Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:4 as permitted sender) client-ip=2620:137:e000::3:4;
MIME-Version: 1.0
References: <20231018153147.167393-1-hamza.mahfooz@amd.com> <CANpmjNPZ0Eii3ZTrVqEL2Ez0Jv23y-emLBCLSZ==xmH--4E65g@mail.gmail.com>
In-Reply-To: <CANpmjNPZ0Eii3ZTrVqEL2Ez0Jv23y-emLBCLSZ==xmH--4E65g@mail.gmail.com>
From:   Nick Desaulniers <ndesaulniers@google.com>
Date:   Wed, 18 Oct 2023 09:38:43 -0700
Message-ID: <CAKwvOdkB8EDpbHjkL=u-5_ri6c+y6EMDx2sET+EEANXoSeDDSQ@mail.gmail.com>
Subject: Re: [PATCH] lib: Kconfig: disable dynamic sanitizers for test builds
To:     Marco Elver <elver@google.com>
Cc:     Hamza Mahfooz <hamza.mahfooz@amd.com>,
        linux-kernel@vger.kernel.org,
        Rodrigo Siqueira <rodrigo.siqueira@amd.com>,
        Harry Wentland <harry.wentland@amd.com>,
        Alex Deucher <alexander.deucher@amd.com>,
        Andrey Ryabinin <ryabinin.a.a@gmail.com>,
        Alexander Potapenko <glider@google.com>,
        Andrey Konovalov <andreyknvl@gmail.com>,
        Dmitry Vyukov <dvyukov@google.com>,
        Vincenzo Frascino <vincenzo.frascino@arm.com>,
        Nathan Chancellor <nathan@kernel.org>,
        Tom Rix <trix@redhat.com>, kasan-dev@googlegroups.com,
        llvm@lists.linux.dev, Arnd Bergmann <arnd@arndb.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Oct 18, 2023 at 9:22=E2=80=AFAM Marco Elver <elver@google.com> wrot=
e:
>
> That being said, we're aware of KASAN having had more issues and there
> are some suboptions that have been disabled because of that (like
> KASAN_STACK). I'm not sure if Clang's KASAN instrumentation has had
> some recent improvements (we did investigate it, but I can't recall
> what the outcome was [1]) - maybe try a more recent compiler? However,
> KCSAN and KMSAN shouldn't have any issues (if KMSAN is enabled,
> FRAME_WARN is 0). And having build tests with them enabled isn't
> useless at all: we're making sure that these tools (even though only
> for debugging), still work. We _want_ them to work during random build
> testing!
>
> Please share the concrete problem you're having, because this change
> will make things worse for everyone in the long run.
>
> [1] https://github.com/llvm/llvm-project/issues/38157

Some recent issues I discovered in clang that exacerbate stack usage
in general, which is then amplified by KASAN:
1. https://github.com/llvm/llvm-project/issues/68746
2. https://github.com/llvm/llvm-project/issues/68747
Those are next up on my to fix list after what I'm working on now.  I
suspect that those aren't the last issues now that I've found a thread
to pull on.

--=20
Thanks,
~Nick Desaulniers