hi Matt, hi Persmule Your attached log and the one in your follow-up
differ in exactly one place — the eighth and last probe, which is the
internal panel:
|working: PCH_GMBUS0 <- 0x03 (LVDS pin pair) -> EDID read, mode set
broken: DDI_AUX_CTL_A (eDP port A AUX) -> Panel.Off, no mode |
The first seven probes are identical (AUX B/C/D for DP1-3, GMBUS 5/4/6/2
for HDMI1-3 and VGA), and so are the detection registers: |DDI_BUF_CTL_A
= 0x18| (bit 2 clear, i.e. no eDP detected) and |PCH_LVDS = 0x02| in
both. So this is not an EDID or GMBUS timing problem — the internal
panel is simply probed as eDP instead of LVDS, the AUX channel does not
answer, and the panel is powered off again.
What makes it odd is that |Config_Helpers.To_Display_Type| maps |LVDS =>
LVDS| unconditionally, and that file is unchanged in the 16 commits
|bdb8f59e| pulls in. Whatever redirects the probe is elsewhere.
The divergence is at the 8th probe. In your log GMBUS pin 3 (LVDS) is
never selected; instead the panel port goes through DDI_AUX_CTL_A
(0x64010), i.e. Read_EDID took the |To_Display_Type (Port) = DP| branch
and fell through to |DP_A|. Ports 1-7 match the 2170p port list exactly,
so only the panel port is affected.
That shouldn't be reachable: |To_Display_Type (LVDS)| is |LVDS|, and
your own detect values (|PCH_LVDS = 0x02|, |DDI_BUF_CTL_A = 0x18|) mean
LVDS is valid and eDP is not. I built the 2170p at both 6d971afd and
6e74c703: the generated |hw-gfx-gma-config.ads| is byte-identical, and
|hw-gfx-gma-config_helpers.o|, |hw-gfx-gma-display_probing.o| and
|hw-gfx-gma-panel.o| are instruction-identical. So the bump changes no
code on this path for your board.
Could you check two things?
1. Rebuild from scratch (|make distclean|) at 7a5f91cae9 with the new
submodule. The bump adds two fields to|Pipe_Config|and
widens|Position_Type|, both shared with coreboot's
own|gma-gfx_init.adb|— an incremental rebuild across the submodule
change could mix old and new object layouts.
2. If it still fails, please
send|build/libgfxinit/common/hw-gfx-gma-config.ads|(the|Panel_Ports|lines)
and your|.config|, plus|git status|in both coreboot
and|3rdparty/libgfxinit|.
On the commit id: bdb8f59e9de777c1ab6d4395cc5b632f6346700a is in
coreboot main — it's "3rdparty/libgfxinit: Update to upstream main"
(CB:94326, merged 2026-07-27), which moves the submodule from 6d971afd
to 6e74c703 and brings in 16 commits. Just the full 40-char hash rather
than the usual short form.
Unrelated: the 57.7 ms GMBUS timeouts appear in both your logs, so
they're not part of this regression — they come from the 500→50 ms
change in |hw-gfx-gma-i2c.adb|.
thank you.
_______________________________________________
coreboot mailing list -- [email protected]
To unsubscribe send an email to [email protected]
