linux-kernel - Re: [PATCH] x86: Alias memset to __builtin

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:   Thu, 26 Mar 2020 10:18:45 -0700
From:   Nick Desaulniers <ndesaulniers@...gle.com>
To:     Clement Courbet <courbet@...gle.com>
Cc:     Nathan Chancellor <natechancellor@...il.com>,
        Kees Cook <keescook@...omium.org>,
        Joe Perches <joe@...ches.com>,
        Bernd Petrovitsch <bernd@...rovitsch.priv.at>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        Michael Ellerman <mpe@...erman.id.au>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@...nel.org>,
        Segher Boessenkool <segher@...nel.crashing.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Allison Randal <allison@...utok.net>,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
        LKML <linux-kernel@...r.kernel.org>,
        clang-built-linux <clang-built-linux@...glegroups.com>
Subject: Re: [PATCH] x86: Alias memset to __builtin_memset.

On Thu, Mar 26, 2020 at 5:38 AM Clement Courbet <courbet@...gle.com> wrote:
>
> I discussed with the original authors who added freestanding to our
> build. It turns out that it was added globally but this was just to
> to workaround powerpc not compiling under clang, but they felt the
> fix was appropriate globally.
>
> Now Nick has dug up https://lkml.org/lkml/2019/8/29/1300, which
> advises against freestanding. Also, I've did some research and
> discovered that the original reason for using freestanding for
> powerpc has been fixed here:
> https://lore.kernel.org/linuxppc-dev/20191119045712.39633-3-natechancellor@gmail.com/
>
> I'm going to remove -ffreestanding from downstream, so we don't really need
> this anymore, sorry for waisting people's time.
>
> I wonder if the freestanding fix from the aforementioned patch is really needed
> though. I think that clang is actually right to point out the issue.
> I don't see any reason why setjmp()/longjmp() are declared as taking longs
> rather than ints. The implementation looks like it only ever propagates the
> value (in longjmp) or sets it to 1 (in setjmp), and we only ever call longjmp
> with integer parameters. But I'm not a PowerPC expert, so I might
> be misreading the code.
>
>
> So it seems that we could just remove freestanding altogether and rewrite the
> code to:
>
> diff --git a/arch/powerpc/include/asm/setjmp.h b/arch/powerpc/include/asm/setjmp.h
> index 279d03a1eec6..7941ae68fe21 100644
> --- a/arch/powerpc/include/asm/setjmp.h
> +++ b/arch/powerpc/include/asm/setjmp.h
> @@ -12,7 +12,9 @@
>
>  #define JMP_BUF_LEN    23
> -extern long setjmp(long *);
> -extern void longjmp(long *, long);
> +typedef long * jmp_buf;
> +
> +extern int setjmp(jmp_buf);
> +extern void longjmp(jmp_buf, int);
>
> I'm happy to send a patch for this, and get rid of more -ffreestanding.
> Opinions ?

Yes, I think the above diff and additionally cleaning up
-ffreestanding from arch/powerpc/kernel/Makefile and
arch/powerpc/xmon/Makefile would be a much better approach.  If that's
good enough to fix the warning, I kind of can't believe we didn't spot
that in the original code review!

Actually, the god awful warning was:
./arch/powerpc/include/asm/setjmp.h:10:13: error: declaration of
built-in function 'setjmp' requires the declaration of the 'jmp_buf'
type, commonly provided in the header <setjmp.h>.
[-Werror,-Wincomplete-setjmp-declaration]
extern long setjmp(long *) __attribute__((returns_twice));
            ^
So jmp_buf was missing, wasn't used, but also the long vs int confusion.

I tested the above diff, all calls to setjmp under arch/powerpc/ just
compare the return value against 0.  Callers of longjmp just pass 1
for the final parameter. So the above changes should be no functional
change.
-- 
Thanks,
~Nick Desaulniers