Re: [PATCH 01/26] drm/i915: Keep engine alive as we retire the context

[Mika Kuoppala <mika.kuoppala@linux.intel.com>]

Chris Wilson <chris@chris-wilson.co.uk> writes:

> Though we pin the context first before taking the pm wakeref, during
> retire we need to unpin before dropping the pm wakeref (breaking the
> "natural" onion). During the unpin, we may need to attach a cleanup
> operation on to the engine wakeref, ergo we want to keep the engine
> awake until after the unpin.
>
> v2: Push the engine wakeref into the barrier so we keep the onion unwind
> ordering in the request itself
>
> Fixes: ce476c80b8bf ("drm/i915: Keep contexts pinned until after the next kernel context switch")
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
>  drivers/gpu/drm/i915/i915_active.c | 22 ++++++++++++++++++++--
>  1 file changed, 20 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
> index 2d019ac6db20..41ed2798687d 100644
> --- a/drivers/gpu/drm/i915/i915_active.c
> +++ b/drivers/gpu/drm/i915/i915_active.c
> @@ -4,6 +4,8 @@
>   * Copyright © 2019 Intel Corporation
>   */
>  
> +#include "gt/intel_engine_pm.h"
> +
>  #include "i915_drv.h"
>  #include "i915_active.h"
>  #include "i915_globals.h"
> @@ -268,8 +270,9 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
>  					    struct intel_engine_cs *engine)
>  {
>  	struct drm_i915_private *i915 = engine->i915;
> +	struct llist_node *pos, *next;
>  	unsigned long tmp;
> -	int err = 0;
> +	int err;
>  
>  	GEM_BUG_ON(!engine->mask);
>  	for_each_engine_masked(engine, i915, engine->mask, tmp) {
> @@ -279,7 +282,7 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
>  		node = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
>  		if (unlikely(!node)) {
>  			err = -ENOMEM;
> -			break;
> +			goto unwind;
>  		}
>  
>  		i915_active_request_init(&node->base,
> @@ -288,10 +291,24 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
>  		node->ref = ref;
>  		ref->count++;
>  
> +		intel_engine_pm_get(engine);
>  		llist_add((struct llist_node *)&node->base.link,
>  			  &ref->barriers);
>  	}
>  
> +	return 0;
> +
> +unwind:
> +	llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
> +		struct active_node *node;
> +
> +		node = container_of((struct list_head *)pos,
> +				    typeof(*node), base.link);
> +		engine = (void *)rcu_access_pointer(node->base.request);
> +
> +		intel_engine_pm_put(engine);
> +		kfree(node);
> +	}
>  	return err;
>  }
>  
> @@ -328,6 +345,7 @@ void i915_active_acquire_barrier(struct i915_active *ref)
>  
>  		llist_add((struct llist_node *)&node->base.link,
>  			  &engine->barrier_tasks);
> +		intel_engine_pm_put(engine);

But you said that you keep the wakeref until unpin?
-Mika


>  	}
>  	i915_active_release(ref);
>  }
> -- 
> 2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx