On Fri, 7 Sep 2018 11:30:55 +0200 Peter Zijlstra <pet...@infradead.org> wrote:
> > That said, I am not against this change. Especially the inlining > > is a good move. Note that lockdep_off()/lockdep_on() must not > > be traced as well. > > Hard to trace an inline funcion; we could make it __always_inline to > feel better. In the old days, you would need to make it __always_inline, because if gcc didn't inline it (and it did so in the past), it would be traced and would re-introduce this bug. But I've hit this issue too many times in the past, that I added "notrace" to the "inline" definition, where all "inline" labeled functions are also "notrace". lockdep.o has the -pg option removed in the Makefile so that all functions in lockdep.c is notrace, which is why it was fine to keep it there. But adding __always_inline may not be a bad idea regardless. -- Steve
