linux-kernel - drm_bridges on fairphone-fp3 are enabled in the wrong order

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-next>] [day] [month] [year] [list]

Message-ID: <c27953274997a56f8e0522f9331e733ae92bf25b.camel@web.de>
Date:   Wed, 05 Jul 2023 09:51:42 +0200
From:   Bert Karwatzki <spasswolf@....de>
To:     linux-kernel@...r.kernel.org
Cc:     Douglas Anderson <dianders@...omium.org>,
        linux-arm-msm@...r.kernel.org,
        Dmitry Baryshkov <dmitry.baryshkov@...aro.org>,
        Luca Weiss <luca@...tu.xyz>
Subject: drm_bridges on fairphone-fp3 are enabled in the wrong order

The fairphone-fp3 has a drm_panel attached to a dsi bridge. This are added to
the bridge_chain in struct drm_encoder in the order dsi, panel. When the
drm_atomic_bridge_chain_pre_enable these get enabled in the order panel, dsi
because of the list_for_each_entry_reverse. But the drm_panel of the fairphone-
fp3 is enabled using mipi_dsi_dcs_write_buffer which only works if the dsi is
enabled before the panel.
 To work around this one can revert

commit 9e15123eca7942caa8a3e1f58ec0df7d088df149
Author: Douglas Anderson <dianders@...omium.org>
Date:   Tue Jan 31 14:18:25 2023 -0800

    drm/msm/dsi: Stop unconditionally powering up DSI hosts at modeset

This leads to a working panel on startup. But when suspending one runs again
into a similar problem. When the bridges are disabled the dsi is disabled first
which leads to a failure in disabling the panel again because
mipi_dsi_dcs_write_buffer fails when the dsi is already switched of.
 As a simple workarund for both problems I have found it works to exchange the
order of the bridge chain in drm_endcoder:

diff --git a/drivers/gpu/drm/msm/dsi/dsi_manager.c
b/drivers/gpu/drm/msm/dsi/dsi_manager.c
index 28b8012a21f2..990f7c68a27c 100644
--- a/drivers/gpu/drm/msm/dsi/dsi_manager.c
+++ b/drivers/gpu/drm/msm/dsi/dsi_manager.c
@@ -550,6 +555,8 @@ int msm_dsi_manager_ext_bridge_init(u8 id)
 		if (ret < 0)
 			return ret;
 	}
+	/* swap bridges in list */
+	list_swap(&encoder->bridge_chain, encoder->bridge_chain.next);

 	/* The pipeline is ready, ping encoders if necessary */
 	msm_dsi_manager_set_split_display(id);

But this does not look like a portable solution so I wonder if there is a better
way to do this.

The linux kernel used here is a linux-next-20220630, with several out-of-tree
patches which are needed for the msm8953 gpu and the display used in the
fairphone-fp3 located here: https://github.com/spasswolf/msm8953-linux.git in
branch msm8953_iommu_rebase_v2_wlan_modem_ipa_cpufreq_display_debug.

Bert Karwatzki