From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752151AbdLEBTV (ORCPT ); Mon, 4 Dec 2017 20:19:21 -0500 Received: from mail-pf0-f194.google.com ([209.85.192.194]:44689 "EHLO mail-pf0-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751846AbdLEBTT (ORCPT ); Mon, 4 Dec 2017 20:19:19 -0500 X-Google-Smtp-Source: AGs4zMaUo5QkyLirm9VQmlOZwuboJLJjKIc3sFpcwD+lJA8cvzZqOoM6f+V/XzLqLKRKAMUAF2JV1g== Date: Mon, 4 Dec 2017 17:19:16 -0800 From: Brian Norris To: Archit Taneja Cc: Nickey Yang , robh+dt@kernel.org, heiko@sntech.de, mark.rutland@arm.com, airlied@linux.ie, hl@rock-chips.com, zyw@rock-chips.comg, linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org, linux-rockchip@lists.infradead.org, xbl@rock-chips.com Subject: Re: [PATCH v3 5/6] dt-bindings: add the rockchip, dual-channel for dw-mipi-dsi Message-ID: <20171205011914.GA59914@google.com> References: <1508903463-7254-1-git-send-email-nickey.yang@rock-chips.com> <1508903463-7254-5-git-send-email-nickey.yang@rock-chips.com> <816a1d2d-d213-709d-f357-63e7c5248ad3@codeaurora.org> <2dab7233-139f-b148-2f5e-40a32b3dc797@rock-chips.com> <0dcf03d3-eda6-1b90-10ed-1d14f07824ab@codeaurora.org> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <0dcf03d3-eda6-1b90-10ed-1d14f07824ab@codeaurora.org> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Archit, I'm a relative n00b here, but I'm trying to follow along and I have some questions: On Fri, Dec 01, 2017 at 06:29:04PM +0530, Archit Taneja wrote: > On 11/30/2017 11:02 PM, Nickey Yang wrote: > >I try to follow as you suggested,use > > > >mipi_dsi: mipi@ff960000 { > >     ... > >     ... > >     clock-master;    /* implies that this DSI instance drivers the clock > >              * for both the DSIs. > >              */ > >     ports { > >         mipi_in: port { > >             ... > >             ... > >         }; > >         /* add extra output ports for both DSIs */ > >         mipi_out: port { > >             mipi_panel_out: endpoint { > >                 remote-endpoint = <&panel_in_channel0>; > >             }; > >         }; > >     }; > >     panel { > >         ... > >         ... > >         /* > >          * panel node can describe its input ports, if both the DSIs output > >          * ports are connected to the same device (i.e, the same DSI panel), > >          * we can assume that the DSIs need to operate in dual DSI mode > >          */ > >         ports { > >             ... > >             port@0 { > >                 panel_in_channel0: endpoint { > >                     remote-endpoint = <&mipi_panel_out>; > >                 }; > >             }; > >             port@1 { > >                 panel_in_channel1: endpoint { > >                     remote-endpoint = <&mipi1_panel_out>; > >                 }; > > > >             }; > >         }; > >     }; > >}; > > > >mipi_dsi1: mipi@ff968000 { > >     ... > >     ... > >     ports { > >         mipi1_in: port { > >             ... > >             ... > >         }; > >         mipi1_out: port { > >             mipi1_panel_out: endpoint { > >                 remote-endpoint = <&panel_in_channel1>; > >             }; > >         }; > >     }; > >} > > > >But it seems we can not use of_drm_find_panel(like below) > > > >/* > >         port = of_graph_get_port_by_id(dev->of_node, 1); > >         if (port) { > >                 endpoint = of_get_child_by_name(port, "endpoint"); > >                 of_node_put(port); > >                 if (!endpoint) { > >                         dev_err(dev, "no output endpoint found\n"); > >                         return -EINVAL; > >                 } > >                 panel_node = of_graph_get_remote_port_parent(endpoint); > >                 of_node_put(endpoint); > >                 if (!panel_node) { > >                         dev_err(dev, "no output node found\n"); > >                         return -EINVAL; > >                 } > >                 panel = of_drm_find_panel(panel_node); > >                 of_node_put(panel_node); > >                 if (!panel) > >                         return -EPROBE_DEFER; > >         } > >*/ > >to get DSI1 outputs,because of_drm_find_panel need compare > > > >if (panel->dev->of_node == np) > > > >in dsi_panel driver innolux->base.dev = &innolux->link->dev; > >dsi->dev > > Yes, we should only have 1 drm_panel in the global panel list. > Shouldn't it be possible to modify the dsi driver such that dsi1 > doesn't care whether it has a drm_panel for it or not, if we are > in dual dsi mode? > > I imagine a sequence like this: > > 1. dsi0 probes, parses the of-graph, finds the panel and saves its device > node. Does this mean we depend on probe order? Or would we be able to -EPROBE_DEFER or similar if dsi1 binds first? > 2. dsi1 probes, parses the of-graph, find the panel's device node > - dsi1 checks if it is the same as the panel attached to dsi0. > - If so, it just takes the drm_panel pointer from dsi0. > - If not, it tries a of_drm_find_panel() on the panel's device node. So, that all means we'd need a new variant of drm_of_find_panel_or_bridge() for "dual" drivers like this? Or else open-code this logic in dw-mipi-dsi.c? > A dual DSI panel driver would also be a bit different. It will be a > mipi_dsi_driver with the master DSI (dsi0) as the mipi_dsi_device. Using > the of-graph helpers, we would get the device node of dsi1 using > of_find_mipi_dsi_host_by_node(), and create another DSI device using > mipi_dsi_device_register_full(). Then, we call mipi_dsi_attach() on > both the dsi devices. That seems...interesting. I guess that sounds like it could work, but someone would have to play with that a bit more. I assume one wouldn't want to do all this in every dual DSI driver that needs this, right? > >struct innolux_panel { > >         struct drm_panel base; > >         struct mipi_dsi_device *link; > >}; > >It means one panel can only be found in his dsi node,(like dsi0 above). > > > >I'm doubting about it, Or  may we follow tegra_dsi_ganged_probe > >(drivers/gpu/drm/tergra/dsi.c) method. > > This method will add a new binding similar to "nvidia,ganged-mode", which > is something we don't want to do. It's unfortunate we have the anti-pattern already merged :( Brian