Re: [PATCH v1 0/3] power: supply: s2mu005-battery: add S2MU106 (Samsung Galaxy XCover5)

From: Kaustabh Chakraborty

Date: Fri Nov 20 2026 - 10:52:35 EST


On 2026-11-17 02:41 -08:00, Adil Hüsrev Tepe wrote:
> I'd like to withdraw this series rather than send a v2.
>
> A review bot (Sashiko AI) flagged that patch 1's fallback compatible
> (samsung,s2mu106-fuel-gauge falling back to samsung,s2mu005-fuel-gauge)
> is unsound: a board using that fallback binds to the S2MU005 driver as-is,
> which still reads the MONOUT average-monitoring registers for
> VOLTAGE_AVG/CURRENT_AVG and uses current_avg to help decide
> POWER_SUPPLY_STATUS. The S2MU106 doesn't have that path, so those reads
> return garbage, and garbage current_avg can flip the reported charging
> status. Patch 3's has_monout flag avoids calling into that path from the
> S2MU106 variant I added, but the binding still allows a device to bind
> through the plain samsung,s2mu005-fuel-gauge compatible with no
> has_monout gate at all, which is the actual hole.
>
> On reflection, extending s2mu005-battery.c was the wrong shape for this
> chip regardless of that specific bug. The S2MU106 fuel gauge is a
> separate, standalone I2C device on the Galaxy XCover5 (own address,
> 0x3b), not a drop-in register-compatible S2MU005 variant sharing a
> device tree node -- it only happens to reuse the same STATUS/RVBAT/
> RCURCC/RSOC register layout and scaling. A dedicated driver avoids the
> fallback-compatible construction entirely, since there's nothing to
> fall back to.
>
> I'll send that as a new series instead: a samsung,s2mu106-fuelgauge
> binding and a standalone drivers/power/supply/s2mu106-fuelgauge.c,
> with no changes to s2mu005-battery.c or its binding. Sorry for the
> noise -- I should have settled on this shape before sending v1.

Actually, this one was just fine. I didn't look at the patches
meticulously but it seems like it does belong here. The correct approach
would be to adjust the driver correctly so that it accepts both.

The compatible patch is, in fact incorrect. The driver clearly uses two
separate compatibles for the two devices, so there should not be any
fallback.

> Adil