On Tue, Nov 11, 2014 at 10:05:05AM +0100, Vojtech Pavlik wrote: > On Mon, Nov 10, 2014 at 11:09:03AM -0600, Josh Poimboeuf wrote: > > > But there are a few (probably much less than 10%) cases like the locking > > > one I used above, where SWITCH_THREAD just isn't going to cut it and for > > > those I would need SWITCH_KERNEL or get very creative with refactoring > > > the patch to do things differently. > > > > I'm not opposed to having both if necessary. But I think the code would > > be _much_ simpler if we could agree on a single consistency model that > > can be used in all cases. Plus there wouldn't be such a strong > > requirement to get incremental patching to work safely (which will add > > more complexity). > > > > I actually agree with you that LEAVE_PATCHED_SET + SWITCH_THREAD is > > pretty nice. > > Cool! Do you see it as the next step consistency model we would focus on > implementing in livepatch after the null model is complete and upstream? Yeah, I'm thinking so. None of the consistency models are perfect, but I think this is a nice hybrid of the kGraft and kpatch models. It allows us to apply the greatest percentage of patches with the highest success rate, while keeping the code complexity at a reasonable level. -- Josh -- To unsubscribe from this list: send the line "unsubscribe live-patching" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
