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  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:   Tue, 25 Sep 2018 13:11:04 +0100
From:   Mark Rutland <mark.rutland@....com>
To:     Christophe LEROY <christophe.leroy@....fr>
Cc:     Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        aneesh.kumar@...ux.vnet.ibm.com, npiggin@...il.com,
        Thomas Gleixner <tglx@...utronix.de>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        John Stultz <john.stultz@...aro.org>,
        Stephen Boyd <sboyd@...nel.org>, linux-fsdevel@...r.kernel.org,
        linuxppc-dev@...ts.ozlabs.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH v1 1/9] timer: fix circular header dependency

On Tue, Sep 25, 2018 at 07:34:15AM +0200, Christophe LEROY wrote:
> 
> 
> Le 24/09/2018 à 17:52, Christophe Leroy a écrit :
> > When switching powerpc to CONFIG_THREAD_INFO_IN_TASK, include/sched.h
> > has to be included in asm/smp.h for the following change, in order
> > to avoid uncomplete definition of task_struct:
> > 
> > -#define raw_smp_processor_id() (current_thread_info()->cpu)
> > +#define raw_smp_processor_id() (current->cpu)
> > 
> > But this generates the following compilation error, due to circular
> > header dependency.
> > 
> >    CC      kernel/time/alarmtimer.o
> > In file included from ./arch/powerpc/include/asm/smp.h:31,
> >                   from ./include/linux/smp.h:64,
> >                   from ./include/linux/percpu.h:7,
> >                   from ./include/linux/hrtimer.h:22,
> >                   from kernel/time/alarmtimer.c:19:
> > ./include/linux/sched.h:558:19: error: field 'dl_timer' has incomplete type
> >    struct hrtimer   dl_timer;
> >                     ^~~~~~~~
> > ./include/linux/sched.h:567:17: error: field 'inactive_timer' has incomplete type
> >    struct hrtimer inactive_timer;
> >                   ^~~~~~~~~~~~~~
> > make[1]: *** [kernel/time/alarmtimer.o] Error 1
> > make: *** [kernel/time/alarmtimer.o] Error 2
> > 
> >    CC      fs/timerfd.o
> > In file included from ./arch/powerpc/include/asm/smp.h:31,
> >                   from ./include/linux/smp.h:64,
> >                   from ./include/linux/percpu.h:7,
> >                   from ./include/linux/hrtimer.h:22,
> >                   from ./include/linux/alarmtimer.h:6,
> >                   from fs/timerfd.c:12:
> > ./include/linux/sched.h:558:19: error: field 'dl_timer' has incomplete type
> >    struct hrtimer   dl_timer;
> >                     ^~~~~~~~
> > ./include/linux/sched.h:567:17: error: field 'inactive_timer' has incomplete type
> >    struct hrtimer inactive_timer;
> >                   ^~~~~~~~~~~~~~
> > make[1]: *** [fs/timerfd.o] Error 1
> > make: *** [fs/timerfd.o] Error 2
> > 
> > This patch fixes it by including linux/hrtimer.h after linux/sched.h
> > 
> > Signed-off-by: Christophe Leroy <christophe.leroy@....fr>
> > ---
> >   Should it be fixed in powerpc instead ? In that case, how ? Any idea ?
> 
> 
> Looks like there are several other places where the problem occurs, so it
> has to be fixed in the powerpc headers instead.
> 
> Seems like RISC arch faced the same issue, and fixed it the following way:
> 
> /*
>  * This is particularly ugly: it appears we can't actually get the
> definition
>  * of task_struct here, but we need access to the CPU this task is running
> on.
>  * Instead of using C we're using asm-offsets.h to get the current processor
>  * ID.
>  */
> #define raw_smp_processor_id() (*((int*)((char*)get_current() +
> TASK_TI_CPU)))
> 
> 
> Unless someone has a better idea, I'll fixed it that way.

To solve this on arm64 we made the cpu number a percpu variable; see
commit:

  57c82954e77fa12c ("arm64: make cpu number a percpu variable")

IIUC, you could do that by placing it in paca_struct.

Thanks,
Mark.

Powered by blists - more mailing lists