Aren't you missing some error paths if you do this? I think we need to
set the state after we've successfully emitted the MI_SET_CONTEXT
command (though really, until we move the ringbuffer tail, it's all
lies, on that note, I'm scared to look at if reset dtrt).
> +
> done:
> i915_gem_context_reference(to);
> ring->last_context = to;
>
> - if (ring->id == RCS && !to->is_initialized && from == NULL) {
> + if (uninitialized) {
> ret = i915_gem_render_state_init(ring);
> if (ret)
> DRM_ERROR("init render state: %d\n", ret);
> }
>
> - to->is_initialized = true;
> -
> return 0;
>
> unpin_out:
I just realized the from == NULL check is something which seems fragile
to me given how much churn we might see with init/reset paths. It'd
probably be good to eventually switch to some global state in dev_priv
which we can track across init/reset/fini
Otherwise it looks correct to me.
--
Ben Widawsky, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
