[Freedreno] [PATCH 2/2] drm/msm: subclass work object for vblank events

[Jeykumar Sankaran]

On 2018-11-05 09:24, Sean Paul wrote:
> On Fri, Nov 02, 2018 at 04:38:48PM -0700, Jeykumar Sankaran wrote:
>> On 2018-11-01 12:18, Sean Paul wrote:
>> > On Wed, Oct 31, 2018 at 05:19:05PM -0700, Jeykumar Sankaran wrote:
>> > > msm maintains a separate structure to define vblank
>> > > work definitions and a list to track events submitted
>> > > to the display worker thread. We can avoid these
>> > > redundant list and its protection mechanism, if we
>> > > subclass the work object to encapsulate vblank
>> > > event parameters.
>> > >
>> > > Signed-off-by: Jeykumar Sankaran <jsanka at codeaurora.org>
>> > > ---
>> > >  drivers/gpu/drm/msm/msm_drv.c | 70
>> > ++++++++++++-------------------------------
>> > >  drivers/gpu/drm/msm/msm_drv.h |  7 -----
>> > >  2 files changed, 19 insertions(+), 58 deletions(-)
>> > >
>> > > diff --git a/drivers/gpu/drm/msm/msm_drv.c
>> > b/drivers/gpu/drm/msm/msm_drv.c
>> > > index 1f384b3..67a96ee 100644
>> > > --- a/drivers/gpu/drm/msm/msm_drv.c
>> > > +++ b/drivers/gpu/drm/msm/msm_drv.c
> 
> /snip
> 
>> > >  static int vblank_ctrl_queue_work(struct msm_drm_private *priv,
>> > >  					int crtc_id, bool enable)
>> > >  {
>> > > -	struct msm_vblank_ctrl *vbl_ctrl = &priv->vblank_ctrl;
>> > > -	struct vblank_event *vbl_ev;
>> > > -	unsigned long flags;
>> > > +	struct msm_vblank_work *vbl_work;
>> > >
>> > > -	vbl_ev = kzalloc(sizeof(*vbl_ev), GFP_ATOMIC);
>> > > -	if (!vbl_ev)
>> > > +	vbl_work = kzalloc(sizeof(*vbl_work), GFP_ATOMIC);
>> > > +	if (!vbl_work)
>> > >  		return -ENOMEM;
>> > >
>> > > -	vbl_ev->crtc_id = crtc_id;
>> > > -	vbl_ev->enable = enable;
>> > > +	kthread_init_work(&vbl_work->work, vblank_ctrl_worker);
>> > >
>> > > -	spin_lock_irqsave(&vbl_ctrl->lock, flags);
>> > > -	list_add_tail(&vbl_ev->node, &vbl_ctrl->event_list);
>> > > -	spin_unlock_irqrestore(&vbl_ctrl->lock, flags);
>> > > +	vbl_work->crtc_id = crtc_id;
>> > > +	vbl_work->enable = enable;
>> > > +	vbl_work->priv = priv;
>> > >
>> > > -	kthread_queue_work(&priv->disp_thread.worker, &vbl_ctrl->work);
>> > > +	kthread_queue_work(&priv->disp_thread.worker, &vbl_work->work);
>> >
>> > So I think this can get even more simplified. In the short term, you
> can
>> > just
>> > use the systemwq to do the enable and disable.
>> 
>> you mean priv->wq?
>> 
> 
> I meant the system workqueue, we probably don't need our own for this.
> 
> 
>> >
>> > In the long term, the enable_vblank/disable_vblank functions should be
>> > optimized so they don't sleep. I took a quick look at them perhaps
> this
>> > is
>> > all because of the crtc_lock mutex? That lock seems a bit suspicious
> to
>> > me,
>> > especially being dropped around the pm_runtime calls in
>> > _dpu_crtc_vblank_enable_no_lock(). I think we could probably rely on
> the
>> > modeset
>> > locks for some of these functions, and perhaps convert it to a
> spinlock
>> > if
>> > we
>> > can't get rid of it entirely.
>> 
>> crtc_lock has a history of usage in the downstream driver. It was
> introduced
>> to protect
>> vblank variables when vblank requests were handled in the user thread
>> (not the display thread). When event threads were introduced to 
>> receive
>> encoder events, the lock was further expanded to protect few more 
>> vars.
> It
>> was
>> also needed to synchronize CRTC accesses between debugfs dump calls
>> and display thread.
> 
> The debugfs case can be solved pretty easily by using the modeset 
> locks. I
> haven't looked closely at the event threads, could we convert crtc_lock 
> to
> a
> spinlock and then make vblank enable/disable synchronous?
Did a little digging into the reason why vblank enable/disable was made 
asynchronous
in the first place. Looks like Rob was also using priv->wq to queue 
vblank requests
before display threads were introduced by the DPU driver.

The only reason I can think of was to support smart panels, where we 
wait for
CTL_START interrupt instead of PING_PONG_DONE which is needed for fence
releases. Need to confirm with Rob for MDP5 behaviour before switcing to 
sync.

For now, I submit a patch to use system wq.
> 
> Sean
> 
>> 
>> Would like to deal with this cleanup bit later once we lose these 
>> extra
>> threads.
>> 
>> Thanks and Regards,
>> Jeykumar S.
>> 
>> >
>> > Sean
>> >
>> > >
>> > >  	return 0;
>> > >  }
>> > > @@ -269,20 +252,8 @@ static int msm_drm_uninit(struct device *dev)
>> > >  	struct msm_drm_private *priv = ddev->dev_private;
>> > >  	struct msm_kms *kms = priv->kms;
>> > >  	struct msm_mdss *mdss = priv->mdss;
>> > > -	struct msm_vblank_ctrl *vbl_ctrl = &priv->vblank_ctrl;
>> > > -	struct vblank_event *vbl_ev, *tmp;
>> > >  	int i;
>> > >
>> > > -	/* We must cancel and cleanup any pending vblank enable/disable
>> > > -	 * work before drm_irq_uninstall() to avoid work re-enabling an
>> > > -	 * irq after uninstall has disabled it.
>> > > -	 */
>> > > -	kthread_flush_work(&vbl_ctrl->work);
>> > > -	list_for_each_entry_safe(vbl_ev, tmp, &vbl_ctrl->event_list, node)
>> > {
>> > > -		list_del(&vbl_ev->node);
>> > > -		kfree(vbl_ev);
>> > > -	}
>> > > -
>> > >  	kthread_flush_worker(&priv->disp_thread.worker);
>> > >  	kthread_stop(priv->disp_thread.thread);
>> > >  	priv->disp_thread.thread = NULL;
>> > > @@ -474,9 +445,6 @@ static int msm_drm_init(struct device *dev,
> struct
>> > drm_driver *drv)
>> > >  	priv->wq = alloc_ordered_workqueue("msm", 0);
>> > >
>> > >  	INIT_LIST_HEAD(&priv->inactive_list);
>> > > -	INIT_LIST_HEAD(&priv->vblank_ctrl.event_list);
>> > > -	kthread_init_work(&priv->vblank_ctrl.work, vblank_ctrl_worker);
>> > > -	spin_lock_init(&priv->vblank_ctrl.lock);
>> > >
>> > >  	drm_mode_config_init(ddev);
>> > >
>> > > diff --git a/drivers/gpu/drm/msm/msm_drv.h
>> > b/drivers/gpu/drm/msm/msm_drv.h
>> > > index e81b1fa..b91e306 100644
>> > > --- a/drivers/gpu/drm/msm/msm_drv.h
>> > > +++ b/drivers/gpu/drm/msm/msm_drv.h
>> > > @@ -77,12 +77,6 @@ enum msm_mdp_plane_property {
>> > >  	PLANE_PROP_MAX_NUM
>> > >  };
>> > >
>> > > -struct msm_vblank_ctrl {
>> > > -	struct kthread_work work;
>> > > -	struct list_head event_list;
>> > > -	spinlock_t lock;
>> > > -};
>> > > -
>> > >  #define MSM_GPU_MAX_RINGS 4
>> > >  #define MAX_H_TILES_PER_DISPLAY 2
>> > >
>> > > @@ -226,7 +220,6 @@ struct msm_drm_private {
>> > >  	struct notifier_block vmap_notifier;
>> > >  	struct shrinker shrinker;
>> > >
>> > > -	struct msm_vblank_ctrl vblank_ctrl;
>> > >  	struct drm_atomic_state *pm_state;
>> > >  };
>> > >
>> > > --
>> > > The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
>> > Forum,
>> > > a Linux Foundation Collaborative Project
>> > >
>> > > _______________________________________________
>> > > Freedreno mailing list
>> > > Freedreno at lists.freedesktop.org
>> > > https://lists.freedesktop.org/mailman/listinfo/freedreno
>> 
>> --
>> Jeykumar S

-- 
Jeykumar S