Re: [2/2] x86/unwind: Make CONFIG_UNWINDER_ORC=y the default in kconfig for 64-bit

From: Josh Poimboeuf <jpoimboe@redhat.com>
To: Andrei Vagin <avagin@virtuozzo.com>
Cc: Ingo Molnar <mingo@kernel.org>,
	linux-kernel@vger.kernel.org, torvalds@linux-foundation.org,
	peterz@infradead.org, hpa@zytor.com, tglx@linutronix.de
Subject: Re: [2/2] x86/unwind: Make CONFIG_UNWINDER_ORC=y the default in kconfig for 64-bit
Date: Thu, 19 Oct 2017 20:28:04 -0500	[thread overview]
Message-ID: <20171020012804.vbqsrajia3ficwzm@treble> (raw)
In-Reply-To: <20171019223522.GA4499@outlook.office365.com>

On Thu, Oct 19, 2017 at 03:35:22PM -0700, Andrei Vagin wrote:
> On Thu, Oct 19, 2017 at 01:16:55PM -0500, Josh Poimboeuf wrote:
> > On Thu, Oct 19, 2017 at 09:51:04AM -0700, Andrei Vagin wrote:
> > > Hi,
> > > 
> > > We run CRIU tests for tip/auto-latest regularly, and a few days ago our
> > > test job started to detect this warning in a kernel log:
> > > 
> > > [   44.235786] WARNING: can't dereference iret registers at ffff8801c5f17fe0 for ip ffffffff95f0d94b
> > > 
> > > What does it mean? How critical is it?
> > > 
> > > Our test job fails if it detects any warning in a kernel log. Maybe we
> > > need to investigate reasons of this warning and try to eliminate it?
> > > 
> > > Here are logs:
> > > https://travis-ci.org/avagin/linux/jobs/289676634
> > 
> > I think it means the unwinder found some bad ORC unwinder metadata.  Any
> > chance you have access to the kernel binary?  I need to know what code
> > corresponds to that ffffffff95f0d94b address.
> > 
> > Or if you can reproduce with the following patch, that should help:
> > 
> > 
> > diff --git a/arch/x86/kernel/unwind_orc.c b/arch/x86/kernel/unwind_orc.c
> > index 570b70d3f604..95b633f0ce51 100644
> > --- a/arch/x86/kernel/unwind_orc.c
> > +++ b/arch/x86/kernel/unwind_orc.c
> > @@ -448,7 +448,7 @@ bool unwind_next_frame(struct unwind_state *state)
> >  
> >  	case ORC_TYPE_REGS_IRET:
> >  		if (!deref_stack_regs(state, sp, &state->ip, &state->sp, false)) {
> > -			orc_warn("can't dereference iret registers at %p for ip %p\n",
> > +			orc_warn("can't dereference iret registers at %p for ip %pB\n",
> >  				 (void *)sp, (void *)orig_ip);
> >  			goto done;
> >  		}
> 
> I applied your patch and rerun tests.
> 
> [   44.947699] WARNING: can't dereference iret registers at ffff880178f5ffe0 for ip int3+0x5b/0x60

Thanks, that was enough for me to figure it out.  Can you test the below fix?

> and now here is a warning from kasan:
> 
> [  477.775676] ==================================================================
> [  477.775845] BUG: KASAN: stack-out-of-bounds in deref_stack_reg+0x11d/0x150

The KASAN warning is a known issue for which the fix is a little more
complicated.  v1 of the patch was here:

  https://lkml.kernel.org/r/cover.1507128293.git.jpoimboe@redhat.com

diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index 49167258d587..f6cdb7a1455e 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -808,7 +808,7 @@ apicinterrupt IRQ_WORK_VECTOR			irq_work_interrupt		smp_irq_work_interrupt
 
 .macro idtentry sym do_sym has_error_code:req paranoid=0 shift_ist=-1
 ENTRY(\sym)
-	UNWIND_HINT_IRET_REGS offset=8
+	UNWIND_HINT_IRET_REGS offset=\has_error_code*8
 
 	/* Sanity check */
 	.if \shift_ist != -1 && \paranoid == 0
