linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Boris Brezillon <boris.brezillon@collabora.com>
To: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: dri-devel@lists.freedesktop.org,
	linux-samsung-soc@vger.kernel.org, linux-kernel@vger.kernel.org,
	Inki Dae <inki.dae@samsung.com>,
	Seung-Woo Kim <sw0312.kim@samsung.com>,
	Andrzej Hajda <a.hajda@samsung.com>,
	Krzysztof Kozlowski <krzk@kernel.org>,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>,
	Laurent Pinchart <laurent.pinchart@ideasonboard.com>,
	Neil Armstrong <narmstrong@baylibre.com>,
	Sam Ravnborg <sam@ravnborg.org>,
	Thierry Reding <thierry.reding@gmail.com>
Subject: Re: [PATCH] drm/bridge: Fix Exynos DSI after making bridge chain a double-linked list
Date: Fri, 27 Dec 2019 13:00:04 +0100	[thread overview]
Message-ID: <20191227130004.69d7dcad@collabora.com> (raw)
In-Reply-To: <20191227110135.4961-1-m.szyprowski@samsung.com>

On Fri, 27 Dec 2019 12:01:35 +0100
Marek Szyprowski <m.szyprowski@samsung.com> wrote:

> Exynos DSI DRM driver uses private calls to out bridge to force certain
> order of operations during init/exit sequences. This no longer works after
> conversion of bridge chain to a double-linked list. To fix the regression
> call bridge related operations manually instead of the generic
> drm_bridge_chain_*() operations.

I think it'd be worth explaining what the problem is (infinite loop
caused by list_for_each_entry() use when the bridge is no longer part
of the chain attached to the encoder).

> 
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Fixes: 05193dc38197 ("drm/bridge: Make the bridge chain a double-linked list")

We also need to fix that in VC4.

> ---
> This patch is a result of the following discussion:
> https://www.spinics.net/lists/dri-devel/msg239256.html
> ---
>  drivers/gpu/drm/exynos/exynos_drm_dsi.c | 21 ++++++++++++---------
>  1 file changed, 12 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/gpu/drm/exynos/exynos_drm_dsi.c b/drivers/gpu/drm/exynos/exynos_drm_dsi.c
> index 3955f84dc893..f5905c239a86 100644
> --- a/drivers/gpu/drm/exynos/exynos_drm_dsi.c
> +++ b/drivers/gpu/drm/exynos/exynos_drm_dsi.c
> @@ -255,7 +255,6 @@ struct exynos_dsi {
>  	struct mipi_dsi_host dsi_host;
>  	struct drm_connector connector;
>  	struct drm_panel *panel;
> -	struct list_head bridge_chain;
>  	struct drm_bridge *out_bridge;
>  	struct device *dev;
>  
> @@ -1391,7 +1390,8 @@ static void exynos_dsi_enable(struct drm_encoder *encoder)
>  		if (ret < 0)
>  			goto err_put_sync;
>  	} else {
> -		drm_bridge_chain_pre_enable(dsi->out_bridge);
> +		if (dsi->out_bridge->funcs->pre_enable)
> +			dsi->out_bridge->funcs->pre_enable(dsi->out_bridge);

Okay, so you're calling ->{pre_enable,enable,disable,post_disable}() on
the first bridge element which only works if the chain contains one
bridge (see below). Maybe you should keep exynos_dsi.bridge_chain and
create custom helpers to iterate over chain elements instead of calling
those hooks only on out_bridge.

>  	}
>  
>  	exynos_dsi_set_display_mode(dsi);
> @@ -1402,7 +1402,8 @@ static void exynos_dsi_enable(struct drm_encoder *encoder)
>  		if (ret < 0)
>  			goto err_display_disable;
>  	} else {
> -		drm_bridge_chain_enable(dsi->out_bridge);
> +		if (dsi->out_bridge->funcs->enable)
> +			dsi->out_bridge->funcs->enable(dsi->out_bridge);
>  	}
>  
>  	dsi->state |= DSIM_STATE_VIDOUT_AVAILABLE;
> @@ -1427,10 +1428,12 @@ static void exynos_dsi_disable(struct drm_encoder *encoder)
>  	dsi->state &= ~DSIM_STATE_VIDOUT_AVAILABLE;
>  
>  	drm_panel_disable(dsi->panel);
> -	drm_bridge_chain_disable(dsi->out_bridge);
> +	if (dsi->out_bridge->funcs->disable)
> +		dsi->out_bridge->funcs->disable(dsi->out_bridge);
>  	exynos_dsi_set_display_enable(dsi, false);
>  	drm_panel_unprepare(dsi->panel);
> -	drm_bridge_chain_post_disable(dsi->out_bridge);
> +	if (dsi->out_bridge->funcs->post_disable)
> +		dsi->out_bridge->funcs->post_disable(dsi->out_bridge);
>  	dsi->state &= ~DSIM_STATE_ENABLED;
>  	pm_runtime_put_sync(dsi->dev);
>  }
> @@ -1521,9 +1524,11 @@ static int exynos_dsi_host_attach(struct mipi_dsi_host *host,
>  
>  	out_bridge  = of_drm_find_bridge(device->dev.of_node);
>  	if (out_bridge) {
> -		drm_bridge_attach(encoder, out_bridge, NULL);
> +		out_bridge->dev = drm;
> +		out_bridge->encoder = encoder;
> +		if (out_bridge->funcs->attach)
> +			out_bridge->funcs->attach(out_bridge);

If you don't want to keep the dsi_exynos.bridge_chain field, you should
probably still use drm_bridge_attach() here and manually reset
encoder->bridge_chain after checking it only contains a single element.
Something like:

		drm_bridge_attach(encoder, out_bridge, NULL);

		/*
		 * FIXME: we need to control the enable/disable
		 * sequence to make sure the DSI host controller is
		 * ready to send DSI commands before the
		 * bridge->pre_enable()/panel->prepare() hooks are
		 * called (those hooks are allowed to send DSI
		 * commands). The only way to do that right now is to
		 * reset the bridge chain so that the core thinks it's
		 * empty and turns drm_bridge_chain_xxx() calls into
		 * NOPs. This only works if the chain contains only one
		 * bridge, otherwise we'd be skipping bridge->xxx()
		 * calls on other bridges which is wrong.
		 */
		WARN_ON(!list_is_singular(&encoder->bridge_chain));
		INIT_LIST_HEAD(&encoder->bridge_chain);

Calling ->attach() directly might lead to an invalid pointer
dereference if out_bridge tries to attach to another bridge
(out_bridge->bridge_node is uninitialized if you don't call
drm_bridge_attach()).

>  		dsi->out_bridge = out_bridge;
> -		list_splice(&encoder->bridge_chain, &dsi->bridge_chain);
>  	} else {
>  		int ret = exynos_dsi_create_connector(encoder);
>  
> @@ -1589,7 +1594,6 @@ static int exynos_dsi_host_detach(struct mipi_dsi_host *host,
>  		if (dsi->out_bridge->funcs->detach)
>  			dsi->out_bridge->funcs->detach(dsi->out_bridge);
>  		dsi->out_bridge = NULL;
> -		INIT_LIST_HEAD(&dsi->bridge_chain);
>  	}
>  
>  	if (drm->mode_config.poll_enabled)
> @@ -1737,7 +1741,6 @@ static int exynos_dsi_probe(struct platform_device *pdev)
>  	init_completion(&dsi->completed);
>  	spin_lock_init(&dsi->transfer_lock);
>  	INIT_LIST_HEAD(&dsi->transfer_list);
> -	INIT_LIST_HEAD(&dsi->bridge_chain);
>  
>  	dsi->dsi_host.ops = &exynos_dsi_ops;
>  	dsi->dsi_host.dev = dev;


  reply	other threads:[~2019-12-27 12:00 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CGME20191227110216eucas1p17cbf91afa905852d3c0b1efeec0f6f8d@eucas1p1.samsung.com>
2019-12-27 11:01 ` [PATCH] drm/bridge: Fix Exynos DSI after making bridge chain a double-linked list Marek Szyprowski
2019-12-27 12:00   ` Boris Brezillon [this message]
2019-12-27 13:23     ` Boris Brezillon
2019-12-27 14:03       ` Boris Brezillon

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:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

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

  git send-email \
    --in-reply-to=20191227130004.69d7dcad@collabora.com \
    --to=boris.brezillon@collabora.com \
    --cc=a.hajda@samsung.com \
    --cc=b.zolnierkie@samsung.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=inki.dae@samsung.com \
    --cc=krzk@kernel.org \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-samsung-soc@vger.kernel.org \
    --cc=m.szyprowski@samsung.com \
    --cc=narmstrong@baylibre.com \
    --cc=sam@ravnborg.org \
    --cc=sw0312.kim@samsung.com \
    --cc=thierry.reding@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).