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;
Date:   Tue, 22 Jun 2021 18:57:39 +1000
From:   Nicholas Piggin <npiggin@gmail.com>
Subject: Re: [PATCH] KVM: PPC: Book3S HV: Workaround high stack usage with
 clang
To:     Michael Ellerman <mpe@ellerman.id.au>,
        Nathan Chancellor <nathan@kernel.org>,
        Paul Mackerras <paulus@ozlabs.org>
Cc:     Arnd Bergmann <arnd@arndb.de>,
        Benjamin Herrenschmidt <benh@kernel.crashing.org>,
        clang-built-linux@googlegroups.com, kvm-ppc@vger.kernel.org,
        linux-kernel@vger.kernel.org, linuxppc-dev@lists.ozlabs.org,
        kernel test robot <lkp@intel.com>,
        Nick Desaulniers <ndesaulniers@google.com>
References: <YNDUEoanTqvayZ5P@archlinux-ax161>
        <20210621182440.990242-1-nathan@kernel.org>
In-Reply-To: <20210621182440.990242-1-nathan@kernel.org>
MIME-Version: 1.0
Message-Id: <1624352117.tss3fnkwt4.astroid@bobo.none>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Excerpts from Nathan Chancellor's message of June 22, 2021 4:24 am:
> LLVM does not emit optimal byteswap assembly, which results in high
> stack usage in kvmhv_enter_nested_guest() due to the inlining of
> byteswap_pt_regs(). With LLVM 12.0.0:
>=20
> arch/powerpc/kvm/book3s_hv_nested.c:289:6: error: stack frame size of
> 2512 bytes in function 'kvmhv_enter_nested_guest' [-Werror,-Wframe-larger=
-than=3D]
> long kvmhv_enter_nested_guest(struct kvm_vcpu *vcpu)
>      ^
> 1 error generated.
>=20
> While this gets fixed in LLVM, mark byteswap_pt_regs() as
> noinline_for_stack so that it does not get inlined and break the build
> due to -Werror by default in arch/powerpc/. Not inlining saves
> approximately 800 bytes with LLVM 12.0.0:
>=20
> arch/powerpc/kvm/book3s_hv_nested.c:290:6: warning: stack frame size of
> 1728 bytes in function 'kvmhv_enter_nested_guest' [-Wframe-larger-than=3D=
]
> long kvmhv_enter_nested_guest(struct kvm_vcpu *vcpu)
>      ^
> 1 warning generated.
>=20
> Link: https://github.com/ClangBuiltLinux/linux/issues/1292
> Link: https://bugs.llvm.org/show_bug.cgi?id=3D49610
> Link: https://lore.kernel.org/r/202104031853.vDT0Qjqj-lkp@intel.com/
> Link: https://gist.github.com/ba710e3703bf45043a31e2806c843ffd
> Reported-by: kernel test robot <lkp@intel.com>
> Signed-off-by: Nathan Chancellor <nathan@kernel.org>

Seems okay to me. If it was something where performance might be=20
signficiant I guess you could ifdef on CC_IS_CLANG, but for this
it shouldn't matter.

Acked-by: Nicholas Piggin <npiggin@gmail.com>

Thanks,
Nick

> ---
>  arch/powerpc/kvm/book3s_hv_nested.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>=20
> diff --git a/arch/powerpc/kvm/book3s_hv_nested.c b/arch/powerpc/kvm/book3=
s_hv_nested.c
> index 60724f674421..1b3ff0af1264 100644
> --- a/arch/powerpc/kvm/book3s_hv_nested.c
> +++ b/arch/powerpc/kvm/book3s_hv_nested.c
> @@ -53,7 +53,8 @@ void kvmhv_save_hv_regs(struct kvm_vcpu *vcpu, struct h=
v_guest_state *hr)
>  	hr->dawrx1 =3D vcpu->arch.dawrx1;
>  }
> =20
> -static void byteswap_pt_regs(struct pt_regs *regs)
> +/* Use noinline_for_stack due to https://bugs.llvm.org/show_bug.cgi?id=
=3D49610 */
> +static noinline_for_stack void byteswap_pt_regs(struct pt_regs *regs)
>  {
>  	unsigned long *addr =3D (unsigned long *) regs;
> =20
>=20
> base-commit: 4a21192e2796c3338c4b0083b494a84a61311aaf
> --=20
> 2.32.0.93.g670b81a890
>=20
>=20