Re: [GIT PULL] security: seccomp changes for v4.21

[Tycho Andersen]

Hi,

On Mon, Jan 07, 2019 at 01:09:09PM -0800, Kees Cook wrote:
> On Mon, Jan 7, 2019 at 2:15 AM Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> >
> >
> > * James Morris <jmorris@xxxxxxxxx> wrote:
> >
> > > From Kees:
> > >
> > > "- Add SECCOMP_RET_USER_NOTIF
> > >
> > > - seccomp fixes for sparse warnings and s390 build (Tycho)"
> > >
> > >
> > >
> > > The following changes since commit 1072bd678547f8663cfb81a22fdb50c589e4976e:
> > >
> > >   security: fs: make inode explicitly non-modular (2018-12-12 14:58:51 -0800)
> > >
> > > are available in the Git repository at:
> > >
> > >   git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security.git next-seccomp
> > >
> > > for you to fetch changes up to 55b8cbe470d103b44104c64dbf89e5cad525d4e0:
> > >
> > >   Merge tag 'seccomp-next-part2' of https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux into next-seccomp (2018-12-17 11:36:26 -0800)
> > >
> > > ----------------------------------------------------------------
> > > James Morris (2):
> > >       Merge tag 'seccomp-next' of https://git.kernel.org/.../kees/linux into next-seccomp
> > >       Merge tag 'seccomp-next-part2' of https://git.kernel.org/.../kees/linux into next-seccomp
> > >
> > > Tycho Andersen (6):
> > >       seccomp: hoist struct seccomp_data recalculation higher
> > >       seccomp: switch system call argument type to void *
> > >       seccomp: add a return code to trap to userspace
> > >       samples: add an example of seccomp user trap
> > >       seccomp: fix poor type promotion
> > >       seccomp, s390: fix build for syscall type change
> > >
> > >  Documentation/ioctl/ioctl-number.txt           |   1 +
> > >  Documentation/userspace-api/seccomp_filter.rst |  84 +++++
> > >  arch/s390/kernel/compat_wrapper.c              |   2 +-
> > >  include/linux/seccomp.h                        |   9 +-
> > >  include/linux/syscalls.h                       |   2 +-
> > >  include/uapi/linux/seccomp.h                   |  40 ++-
> > >  kernel/seccomp.c                               | 467 ++++++++++++++++++++++++-
> > >  samples/seccomp/.gitignore                     |   1 +
> > >  samples/seccomp/Makefile                       |   7 +-
> > >  samples/seccomp/user-trap.c                    | 375 ++++++++++++++++++++
> > >  tools/testing/selftests/seccomp/seccomp_bpf.c  | 447 ++++++++++++++++++++++-
> > >  11 files changed, 1411 insertions(+), 24 deletions(-)
> > >  create mode 100644 samples/seccomp/user-trap.c
> >
> > 32-bit x86 allyesconfig doesn't build:
> >
> >  /usr/bin/ld: i386:x86-64 architecture of input file `samples/seccomp/user-trap.o' is incompatible with i386 output
> >  /usr/bin/ld: samples/seccomp/user-trap.o: file class ELFCLASS64 incompatible with ELFCLASS32
> >  /usr/bin/ld: final link failed: File in wrong format
> >  collect2: error: ld returned 1 exit status
> >  scripts/Makefile.host:99: recipe for target 'samples/seccomp/user-trap' failed
> >  make[2]: *** [samples/seccomp/user-trap] Error 1
> >
> > Is this a known regression?
> 
> That looks like something is busted in the samples Makefile? Tycho,
> are you able to reproduce this?

Given that the samples build only happens when CROSS_COMPILE is not
set, I'll see about digging up a 32-bit box. But I think the patch
below should fix it, I'll send it out when I actually get it tested.

Sorry for the breakage!

Tycho