Date:   Mon, 13 Apr 2020 12:12:58 -0700
From:   Nick Desaulniers <ndesaulniers@...gle.com>
To:     Al Viro <viro@...iv.linux.org.uk>
Cc:     Oleg Nesterov <oleg@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        Peter Zijlstra <peterz@...radead.org>,
        clang-built-linux <clang-built-linux@...glegroups.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        "H. Peter Anvin" <hpa@...or.com>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@...nel.org>,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Andy Lutomirski <luto@...nel.org>,
        Sami Tolvanen <samitolvanen@...gle.com>,
        Marco Elver <elver@...gle.com>,
        Brian Gerst <brgerst@...il.com>, Arnd Bergmann <arnd@...db.de>,
        Andrew Morton <akpm@...ux-foundation.org>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] x86: signal: move save_altstack_ex out of generic headers

On Sat, Apr 4, 2020 at 10:06 AM Al Viro <viro@...iv.linux.org.uk> wrote:
>
> On Sat, Apr 04, 2020 at 06:01:00PM +0200, Oleg Nesterov wrote:
> > On 04/03, Nick Desaulniers wrote:
> > >
> > > --- a/arch/x86/kernel/signal.c
> > > +++ b/arch/x86/kernel/signal.c
> > > @@ -416,6 +416,7 @@ static int __setup_rt_frame(int sig, struct ksignal *ksig,
> > >     return 0;
> > >  Efault:
> > >     user_access_end();
> > > +   reset_altstack();
> > >     return -EFAULT;
> > >  }
> > >  #else /* !CONFIG_X86_32 */
> > > @@ -507,6 +508,7 @@ static int __setup_rt_frame(int sig, struct ksignal *ksig,
> > >
> > >  Efault:
> > >     user_access_end();
> > > +   reset_altstack();
> > >     return -EFAULT;
> > >  }
> >
> > I must have missed something, but this looks just wrong.
> >
> > reset_altstack() should be called when __setup_rt_frame() (and
> > unsafe_save_altstack() in particular) succeeds, not when it fails.
> >
> > Nevermind, Al has already suggested to use signal_delivered()...
>
> FWIW, I propose to do is the patch below (against the current mainline);
> objections?
>
> Don't do sas_ss_reset() until we are certain that sigframe won't be abandoned
>
> Currently we handle SS_AUTODISARM as soon as we have stored the
> altstack settings into sigframe - that's the point when we have
> set the things up for eventual sigreturn to restore the old settings.
> And if we manage to set the sigframe up (we are not done with that
> yet), everything's fine.  However, in case of failure we end up
> with sigframe-to-be abandoned and SIGSEGV force-delivered.  And
> in that case we end up with inconsistent rules - late failures
> have altstack reset, early ones do not.
>
> It's trivial to get consistent behaviour - just handle SS_AUTODISARM
> once we have set the sigframe up and are committed to entering
> the handler, i.e. in signal_delivered().
>
> Signed-off-by: Al Viro <viro@...iv.linux.org.uk>

Tested-by: Nick Desaulniers <ndesaulniers@...gle.com>
I'll follow up on our KASAN implementation turning two consecutive
longs into memset, which is odd.  Some of the recent changes to
unsafe_put_user to pass a label are triggering new
error: inline assembly requires more registers than available
in arch/x86/ia32/ia32_signal.c for me, but they're unrelated to this patch.

> ---
> diff --git a/include/linux/compat.h b/include/linux/compat.h
> index 0480ba4db592..f614967374f5 100644
> --- a/include/linux/compat.h
> +++ b/include/linux/compat.h
> @@ -461,8 +461,6 @@ int __compat_save_altstack(compat_stack_t __user *, unsigned long);
>                         &__uss->ss_sp, label); \
>         unsafe_put_user(t->sas_ss_flags, &__uss->ss_flags, label); \
>         unsafe_put_user(t->sas_ss_size, &__uss->ss_size, label); \
> -       if (t->sas_ss_flags & SS_AUTODISARM) \
> -               sas_ss_reset(t); \
>  } while (0);
>
>  /*
> diff --git a/include/linux/signal.h b/include/linux/signal.h
> index 05bacd2ab135..28fe9cc134f7 100644
> --- a/include/linux/signal.h
> +++ b/include/linux/signal.h
> @@ -450,8 +450,6 @@ int __save_altstack(stack_t __user *, unsigned long);
>         unsafe_put_user((void __user *)t->sas_ss_sp, &__uss->ss_sp, label); \
>         unsafe_put_user(t->sas_ss_flags, &__uss->ss_flags, label); \
>         unsafe_put_user(t->sas_ss_size, &__uss->ss_size, label); \
> -       if (t->sas_ss_flags & SS_AUTODISARM) \
> -               sas_ss_reset(t); \
>  } while (0);
>
>  #ifdef CONFIG_PROC_FS
> diff --git a/kernel/signal.c b/kernel/signal.c
> index e58a6c619824..4cfe0b9af588 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -2769,6 +2769,8 @@ static void signal_delivered(struct ksignal *ksig, int stepping)
>         if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
>                 sigaddset(&blocked, ksig->sig);
>         set_current_blocked(&blocked);
> +       if (current->sas_ss_flags & SS_AUTODISARM)
> +               sas_ss_reset(current);
>         tracehook_signal_handler(stepping);
>  }
>
> @@ -4070,11 +4072,7 @@ int __save_altstack(stack_t __user *uss, unsigned long sp)
>         int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
>                 __put_user(t->sas_ss_flags, &uss->ss_flags) |
>                 __put_user(t->sas_ss_size, &uss->ss_size);
> -       if (err)
> -               return err;
> -       if (t->sas_ss_flags & SS_AUTODISARM)
> -               sas_ss_reset(t);
> -       return 0;
> +       return err;
>  }
>
>  #ifdef CONFIG_COMPAT
> @@ -4129,11 +4127,7 @@ int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
>                          &uss->ss_sp) |
>                 __put_user(t->sas_ss_flags, &uss->ss_flags) |
>                 __put_user(t->sas_ss_size, &uss->ss_size);
> -       if (err)
> -               return err;
> -       if (t->sas_ss_flags & SS_AUTODISARM)
> -               sas_ss_reset(t);
> -       return 0;
> +       return err;
>  }
>  #endif
>


-- 
Thanks,
~Nick Desaulniers

Powered by blists - more mailing lists