Message-ID: <CAFPSGXYdA=09xWZJi2oCQsbr3eH1GBgzjD8DRm5=beiKAtx37Q@mail.gmail.com>
Date:   Mon, 19 Apr 2021 07:01:00 +0800
From:   Kevin Tang <kevin3.tang@...il.com>
To:     Maxime Ripard <maxime@...no.tech>
Cc:     Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>,
        Sean Paul <sean@...rly.run>, David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Rob Herring <robh+dt@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Orson Zhai <orsonzhai@...il.com>,
        Chunyan Zhang <zhang.lyra@...il.com>,
        "Linux-Kernel@...r. Kernel. Org" <linux-kernel@...r.kernel.org>,
        ML dri-devel <dri-devel@...ts.freedesktop.org>,
        devicetree@...r.kernel.org
Subject: Re: [PATCH v4 4/6] drm/sprd: add Unisoc's drm display controller driver

Hi,

Maxime Ripard <maxime@...no.tech> 于2021年4月15日周四 下午5:03写道：
>
> Hi,
>
> On Thu, Apr 15, 2021 at 08:18:52AM +0800, Kevin Tang wrote:
> > Maxime Ripard <maxime@...no.tech> 于2021年3月24日周三 下午7:10写道：
> > > > +static struct sprd_dpu *sprd_crtc_init(struct drm_device *drm,
> > > > +                      struct drm_plane *primary)
> > > > +{
> > > > +     struct device_node *port;
> > > > +     struct sprd_dpu *dpu;
> > > > +
> > > > +     /*
> > > > +      * set crtc port so that drm_of_find_possible_crtcs call works
> > > > +      */
> > > > +     port = of_parse_phandle(drm->dev->of_node, "ports", 0);
> > > > +     if (!port) {
> > > > +             drm_err(drm, "find 'ports' phandle of %s failed\n",
> > > > +                       drm->dev->of_node->full_name);
> > > > +             return ERR_PTR(-EINVAL);
> > > > +     }
> > > > +     of_node_put(port);
> > >
> > > The YAML binding should already make sure that your binding is sane, and
> > > if you still get a DT that doesn't follow it, you have a whole lot of
> > > other issues than whether ports is there :)
> > >
> > > > +     dpu = drmm_crtc_alloc_with_planes(drm, struct sprd_dpu, base,
> > > > +                                     primary, NULL,
> > > > +                                     &sprd_crtc_funcs, NULL);
> > > > +     if (IS_ERR(dpu)) {
> > > > +             drm_err(drm, "failed to init crtc.\n");
> > > > +             return dpu;
> > > > +     }
> > > > +
> > > > +     dpu->base.port = port;
> > >
> > > But you're still referencing it here, while you called of_node_put on it
> > > already? You should only call it once you're done with it.
> >
> >  of_node_put should be called after done with it, this maybe indeed be a bug.
> > i will fix it.
> > >
> > >
> > > I'm not really sure why you would need drm_of_find_possible_crtcs to
> > > work then if you don't follow the OF-Graph bindings.
> >
> > it scan all endports of encoder, if a matching crtc is found by
> > OF-Graph bindings
> > and then genarate the crtc mask, here is description:
> > 41  /**
> > 42   * drm_of_find_possible_crtcs - find the possible CRTCs for an encoder port
> > 43   * @dev: DRM device
> > 44   * @port: encoder port to scan for endpoints
> > 45   *
> > 46   * Scan all endpoints attached to a port, locate their attached CRTCs,
> > 47   * and generate the DRM mask of CRTCs which may be attached to this
> > 48   * encoder.
> > 49   *
> > if we don't follow the OF-Graph bindings, crtc can't attched to encoder.
>
> Yeah, what I'm actually confused about is why you would need the
> of_parse_phandle call. You usually call drm_of_find_possible_crtcs with
> the encoder device node, so from your MIPI-DSI driver in your case, and
> with it's device->of_node pointer and it should work perfectly fine?
I still confused about use drm_of_find_possible_crtcs to bind crtc and
encoder, the port of drm_crtc, default is null?

709  /**
710   * struct drm_crtc - central CRTC control structure
711   * @dev: parent DRM device
712   * @port: OF node used by drm_of_find_possible_crtcs()
-----------------------------------------------------------------------------------------------------
25  static uint32_t drm_crtc_port_mask(struct drm_device *dev,
26     struct device_node *port)
27  {
28   unsigned int index = 0;
29   struct drm_crtc *tmp;
30
31   drm_for_each_crtc(tmp, dev) {
32   if (tmp->port == port)
33   return 1 << index;
34
35   index++;
36   }
37
38   return 0;
39  }

I did not see any place to initialize the port of drm_crtc in the drm
framework, if not setup it.
it looks like the port of drm_crtc is undefined.

https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/rockchip/rockchip_drm_vop.c#L1851
https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/tilcdc/tilcdc_crtc.c#L1051

The difference from others is that we use of_parse_phandle,
as you said, follow the OF-Graph bindings , should be also use
of_get_child_by_name replace it?
like this:
-----------------------------------------------------------------------------------
    struct device_node *port;
    port = of_get_child_by_name(dpu->dev->of_node, "port");
    if (!port) {
        DRM_ERROR("find 'ports' phandle of %s failed\n",
              dpu->dev->of_node->full_name);
        return -EINVAL;
    }
    crtc->port = port;
    of_node_put(port);
------------------------------------------------------------------------------------

If we have multiple display controllers, use of_parse_phandle by
different index also seem is ok.
display-subsystem {
                compatible = "sprd,display-subsystem";
                ports = <&dpu0_port, &dpu1_port, &dpu2_port......>;
 };

dpu0_port = of_parse_phandle(drm->dev->of_node, "ports", 0);
dpu1_port = of_parse_phandle(drm->dev->of_node, "ports", 1);
dpu2_port = of_parse_phandle(drm->dev->of_node, "ports", 2);


>
> Maxime

Powered by blists - more mailing lists