linux-kernel - Re: [PATCH] drm/bridge/synopsys: dsi: Don't blindly call post

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Thu, 25 Apr 2019 21:02:51 +0300
From:   Laurent Pinchart <laurent.pinchart@...asonboard.com>
To:     Matt Redfearn <matt.redfearn@...nci.com>
Cc:     Andrzej Hajda <a.hajda@...sung.com>,
        Philippe Cornu <philippe.cornu@...com>,
        "dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
        Matthew Redfearn <matthew.redfearn@...nci.com>,
        Nickey Yang <nickey.yang@...k-chips.com>,
        Heiko Stuebner <heiko@...ech.de>,
        Archit Taneja <architt@...eaurora.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>
Subject: Re: [PATCH] drm/bridge/synopsys: dsi: Don't blindly call post_disable

On Thu, Apr 25, 2019 at 08:59:15PM +0300, Laurent Pinchart wrote:
> On Thu, Apr 25, 2019 at 12:39:27PM +0000, Matt Redfearn wrote:
> > On 25/04/2019 13:13, Andrzej Hajda wrote:
> >> On 24.04.2019 16:22, Matt Redfearn wrote:
> >>> The DRM documentation states that post_disable is an optional callback.
> >>> As such an implementing device may not populate it. To avoid panicing
> >>> the kernel by calling a NULL function pointer, we should NULL check it
> >>> before blindy calling it.
> >>>
> >>> Signed-off-by: Matt Redfearn <matt.redfearn@...nci.com>
> >> 
> >>> ---
> >>>
> >>>   drivers/gpu/drm/bridge/synopsys/dw-mipi-dsi.c | 3 ++-
> >>>   1 file changed, 2 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/bridge/synopsys/dw-mipi-dsi.c b/drivers/gpu/drm/bridge/synopsys/dw-mipi-dsi.c
> >>> index 38e88071363..0ee440216b8 100644
> >>> --- a/drivers/gpu/drm/bridge/synopsys/dw-mipi-dsi.c
> >>> +++ b/drivers/gpu/drm/bridge/synopsys/dw-mipi-dsi.c
> >>> @@ -805,7 +805,8 @@ static void dw_mipi_dsi_bridge_post_disable(struct drm_bridge *bridge)
> >>>   	 * This needs to be fixed in the drm_bridge framework and the API
> >>>   	 * needs to be updated to manage our own call chains...
> >>>   	 */
> >>> -	dsi->panel_bridge->funcs->post_disable(dsi->panel_bridge);
> >>> +	if (dsi->panel_bridge->funcs->post_disable)
> >>> +		dsi->panel_bridge->funcs->post_disable(dsi->panel_bridge);
> >>>   
> >> 
> >> Why not drm_bridge_post_disable ?
> > 
> > Ah - that seems like a nicer fix! Do you think the comment above 
> > describing why this function pointer is called directly can be removed 
> > as well if we go this route?
> 
> It shouldn't be necessary to call ->post_disable manually here as the
> bridge core handles it internally. This is a hack to work around a
> problem, and should be fixed properly.
> 
> > If someone calls drm_bridge_post_disable() on the Synposys DSI 
> > drm_bridge it will go on to call post_disable on all other bridges in 
> > the chain, in addition to us calling them here. Is it an issue to call 
> > it multiple times?
> 
> It depends on the panel implementation, but in general it's not a good
> idea. It may happen to work, but could break at any time in the future.

Double-checking the driver, the .attach() operation doesn't propagate to
the next bridge, so the bridge core will not know about it, and will not
propagate .post_disable() either. I think this should be fixed in a way
that uses the drm bridge core infrastructure.

> >>>   	if (dsi->slave) {
> >>>   		dw_mipi_dsi_disable(dsi->slave);

-- 
Regards,

Laurent Pinchart