On Thu, Jan 21, 2016 at 10:29:13PM +0100, Jiri Kosina wrote: > On Thu, 21 Jan 2016, Torsten Duwe wrote: > > > > mcount call sites looks normal on first sight... > > > > Not quite. > > LR is not saved on the stack before the call. > > Argh! > > > > Petr, this looks like 12 bytes offset for gcc-6. > > I think I can work around the rest. > > Are we sure that gcc is doing the right thing here? > > I am far from claiming understanding of ppc64 ABI, but from what Vojtech > told me I understood that saving link register is necessary for (at least) > graph tracer to work properly. It is held in R0 only, and saved right after _mcount. Thus, _mcount just must not clobber R0 or save it the same way as it's done afterwards or like gcc4 does it. I'll make a v6 that's compiler agnostic. It's a few lines to change for the kernel proper, and I'll have to have a look at the trampolines for modules. Torsten -- 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
