Commit 13f1bfd3 authored by Chris Wilson's avatar Chris Wilson
Browse files

drm/i915: Make object/vma allocation caches global 



As our allocations are not device specific, we can move our slab caches
to a global scope.

Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190228102035.5857-2-chris@chris-wilson.co.uk
parent 32eb6bcf

drivers/gpu/drm/i915/gvt/dmabuf.c

+1 −1
Original line number Diff line number Diff line
@@ -153,7 +153,7 @@ static struct drm_i915_gem_object *vgpu_create_gem(struct drm_device *dev, 
	struct drm_i915_private *dev_priv = to_i915(dev);

	struct drm_i915_gem_object *obj;



	obj = i915_gem_object_alloc(dev_priv);

	obj = i915_gem_object_alloc();

	if (obj == NULL)

		return NULL;

drivers/gpu/drm/i915/i915_drv.h

+0 −6
Original line number Diff line number Diff line
@@ -1470,10 +1470,6 @@ struct intel_cdclk_state { 
struct drm_i915_private {

	struct drm_device drm;



	struct kmem_cache *objects;

	struct kmem_cache *vmas;

	struct kmem_cache *luts;



	const struct intel_device_info __info; /* Use INTEL_INFO() to access. */

	struct intel_runtime_info __runtime; /* Use RUNTIME_INFO() to access. */

	struct intel_driver_caps caps;
@@ -2802,8 +2798,6 @@ void i915_gem_load_init_fences(struct drm_i915_private *dev_priv); 
int i915_gem_freeze(struct drm_i915_private *dev_priv);

int i915_gem_freeze_late(struct drm_i915_private *dev_priv);



void *i915_gem_object_alloc(struct drm_i915_private *dev_priv);

void i915_gem_object_free(struct drm_i915_gem_object *obj);

void i915_gem_object_init(struct drm_i915_gem_object *obj,

			 const struct drm_i915_gem_object_ops *ops);

struct drm_i915_gem_object *

drivers/gpu/drm/i915/i915_gem.c

+3 −44
Original line number Diff line number Diff line
@@ -624,17 +624,6 @@ i915_gem_phys_pwrite(struct drm_i915_gem_object *obj, 
	return 0;

}



void *i915_gem_object_alloc(struct drm_i915_private *dev_priv)

{

	return kmem_cache_zalloc(dev_priv->objects, GFP_KERNEL);

}



void i915_gem_object_free(struct drm_i915_gem_object *obj)

{

	struct drm_i915_private *dev_priv = to_i915(obj->base.dev);

	kmem_cache_free(dev_priv->objects, obj);

}



static int

i915_gem_create(struct drm_file *file,

		struct drm_i915_private *dev_priv,
@@ -2895,10 +2884,6 @@ static void shrink_caches(struct drm_i915_private *i915) 
	 * filled slabs to prioritise allocating from the mostly full slabs,

	 * with the aim of reducing fragmentation.

	 */

	kmem_cache_shrink(i915->luts);

	kmem_cache_shrink(i915->vmas);

	kmem_cache_shrink(i915->objects);



	i915_globals_park();

}
@@ -3094,7 +3079,7 @@ void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file) 
		list_del(&lut->obj_link);

		list_del(&lut->ctx_link);



		kmem_cache_free(i915->luts, lut);

		i915_lut_handle_free(lut);

		__i915_gem_object_release_unless_active(obj);

	}
@@ -4199,7 +4184,7 @@ i915_gem_object_create(struct drm_i915_private *dev_priv, u64 size) 
	if (overflows_type(size, obj->base.size))

		return ERR_PTR(-E2BIG);



	obj = i915_gem_object_alloc(dev_priv);

	obj = i915_gem_object_alloc();

	if (obj == NULL)

		return ERR_PTR(-ENOMEM);
@@ -5225,19 +5210,7 @@ static void i915_gem_init__mm(struct drm_i915_private *i915) 


int i915_gem_init_early(struct drm_i915_private *dev_priv)

{

	int err = -ENOMEM;



	dev_priv->objects = KMEM_CACHE(drm_i915_gem_object, SLAB_HWCACHE_ALIGN);

	if (!dev_priv->objects)

		goto err_out;



	dev_priv->vmas = KMEM_CACHE(i915_vma, SLAB_HWCACHE_ALIGN);

	if (!dev_priv->vmas)

		goto err_objects;



	dev_priv->luts = KMEM_CACHE(i915_lut_handle, 0);

	if (!dev_priv->luts)

		goto err_vmas;

	int err;



	INIT_LIST_HEAD(&dev_priv->gt.active_rings);

	INIT_LIST_HEAD(&dev_priv->gt.closed_vma);
@@ -5262,13 +5235,6 @@ int i915_gem_init_early(struct drm_i915_private *dev_priv) 
		DRM_NOTE("Unable to create a private tmpfs mount, hugepage support will be disabled(%d).\n", err);



	return 0;



err_vmas:

	kmem_cache_destroy(dev_priv->vmas);

err_objects:

	kmem_cache_destroy(dev_priv->objects);

err_out:

	return err;

}



void i915_gem_cleanup_early(struct drm_i915_private *dev_priv)
@@ -5280,13 +5246,6 @@ void i915_gem_cleanup_early(struct drm_i915_private *dev_priv) 


	cleanup_srcu_struct(&dev_priv->gpu_error.reset_backoff_srcu);



	kmem_cache_destroy(dev_priv->luts);

	kmem_cache_destroy(dev_priv->vmas);

	kmem_cache_destroy(dev_priv->objects);



	/* And ensure that our DESTROY_BY_RCU slabs are truly destroyed */

	rcu_barrier();



	i915_gemfs_fini(dev_priv);

}

drivers/gpu/drm/i915/i915_gem_context.c

+34 −1
Original line number Diff line number Diff line
@@ -94,6 +94,20 @@ 


#define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1



static struct i915_global_context {

	struct kmem_cache *slab_luts;

} global;



struct i915_lut_handle *i915_lut_handle_alloc(void)

{

	return kmem_cache_alloc(global.slab_luts, GFP_KERNEL);

}



void i915_lut_handle_free(struct i915_lut_handle *lut)

{

	return kmem_cache_free(global.slab_luts, lut);

}



static void lut_close(struct i915_gem_context *ctx)

{

	struct i915_lut_handle *lut, *ln;
@@ -102,7 +116,7 @@ static void lut_close(struct i915_gem_context *ctx) 


	list_for_each_entry_safe(lut, ln, &ctx->handles_list, ctx_link) {

		list_del(&lut->obj_link);

		kmem_cache_free(ctx->i915->luts, lut);

		i915_lut_handle_free(lut);

	}



	rcu_read_lock();
@@ -1408,3 +1422,22 @@ out_unlock: 
#include "selftests/mock_context.c"

#include "selftests/i915_gem_context.c"

#endif



int __init i915_global_context_init(void)

{

	global.slab_luts = KMEM_CACHE(i915_lut_handle, 0);

	if (!global.slab_luts)

		return -ENOMEM;



	return 0;

}



void i915_global_context_shrink(void)

{

	kmem_cache_shrink(global.slab_luts);

}



void i915_global_context_exit(void)

{

	kmem_cache_destroy(global.slab_luts);

}

drivers/gpu/drm/i915/i915_gem_context.h

+8 −0
Original line number Diff line number Diff line
@@ -32,6 +32,7 @@ 
#include "i915_gem.h"

#include "i915_scheduler.h"

#include "intel_device_info.h"

#include "intel_ringbuffer.h"



struct pid;
@@ -407,4 +408,11 @@ void intel_context_init(struct intel_context *ce, 
			struct i915_gem_context *ctx,

			struct intel_engine_cs *engine);



struct i915_lut_handle *i915_lut_handle_alloc(void);

void i915_lut_handle_free(struct i915_lut_handle *lut);



int i915_global_context_init(void);

void i915_global_context_shrink(void);

void i915_global_context_exit(void);



#endif /* !__I915_GEM_CONTEXT_H__ */

CRA Git | Maintained and supported by SUSTech CRA and CCSE