Messages in this thread | | | Date | Thu, 13 Nov 2008 09:39:35 +0100 | From | Ingo Molnar <> | Subject | Re: [PATCH 1/1] ftrace: do not update max buffer with no users |
| |
* Steven Rostedt <rostedt@goodmis.org> wrote:
> > > On Wed, 12 Nov 2008, Andrew Morton wrote: > > > > > > This sort of thing: > > > > +#ifdef CONFIG_TRACER_MAX_TRACE > > if (current_trace && current_trace->print_max) > > iter->tr = &max_tr; > > else > > +#endif > > iter->tr = inode->i_private; > > > > is particularly regrettable. > > Another answer is to not allocate it and check for NULL before using it. > But this would mean that I would have to find all the places that touch it > (the same as the ifdefs) and make sure they are safe. > > The reason I picked the ifdefs was to guarantee that there were no users > if it was not being used.
Since this is a regression, the best answer is usually the smallest possible patch. Which by all likelyhood would be the removal of CONFIG_TRACER_MAX_TRACE uses:
./kernel/trace/trace.c:#ifdef CONFIG_TRACER_MAX_TRACE ./trace.c:#ifdef CONFIG_TRACER_MAX_TRACE ./trace.c:#ifdef CONFIG_TRACER_MAX_TRACE
Note, please dont touch the Kconfig - that's just unnecessary churn that does nothing to resolve the regression. We'll have an unused symbol (and we'll fix it for real in the devel branches) but that's not an issue for .28.
Ok?
Ingo
| |