On Wed, 26 Aug 2026 19:54:04 +0200 Jorijn van der Graaf <[email protected]> wrote:
> The Sensortek STK36C61 is a 3-in-1 ambient light / proximity / RGB > colour sensor whose register interface is compatible with the stk3310 > across the ALS and proximity data, threshold, gain and > integration-time registers this driver uses. The part name comes from > the Fairphone 6 manual, which lists the component as "3IN1 ALPS/RGB > sensor/STK36C61". There is no public datasheet; the layout was probed > on the device: chip ID 0x95, the stk3310 STATE/FLAG bit layout, data > and threshold registers, gain steps that scale the ALS reading > accordingly, and an RGBC data block directly after the ALS data. > > Since v1 the colour block's controls were located, confirming > Jonathan's guess on v1's patch 3 that they sit in a register we had > not found: the STK37660 datasheet, a documented sibling with the same > data register layout, places a clear-channel gain field in a GAINCTRL > register at 0x4E, split from the RGB gain in ALSCTRL, and stepping it > on the STK36C61 multiplies the clear count by ~4 per step while the > other channels hold still. The STK36C61 patch therefore now exposes > scale and integration time on the colour channels. > > Tested on a Fairphone 6 running a 7.2-based tree: this series' > stk3310.c, byte-identical, built as a module against that tree (the > i2c core files the binding behaviour depends on are identical between > the trees). Verified there: DT boot instantiation, the colour > scale/integration-time interface, and the sysfs name behaviours this > series changes: a lower-case name selects the full profile through > the id table, capitals no longer bind, and a full-compatible-string > client now fails probe with the missing-driver-data error. A board > DTS node using the new compatible follows separately via > linux-arm-msm. > > Changes in v2: > - new first patch lower-casing the i2c device ID names (Andy), and > probe now fails on missing match data instead of falling back to > the stk3310 profile (Andy) > - new precursor patch moving the data registers into the channel > .address field (Jonathan), split out of the match-data patch; the > bulk read gains a regmap local variable (Andy) > - the ACPI table entries use named initializers (Jonathan) > - the STK36C61 patch exposes scale and integration time on the colour > channels (Jonathan's suggestion on v1 patch 3; the clear-gain > register located via the STK37660 datasheet and verified on the > device). With those branches of read_raw now handling the intensity > channels, the ret = -EINVAL single-exit form promised on the v1 > thread (Andy) has no unreachable branch left to apply to > - picked up Krzysztof's Reviewed-by on the binding patch > > v1: > https://lore.kernel.org/all/[email protected]/ > FWIW I took another look through and with those small things Andy raised tidied up (including a more concise commit message for the final patch) this looks good to go to me. Thanks, Jonathan
