On Mon 23 Sep 2024 at 14:21, Mark Brown <broonie@xxxxxxxxxx> wrote: > On Sat, Sep 21, 2024 at 06:49:34PM +0200, Jerome Brunet wrote: >> On Sat 21 Sep 2024 at 08:22, Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > >> > In other words, if always-on is _not_ set in >> > regulator constraints, I'd see that as request to override write-protect >> > in the driver if there is a change request from regulator code. > >> That's very much different from what we initially discussed. It can >> certainly be done, what is proposed here already does 90% of the job in >> that direction. However, I'm not sure that is what people intended when >> they did not put anything. A chip that was previously locked, would be >> unlocked following such change. It's an important behaviour change. > > The general approach we take for regulators is to not touch the hardware > state unless we were explicitly asked to do something by firmware > configuration. The theory is that this avoids us doing anything that > causes physical damage by mistake. > >> >> This is something that might get fix with this change [1]. Even with that >> >> fixed, passing init_data systematically would be convenient only if you >> >> plan on skipping DT provided constraints (there are lot of those), or >> >> redo the parsing in PMBus. > >> > I disagree. I am perfectly fine with DT overriding constraints provided >> > by the driver. The driver can provide its own constraints, and if dt >> > overrides them, so be it. > >> That's not what the regulator framework does. At the moment, it is DT >> and nothing else. After the linked change, it would be DT if no >> init_data is passed - otherwise, the init_data. > >> If a something in between, whichever the one you want to give priority >> to, that will have to re-implemented on the caller side. >> This is what I meant by redo the parsing on pmbus side. > > Right, and I've got a feeling that any attempt to combine constraints is > going to need to be done in a case by case manner since what's tasteful > is going to vary depending on how much we trust the various sources of > information. > >> It goes way beyond what I'm proposing. >> The only thing done here is something you simply cannot put in DT >> because DT is static. Following init, if the chip write protected, >> REGULATOR_CHANGE_STATUS should not be set, regardless of what is in DT. >> If it is not set, I'm not adding it. > >> Also, what I'm proposing does not get in the way of DT, or anything >> else, providing constraints. What I propose allow to make adjustement in >> the HW based on the constraint, if this is what you want to do. It also >> allows to update the constaints based on what the HW actually is. >> If the chip cannot be written, regulator needs to know. > > So, I know we talked about this a bit on IRC but I didn't register the > specific use case. Now I see that it's coming down to the fact that the > chip is simply write protected I'm wondering if it might not be simpler > to handle this at the ops level rather than the constraints level. When > registering the driver could check for write protection and then instead > of registering the normal operations register an alternative set with > the relevant write operations removed. That would have the same effect > that you're going for AFAICT. Sorry for not thinking of it earlier. Actually I thaught about it, that was my first idea. There is tiny difference between the 2 approach: * A regulator that does not provide enable()/disable() is considered always-on, so it is not really checked if it is enabled or not * A regulator that does provide enable()/disable() but disallow status change will be checked with is_enabled() In the second case, we will pick up on regulator that is disabled and we can't enable. In the first case, I don't think we do. I don't know if it is a bug of not but that makes the 2nd case more correct for what we do with pmbus regs I think The other thing, although is more a pmbus implemantion consideration, is that the type regulator is opaque in pmbus_regulator_register. Different types could be registered so manipulating the ops is tricky. That's where a callback is helpful If building the ops dynamically is the preferred way, I'll find a way to make it happen. I'll need to way to identify which one need it, loose the 'const' qualifier in a lot of place, etc ... that will be a bit hackish I'm afraid. > >> > We should not try to override devicetree constraints. > >> I don't think I am. I'm just reading the chip state and adjusting the >> constraint. Even after implementing what is suggested above, it will >> still be necessary to readback and adjust the constraint based the >> read protection. Unlock is not guranteed to succeed, the chip may be >> permanently lock. Some provide the option to do that. > > I'm not familiar with this hardware so I'll defer to the two of you on > what's tasteful with regard to handling this, based on the above it > might be a per device thing depending on how reversable the write > protection is. It looks like currently we don't change this at runtime > but I might not be looking properly? At the moment, it does not. With this patch it might but only a registration. We've been discussing other modes but it would not impact regulator I think. -- Jerome