Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Michael Ellerman <mpe@ellerman.id.au>
To:     "Naveen N. Rao" <naveen.n.rao@linux.ibm.com>,
        Qian Cai <cai@lca.pw>, Russell Currey <ruscur@russell.cc>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        linuxppc-dev <linuxppc-dev@lists.ozlabs.org>,
        Nicholas Piggin <npiggin@gmail.com>,
        Steven Rostedt <rostedt@goodmis.org>
Subject: Re: POWER9 crash due to STRICT_KERNEL_RWX (WAS: Re: Linux-next POWER9 NULL pointer NIP...)
In-Reply-To: <1587106774.1oa2whm69m.naveen@linux.ibm.com>
References: <15AC5B0E-A221-4B8C-9039-FA96B8EF7C88@lca.pw> <87eeszlb6u.fsf@mpe.ellerman.id.au> <0675B22E-8F32-432C-9378-FDE159DD1729@lca.pw> <20200407093054.3eb23e45@gandalf.local.home> <EA9F9A54-87BC-477A-BE8A-7D53F80C5223@lca.pw> <20200409101413.35d9c72d@gandalf.local.home> <06A2EA93-B730-4DB1-819F-D27E7032F0B3@lca.pw> <161662E3-5D9C-4C15-919C-CFEFE4CC35CB@lca.pw> <69F0448F-CA5B-497D-B8AF-2848175B9477@lca.pw> <1587106774.1oa2whm69m.naveen@linux.ibm.com>
Date:   Fri, 17 Apr 2020 21:49:02 +1000
Message-ID: <87blnqib81.fsf@mpe.ellerman.id.au>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

"Naveen N. Rao" <naveen.n.rao@linux.ibm.com> writes:
> Hi Qian,
>
> Qian Cai wrote:
>> OK, reverted the commit,
>>=20
>> c55d7b5e6426 (=E2=80=9Cpowerpc: Remove STRICT_KERNEL_RWX incompatibility=
 with RELOCATABLE=E2=80=9D)
>>=20
>> or set STRICT_KERNEL_RWX=3Dn fixed the crash below and also mentioned in=
 this thread,
>>=20
>> https://lore.kernel.org/lkml/15AC5B0E-A221-4B8C-9039-FA96B8EF7C88@lca.pw/
>
> Do you see any errors logged in dmesg when you see the crash?=20=20
> STRICT_KERNEL_RWX changes how patch_instruction() works, so it would be=20
> interesting to see if there are any ftrace-related errors thrown before=20
> the crash.

I've been able to reproduce with STRICT_KERNEL_RWX=3Dy and concurrently
running:

# while true; do echo function > /sys/kernel/debug/tracing/current_tracer ;=
 echo nop > /sys/kernel/debug/tracing/current_tracer ; done

and:

# while true; do find /lib/modules/$(uname -r) -name '*.ko' -printf "%f\n" =
| sed -e "s/\.ko//" | xargs -i modprobe -va {}; lsmod | awk '{print $1}' | =
xargs -i modprobe -vr {}; done

ie. stressing module loading/unloading and ftrace at the same time.


It's not 100% but it usually reproduces within 10-20 minutes.

It looks like sometimes our __patch_instruction() fails, and then that
somehow leads to things getting further messed up. Presumably we have
some bad error handling somewhere.

cheers