linux-kernel - Re: [PATCH] arm64: Restrict ARM64_BTI

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 12 Jul 2021 10:04:54 -0700
From:   Nick Desaulniers <ndesaulniers@...gle.com>
To:     Will Deacon <will@...nel.org>,
        Nathan Chancellor <nathan@...nel.org>
Cc:     Catalin Marinas <catalin.marinas@....com>,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        clang-built-linux@...glegroups.com
Subject: Re: [PATCH] arm64: Restrict ARM64_BTI_KERNEL to clang 12.0.0 and newer

On Mon, Jul 12, 2021 at 5:24 AM Will Deacon <will@...nel.org> wrote:
>
> Hi Nathan,
>
> On Thu, Jul 08, 2021 at 05:06:27PM -0700, Nathan Chancellor wrote:
> > Commit 97fed779f2a6 ("arm64: bti: Provide Kconfig for kernel mode BTI")
> > disabled CONFIG_ARM64_BTI_KERNEL when CONFIG_GCOV_KERNEL was enabled and
> > compilign with clang because of warnings that occur because LLVM was not
>
> Typo: "compilign"
>
> > emitting PAC/BTI instructions for compiler generated functions.
> >
> > This was an adequate fix to avoid the warnings with allmodconfig until
> > commit 51c2ee6d121c ("Kconfig: Introduce ARCH_WANTS_NO_INSTR and
> > CC_HAS_NO_PROFILE_FN_ATTR"), which prevented CONFIG_GCOV_KERNEL from
> > being selected with clang 12.0.0 and older because it does not support
> > the no_profile_instrument_function attribute. As a result,
> > CONFIG_ARM64_BTI_KERNEL gets enabled and there are more warnings of this
> > nature.
>
> I'm half asleep today, but I'm struggling to follow what GCOV_KERNEL has
> to do with the warnings here. Prior to 51c2ee6d121c, you could still go
> and turn that option off and so wouldn't the same warnings be triggered
> in that configuration?

Nathan, consider if any of the below can be reused to reword the commit message?
```
Prior to 51c2ee6d121c, allmodconfig with clang-11 would not select
ARM64_BTI_KERNEL because GCOV_KERNEL already had been.

After 51c2ee6d121c, allmodconfig with clang-11 does select
ARM64_BTI_KERNEL because GCOV_KERNEL cannot be selected.

clang-11 had basic support for BTI, but failed to instrument compiler
generated functions properly.  This produces the following stream of
warnings for clang-11 allmodconfig builds of ARCH=arm64:

warning: some functions compiled with BTI and some compiled without BTI
warning: not setting BTI in feature flags

This was fixed in clang-12 via https://reviews.llvm.org/D85649.
```
Perhaps its of interest to find which config in (allmodconfig minus
GCOV_KERNEL) but not in (defconfig) produces these, since I don't
think we see them in defconfig builds of clang-11?
https://github.com/ClangBuiltLinux/continuous-integration2/runs/3009146591?check_suite_focus=true

>
> In other words, I think there are two aspects to this patch:
>
>         1. Removing the (now useless) !GCOV_KERNEL dependency
>         2. Requiring Clang >= version 12
>
> but the commit message doesn't really say where these warnings are coming
> from or why Clang 12 gets rid of them.
>
> Thanks,
>
> Will



-- 
Thanks,
~Nick Desaulniers