linux-kernel - Re: [RFC][PATCH 2/2] sched: Use fancy new guards

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]

Message-ID: <20230526164130.GA4053578@hirez.programming.kicks-ass.net>
Date:   Fri, 26 May 2023 18:41:30 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Greg KH <gregkh@...uxfoundation.org>
Cc:     torvalds@...ux-foundation.org, keescook@...omium.org,
        pbonzini@...hat.com, linux-kernel@...r.kernel.org,
        ojeda@...nel.org, ndesaulniers@...gle.com, mingo@...hat.com,
        will@...nel.org, longman@...hat.com, boqun.feng@...il.com,
        juri.lelli@...hat.com, vincent.guittot@...aro.org,
        dietmar.eggemann@....com, rostedt@...dmis.org, bsegall@...gle.com,
        mgorman@...e.de, bristot@...hat.com, vschneid@...hat.com,
        paulmck@...nel.org, frederic@...nel.org, quic_neeraju@...cinc.com,
        joel@...lfernandes.org, josh@...htriplett.org,
        mathieu.desnoyers@...icios.com, jiangshanlai@...il.com,
        qiang1.zhang@...el.com, rcu@...r.kernel.org, tj@...nel.org,
        tglx@...utronix.de
Subject: Re: [RFC][PATCH 2/2] sched: Use fancy new guards

On Fri, May 26, 2023 at 05:25:58PM +0100, Greg KH wrote:
> On Fri, May 26, 2023 at 05:05:51PM +0200, Peter Zijlstra wrote:
> > Convert kernel/sched/core.c to use the fancy new guards to simplify
> > the error paths.
> 
> That's slightly crazy...
> 
> I like the idea, but is this really correct:
> 
> 
> > 
> > Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
> > ---
> >  kernel/sched/core.c  | 1223 +++++++++++++++++++++++----------------------------
> >  kernel/sched/sched.h |   39 +
> >  2 files changed, 595 insertions(+), 667 deletions(-)
> > 
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -1097,24 +1097,21 @@ int get_nohz_timer_target(void)
> >  
> >  	hk_mask = housekeeping_cpumask(HK_TYPE_TIMER);
> >  
> > -	rcu_read_lock();
> > -	for_each_domain(cpu, sd) {
> > -		for_each_cpu_and(i, sched_domain_span(sd), hk_mask) {
> > -			if (cpu == i)
> > -				continue;
> > +	void_scope(rcu) {
> > +		for_each_domain(cpu, sd) {
> > +			for_each_cpu_and(i, sched_domain_span(sd), hk_mask) {
> > +				if (cpu == i)
> > +					continue;
> >  
> > -			if (!idle_cpu(i)) {
> > -				cpu = i;
> > -				goto unlock;
> > +				if (!idle_cpu(i))
> > +					return i;
> 
> You can call return from within a "scope" and it will clean up properly?

Yep, that's the main feature here.

> I tried to read the cpp "mess" but couldn't figure out how to validate
> this at all, have a set of tests for this somewhere?

I have it in userspace with printf, but yeah, I'll go make a selftest
somewhere.

One advantage of using the scheduler locks as testbed is that if you get
it wrong it burns *real* fast -- been there done that etc.

> Anyway, the naming is whack, but I don't have a proposed better name,
> except you might want to put "scope_" as the prefix not the suffix, but
> then that might look odd to, so who knows.

Yeah, naming is certainly crazy, but I figured I should get it all
working before spending too much time on that.

I can certainly do 's/lock_scope/scope_lock/g' on it all.

> But again, the idea is good, it might save us lots of "you forgot to
> clean this up on the error path" mess that we are getting constant churn
> for these days...

That's the goal...