Re: [PATCH] x86/mm: annotate no_context with UNWIND_HINTS

From: Josh Poimboeuf <jpoimboe@redhat.com>
To: Nathan Chancellor <natechancellor@gmail.com>
Cc: Andy Lutomirski <luto@kernel.org>,
	nick.desaulniers@gmail.com,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	"H. Peter Anvin" <hpa@zytor.com>, X86 ML <x86@kernel.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] x86/mm: annotate no_context with UNWIND_HINTS
Date: Mon, 15 Oct 2018 09:22:52 -0500	[thread overview]
Message-ID: <20181015142252.6p2zgkznj26jzdwl@treble> (raw)
In-Reply-To: <20181015051705.GA20687@flashbox>

On Sun, Oct 14, 2018 at 10:17:05PM -0700, Nathan Chancellor wrote:
> On Sun, Oct 14, 2018 at 08:43:18PM -0700, Andy Lutomirski wrote:
> > On Sun, Oct 14, 2018 at 5:37 PM Nick Desaulniers
> > <nick.desaulniers@gmail.com> wrote:
> > >
> > > Fixes the objtool warning:
> > > arch/x86/mm/fault.o: warning: objtool: no_context()+0x220: unreachable
> > > instruction
> > >
> > > Link: https://github.com/ClangBuiltLinux/linux/issues/204
> > > Signed-off-by: Nick Desaulniers <nick.desaulniers@gmail.com>
> > > ---
> > >  arch/x86/mm/fault.c | 6 ++++--
> > >  1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
> > > index 47bebfe6efa7..057d2178fa19 100644
> > > --- a/arch/x86/mm/fault.c
> > > +++ b/arch/x86/mm/fault.c
> > > @@ -760,9 +760,11 @@ no_context(struct pt_regs *regs, unsigned long error_code,
> > >                  * and then double-fault, though, because we're likely to
> > >                  * break the console driver and lose most of the stack dump.
> > >                  */
> > > -               asm volatile ("movq %[stack], %%rsp\n\t"
> > > +               asm volatile (UNWIND_HINT_SAVE
> > > +                             "movq %[stack], %%rsp\n\t"
> > >                               "call handle_stack_overflow\n\t"
> > > -                             "1: jmp 1b"
> > > +                             "1: jmp 1b\n\t"
> > > +                             UNWIND_HINT_RESTORE
> > >                               : ASM_CALL_CONSTRAINT
> > >                               : "D" ("kernel stack overflow (page fault)"),
> > >                                 "S" (regs), "d" (address),
> > 
> > NAK.  Just below this snippet is unreachable();
> > 
> > Can you reply with objtool -dr output on a problematic fault.o?  Josh,
> > it *looks* like annotate_unreachable() should be doing the right
> > thing, but something is clearly busted.
> > 
> > Also, shouldn't compiler-clang.h contain a reasonable definition of
> > unreachable()?
> > 
> > --Andy
> 
> Hi Andy,
> 
> Did you mean 'objdump -dr'? If so, here you go (rather long, sorry if I
> should have pasted it here instead):
> https://gist.github.com/nathanchance/f038bb0a6653b975bb8a4e64fcd5503e

As Andy said, I think compiler-clang.h needs an unreachable()
definition.  Without that, I'm guessing you're getting this:

  #ifndef unreachable
  # define unreachable() do { annotate_reachable(); do { } while (1); } while (0)
  #endif

which converts "unreachable" to "reachable".  That used to be needed for
pre-4.5 versions of GCC when we supported them, but it's the wrong
behavior for this piece of code (and I should remove all those old GCC
hacks soon...)

-- 
Josh