Re: [PATCH v6 0/3] Introduce EC driver for Snapdragon X1E based Dell XPS 13 9345

[Liviu Nicoara <lnicoara@thinkoid.org>]

On Sun, 11 Oct 2026 19:50:07 +0200, Aleksandrs Vinarskis wrote:
> This series adds Embedded Controller (EC) driver for Dell XPS 13 9345.
> While EC appears to control most of device's peripherals, particular
> driver addresses power and thermal managment issues.

Tested v6 on a second XPS 13 9345 (BIOS 2.14.0), on 7.2.6 with the
PMIC5 Gen3 ADC and ADC_TM drivers and the board thermal zones [1]
backported. The driver probes, gets all seven channels and reports to
the EC without an error. Two observations from this unit:

1. The two GPIO thermistor channels. Here AMUX3_GPIO (0x14c,
   lpddr5_therm) reads the top of the conversion table, ~130 C, from
   boot on, and AMUX4_GPIO (0x14d, usb_charger_left_therm) does not
   move when a charger warms the left port. AMUX2_GPIO (0x14b) does,
   1.6 C in three minutes, and AMUX1_GPIO (0x14a) reads a plausible
   board temperature. Windows' ADC1.bin puts SYS_THERM1/2 on
   0x14a/0x14b, and hamoa-pmics.dtsi in next defines those two. The
   patch configures no PMIC GPIO for AMUX3/4, where mainline sets
   pm8550 gpio3 and gpio4 for AMUX1/2; so either the two thermistors
   sit on 0x14a/0x14b on this board, or AMUX3/4 want the pin
   configuration. As it stands the EC is told the memory is at 130 C.
   The fans did not react to that here, at idle on battery.

2. Resume. With the driver loaded, a key press wakes the machine
   through the PMIC power key: the EC has powered the keyboard off, as
   the comment in the driver says, and the keyboard delivers no event
   for the key that woke the machine; instead pm8941-pwrkey reports a
   KEY_POWER press and release at resume. With logind's default
   HandlePowerKey=poweroff and no desktop holding the power-key
   inhibitor, the machine shuts down right after it wakes. With the
   inhibitor held, the resume is clean and the machine stays up, so
   the resume command does its job. Without the driver the keyboard
   wakes the host as an i2c-hid device and no power-key event appears.
   Reproduced twice; GNOME and KDE take that inhibitor, so they would
   not see it.

[1] https://lore.kernel.org/all/20261006185557.23016-1-lnicoara@thinkoid.org/

Liviu