Re: linux-next: build failure after merge of the kspp tree

From: Will Deacon
Date: Fri Jul 27 2018 - 09:27:45 EST


On Fri, Jul 27, 2018 at 02:01:06PM +0100, Will Deacon wrote:
> On Fri, Jul 27, 2018 at 01:55:22PM +0100, Will Deacon wrote:
> > On Fri, Jul 27, 2018 at 08:55:11PM +1000, Stephen Rothwell wrote:
> > > On Fri, 27 Jul 2018 19:06:47 +1000 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> > > >
> > > > On Fri, 27 Jul 2018 19:02:07 +1000 Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > After merging the kspp tree, today's linux-next build (x86_64
> > > > > allmodconfig) failed like this:
> > > > >
> > > > > cc1: error: plugin stackleak_plugin should be specified before -fplugin-arg-stackleak_plugin-disable in the command line
> > > > >
> > > > > Maybe caused by commit
> > > > >
> > > > > a8b9eaddb9c0 ("gcc-plugins: Add STACKLEAK plugin for tracking the kernel stack")
> > > > >
> > > > > I have used the kspp tree from next-20180726 for today.
> > > >
> > > > Well, that obviously didn't work since the tree hasn't changed for a
> > > > few days.
> > > >
> > > > I can't see what has interacted to make this happen, so I have dropped
> > > > the kspp tree for today.
> > >
> > > Actually, it may have been caused by commit
> > >
> > > 0b3e336601b8 ("arm64: Add support for STACKLEAK gcc plugin")
> > >
> > > from the arm64 tree.
> >
> > Thanks, Stephen. I managed to reproduce this by merging for-next/kspp from
> > Kees's tree and for-next/core from the arm64 tree. The failure happens when
> > building the EFI stub, so the commit you mention above is almost certainly
> > the culprit.
> >
> > We build the stub with the following GCC invocation:
> >
> > gcc -Wp,-MD,drivers/firmware/efi/libstub/.efi-stub-helper.o.d -nostdinc -isystem /usr/lib/gcc/x86_64-linux-gnu/4.9/include -I./arch/x86/include -I./arch/x86/include/generated -I./include -I./arch/x86/include/uapi -I./arch/x86/include/generated/uapi -I./include/uapi -I./include/generated/uapi -include ./include/linux/kconfig.h -include ./include/linux/compiler_types.h -D__KERNEL__ -mcmodel=small -m64 -D__KERNEL__ -O2 -fPIC -fno-strict-aliasing -mno-red-zone -mno-mmx -mno-sse -fshort-wchar -DDISABLE_BRANCH_PROFILING -D__NO_FORTIFY -ffreestanding -fno-stack-protector -fplugin-arg-stackleak_plugin-disable -fno-builtin -DKBUILD_BASENAME='"efi_stub_helper"' -DKBUILD_MODNAME='"efi_stub_helper"' -c -o drivers/firmware/efi/libstub/.tmp_efi-stub-helper.o drivers/firmware/efi/libstub/efi-stub-helper.c
> >
> > so given that we're not passing any -fplugin= option anyway (because we
> > override KBUILD_CFLAGS for the stub), I don't understand why we need
> > to the disable option at all.
> >
> > Laura?
>
> ... ah, but arm and arm64 inherit the old KBUILD_CFLAGS via the
> cflags-$(CONFIG_ARM64) and cflags-$(CONFIG_ARM) definitions, so they
> would be the places where we need to disable the plugin.

i.e. something like the diff below.

Will

--->8

diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
index 25dd2a14560d..f3700fe08908 100644
--- a/drivers/firmware/efi/libstub/Makefile
+++ b/drivers/firmware/efi/libstub/Makefile
@@ -11,9 +11,11 @@ cflags-$(CONFIG_X86) += -m$(BITS) -D__KERNEL__ -O2 \
-fPIC -fno-strict-aliasing -mno-red-zone \
-mno-mmx -mno-sse -fshort-wchar

-cflags-$(CONFIG_ARM64) := $(subst -pg,,$(KBUILD_CFLAGS)) -fpie
+cflags-$(CONFIG_ARM64) := $(subst -pg,,$(KBUILD_CFLAGS)) -fpie \
+ $(DISABLE_STACKLEAK_PLUGIN)
cflags-$(CONFIG_ARM) := $(subst -pg,,$(KBUILD_CFLAGS)) \
- -fno-builtin -fpic -mno-single-pic-base
+ -fno-builtin -fpic -mno-single-pic-base \
+ $(DISABLE_STACKLEAK_PLUGIN)

cflags-$(CONFIG_EFI_ARMSTUB) += -I$(srctree)/scripts/dtc/libfdt

@@ -21,7 +23,6 @@ KBUILD_CFLAGS := $(cflags-y) -DDISABLE_BRANCH_PROFILING \
-D__NO_FORTIFY \
$(call cc-option,-ffreestanding) \
$(call cc-option,-fno-stack-protector) \
- $(DISABLE_STACKLEAK_PLUGIN)

GCOV_PROFILE := n
KASAN_SANITIZE := n