archive mirror
 help / color / mirror / Atom feed
From: Andrzej Hajda <>
To: Jagan Teki <>
Cc: Maxime Ripard <>,
	Sam Ravnborg <>,
	Daniel Vetter <>,
	David Airlie <>,
	Thierry Reding <>,
	Laurent Pinchart <>,
	Maarten Lankhorst <>,
	Thomas Zimmermann <>,
	Jernej Skrabec <>,
	Neil Armstrong <>,
	Jonas Karlman <>,
	Robert Foss <>,
	linux-kernel <>,
	dri-devel <>
Subject: Re: [PATCH v2 0/8] drm/bridge: Make panel and bridge probe order consistent
Date: Mon, 9 Aug 2021 14:03:59 +0200	[thread overview]
Message-ID: <> (raw)
In-Reply-To: <>

Hi Jagan,

W dniu 09.08.2021 o 10:00, Jagan Teki pisze:
> Hi Andrzej,
> On Wed, Aug 4, 2021 at 7:48 PM a.hajda <> wrote:
>> Hi Maxime,
>> I have been busy with other tasks, and I did not follow the list last
>> time, so sorry for my late response.
>> On 28.07.2021 15:32, Maxime Ripard wrote:
>>> Hi,
>>> We've encountered an issue with the RaspberryPi DSI panel that prevented the
>>> whole display driver from probing.
>>> The issue is described in detail in the commit 7213246a803f ("drm/vc4: dsi:
>>> Only register our component once a DSI device is attached"), but the basic idea
>>> is that since the panel is probed through i2c, there's no synchronization
>>> between its probe and the registration of the MIPI-DSI host it's attached to.
>>> We initially moved the component framework registration to the MIPI-DSI Host
>>> attach hook to make sure we register our component only when we have a DSI
>>> device attached to our MIPI-DSI host, and then use lookup our DSI device in our
>>> bind hook.
>>> However, all the DSI bridges controlled through i2c are only registering their
>>> associated DSI device in their bridge attach hook, meaning with our change
>> I guess this is incorrect. I have promoted several times the pattern
>> that device driver shouldn't expose its interfaces (for example
>> component_add, drm_panel_add, drm_bridge_add) until it gathers all
>> required dependencies. In this particular case bridges should defer
>> probe until DSI bus becomes available. I guess this way the patch you
>> reverts would work.
>> I advised few times this pattern in case of DSI hosts, apparently I
>> didn't notice the similar issue can appear in case of bridges. Or there
>> is something I have missed???
> Look like Maxime is correct. I2C based DSI bridge will get probe
> during bridge_attach which usually called from bridge driver
> bridge_attach call. Non-I2C bridges and DSI panels will get probe
> during host.attach.
> We do get similar situation for dw-mipi-dsi bridges, where icn6211
> bridge is not I2C-based bridge and it gets probed in host_attach and
> sn65dsi83 is I2C based bridge and it gets probed in bridge_attach.
> Here is the simple call trace we have observed with dw-mipi-dsi bridge
> when all possible DSI device are trying to probe.
> 1. DSI panels and bridges will invoke the host attach
>     from probe in order to find the panel_or_bridge.
>     chipone_probe()
>         dw_mipi_dsi_host_attach().start
>             dw_mipi_dsi_panel_or_bridge()
>                  ...found the panel_or_bridge...
>     ltdc_encoder_init().start
>         dw_mipi_dsi_bridge_attach().start
>                     dw_mipi_dsi_host_attach().start
>                         chipone_attach(). start
>                         chipone_attach(). done
>                     dw_mipi_dsi_host_attach().done
>         dw_mipi_dsi_bridge_attach(). done
>     ltdc_encoder_init().done
> 2. I2C based DSI bridge will invoke the drm_bridge_attach
>     from bridge attach in order to find the panel_or_bridge.
>     ltdc_encoder_init().start
>         dw_mipi_dsi_bridge_attach().start
>             dw_mipi_dsi_panel_or_bridge()
>                  ...found the panel_or_bridge...
>                     dw_mipi_dsi_host_attach().start
>                         sn65dsi83_attach(). start
>                         sn65dsi83_attach(). done
>                     dw_mipi_dsi_host_attach().done
>         dw_mipi_dsi_bridge_attach(). done
>     ltdc_encoder_init().done
> It is correct that the I2C-based bridges will attach the host via
> mipi_dsi_attach in driver bridge API where as it done in probe for
> Non-I2C bridges and DSI panels.

The call order depends on the registration time of DSI host. In case of 
dw-mipi-dsi it is called from .component_bind callback (dsi_bind-> 
dsi_host_init -> mipi_dsi_host_register). And this is "the original sin" :)

dw-mipi-dsi calls component_add without prior acquiring its dependency - 
drm_bridge and before DSI host registration. In such situation bridge 
author should follow this pattern and perform similar initialization: 
first drm_bridge_add, then mipi_dsi_attach.

And now authors of bridges are in dead end in case they want their 
bridge/panel drivers cooperate with dw-mipi-dsi host (with pattern look 
for sink - bridge/panel, then register DSI bus) and with other DSI hosts 
(most of then use pattern - register DSI bus then look for the sink -  
panel or bridge).

Quick look at the DSI hosts suggests the pattern 
get-sink-then-register-bus are used only by kirin/dw_drm_dsi.c and msm/dsi.

All other DSI hosts uses apparently register-bus-then-get-sink pattern - 
as I said it was not profound analysis - just few greps of some keywords.

>> Anyway there are already eleven(?) bridge drivers using this pattern. I
>> wonder if fixing it would be difficult, or if it expose other issues???
>> The patches should be quite straightforward - move
>> of_find_mipi_dsi_host_by_node and mipi_dsi_device_register_full to probe
>> time.
>> Finally I think that if we will not fix these bridge drivers we will
>> encounter another set of issues with new platforms connecting "DSI host
>> drivers assuming this pattern" and "i2c/dsi device drivers assuming
>> pattern already present in the bridges".
> Agreed, I'm trying to understand the several ways to fix this. Right
> now I'm trying this on sun6i_mipi_dsi and exynos_drm_dsi. Will let you
> know for any update and suggestions on the same.

Quick look at sun6i suggests it uses register-bus-then-get-sink pattern 
(incompatible with kirin), only issue is that currently it support only 
panel sink.

Exynos uses also register-bus-then-get-sink pattern, but it slightly 
different as it supports dynamic attach/detach of sinks.



> Thanks,
> Jagan.

  reply	other threads:[~2021-08-09 12:04 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-28 13:32 Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 1/8] Revert "drm/vc4: dsi: Only register our component once a DSI device is attached" Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 2/8] drm/bridge: Add a function to abstract away panels Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 3/8] drm/bridge: Add documentation sections Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 4/8] drm/bridge: Document the probe issue with MIPI-DSI bridges Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 5/8] drm/panel: Create attach and detach callbacks Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 6/8] drm/vc4: dsi: Switch to drm_of_get_bridge Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 7/8] drm/panel: raspberrypi-touchscreen: Use the attach hook Maxime Ripard
2021-07-28 13:32 ` [PATCH v2 8/8] drm/panel: raspberrypi-touchscreen: Remove MIPI-DSI driver Maxime Ripard
     [not found] ` <>
2021-08-04 14:09   ` [PATCH v2 0/8] drm/bridge: Make panel and bridge probe order consistent a.hajda
2021-08-09  8:00     ` Jagan Teki
2021-08-09 12:03       ` Andrzej Hajda [this message]
2021-08-20 16:49     ` Maxime Ripard

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \
    --subject='Re: [PATCH v2 0/8] drm/bridge: Make panel and bridge probe order consistent' \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).