Messages in this thread | | | Date | Fri, 01 Jun 2012 21:52:05 +0900 | From | Masami Hiramatsu <> | Subject | Re: [PATCH 1/5] ftrace: Synchronize variable setting with breakpoints |
| |
(2012/06/01 20:37), Steven Rostedt wrote: > On Fri, 2012-06-01 at 13:53 +0900, Masami Hiramatsu wrote: > >> Hmm, why don't we have two text_poke interfaces for single and >> batch? As like dyn_ftrace, since modifying massive points takes >> a lot time, so we may have additional kconfig something like >> CONFIG_QUICK_BATCH_TEXT_POKE which switches text area to rw while >> batch text_poke. >> > > I'll be working on patches to consolidate the two after I get everything > else working :-) I still need to work on the ftrace kprobe stuff too. > > I hate having a config option to switch between the two, except for > something that can be there if we find the new approach is buggy (like > we have with lockdep). That may be a solution for this if we don't agree > on one now. That is, bring back stop_machine() when LOCKDEP is enabled. > But that should only be a temporary work around not a true fix. > > I have no problem with having most of the modifying code be shared > between text_poke and ftrace. I guess the question is, do we want to do > it only with the FIXMAP or do we want text_poke and ftrace to use the rw > method for large batches. Heck, we can set a limit. If we are going to > update more that 100 locations, we switch the kernel to rw, otherwise we > do the FIXMAP update.
That's reasonable for me :). You can feel free to change/remove text_poke_smp_batch which is already old-style interface.
Thanks!
-- Masami HIRAMATSU Software Platform Research Dept. Linux Technology Center Hitachi, Ltd., Yokohama Research Laboratory E-mail: masami.hiramatsu.pt@hitachi.com
| |