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:   Mon, 31 Jan 2022 13:09:41 -0800
From:   Kees Cook <keescook@chromium.org>
To:     Nick Desaulniers <ndesaulniers@google.com>
Cc:     Nathan Chancellor <nathan@kernel.org>,
        linux-hardening@vger.kernel.org, llvm@lists.linux.dev,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] fortify: Update compile-time tests for Clang 14
Message-ID: <202201311306.0FC6E303@keescook>
References: <20220130181616.420092-1-keescook@chromium.org>
 <CAKwvOdmETY-j9B9vms4d31kAc0xb1yE902KijiH2baY5-M=fLQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAKwvOdmETY-j9B9vms4d31kAc0xb1yE902KijiH2baY5-M=fLQ@mail.gmail.com>
Precedence: bulk

On Mon, Jan 31, 2022 at 11:09:27AM -0800, Nick Desaulniers wrote:
> On Sun, Jan 30, 2022 at 10:16 AM Kees Cook <keescook@chromium.org> wrote:
> >
> > Clang 14 introduces support for compiletime_assert(). Update the
> > compile-time warning regex to catch Clang's variant of the warning text
> > in preparation for Clang supporting CONFIG_FORTIFY_SOURCE.
> 
> https://twitter.com/ifosteve/status/1190348262500421634?lang=en
> error messages can change over time. More thoughts below.

Sure, but I don't want the compile-time checks to silently regress,
which requires looking specifically for the error.

> > Cc: Nathan Chancellor <nathan@kernel.org>
> > Cc: Nick Desaulniers <ndesaulniers@google.com>
> > Cc: linux-hardening@vger.kernel.org
> > Cc: llvm@lists.linux.dev
> > Signed-off-by: Kees Cook <keescook@chromium.org>
> > ---
> > I'm splitting this patch out of the main Clang FORTIFY enabling patch.
> > ---
> >  scripts/test_fortify.sh | 8 ++++++--
> >  1 file changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/scripts/test_fortify.sh b/scripts/test_fortify.sh
> > index a4da365508f0..c2688ab8281d 100644
> > --- a/scripts/test_fortify.sh
> > +++ b/scripts/test_fortify.sh
> > @@ -46,8 +46,12 @@ if "$@" -Werror -c "$IN" -o "$OUT".o 2> "$TMP" ; then
> >                 status="warning: unsafe ${FUNC}() usage lacked '$WANT' symbol in $IN"
> >         fi
> >  else
> > -       # If the build failed, check for the warning in the stderr (gcc).
> > -       if ! grep -q -m1 "error: call to .\b${WANT}\b." "$TMP" ; then
> > +       # If the build failed, check for the warning in the stderr.
> > +       # GCC:
> > +       # ./include/linux/fortify-string.h:316:25: error: call to '__write_overflow_field' declared with attribute warning: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror=attribute-warning]
> > +       # Clang 14:
> > +       # ./include/linux/fortify-string.h:316:4: error: call to __write_overflow_field declared with 'warning' attribute: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror,-Wattribute-warning]
> > +       if ! grep -Eq -m1 "error: call to .?\b${WANT}\b.?" "$TMP" ; then
> 
> Doesn't this depend on -Werror being set? I guess it did so before
> hand, too, but couldn't I unset CONFIG_WERROR then this check would
> still fail (since instead of `error:` we'd have `warning:`)? If we
> used __attribute__((error(""))) then this would always be an error.
> Right now, it is only because -Werror is set promoting the warning
> diagnostic to an error.

Right, see earlier up in the script. "-Werror" is explicitly set:

line 40: if "$@" -Werror -c "$IN" -o "$OUT".o 2> "$TMP" ; then

-- 
Kees Cook