Silan SC7A20 accelerometer

Identity

Status — ◐ partial

The local tree now has a candidate iicbus driver at src/sys/dev/iicbus/sc7a20.c, and PINETAB2 includes device sc7a20. It probes silan,sc7a20, checks WHO_AM_I == 0x11, enables normal 100 Hz XYZ sampling, and exposes raw axis sysctls under dev.sc7a20.0.*.

This is still bench-pending. No PineTab2 run has confirmed that the chip responds on i2c5, that the board axis signs match the Linux orientation policy, or that the GPIO3_A2 interrupt mapping works.

Driver Shape

• dev.sc7a20.0.whoami — identity byte; expected 0x11.
• dev.sc7a20.0.enable — write 0 / 1 to power down or resume 100 Hz sampling.
• dev.sc7a20.0.status — raw STATUS_REG.
• dev.sc7a20.0.accel_{x,y,z}_raw — signed 12-bit sample shifted out of the left-justified output registers.
• dev.sc7a20.0.accel_{x,y,z}_mg — same first-pass value, named for the high-resolution +/-2 g mode’s approximate 1 mg/count scale.

The first slice intentionally avoids IRQs and evdev. Polled sysctls are enough to prove the I2C path and board orientation without adding a tablet policy surface before the hardware is on the bench.

Bring-Up Predicate

After a PineTab2 boot with the first-boot DTB:

sysctl dev.sc7a20.0

Expected first-pass evidence:

1. dev.sc7a20.0.whoami reports 17 (0x11).
2. One of accel_*_mg sits near +/-1000 when the tablet is flat and stationary.
3. Rotating the tablet moves that +/-1000 g-vector between axes with a stable sign.
4. status changes while the tablet is being moved.

Open Work

• Wire the GPIO3_A2 interrupt and cache samples from a taskqueue worker, following the PinePhone Pro mpu6500 pattern.
• Confirm axis orientation against the panel’s native rotation before adding a compositor rotation daemon.
• Decide whether this stays as sysctls plus userland policy or grows an evdev / IIO-equivalent interface.

Related

• PineTab2 component matrix
• PineTab2 source map