It fails due to things I've just described - bq27x00_battery module is loaded in your system, and for some reasons, your chip started to report itself as requiring calibration - either device was without battery for a while, or you had more than 32 cycles since last calibration.

Due to - IMO, and in opinion of anyone even remotely related to BME replacement, that I've talked to - bad design decisions by Pali, bq27x00_battery stop exporting few sysf nodes *at all*, when battery isn't calibrated - causing "read errors", despite, that chip itself is still providing proper values. It means, that after 32 charging cycles, perfectly fine data will get dropped all-together.

I tried (and other people too - in fact, there is no one, except Pali, that think current approach is good), to convince Pali that in linux world, sysfnodes should *always* report what hardware reports, instead of censoring it for whatever reasons (not that here we have any good reasons, too). Sadly, he is very stubborn, and no luck here. You may try your luck at pestering him.
---

For now, only workaround is to modprobe -r bq27x00_battery (BNF will switch to i2cget access), or calibrate your battery.


As said, in next version, BNF will switch back to i2c-only access, that will work even with bq27x00_battery module loaded, though (due to I2C_FORCE_SLAVE and bq27x00_battery reads being atomic)

/Estel