On Fri, 18 May 2012, Mark Brown wrote: > On Fri, May 18, 2012 at 11:28:31AM +0200, Guennadi Liakhovetski wrote: > > > 1. A regulator that can (in principle) change state, i.e., switch on and > > off. Such a regulator is good to keep for the runtime to power up and down > > the card. > > > 2. A regulator, that cannot switch, but at least can tell its supplied > > voltage is used ones to read out supported voltages and released again. > > This doesn't make sense, you can only change the voltages if you hold a > reference to the regulator. Yes, that's obviously a thinko. I forgot, that mmc also switches between different voltages, not only powers on and off. > > 3. A regulator that can do none of the above is defined as "useless" > > > > You may also run into trouble on boards that use the ability to disable > > > unused regulators at the end of boot - they'll power things off even > > > without the ability to change status at runtime. > > > Aha, you mean, I shouldn't put() the regulator, even if it cannot change > > status itself? > > Yes, you ought to to be safe and like I say if you want to manage the > voltage then you need to keep a reference to the regulator. > > > Can this also happen with a dummy regulator? > > Obviously not, Well, I certainly understand, that a dummy regulator cannot switch power by itself:-) Well, theoretically it can happen, that the board code / OF is missing the required entry and a shared regulator will be powered off, that also powers mmc, but that's obviously a bug. > though if you've got explicit code in to handle dummy > regulators you're rather missing the point. Their entire purpose is to > provide a crutch to keep the system going if bits are missing from the > bindings, they're not intended to be used in production. I was thinking about what to use to power off the card if both a platform callback (legacy) and a regulator are available (see, e.g., "[PATCH 08/29] mmc: sh_mmcif: add regulator support"). But perhaps giving preference to the regulator is bogus. Maybe I should just switch both, or even give preference to the call-back. What do you think? Switching both is, perhaps, the safest bet. Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer http://www.open-technology.de/ -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
