linux-hardening - Re: [PATCH] lib/test_string.c: Add test for strlen()

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 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:   Thu, 3 Feb 2022 11:50:34 -0800
From:   Nick Desaulniers <ndesaulniers@...gle.com>
To:     Geert Uytterhoeven <geert@...ux-m68k.org>,
        Kees Cook <keescook@...omium.org>
Cc:     Guenter Roeck <linux@...ck-us.net>, Peter Rosin <peda@...ntia.se>,
        Andy Shevchenko <andy@...nel.org>,
        Matteo Croce <mcroce@...rosoft.com>,
        Nathan Chancellor <nathan@...nel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        llvm@...ts.linux.dev, linux-hardening@...r.kernel.org
Subject: Re: [PATCH] lib/test_string.c: Add test for strlen()

On Thu, Feb 3, 2022 at 10:10 AM Geert Uytterhoeven <geert@...ux-m68k.org> wrote:
>
> Hi Kees,
>
> On Thu, Feb 3, 2022 at 6:15 PM Kees Cook <keescook@...omium.org> wrote:
> > On Thu, Feb 03, 2022 at 09:04:22AM +0100, Geert Uytterhoeven wrote:
> > > Not if -ffreestanding, which is what several architectures are
> > > using nowadays, to a.o. prevent gcc from replacing calls to stdlib
> > > functions to other stdlib functions (e.g. strncat() -> strlen() +
> > > store, strncmp() -> strcmp()), which breaks linking if the latter is
> > > only provided inline.
> >
> > Hah, for i386:
> >
> > arch/x86/Makefile
> >         # temporary until string.h is fixed
> >         KBUILD_CFLAGS += -ffreestanding
> >
> > This "temporary" is from 2006. ;)\

IIRC I sent a patch removing that. Yeah, Kees even signed off on it.
https://lore.kernel.org/lkml/20200817220212.338670-5-ndesaulniers@google.com/#t
I still think that's the right way to go, perhaps worth a resend to
rekick discussions.

>
> And before that, we had it in the main Makefile.
>
> > 6edfba1b33c7 ("[PATCH] x86_64: Don't define string functions to builtin")
> >
> > Removing that appears to solve it, and appears to build correctly. I'll
> > continue testing.
> >
> > > It works after dropping -ffreestanding.
> >
> > I wonder if the other architectures were just copying x86?
>
> At least on m68k it was added because gcc added new optimizations
> that broke if an architecture provides some functions as inline.
> As the kernel is not supposed to be linked with the standard C
> library, -ffreestanding should be correct, isn't it?

The kernel does not link against a libc; but it does provide many
symbols that libc would provide, with the same or similar enough
semantics that I would strongly recommend we _don't_ use
-ffreestanding in order to get such libcall optimizations (there are a
lot; see https://github.com/llvm/llvm-project/blob/main/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
for some examples) and simply use -fno-builtin-* when necessary, or
fix the kernel implementations individually.

-- 
Thanks,
~Nick Desaulniers