> So, maybe the easier thing to do is change i2c_lock_adapter to only > lock the segment, and then have the callers beneath drivers/i2c/ > (plus the above mlx90614 driver) that really want to lock the root > adapter instead of the segment adapter call a new function named > i2c_lock_root (or something like that). Admittedly, that will be > a few more trivial changes, but all but one will be under the I2C > umbrella and thus require less interaction. > > Wolfram, what do you think? It sounds tempting, yet I am concerned about regressions. From that point of view, it is safer to introduce i2c_lock_segment() and convert the users which would benefit from that. How many drivers would be affected?
Attachment:
signature.asc
Description: PGP signature
