On Thu, May 28, 2009 at 07:00:13PM +0100, Mark Brown wrote: > On Thu, May 28, 2009 at 10:33:20AM -0700, Tony Lindgren wrote: > > * Russell King - ARM Linux <linux@xxxxxxxxxxxxxxxx> [090528 09:20]: > > > > I really hate it when I see this kind of initialization after registration. > > > It feels totally wrong and fragile. > > > > At one point, I had clkdev matching using struct device pointer as well, > > > but it was realised that was far too limiting - you couldn't declare > > > clock entries without first having all devices setup, and then you run > > > into problems with ordering. > > > > It looks like the regulator stuff is suffering this same problem - it > > > wants to match by struct device pointer. That's fine if all your > > > struct device's are statically allocated, but as soon as you start > > > having dynamic ones, it gets _much_ harder to cope with. > > > Yeah. I believe Mark is working on sorting out the regulator fwk issues > > regarding this. > > Eh, no, not really. It's the same as the clock API in this regard - if > you want the struct device can be NULL and you can do a name based > lookup only but then you have to pass the name around in platform data > to support configurability. If the clock API implements some other > solutions we'll probably follow them but I'm not aware of any. That's not quite what we're talking about. With clkdev, you specify device name itself, rather than address-of-struct-device. Using the device name gives you independence from the initialization or creation of the struct device (if indeed the struct device is created at all.) -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
