linux-kernel - Re: [PATCH v2] sched/sysctl: Fix attributes of some extern declarations

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: <20171031223157.GF96615@google.com>
Date:   Tue, 31 Oct 2017 15:31:57 -0700
From:   Matthias Kaehlcke <mka@...omium.org>
To:     Ingo Molnar <mingo@...nel.org>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>
Cc:     Shile Zhang <shile.zhang@...ia.com>,
        Peter Zijlstra <peterz@...radead.org>,
        linux-kernel@...r.kernel.org,
        Nick Desaulniers <nick.desaulniers@...il.com>,
        Douglas Anderson <dianders@...omium.org>,
        Guenter Roeck <groeck@...omium.org>
Subject: Re: [PATCH v2] sched/sysctl: Fix attributes of some extern
 declarations

El Tue, Oct 30, 2017 at 10:57:58AM +0100 Ingo Molnar ha dit:

> * Matthias Kaehlcke <mka@...omium.org> wrote:
> 
> > The definition of sysctl_sched_migration_cost, sysctl_sched_nr_migrate
> > and sysctl_sched_time_avg includes the attribute const_debug. This
> > attribute is not part of the extern declaration of these variables in
> > include/linux/sched/sysctl.h, as a result clang generates warnings like
> > this:
> > 
> >   kernel/sched/sched.h:1618:33: warning: section attribute is specified on
> >     redeclared variable [-Wsection]
> >   extern const_debug unsigned int sysctl_sched_time_avg;
> >                                 ^
> >   ./include/linux/sched/sysctl.h:42:21: note: previous declaration is here
> >   extern unsigned int sysctl_sched_time_avg;
> > 
> > The header only declares the variables when CONFIG_SCHED_DEBUG is defined,
> > therefore it is not necessary to duplicate the definition of const_debug.
> > Instead we can use the attribute __read_mostly, which is the expansion of
> > const_debug when CONFIG_SCHED_DEBUG is set.
> > 
> > Signed-off-by: Matthias Kaehlcke <mka@...omium.org>
> > Reviewed-by: Nick Desaulniers <nick.desaulniers@...il.com>
> > ---
> > Changes in v2:
> > - removed pointless include of linux/static_key.h
> > - added Reviewed-by tag
> > 
> >  include/linux/sched/sysctl.h | 6 +++---
> >  1 file changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
> > index 0f5ecd4d298e..d34c823f3d36 100644
> > --- a/include/linux/sched/sysctl.h
> > +++ b/include/linux/sched/sysctl.h
> > @@ -37,9 +37,9 @@ extern unsigned int sysctl_numa_balancing_scan_period_max;
> >  extern unsigned int sysctl_numa_balancing_scan_size;
> >  
> >  #ifdef CONFIG_SCHED_DEBUG
> > -extern unsigned int sysctl_sched_migration_cost;
> > -extern unsigned int sysctl_sched_nr_migrate;
> > -extern unsigned int sysctl_sched_time_avg;
> > +extern __read_mostly unsigned int sysctl_sched_migration_cost;
> > +extern __read_mostly unsigned int sysctl_sched_nr_migrate;
> > +extern __read_mostly unsigned int sysctl_sched_time_avg;
> 
> So I hate this change, because it pointlessly duplicates an attribute that should 
> only matter at the definition site.

It's certainly not ideal, and then again essentially the same is done
in kernel/sched/sched.h, just that here the specific attribute is
hidden behind const_debug.

> The Clang warning:
> 
> >   kernel/sched/sched.h:1618:33: warning: section attribute is specified on
> >     redeclared variable [-Wsection]
> 
> suggests that the -Wsection warning can be turned off. The Clang build should
> probably do that.

That can definitely be done. I don't have a really strong opinion on
this, just wonder if the warning could be useful in other circumstances
to spot conflicting declarations. This instance looks like a somewhat
special case with the declarations in two header files and it might be
worth to pay the price of the pointlessly duplicate attribute to get a
useful warning in others.

Adding kbuild maintainer Masahiro Yamada to the thread, who tends to
be reluctant about disabling warnings globally.