linux-kernel - Re: [PATCH v2 2/5] preempt/dynamic: Introduce preempt mode accessors

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 11 Nov 2021 10:36:17 +0100
From:   Marco Elver <elver@...gle.com>
To:     Mike Galbraith <efault@....de>
Cc:     Valentin Schneider <valentin.schneider@....com>,
        linux-kernel@...r.kernel.org, kasan-dev@...glegroups.com,
        linuxppc-dev@...ts.ozlabs.org, linux-kbuild@...r.kernel.org,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...nel.org>,
        Frederic Weisbecker <frederic@...nel.org>,
        Dmitry Vyukov <dvyukov@...gle.com>,
        Michael Ellerman <mpe@...erman.id.au>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Michal Marek <michal.lkml@...kovi.net>,
        Nick Desaulniers <ndesaulniers@...gle.com>
Subject: Re: [PATCH v2 2/5] preempt/dynamic: Introduce preempt mode accessors

On Thu, 11 Nov 2021 at 04:47, Mike Galbraith <efault@....de> wrote:
>
> On Thu, 2021-11-11 at 04:35 +0100, Mike Galbraith wrote:
> > On Thu, 2021-11-11 at 04:16 +0100, Mike Galbraith wrote:
> > > On Wed, 2021-11-10 at 20:24 +0000, Valentin Schneider wrote:
> > > >
> > > > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > > > index 5f8db54226af..0640d5622496 100644
> > > > --- a/include/linux/sched.h
> > > > +++ b/include/linux/sched.h
> > > > @@ -2073,6 +2073,22 @@ static inline void cond_resched_rcu(void)
> > > >  #endif
> > > >  }
> > > >
> > > > +#ifdef CONFIG_PREEMPT_DYNAMIC
> > > > +
> > > > +extern bool is_preempt_none(void);
> > > > +extern bool is_preempt_voluntary(void);
> > > > +extern bool is_preempt_full(void);
> > > > +
> > > > +#else
> > > > +
> > > > +#define is_preempt_none() IS_ENABLED(CONFIG_PREEMPT_NONE)
> > > > +#define is_preempt_voluntary()
> > > > IS_ENABLED(CONFIG_PREEMPT_VOLUNTARY)
> > > > +#define is_preempt_full() IS_ENABLED(CONFIG_PREEMPT)
> > >
> > > I think that should be IS_ENABLED(CONFIG_PREEMPTION), see
> > > c1a280b68d4e.
> > >
> > > Noticed while applying the series to an RT tree, where tglx
> > > has done that replacement to the powerpc spot your next patch
> > > diddles.
> >
> > Damn, then comes patch 5 properly differentiating PREEMPT/PREEMPT_RT.
>
> So I suppose the powerpc spot should remain CONFIG_PREEMPT and become
> CONFIG_PREEMPTION when the RT change gets merged, because that spot is
> about full preemptibility, not a distinct preemption model.
>
> That's rather annoying :-/

I guess the question is if is_preempt_full() should be true also if
is_preempt_rt() is true?

Not sure all cases are happy with that, e.g. the kernel/trace/trace.c
case, which wants to print the precise preemption level.

To avoid confusion, I'd introduce another helper that says true if the
preemption level is "at least full", currently that'd be "full or rt".
Something like is_preempt_full_or_rt() (but might as well write
"is_preempt_full() || is_preempt_rt()"), or is_preemption() (to match
that Kconfig variable, although it's slightly confusing). The
implementation of that helper can just be a static inline function
returning "is_preempt_full() || is_preempt_rt()".

Would that help?