linux-kernel - fms extension (Was: [PATCH] fs/pipe: stop duplicating union pipe

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]

Message-ID: <20251030-meerjungfrau-getrocknet-7b46eacc215d@brauner>
Date: Thu, 30 Oct 2025 14:23:01 +0100
From: Christian Brauner <brauner@...nel.org>
To: Nathan Chancellor <nathan@...nel.org>, 
	Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Ard Biesheuvel <ardb@...nel.org>, linux-efi@...r.kernel.org, 
	Rasmus Villemoes <linux@...musvillemoes.dk>, Alexander Viro <viro@...iv.linux.org.uk>, 
	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org, linux-kbuild@...r.kernel.org, 
	David Sterba <dsterba@...e.com>
Subject: fms extension (Was: [PATCH] fs/pipe: stop duplicating union
 pipe_index declaration)

On Wed, Oct 29, 2025 at 04:30:57PM -0700, Nathan Chancellor wrote:
> On Thu, Oct 30, 2025 at 12:13:11AM +0100, Christian Brauner wrote:
> > I'm fine either way. @Nathan, if you just want to give Linus the patch
> > if it's small enough or just want to give me a stable branch I can pull
> > I'll be content. Thanks!
> 
> I do not care either way but I created a shared branch/tag since it was
> easy enough to do. If Linus wants to take these directly for -rc4, I am
> fine with that as well.
> 
> Cheers,
> Nathan
> 
> The following changes since commit 3a8660878839faadb4f1a6dd72c3179c1df56787:
> 
>   Linux 6.18-rc1 (2025-10-12 13:42:36 -0700)
> 
> are available in the Git repository at:
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/kbuild/linux.git tags/kbuild-ms-extensions-6.19

Thanks, I pulled this and placed it into a branch that I can base other
branches on.

_But_, I'm already running into problems. :)

I'm changing a struct ns_common in ns_common.h (struct ns_common) and
wanted to make use of the fms extensions. ns_common.h is heavily
included by virtue of the namespace stuff. So we get an include chain
like the following:

In file included from ./include/linux/cgroup.h:23,
                 from ./include/linux/memcontrol.h:13,
                 from ./include/linux/swap.h:9,
                 from ./include/asm-generic/tlb.h:15,
                 from ./arch/x86/include/asm/tlb.h:8,
                 from ./arch/x86/include/asm/efi.h:7,
                 from drivers/firmware/efi/libstub/x86-stub.c:13:
./include/linux/ns_common.h:132:31: error: declaration does not declare anything [-Werror]
  132 |                 struct ns_tree;
      |                               ^
./include/linux/ns_common.h: In function '__ns_ref_active_read':
./include/linux/ns_common.h:228:31: error: 'const struct ns_common' has no member named '__ns_ref_active'
  228 |         return atomic_read(&ns->__ns_ref_active);
      |                               ^~
In file included from ./arch/x86/include/asm/bug.h:108,
                 from ./arch/x86/include/asm/alternative.h:9,
                 from ./arch/x86/include/asm/segment.h:6,
                 from ./arch/x86/include/asm/ptrace.h:5,
                 from ./arch/x86/include/asm/math_emu.h:5,
                 from ./arch/x86/include/asm/processor.h:13,
                 from ./arch/x86/include/asm/timex.h:5,
                 from ./include/linux/timex.h:67,
                 from ./include/linux/time32.h:13,
                 from ./include/linux/time.h:60,
                 from ./include/linux/efi.h:17,
                 from drivers/firmware/efi/libstub/x86-stub.c:9:

Because struct cgroup_namespace embeddds struct ns_common and it
proliferates via mm stuff into the efi code.

So the EFI cod has it's own KBUILD_CFLAGS. It does:

# non-x86 reuses KBUILD_CFLAGS, x86 does not
cflags-y			:= $(KBUILD_CFLAGS)

<snip>

KBUILD_CFLAGS			:= $(subst $(CC_FLAGS_FTRACE),,$(cflags-y)) \
				   -Os -DDISABLE_BRANCH_PROFILING \
				   -include $(srctree)/include/linux/hidden.h \
				   -D__NO_FORTIFY \
				   -ffreestanding \
				   -fno-stack-protector \
				   $(call cc-option,-fno-addrsig) \
				   -D__DISABLE_EXPORTS

which means x86 doesn't get -fms-extension breaking the build. If I
manually insert:

diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
index 94b05e4451dd..4ad2f8f42134 100644
--- a/drivers/firmware/efi/libstub/Makefile
+++ b/drivers/firmware/efi/libstub/Makefile
@@ -42,6 +42,8 @@ KBUILD_CFLAGS                 := $(subst $(CC_FLAGS_FTRACE),,$(cflags-y)) \
                                   -ffreestanding \
                                   -fno-stack-protector \
                                   $(call cc-option,-fno-addrsig) \
+                                  -fms-extensions \
+                                  -Wno-microsoft-anon-tag \
                                   -D__DISABLE_EXPORTS

The build works...

I think we need to decide how to fix this now because as soon as someone
makes use of the extension that is indirectly included by that libstub
thing we're fscked.