LT9611 board probe, 2026-09-29 (UTC)

Evidence class: physical board, serial console (tools/console.py /dev/ttyACM0 under flock /tmp/k230-board.lock). The board was booted from the normal panel DTB with touch running, on system fv2gvvq6 (test activation). No HDMI change was installed. Transcripts are trimmed of prompt escape codes only; they contain no addresses or credentials.

1.1 Bus and adapter number (03:32:00Z)

/sys/class/i2c-adapter/i2c-0 Synopsys DesignWare I2C adapter .../soc/i2c@91409000
/sys/class/i2c-adapter/i2c-1 Synopsys DesignWare I2C adapter .../soc/i2c@91408000

i2c@91408000 is the K230’s I2C3, so Linux assigned adapter 1 to &i2c3, not 3. i2cdetect -y -r 1:

30: -- -- -- -- -- -- -- -- -- -- -- 3b -- -- -- --
50: -- -- -- -- -- -- -- -- -- -- -- -- -- UU -- --

0x3b answers (no driver bound). 0x5d shows UU, bound to gt9895 (/sys/bus/i2c/devices/1-005d).

1.2 Register reads (03:31:21Z)

The plain reads on page 0: i2cget -y 1 0x3b for 0x00, 0x01, 0x02 and 0xff each returned 0x00. The chip acknowledged every read.
Deviation from task 1.2’s read-only rule: to read the chip ID, the same probe session also sent the register writes that mainline drivers/gpu/drm/bridge/lontium-lt9611.c itself uses: page select 0xff=0x80, I2C-enable 0xee=0x01, page 0x82, then back to page 0x80. None of these writes touches the video path, the MIPI RX or the HDMI TX. The console showed:

ID=0x17 0x02 0xe2        # page 0x80 regs 0x00,0x01,0x02
HPD5E=0x78               # page 0x82 reg 0x5e

0x17 0x02 matches the mainline driver’s LT9611 chip-ID check. 0xe2 is the revision byte.
Reg 0x825e bit 2 is the HPD bit the mainline driver reads. It was 0, so no HDMI sink was detected at that moment. It is unknown whether a cable was attached; this probe does not establish that HPD sensing works.

1.3 GPIO23/GPIO24 (from `/sys/kernel/debug/gpio`)

gpiochip0: GPIOs 512-543, parent: platform/9140b000.gpio
 gpio-534 (                    |dsi_reset           ) out hi
 gpio-536 (                    |reset               ) out hi ACTIVE LOW
 gpio-537 (                    |backlight_gpio      ) out hi

GPIO24 (gpio-536) is an output, physically high, and held by the touch driver as an active-low reset (deasserted). Because the net is shared, this holds the LT9611 out of reset too, which is consistent with its answering on I2C.
GPIO23 (gpio-535) does not appear as a requested GPIO line. The touch driver takes it as an interrupt only.

1.4 Open-drain question

Not determinable from software on this kernel. The node /sys/kernel/debug/pinctrl/*/pinconf-pins does not exist (the only entries are pinctrl-devices, pinctrl-handles and pinctrl-maps), so the pull or drive configuration of GPIO23 could not be read. The question stays open in design.md.

Not verified here