Re: [PATCH 1/9] scripts/link-vmlinux.sh: Delay orphan handling warnings until final link

From: Kees Cook <keescook@chromium.org>
To: Nicholas Piggin <npiggin@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>, Borislav Petkov <bp@suse.de>,
	Catalin Marinas <catalin.marinas@arm.com>,
	clang-built-linux@googlegroups.com,
	"H.J. Lu" <hjl.tools@gmail.com>,
	James Morse <james.morse@arm.com>,
	linux-arch@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	Russell King <linux@armlinux.org.uk>,
	linux-kbuild@vger.kernel.org, linux-kernel@vger.kernel.org,
	Mark Rutland <mark.rutland@arm.com>,
	Masahiro Yamada <masahiroy@kernel.org>,
	Peter Collingbourne <pcc@google.com>,
	Will Deacon <will@kernel.org>,
	x86@kernel.org
Subject: Re: [PATCH 1/9] scripts/link-vmlinux.sh: Delay orphan handling warnings until final link
Date: Sun, 22 Mar 2020 09:00:22 -0700	[thread overview]
Message-ID: <202003220859.E54327D98C@keescook> (raw)
In-Reply-To: <1584868418.o62lxee8k1.astroid@bobo.none>

On Sun, Mar 22, 2020 at 07:16:29PM +1000, Nicholas Piggin wrote:
> Kees Cook's on March 21, 2020 4:24 am:
> > On Fri, Mar 20, 2020 at 12:47:54PM +1000, Nicholas Piggin wrote:
> >> Kees Cook's on February 28, 2020 10:22 am:
> >> > Right now, powerpc adds "--orphan-handling=warn" to LD_FLAGS_vmlinux
> >> > to detect when there are unexpected sections getting added to the kernel
> >> > image. There is no need to report these warnings more than once, so it
> >> > can be removed until the final link stage.
> >> > 
> >> > This helps pave the way for other architectures to enable this, with the
> >> > end goal of enabling this warning by default for vmlinux for all
> >> > architectures.
> >> > 
> >> > Signed-off-by: Kees Cook <keescook@chromium.org>
> >> > ---
> >> >  scripts/link-vmlinux.sh | 6 ++++++
> >> >  1 file changed, 6 insertions(+)
> >> > 
> >> > diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
> >> > index 1919c311c149..416968fea685 100755
> >> > --- a/scripts/link-vmlinux.sh
> >> > +++ b/scripts/link-vmlinux.sh
> >> > @@ -255,6 +255,11 @@ info GEN modules.builtin
> >> >  tr '\0' '\n' < modules.builtin.modinfo | sed -n 's/^[[:alnum:]:_]*\.file=//p' |
> >> >  	tr ' ' '\n' | uniq | sed -e 's:^:kernel/:' -e 's/$/.ko/' > modules.builtin
> >> >  
> >> > +
> >> > +# Do not warn about orphan sections until the final link stage.
> >> > +saved_LDFLAGS_vmlinux="${LDFLAGS_vmlinux}"
> >> > +LDFLAGS_vmlinux="$(echo "${LDFLAGS_vmlinux}" | sed -E 's/ --orphan-handling=warn( |$)/ /g')"
> >> > +
> >> >  btf_vmlinux_bin_o=""
> >> >  if [ -n "${CONFIG_DEBUG_INFO_BTF}" ]; then
> >> >  	if gen_btf .tmp_vmlinux.btf .btf.vmlinux.bin.o ; then
> >> > @@ -306,6 +311,7 @@ if [ -n "${CONFIG_KALLSYMS}" ]; then
> >> >  	fi
> >> >  fi
> >> >  
> >> > +LDFLAGS_vmlinux="${saved_LDFLAGS_vmlinux}"
> >> >  vmlinux_link vmlinux "${kallsymso}" ${btf_vmlinux_bin_o}
> >> >  
> >> >  if [ -n "${CONFIG_BUILDTIME_TABLE_SORT}" ]; then
> >> 
> >> That's ugly. Why not just enable it for all archs?
> > 
> > It is ugly; I agree.
> > 
> > I can try to do this for all architectures, but I worry there are a
> > bunch I can't test. But I guess it would stand out. ;)
> 
> It's only warn, so it doesn't break their builds (unless there's a 
> linker error on warn option I don't know about?). We had a powerpc bug 
> that would have been caught with it as well, so it's not a bad idea to
> get everyone using it.

Well, it's bad form to add warnings to a build. I am expected to fix any
warnings before I enable a warning flag.

> I would just do it. Doesn't take much to fix.

I will do my best on the archs I can't test. :)

-- 
Kees Cook