Re: [PATCH v2 2/2] scripts: ftrace - move the nop-processing in ftrace_init to compile time

[Yinan Liu]

在 2021/11/16 下午9:07, Steven Rostedt 写道:
> On Tue, 16 Nov 2021 09:10:20 +0100
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> > On Tue, Nov 16, 2021 at 10:49:42AM +0800, Yinan Liu wrote:
> > > In some business scenarios, GCC versions are so old that
> > > optimizations in ftrace cannot be completed, such as
> > > -mrecord-mcount and -mnop-mcount. The recordmCount in the
> > > kernel is actually used. In this case, ftrace_init will
> > > consume a period of time, usually around 9~12ms. Do nop
> > > substitution in recordmcount.c to speed up ftrace_init.
> >
> > I really don't buy this.. if you can build a fresh kernel, you can
> > install a fresh gcc too -- and if you care about performance that's a
> > very good idea anyway.
>
> I'm not sure this is true for all archs, is it? That is, is the nop
> substitution available in all archs that support mcount updates. Some
> (most) archs are special, because they have to deal with link registers and
> such.
>
> And because of that, I'm not sure the patch works for all those archs.
>
> -- Steve

At present, I have only verified it under x86. In other cases, I am 
short of arch or scenarios. I am not sure whether the patch is 
applicable to all the architectures, perhaps adding something to x86?
Do you think it makes sense?

In fact, many companies cannot upgrade GCC due to online business 
reasons. This patch will help.




Best regards!
--Yinan liu