On 27/07/2020 22:13:42+0100, Miguel Borges de Freitas wrote: > Russell King - ARM Linux admin <linux@xxxxxxxxxxxxxxx> escreveu no dia > segunda, 27/07/2020 à(s) 18:30: > > > > On Mon, Jul 27, 2020 at 06:16:32PM +0200, Alexandre Belloni wrote: > > > On 27/07/2020 17:55:50+0200, Jon Nettleton wrote: > > > > > So, can we please have that discussion, it is pertinent to this patch. > > > > > > > > > > > > > Thinking about this some more, I believe whether or not an IOCTL > > > > interface is in the works or needed is irrelevant. This patch > > > > describes the hardware and how it is designed and the topic of > > > > discussion is if we need a simple boolean state, or if we need > > > > something that could be used to support dynamic configuration in the > > > > future. I would rather make this decision now rather than keep > > > > tacking on boolean config options, or revisit a bunch of device-tree > > > > changes. > > For what it's worth I also tend to agree. > The patchset, regardless of the property name (that I admit might be > misleading), is intended at enforcing a mode that the RTC/driver > should use by default. This mode is strongly related to the hardware > definition/implementation and its capabilities. While I understand the > need for the IOCTL interface to solve issues exactly like the > aforementioned factory example, I fail to see how it can be of any > help to solve the problem at hand - as it won't likely configure the > driver to use a different default mode depending on the board. The > IOCTL interface might also allow the userspace to change this property > back to the default mode (000) and end up breaking the RTC operation, > but I guess that's the cost of configurability and, in the end, the > user's responsibility. > Any pointers on how to proceed are appreciated. > I would think the simpler way to proceed is to have a device specific property indicating that standard mode is not available. From the driver, you can then switch from standard to DSM when this property is present. I think it is difficult to come up with a generic property for that as most other RTCs with level/threshold switching have a fast edge detection feature that is usually enabled by default. So what they would require instead is a property indicating that the voltage is ramping down at a certain rate allowing to disable fast edge detection and saving a bit of power. -- Alexandre Belloni, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
