diff mbox

[v2] drm/i915: Move the mb() following release-mmap into release-mmap

Message ID 1448015498-24846-1-git-send-email-chris@chris-wilson.co.uk
State New, archived
Headers show

Commit Message

Chris Wilson Nov. 20, 2015, 10:31 a.m. UTC
As paranoia, we want to ensure that the CPU's PTEs have been revoked for
the object before we return from i915_gem_release_mmap(). This allows us
to rely on there being no outstanding memory accesses and guarantees
serialisation of the code against concurrent access just by calling
i915_gem_release_mmap().

v2: Reduce the mb() into a wmb() following the revoke.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: "Goel, Akash" <akash.goel@intel.com
---
 drivers/gpu/drm/i915/i915_gem.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

Comments

Daniel Vetter Nov. 24, 2015, 1:36 p.m. UTC | #1
On Fri, Nov 20, 2015 at 10:31:38AM +0000, Chris Wilson wrote:
> As paranoia, we want to ensure that the CPU's PTEs have been revoked for
> the object before we return from i915_gem_release_mmap(). This allows us
> to rely on there being no outstanding memory accesses and guarantees
> serialisation of the code against concurrent access just by calling
> i915_gem_release_mmap().
> 
> v2: Reduce the mb() into a wmb() following the revoke.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
> Cc: "Goel, Akash" <akash.goel@intel.com
> ---
>  drivers/gpu/drm/i915/i915_gem.c | 13 ++++++++++---
>  1 file changed, 10 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index df234d00b376..09c829f38786 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1903,11 +1903,21 @@ out:
>  void
>  i915_gem_release_mmap(struct drm_i915_gem_object *obj)
>  {
> +	/* Serialisation between user GTT access and our code depends upon
> +	 * revoking the CPU's PTE whilst the mutex is held. The next user
> +	 * pagefault then has to wait until we release the mutex.
> +	 */
> +	lockdep_assert_held(&obj->base.dev->struct_mutex);

lockdep_assert_held is a nop without lockdep, that's why I prefer
WARN_ON(!mutex_is_locked). Either way:

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

> +
>  	if (!obj->fault_mappable)
>  		return;
>  
>  	drm_vma_node_unmap(&obj->base.vma_node,
>  			   obj->base.dev->anon_inode->i_mapping);
> +
> +	/* Ensure that the CPU's PTE are revoked before we return */
> +	wmb();
> +
>  	obj->fault_mappable = false;
>  }
>  
> @@ -3212,9 +3222,6 @@ static void i915_gem_object_finish_gtt(struct drm_i915_gem_object *obj)
>  	if ((obj->base.read_domains & I915_GEM_DOMAIN_GTT) == 0)
>  		return;
>  
> -	/* Wait for any direct GTT access to complete */
> -	mb();
> -
>  	old_read_domains = obj->base.read_domains;
>  	old_write_domain = obj->base.write_domain;
>  
> -- 
> 2.6.2
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Ville Syrjala Nov. 24, 2015, 1:57 p.m. UTC | #2
On Tue, Nov 24, 2015 at 02:36:43PM +0100, Daniel Vetter wrote:
> On Fri, Nov 20, 2015 at 10:31:38AM +0000, Chris Wilson wrote:
> > As paranoia, we want to ensure that the CPU's PTEs have been revoked for
> > the object before we return from i915_gem_release_mmap(). This allows us
> > to rely on there being no outstanding memory accesses and guarantees
> > serialisation of the code against concurrent access just by calling
> > i915_gem_release_mmap().
> > 
> > v2: Reduce the mb() into a wmb() following the revoke.
> > 
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
> > Cc: "Goel, Akash" <akash.goel@intel.com
> > ---
> >  drivers/gpu/drm/i915/i915_gem.c | 13 ++++++++++---
> >  1 file changed, 10 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> > index df234d00b376..09c829f38786 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -1903,11 +1903,21 @@ out:
> >  void
> >  i915_gem_release_mmap(struct drm_i915_gem_object *obj)
> >  {
> > +	/* Serialisation between user GTT access and our code depends upon
> > +	 * revoking the CPU's PTE whilst the mutex is held. The next user
> > +	 * pagefault then has to wait until we release the mutex.
> > +	 */
> > +	lockdep_assert_held(&obj->base.dev->struct_mutex);
> 
> lockdep_assert_held is a nop without lockdep, that's why I prefer
> WARN_ON(!mutex_is_locked). Either way:

I thought someone proposed adding i915_assert_held() or something
at one point? Or did I just imagine it?

> 
> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> 
> > +
> >  	if (!obj->fault_mappable)
> >  		return;
> >  
> >  	drm_vma_node_unmap(&obj->base.vma_node,
> >  			   obj->base.dev->anon_inode->i_mapping);
> > +
> > +	/* Ensure that the CPU's PTE are revoked before we return */
> > +	wmb();
> > +
> >  	obj->fault_mappable = false;
> >  }
> >  
> > @@ -3212,9 +3222,6 @@ static void i915_gem_object_finish_gtt(struct drm_i915_gem_object *obj)
> >  	if ((obj->base.read_domains & I915_GEM_DOMAIN_GTT) == 0)
> >  		return;
> >  
> > -	/* Wait for any direct GTT access to complete */
> > -	mb();
> > -
> >  	old_read_domains = obj->base.read_domains;
> >  	old_write_domain = obj->base.write_domain;
> >  
> > -- 
> > 2.6.2
> > 
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx@lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> 
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> http://blog.ffwll.ch
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index df234d00b376..09c829f38786 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1903,11 +1903,21 @@  out:
 void
 i915_gem_release_mmap(struct drm_i915_gem_object *obj)
 {
+	/* Serialisation between user GTT access and our code depends upon
+	 * revoking the CPU's PTE whilst the mutex is held. The next user
+	 * pagefault then has to wait until we release the mutex.
+	 */
+	lockdep_assert_held(&obj->base.dev->struct_mutex);
+
 	if (!obj->fault_mappable)
 		return;
 
 	drm_vma_node_unmap(&obj->base.vma_node,
 			   obj->base.dev->anon_inode->i_mapping);
+
+	/* Ensure that the CPU's PTE are revoked before we return */
+	wmb();
+
 	obj->fault_mappable = false;
 }
 
@@ -3212,9 +3222,6 @@  static void i915_gem_object_finish_gtt(struct drm_i915_gem_object *obj)
 	if ((obj->base.read_domains & I915_GEM_DOMAIN_GTT) == 0)
 		return;
 
-	/* Wait for any direct GTT access to complete */
-	mb();
-
 	old_read_domains = obj->base.read_domains;
 	old_write_domain = obj->base.write_domain;