Chad Monroe 825e0adcf3 mediatek: leds-smartrg: pace and verify MCU writes ...

Some LED state changes are silently missed on mt7622 boards such
as the SDG-841. The MCU ACKs but silently discards register writes
that arrive while it is still processing a previous one. i2c from
shell reproduced drops at write gaps below ~3ms, while 5ms spacing
was reliable across hundreds of writes. Paced writes latch
correctly regardless of channel state and write order, ruling out
register write ordering as the cause.

The SmartRG vendor driver paced its control writes with msleep(1),
which is why this never showed there. Boards based on
mt7987/mt7988 tolerate the unpaced bursts with identical MCU
firmware (2.4).

Since a discarded write is ACKed, the return code cannot detect
it. After each write, wait out the measured consumption time, read
the register back, and rewrite until the value actually latched,
warning and returning -EIO after five attempts.

The msleep(5) between channel inits in srg_led_probe() paced the
same writes by hand. Drop it, so that srg_led_i2c_write() is the
only place that owns the MCU timing.

Signed-off-by: Chad Monroe <chad@monroe.io>
Link: https://github.com/openwrt/openwrt/pull/24354
Signed-off-by: Jonas Jelonek <jelonek.jonas@gmail.com>

2026-08-31 20:55:21 +02:00

..

leds-smartrg-system.c

mediatek: leds-smartrg: pace and verify MCU writes

2026-08-31 20:55:21 +02:00