Re: [PATCH v5 seccomp 5/5] seccomp/cache: Report cache data through /proc/pid/seccomp_cache

[YiFei Zhu]

On Thu, Dec 17, 2020 at 6:14 AM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> Should there be a dependency on SECCOMP_ARCH_NATIVE?
> Should all architectures that implement seccomp have this?
>
> E.g. mips does select HAVE_ARCH_SECCOMP_FILTER, but doesn't
> have SECCOMP_ARCH_NATIVE?
>
> (noticed with preliminary out-of-tree seccomp implementation for m68k,
>  which doesn't have SECCOMP_ARCH_NATIVE

Hi Geert

You are correct. This specific patch in this series was not applied,
and this was addressed in a follow up patch series [1]. MIPS does not
define SECCOMP_ARCH_NATIVE because the bitmap expects syscall numbers
to start from 0, whereas MIPS does not (defines
CONFIG_HAVE_SPARSE_SYSCALL_NR). The follow up patch makes it so that
any arch with HAVE_SPARSE_SYSCALL_NR (currently just MIPS) cannot have
CONFIG_SECCOMP_CACHE_DEBUG on, by the depend on clause.

I see that you are doing an out of tree seccomp implementation for
m68k. Assuming unchanged arch/xtensa/include/asm/syscall.h, something
like this to arch/m68k/include/asm/seccomp.h should make it work:

#define SECCOMP_ARCH_NATIVE        AUDIT_ARCH_M68K
#define SECCOMP_ARCH_NATIVE_NR        NR_syscalls
#define SECCOMP_ARCH_NATIVE_NAME    "m68k"

If the file does not exist already, arch/xtensa/include/asm/seccomp.h
is a good example of how the file should look like, and remember to
remove `generic-y += seccomp.h` from arch/m68k/include/asm/Kbuild.

[1] https://lore.kernel.org/lkml/cover.1605101222.git.yifeifz2@xxxxxxxxxxxx/T/

YiFei Zhu