[PATCH v2 1/3] drm/i915/shrinker: Only report objects with extra pinned pages as pinned

[PATCH v2 1/3] drm/i915/shrinker: Only report objects with extra pinned pages as pinned

[Chris Wilson]

When iterating over the bound list, we expect all objects there to have
their pages pinned (by the bound VMA). So only report those objects with
additional pin count on their pages as "pinned". These should be those
objects used for display and hardware access.

Reported-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Akash Goel <akash.goel@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_shrinker.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index d46388f25e04..4e4fcfa76b4c 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -361,7 +361,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
 		if (!obj->base.filp)
 			continue;
 
-		if (obj->pages_pin_count)
+		if (obj->pages_pin_count > num_vma_bound(obj))
 			pinned += obj->base.size;
 		else
 			bound += obj->base.size;
-- 
2.8.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 2/3] drm/i915/shrinker: Report "unevictable" pages

[Chris Wilson]

Inside the shrinker we call can_release_pages() to indicate whether or
not we can make forward progress in freeing up memory by unbinding that
object. When adding our report to oom, we should be using the same
logic.

Whilst here, change the reporting from bytes to pages so that it looks
smaller to the user!, is consistent with the neighbouring oom report
itself which displays counts in pages, and makes the unsigned long
overflow less likely.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_shrinker.c | 22 +++++++++++-----------
 1 file changed, 11 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index 4e4fcfa76b4c..cb225e039d48 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -336,7 +336,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
 		container_of(nb, struct drm_i915_private, mm.oom_notifier);
 	struct shrinker_lock_uninterruptible slu;
 	struct drm_i915_gem_object *obj;
-	unsigned long pinned, bound, unbound, freed_pages;
+	unsigned long unevictable, bound, unbound, freed_pages;
 
 	if (!i915_gem_shrinker_lock_uninterruptible(dev_priv, &slu, 5000))
 		return NOTIFY_DONE;
@@ -347,33 +347,33 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
 	 * assert that there are no objects with pinned pages that are not
 	 * being pointed to by hardware.
 	 */
-	unbound = bound = pinned = 0;
+	unbound = bound = unevictable = 0;
 	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
 		if (!obj->base.filp) /* not backed by a freeable object */
 			continue;
 
-		if (obj->pages_pin_count)
-			pinned += obj->base.size;
+		if (!can_release_pages(obj))
+			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
-			unbound += obj->base.size;
+			unbound += obj->base.size >> PAGE_SHIFT;
 	}
 	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
 		if (!obj->base.filp)
 			continue;
 
-		if (obj->pages_pin_count > num_vma_bound(obj))
-			pinned += obj->base.size;
+		if (!can_release_pages(obj))
+			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
-			bound += obj->base.size;
+			bound += obj->base.size >> PAGE_SHIFT;
 	}
 
 	i915_gem_shrinker_unlock_uninterruptible(dev_priv, &slu);
 
 	if (freed_pages || unbound || bound)
-		pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n",
-			freed_pages << PAGE_SHIFT, pinned);
+		pr_info("Purging GPU memory, %lu pages freed, %lu pages still pinned.\n",
+			freed_pages, unevictable);
 	if (unbound || bound)
-		pr_err("%lu and %lu bytes still available in the "
+		pr_err("%lu and %lu pages still available in the "
 		       "bound and unbound GPU page lists.\n",
 		       bound, unbound);
 
-- 
2.8.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 3/3] drm/i915/shrinker: Only shmemfs objects are backed by swap

[Chris Wilson]

Since we can only swap out shmemfs objects, those are the only ones that
influence the ability of the shrinker to can free pages. Currently, all
non-shmemfs objects have a raised pages_pin_count to protect them from
the shrinker, so this just makes the logic for can_release_pages()
clearer (and safer in future so that we don't over estimate our ability
to free up pages from future non-swappable objects).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_shrinker.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index cb225e039d48..e44c6358bd5a 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -70,6 +70,10 @@ static bool swap_available(void)
 
 static bool can_release_pages(struct drm_i915_gem_object *obj)
 {
+	/* Only shmemfs objects are backed by swapped */
+	if (!obj->base.filp)
+		return false;
+
 	/* Only report true if by unbinding the object and putting its pages
 	 * we can actually make forward progress towards freeing physical
 	 * pages.
@@ -349,18 +353,12 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
 	 */
 	unbound = bound = unevictable = 0;
 	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
-		if (!obj->base.filp) /* not backed by a freeable object */
-			continue;
-
 		if (!can_release_pages(obj))
 			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
 			unbound += obj->base.size >> PAGE_SHIFT;
 	}
 	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
-		if (!obj->base.filp)
-			continue;
-
 		if (!can_release_pages(obj))
 			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
-- 
2.8.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 2/3] drm/i915/shrinker: Report "unevictable" pages

[Mika Kuoppala]

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

> [ text/plain ]
> Inside the shrinker we call can_release_pages() to indicate whether or
> not we can make forward progress in freeing up memory by unbinding that
> object. When adding our report to oom, we should be using the same
> logic.
>
> Whilst here, change the reporting from bytes to pages so that it looks
> smaller to the user!, is consistent with the neighbouring oom report
> itself which displays counts in pages, and makes the unsigned long
> overflow less likely.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

> ---
>  drivers/gpu/drm/i915/i915_gem_shrinker.c | 22 +++++++++++-----------
>  1 file changed, 11 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> index 4e4fcfa76b4c..cb225e039d48 100644
> --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
> +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> @@ -336,7 +336,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
>  		container_of(nb, struct drm_i915_private, mm.oom_notifier);
>  	struct shrinker_lock_uninterruptible slu;
>  	struct drm_i915_gem_object *obj;
> -	unsigned long pinned, bound, unbound, freed_pages;
> +	unsigned long unevictable, bound, unbound, freed_pages;
>  
>  	if (!i915_gem_shrinker_lock_uninterruptible(dev_priv, &slu, 5000))
>  		return NOTIFY_DONE;
> @@ -347,33 +347,33 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
>  	 * assert that there are no objects with pinned pages that are not
>  	 * being pointed to by hardware.
>  	 */
> -	unbound = bound = pinned = 0;
> +	unbound = bound = unevictable = 0;
>  	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
>  		if (!obj->base.filp) /* not backed by a freeable object */
>  			continue;
>  
> -		if (obj->pages_pin_count)
> -			pinned += obj->base.size;
> +		if (!can_release_pages(obj))
> +			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
> -			unbound += obj->base.size;
> +			unbound += obj->base.size >> PAGE_SHIFT;
>  	}
>  	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
>  		if (!obj->base.filp)
>  			continue;
>  
> -		if (obj->pages_pin_count > num_vma_bound(obj))
> -			pinned += obj->base.size;
> +		if (!can_release_pages(obj))
> +			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
> -			bound += obj->base.size;
> +			bound += obj->base.size >> PAGE_SHIFT;
>  	}
>  
>  	i915_gem_shrinker_unlock_uninterruptible(dev_priv, &slu);
>  
>  	if (freed_pages || unbound || bound)
> -		pr_info("Purging GPU memory, %lu bytes freed, %lu bytes still pinned.\n",
> -			freed_pages << PAGE_SHIFT, pinned);
> +		pr_info("Purging GPU memory, %lu pages freed, %lu pages still pinned.\n",
> +			freed_pages, unevictable);
>  	if (unbound || bound)
> -		pr_err("%lu and %lu bytes still available in the "
> +		pr_err("%lu and %lu pages still available in the "
>  		       "bound and unbound GPU page lists.\n",
>  		       bound, unbound);
>  
> -- 
> 2.8.1
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 3/3] drm/i915/shrinker: Only shmemfs objects are backed by swap

[Mika Kuoppala]

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

> [ text/plain ]
> Since we can only swap out shmemfs objects, those are the only ones that
> influence the ability of the shrinker to can free pages. Currently, all
> non-shmemfs objects have a raised pages_pin_count to protect them from
> the shrinker, so this just makes the logic for can_release_pages()
> clearer (and safer in future so that we don't over estimate our ability
> to free up pages from future non-swappable objects).
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

> ---
>  drivers/gpu/drm/i915/i915_gem_shrinker.c | 10 ++++------
>  1 file changed, 4 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> index cb225e039d48..e44c6358bd5a 100644
> --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
> +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> @@ -70,6 +70,10 @@ static bool swap_available(void)
>  
>  static bool can_release_pages(struct drm_i915_gem_object *obj)
>  {
> +	/* Only shmemfs objects are backed by swapped */
> +	if (!obj->base.filp)
> +		return false;
> +
>  	/* Only report true if by unbinding the object and putting its pages
>  	 * we can actually make forward progress towards freeing physical
>  	 * pages.
> @@ -349,18 +353,12 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
>  	 */
>  	unbound = bound = unevictable = 0;
>  	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
> -		if (!obj->base.filp) /* not backed by a freeable object */
> -			continue;
> -
>  		if (!can_release_pages(obj))
>  			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
>  			unbound += obj->base.size >> PAGE_SHIFT;
>  	}
>  	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
> -		if (!obj->base.filp)
> -			continue;
> -
>  		if (!can_release_pages(obj))
>  			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
> -- 
> 2.8.1
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 3/3] drm/i915/shrinker: Only shmemfs objects are backed by swap

[Joonas Lahtinen]

On ke, 2016-04-20 at 12:04 +0100, Chris Wilson wrote:
> Since we can only swap out shmemfs objects, those are the only ones that
> influence the ability of the shrinker to can free pages. Currently, all
> non-shmemfs objects have a raised pages_pin_count to protect them from
> the shrinker, so this just makes the logic for can_release_pages()
> clearer (and safer in future so that we don't over estimate our ability
> to free up pages from future non-swappable objects).
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

> ---
>  drivers/gpu/drm/i915/i915_gem_shrinker.c | 10 ++++------
>  1 file changed, 4 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> index cb225e039d48..e44c6358bd5a 100644
> --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
> +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
> @@ -70,6 +70,10 @@ static bool swap_available(void)
>  
>  static bool can_release_pages(struct drm_i915_gem_object *obj)
>  {
> +	/* Only shmemfs objects are backed by swapped */
> +	if (!obj->base.filp)
> +		return false;
> +
>  	/* Only report true if by unbinding the object and putting its pages
>  	 * we can actually make forward progress towards freeing physical
>  	 * pages.
> @@ -349,18 +353,12 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
>  	 */
>  	unbound = bound = unevictable = 0;
>  	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
> -		if (!obj->base.filp) /* not backed by a freeable object */
> -			continue;
> -
>  		if (!can_release_pages(obj))
>  			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
>  			unbound += obj->base.size >> PAGE_SHIFT;
>  	}
>  	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
> -		if (!obj->base.filp)
> -			continue;
> -
>  		if (!can_release_pages(obj))
>  			unevictable += obj->base.size >> PAGE_SHIFT;
>  		else
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 3/3] drm/i915/shrinker: Only shmemfs objects are backed by swap

[Chris Wilson]

Since we can only swap out shmemfs objects, those are the only ones that
influence the ability of the shrinker to can free pages. Currently, all
non-shmemfs objects have a raised pages_pin_count to protect them from
the shrinker, so this just makes the logic for can_release_pages()
clearer (and safer in future so that we don't over estimate our ability
to free up pages from future non-swappable objects).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_shrinker.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index cb225e039d48..e44c6358bd5a 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -70,6 +70,10 @@ static bool swap_available(void)
 
 static bool can_release_pages(struct drm_i915_gem_object *obj)
 {
+	/* Only shmemfs objects are backed by swapped */
+	if (!obj->base.filp)
+		return false;
+
 	/* Only report true if by unbinding the object and putting its pages
 	 * we can actually make forward progress towards freeing physical
 	 * pages.
@@ -349,18 +353,12 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
 	 */
 	unbound = bound = unevictable = 0;
 	list_for_each_entry(obj, &dev_priv->mm.unbound_list, global_list) {
-		if (!obj->base.filp) /* not backed by a freeable object */
-			continue;
-
 		if (!can_release_pages(obj))
 			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
 			unbound += obj->base.size >> PAGE_SHIFT;
 	}
 	list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
-		if (!obj->base.filp)
-			continue;
-
 		if (!can_release_pages(obj))
 			unevictable += obj->base.size >> PAGE_SHIFT;
 		else
-- 
2.8.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx