On 13/10/23 20:11, Wolfram Sang wrote: >> mmmhhh... still a bit skeptical about waiting 100 times 10us in >> atomic. > Has it been discussed already why the non-atomic version of > read_poll_timeout is not enough? > For mv64xxx i2c_recovery() is called from two places. One would be fine with read_poll_timeout() but the other is in an interrupt handler so needs the atomic version (or something else that doesn't schedule).
