public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Stefan Agner <stefan@agner.ch>
To: p.zabel@pengutronix.de
Cc: airlied@linux.ie, rmk+kernel@armlinux.org.uk,
	l.stach@pengutronix.de, dri-devel@lists.freedesktop.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] Revert "drm/imx: don't destroy mode objects manually on driver unbind"
Date: Wed, 17 Oct 2018 13:25:21 +0200	[thread overview]
Message-ID: <d55ebfae98b0ccf30895700be0881264@agner.ch> (raw)
In-Reply-To: <20181016160923.2042-1-stefan@agner.ch>

On 16.10.2018 18:09, Stefan Agner wrote:
> This reverts commit 8e3b16e2117409625b89807de3912ff773aea354.
> 
> Using the component framework requires all components to undo in
> ->unbind what ->bind does. Unfortunately that particular commit
> broke this rule. In particular, this is an issue if a single
> component during probe fails. In that case, component_bind_all()
> calls unbind on already succussfully bound components, and then
> frees memory allocated using devm. If one of those components
> registered e.g. an encoder with the framework then this leads to
> use after free situations.
> 
> Revert the commit to ensure that all components properly undo
> what ->bind does.

After Lucas comment mentioning HDMI unbind is not proper I looked
through all the unbind again. The other unbind functions need some
fixing too. I did not bother checking whether those were always broken
or just because things changed (the commit this is reverting was in
2016).... Here is what I found:

> 
> Link:
> https://www.mail-archive.com/dri-devel@lists.freedesktop.org/msg233327.html
> Suggested-by: Russell King <rmk+kernel@armlinux.org.uk>
> Signed-off-by: Stefan Agner <stefan@agner.ch>
> ---
>  drivers/gpu/drm/imx/imx-drm-core.c     | 4 ++--
>  drivers/gpu/drm/imx/imx-ldb.c          | 6 ++++++
>  drivers/gpu/drm/imx/imx-tve.c          | 3 +++
>  drivers/gpu/drm/imx/parallel-display.c | 3 +++
>  4 files changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/imx/imx-drm-core.c
> b/drivers/gpu/drm/imx/imx-drm-core.c
> index 5ea0c82f9957..caa6061a98ba 100644
> --- a/drivers/gpu/drm/imx/imx-drm-core.c
> +++ b/drivers/gpu/drm/imx/imx-drm-core.c
> @@ -305,11 +305,11 @@ static void imx_drm_unbind(struct device *dev)
>  
>  	drm_fb_cma_fbdev_fini(drm);
>  
> -	drm_mode_config_cleanup(drm);
> -
>  	component_unbind_all(drm->dev, drm);
>  	dev_set_drvdata(dev, NULL);
>  
> +	drm_mode_config_cleanup(drm);
> +
>  	drm_dev_put(drm);
>  }
>  
> diff --git a/drivers/gpu/drm/imx/imx-ldb.c b/drivers/gpu/drm/imx/imx-ldb.c
> index 3bd0f8a18e74..592aabc4a262 100644
> --- a/drivers/gpu/drm/imx/imx-ldb.c
> +++ b/drivers/gpu/drm/imx/imx-ldb.c
> @@ -723,6 +723,12 @@ static void imx_ldb_unbind(struct device *dev,
> struct device *master,
>  		if (channel->panel)
>  			drm_panel_detach(channel->panel);
>  
> +		if (!channel->connector.funcs)
> +			continue;
> +
> +		channel->connector.funcs->destroy(&channel->connector);
> +		channel->encoder.funcs->destroy(&channel->encoder);

There can be an encoder and bridge, or an encoder and connector. All of
them should be properly cleaned up.

So I guess this should look like this:

  		if (channel->panel)
  			drm_panel_detach(channel->panel);

  		if (channel->bridge)
  			drm_bridge_detach(channel->bridge);

		if (channel->connector.funcs)
			channel->connector.funcs->destroy(&channel->connector);

		if (channel->encoder.funcs)
			channel->encoder.funcs->destroy(&channel->encoder);

		kfree(channel->edid);
		i2c_put_adapter(channel->ddc);

The last two functions following are only strictly necessary when
connector is initialized. But its safe to call them with null, so I
would just call them always.


> +
>  		kfree(channel->edid);
>  		i2c_put_adapter(channel->ddc);
>  	}
> diff --git a/drivers/gpu/drm/imx/imx-tve.c b/drivers/gpu/drm/imx/imx-tve.c
> index cffd3310240e..8d6e89ce1edb 100644
> --- a/drivers/gpu/drm/imx/imx-tve.c
> +++ b/drivers/gpu/drm/imx/imx-tve.c
> @@ -673,6 +673,9 @@ static void imx_tve_unbind(struct device *dev,
> struct device *master,
>  {
>  	struct imx_tve *tve = dev_get_drvdata(dev);
>  
> +	tve->connector.funcs->destroy(&tve->connector);
> +	tve->encoder.funcs->destroy(&tve->encoder);
> +

Cleanup of tve->ddc missing.

>  	if (!IS_ERR(tve->dac_reg))
>  		regulator_disable(tve->dac_reg);
>  }
> diff --git a/drivers/gpu/drm/imx/parallel-display.c
> b/drivers/gpu/drm/imx/parallel-display.c
> index aefd04e18f93..6f11bffcde37 100644
> --- a/drivers/gpu/drm/imx/parallel-display.c
> +++ b/drivers/gpu/drm/imx/parallel-display.c
> @@ -258,6 +258,9 @@ static void imx_pd_unbind(struct device *dev,
> struct device *master,
>  	if (imxpd->panel)
>  		drm_panel_detach(imxpd->panel);
>  

And in this case a bridge detach is missing.

Will send a v2 with that addressed.

--
Stefan

> +	imxpd->encoder.funcs->destroy(&imxpd->encoder);
> +	imxpd->connector.funcs->destroy(&imxpd->connector);
> +
>  	kfree(imxpd->edid);
>  }

  parent reply	other threads:[~2018-10-17 11:25 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-10-16 16:09 [PATCH] Revert "drm/imx: don't destroy mode objects manually on driver unbind" Stefan Agner
2018-10-16 16:45 ` Daniel Vetter
2018-10-16 16:51 ` Lucas Stach
2018-10-17 10:52   ` Stefan Agner
2018-10-17 11:25 ` Stefan Agner [this message]
2018-10-17 12:00   ` Stefan Agner

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=d55ebfae98b0ccf30895700be0881264@agner.ch \
    --to=stefan@agner.ch \
    --cc=airlied@linux.ie \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=l.stach@pengutronix.de \
    --cc=linux-kernel@vger.kernel.org \
    --cc=p.zabel@pengutronix.de \
    --cc=rmk+kernel@armlinux.org.uk \
    /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