Date: Mon, 7 Dec 2015 11:47:29 +0000
From: Will Deacon <will.deacon@arm.com>
To: Oleg Nesterov <oleg@redhat.com>
Cc: John Blackwood <john.blackwood@ccur.com>, linux-kernel@vger.kernel.org,
        fweisbec@gmail.com
Subject: Re: [PATCH] ARM64: Clear out any singlestep state on a ptrace detach
 operation
Message-ID: <20151207114728.GC23430@arm.com>
References: <5660A08B.6030204@ccur.com>
 <20151204100357.GA26172@arm.com>
 <20151205184840.GA6847@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20151205184840.GA6847@redhat.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1919
Lines: 57

On Sat, Dec 05, 2015 at 07:48:40PM +0100, Oleg Nesterov wrote:
> On 12/04, Will Deacon wrote:
> >
> > I hacked up a quick patch below (not even compile-tested), but I'm not
> > sure what to do about hardware {break,watch}points. Some architectures
> > explicitly clear those on detach, whereas others appear to leave them
> > alone. Thoughts?
> 
> Heh ;)
> 
> Please see fab840fc2d542fabcab "ptrace: PTRACE_DETACH should do
> flush_ptrace_hw_breakpoint(child)".
> 
> And the next "revert" commit, 35114fcbe0b9b0fa3f6653a2.

Oh, joy!

> > --- a/kernel/ptrace.c
> > +++ b/kernel/ptrace.c
> > @@ -454,13 +454,20 @@ static bool __ptrace_detach(struct task_struct *tracer, struct task_struct *p)
> >  	return dead;
> >  }
> >
> > +#ifndef arch_ptrace_detach
> > +#define arch_ptrace_detach(child)	do { } while (0)
> > +#endif
> > +
> >  static int ptrace_detach(struct task_struct *child, unsigned int data)
> >  {
> >  	if (!valid_signal(data))
> >  		return -EIO;
> >
> > -	/* Architecture-specific hardware disable .. */
> > -	ptrace_disable(child);
> > +	arch_ptrace_detach(child);
> > +	user_disable_single_step(child);
> > +#ifdef TIF_SYSCALL_EMU
> > +	clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
> > +#endif
> >  	clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
> 
> Well, personally I'd prefer to keep the arch-dependent ptrace_disable(), this
> just looks safer to me. Although I agree that its name is bad and
> arch_ptrace_detach() looks much better.

Fair enough. I don't think my patch changed any behaviour, but I can't
test it for all the architectures I touched and this area is horribly
fragile wrt userspace.

I'll merge the original patch from John.

Will
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/