Messages in this thread | | | Date | Sun, 30 Dec 2007 09:52:26 -0500 | From | Masami Hiramatsu <> | Subject | Re: [PATCH] x86: kprobes remove fix_riprel #ifdef |
| |
Hello Ingo,
Ingo Molnar wrote: > * Masami Hiramatsu <mhiramat@redhat.com> wrote: > >> Hello Harvey, >> >> A similar idea was already nack-ed by Ananth. >> http://sources.redhat.com/ml/systemtap/2007-q4/msg00468.html >> And I agree his thought. >> >> Especially, "riprel" does not exist on x86_32, so fix_riprel() >> is meaningless on it. >> Thus, I think it would better be ifdef'd in call-site. > > but we regularly do this in generic code: we add calls that are NOPs on > some architectures. For example flush_cache_page() makes no sense on the > x86 architecture.
Indeed. By the way, flush_cache_page() is defined as a do-while(0) on x86. Would it better to define fix_riprel() as a do-while(0) on x86-32? I think this obviously indicates that function has no effect.
> So i'm inclined to apply Harvey's cleanup - less > #ifdef complexity in higher-level code is very much favored, even > if "riprel" is a NOP concept on 32-bit.
OK, I agree about that fix_riprel() which is ifdef'd twice is too much ifdef'd. Reducing ifdef is good to me.
Thanks, > > Ingo
-- Masami Hiramatsu
Software Engineer Hitachi Computer Products (America) Inc. Software Solutions Division
e-mail: mhiramat@redhat.com, masami.hiramatsu.pt@hitachi.com
| |