On 9/23/24 06:21, Mark Brown wrote:
On Fri, Sep 20, 2024 at 06:47:05PM +0200, Jerome Brunet wrote:+int pmbus_regulator_init_cb(struct regulator_dev *rdev, + struct regulator_config *config) +{ + struct pmbus_data *data = config->driver_data; + struct regulation_constraints *constraints = rdev->constraints; + + if (data->flags & PMBUS_OP_PROTECTED) + constraints->valid_ops_mask &= ~REGULATOR_CHANGE_STATUS; + + if (data->flags & PMBUS_VOUT_PROTECTED) + constraints->valid_ops_mask &= ~REGULATOR_CHANGE_VOLTAGE; + + return 0; +} +EXPORT_SYMBOL_NS_GPL(pmbus_regulator_init_cb, PMBUS);I'm fairly comfortable with this from a regulator point of view, modulo the suggestion I posted in the other message about registering separate ops. The fact that there's three combinations of ops is annoying but doesn't feel too bad, though I didn't actually write it out so perhaps it looks horrible. In general removing permissions is safe, and without separate steps to remove write protect (which I see in your patch 5) the writes wouldn't actually work anyway.
I still consider the callback to be unnecessary, but I don't really have time to implement a better solution myself. If you accept the regulator patches, I'll have another look at the series as-is. Guenter
