Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Message-ID: <2e29ba5490815f2098d9aa50bb84470aac7ba08b.camel@redhat.com>
Subject: Re: [PATCH 06/26] drm/dp_mst: Move PDT teardown for ports into
 destroy_connector_work
From:   Lyude Paul <lyude@redhat.com>
To:     Daniel Vetter <daniel@ffwll.ch>
Cc:     dri-devel@lists.freedesktop.org, Juston Li <juston.li@intel.com>,
        Imre Deak <imre.deak@intel.com>,
        Ville =?ISO-8859-1?Q?Syrj=E4l=E4?= 
        <ville.syrjala@linux.intel.com>, Harry Wentland <hwentlan@amd.com>,
        Maarten Lankhorst <maarten.lankhorst@linux.intel.com>,
        Maxime Ripard <maxime.ripard@bootlin.com>,
        Sean Paul <sean@poorly.run>, David Airlie <airlied@linux.ie>,
        linux-kernel@vger.kernel.org
Date:   Fri, 30 Aug 2019 19:46:03 -0400
In-Reply-To: <20190813145242.GW7444@phenom.ffwll.local>
References: <20190718014329.8107-1-lyude@redhat.com>
         <20190718014329.8107-7-lyude@redhat.com>
         <20190813145242.GW7444@phenom.ffwll.local>
Organization: Red Hat
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.32.4 (3.32.4-1.fc30) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, 2019-08-13 at 16:52 +0200, Daniel Vetter wrote:
> On Wed, Jul 17, 2019 at 09:42:29PM -0400, Lyude Paul wrote:
> > This will allow us to add some locking for port PDTs, which can't be
> > done from drm_dp_destroy_port() since we don't know what locks the
> > caller might be holding. Also, this gets rid of a good bit of unneeded
> > code.
> > 
> > Cc: Juston Li <juston.li@intel.com>
> > Cc: Imre Deak <imre.deak@intel.com>
> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > Cc: Harry Wentland <hwentlan@amd.com>
> > Signed-off-by: Lyude Paul <lyude@redhat.com>
> > ---
> >  drivers/gpu/drm/drm_dp_mst_topology.c | 42 +++++++++++----------------
> >  1 file changed, 17 insertions(+), 25 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c
> > b/drivers/gpu/drm/drm_dp_mst_topology.c
> > index defc5e09fb9a..0295e007c836 100644
> > --- a/drivers/gpu/drm/drm_dp_mst_topology.c
> > +++ b/drivers/gpu/drm/drm_dp_mst_topology.c
> > @@ -1509,31 +1509,22 @@ static void drm_dp_destroy_port(struct kref *kref)
> >  		container_of(kref, struct drm_dp_mst_port, topology_kref);
> >  	struct drm_dp_mst_topology_mgr *mgr = port->mgr;
> >  
> > -	if (!port->input) {
> > -		kfree(port->cached_edid);
> > -
> > -		/*
> > -		 * The only time we don't have a connector
> > -		 * on an output port is if the connector init
> > -		 * fails.
> > -		 */
> > -		if (port->connector) {
> > -			/* we can't destroy the connector here, as
> > -			 * we might be holding the mode_config.mutex
> > -			 * from an EDID retrieval */
> > -
> > -			mutex_lock(&mgr->destroy_connector_lock);
> > -			list_add(&port->next, &mgr->destroy_connector_list);
> > -			mutex_unlock(&mgr->destroy_connector_lock);
> > -			schedule_work(&mgr->destroy_connector_work);
> > -			return;
> > -		}
> > -		/* no need to clean up vcpi
> > -		 * as if we have no connector we never setup a vcpi */
> > -		drm_dp_port_teardown_pdt(port, port->pdt);
> > -		port->pdt = DP_PEER_DEVICE_NONE;
> > +	/* There's nothing that needs locking to destroy an input port yet */
> > +	if (port->input) {
> > +		drm_dp_mst_put_port_malloc(port);
> > +		return;
> >  	}
> > -	drm_dp_mst_put_port_malloc(port);
> > +
> > +	kfree(port->cached_edid);
> > +
> > +	/*
> > +	 * we can't destroy the connector here, as we might be holding the
> > +	 * mode_config.mutex from an EDID retrieval
> > +	 */
> > +	mutex_lock(&mgr->destroy_connector_lock);
> > +	list_add(&port->next, &mgr->destroy_connector_list);
> > +	mutex_unlock(&mgr->destroy_connector_lock);
> > +	schedule_work(&mgr->destroy_connector_work);
> 
> So if I'm not completely blind this just flattens the above code flow (by
> inverting the if (port->input)).

Now I'm really remembering why I refactored this. The control flow on the
previous version of this is pretty misleading. To summarize so it's a bit more
obvious:

if (port->input) {
	drm_dp_mst_put_port_malloc(port);
	return;
} else if (port->connector) {
	add_connector_to_destroy_list();
	return;
	/* ^ now, this is where PDT teardown happens */
} else {
	drm_dp_port_teardown_pdt(port, port->pdt);
}
/* free edid etc etc */

So, I suppose the title of this patch would be more accurate if it was
"drm/dp_mst: Remove PDT teardown in destroy_port() and refactor"
I'll go ahead and change that

> 
> >  }
> >  
> >  /**
> > @@ -3881,7 +3872,8 @@ drm_dp_finish_destroy_port(struct drm_dp_mst_port
> > *port)
> >  {
> >  	INIT_LIST_HEAD(&port->next);
> >  
> > -	port->mgr->cbs->destroy_connector(port->mgr, port->connector);
> > +	if (port->connector)
> 
> And this here I can't connect with the commit message. I'm confused, did
> something go wrong with some rebase here, and this patch should have a
> different title/summary?
> -Daniel

No, this is correct. In the previous drm_dp_destroy_port() function we only
added a port to the delayed destroy work if it had a connector on it. Now that
we add ports unconditionally, we have to check port->connector before trying
to call port->mgr->cbs->destroy_connector() since port->connector is no longer
guaranteed to be != NULL.

> 
> > +		port->mgr->cbs->destroy_connector(port->mgr, port->connector);
> >  
> >  	drm_dp_port_teardown_pdt(port, port->pdt);
> >  	port->pdt = DP_PEER_DEVICE_NONE;
> > -- 
> > 2.21.0
> > 
-- 
Cheers,
	Lyude Paul