On Thu, Mar 19, 2015 at 04:35:06PM +0100, Borislav Petkov wrote: > And I don't want to leave the driver loaded when there's nothing to > be loaded for. One instance in this driver's specific case is one > northbridge and there are numascale boxes with hundreds of northbridges. > > If you have a better idea about how to unload the driver, asynchronously > or not, after all probe() calls have failed, I'm all ears.
We don't go around and implement random hacks ignoring layering and conventions even if that one off case seems to benefit whatever corner case minutely, because those kind of hacks accumulate and hinder with improvements at much larger scale and the benefit here is minute. The driver model has been moving onto separating module load and probing because that makes far more sense to most drivers that we use nowadays and also gives userland a cleaner way to manage modules by separating the two operations - loading and probing. If you think this is a big enough problem, please go ahead and build proper infrastructure for it. Be it a module attribute or even just a flag telling userland to indicate that the module can be auto-unloaded. Sure, it might look like a smart thing from a very confined viewpoint but you're seeing the problem unfolding right now. This code is standing in the way of a much more impactful generic driver layer improvement. Thanks. -- tejun -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
