| Message ID | E1ZgzLt-0003lG-Aq@rmk-PC.arm.linux.org.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Russell, On 29 September 2015 at 19:10, Russell King <rmk+kernel@arm.linux.org.uk> wrote: > Rather than using a spinlock, use xchg() to atomically update > dplane->old_fb. This allows us to eliminate dplane->lock. > > Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> > --- [...] > @@ -76,13 +75,10 @@ static void armada_ovl_retire_fb(struct armada_ovl_plane *dplane, > { > struct drm_framebuffer *old_fb; > > - spin_lock(&dplane->lock); > - old_fb = dplane->old_fb; > - dplane->old_fb = fb; > - spin_unlock(&dplane->lock); > + old_fb = xchg(&dplane->old_fb, fb); > > if (old_fb) > - armada_drm_queue_unref_work(dplane->base.dev, old_fb); > + armada_drm_queue_unref_work(dplane->base.base.dev, old_fb); Shouldn't this be part of another patch ? -Emil
On Wed, Sep 30, 2015 at 11:49:44AM +0100, Emil Velikov wrote: > Hi Russell, > > On 29 September 2015 at 19:10, Russell King <rmk+kernel@arm.linux.org.uk> wrote: > > Rather than using a spinlock, use xchg() to atomically update > > dplane->old_fb. This allows us to eliminate dplane->lock. > > > > Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> > > --- > [...] > > @@ -76,13 +75,10 @@ static void armada_ovl_retire_fb(struct armada_ovl_plane *dplane, > > { > > struct drm_framebuffer *old_fb; > > > > - spin_lock(&dplane->lock); > > - old_fb = dplane->old_fb; > > - dplane->old_fb = fb; > > - spin_unlock(&dplane->lock); > > + old_fb = xchg(&dplane->old_fb, fb); > > > > if (old_fb) > > - armada_drm_queue_unref_work(dplane->base.dev, old_fb); > > + armada_drm_queue_unref_work(dplane->base.base.dev, old_fb); > Shouldn't this be part of another patch ? Yes, you're right, it should be in: drm/armada: introduce generic armada_plane struct Now moved there, thanks.
diff --git a/drivers/gpu/drm/armada/armada_overlay.c b/drivers/gpu/drm/armada/armada_overlay.c index 093c2d4f2b79..8738b590abc2 100644 --- a/drivers/gpu/drm/armada/armada_overlay.c +++ b/drivers/gpu/drm/armada/armada_overlay.c @@ -31,7 +31,6 @@ struct armada_ovl_plane_properties { struct armada_ovl_plane { struct drm_plane base; - spinlock_t lock; struct drm_framebuffer *old_fb; uint32_t src_hw; uint32_t dst_hw; @@ -76,13 +75,10 @@ static void armada_ovl_retire_fb(struct armada_ovl_plane *dplane, { struct drm_framebuffer *old_fb; - spin_lock(&dplane->lock); - old_fb = dplane->old_fb; - dplane->old_fb = fb; - spin_unlock(&dplane->lock); + old_fb = xchg(&dplane->old_fb, fb); if (old_fb) - armada_drm_queue_unref_work(dplane->base.dev, old_fb); + armada_drm_queue_unref_work(dplane->base.base.dev, old_fb); } /* === Plane support === */ @@ -289,10 +285,7 @@ static int armada_ovl_plane_disable(struct drm_plane *plane) if (plane->fb) drm_framebuffer_unreference(plane->fb); - spin_lock_irq(&dplane->lock); - fb = dplane->old_fb; - dplane->old_fb = NULL; - spin_unlock_irq(&dplane->lock); + fb = xchg(&dplane->old_fb, NULL); if (fb) drm_framebuffer_unreference(fb); @@ -464,7 +457,6 @@ int armada_overlay_plane_create(struct drm_device *dev, unsigned long crtcs) if (!dplane) return -ENOMEM; - spin_lock_init(&dplane->lock); init_waitqueue_head(&dplane->vbl.wait); armada_drm_vbl_event_init(&dplane->vbl.update, armada_ovl_plane_vbl, dplane);
Rather than using a spinlock, use xchg() to atomically update dplane->old_fb. This allows us to eliminate dplane->lock. Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> --- drivers/gpu/drm/armada/armada_overlay.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-)