Re: [PATCH 1/8] selftests: livepatch: test-syscall: Check for ARCH_HAS_SYSCALL_WRAPPER

[Miroslav Benes <mbenes@suse.cz>]

[-- Attachment #1: Type: text/plain, Size: 3208 bytes --]

On Tue, 31 Mar 2026, Marcos Paulo de Souza wrote:

> On Mon, 2026-03-16 at 16:12 -0400, Joe Lawrence wrote:
> > On Fri, Mar 13, 2026 at 05:58:32PM -0300, Marcos Paulo de Souza
> > wrote:
> > > Instead of checking if the architecture running the test was
> > > powerpc,
> > > check if CONF_ARCH_HAS_SYSCALL_WRAPPER is defined or not.
> > > 
> > > No functional changes.
> > > 
> > > Signed-off-by: Marcos Paulo de Souza <mpdesouza@suse.com>
> > > ---
> > >  tools/testing/selftests/livepatch/test_modules/test_klp_syscall.c
> > > | 7 +++----
> > >  1 file changed, 3 insertions(+), 4 deletions(-)
> > > 
> > > diff --git
> > > a/tools/testing/selftests/livepatch/test_modules/test_klp_syscall.c
> > > b/tools/testing/selftests/livepatch/test_modules/test_klp_syscall.c
> > > index dd802783ea849..c01a586866304 100644
> > > ---
> > > a/tools/testing/selftests/livepatch/test_modules/test_klp_syscall.c
> > > +++
> > > b/tools/testing/selftests/livepatch/test_modules/test_klp_syscall.c
> > > @@ -12,15 +12,14 @@
> > >  #include <linux/slab.h>
> > >  #include <linux/livepatch.h>
> > >  
> > > -#if defined(__x86_64__)
> > > +#if !defined(CONFIG_ARCH_HAS_SYSCALL_WRAPPER)
> > > +#define FN_PREFIX
> > > +#elif defined(__x86_64__)
> > >  #define FN_PREFIX __x64_
> > >  #elif defined(__s390x__)
> > >  #define FN_PREFIX __s390x_
> > >  #elif defined(__aarch64__)
> > >  #define FN_PREFIX __arm64_
> > > -#else
> > > -/* powerpc does not select ARCH_HAS_SYSCALL_WRAPPER */
> > > -#define FN_PREFIX
> > 
> > The patch does maintain the previous behavior, but I'm wondering if
> > the
> > original assertion about ARCH_HAS_SYSCALL_WRAPPER on Power was
> > correct:
> > 
> >   $ grep ARCH_HAS_SYSCALL_WRAPPER arch/powerpc/Kconfig
> >           select ARCH_HAS_SYSCALL_WRAPPER         if !SPU_BASE &&
> > !COMPAT
> >           depends on PPC64 && ARCH_HAS_SYSCALL_WRAPPER
> > 
> > Perhaps I just forgot what that additional piece of information that
> > explains the comment (highly probable these days), and if so, might
> > be
> > nice to add to this commit since I don't see it in 6a71770442b5
> > ("selftests: livepatch: Test livepatching a heavily called syscall").
> 
> Looking again at the code and at the symbols for SLE for ppc64le, I can
> say that, even with ARCH_HAS_SYSCALL_WRAPPER being set, the syscall
> names are not changed for ppc64le. Looking at
> arch/powerpc/kernel/systbl.c:
> 
> #ifdef CONFIG_ARCH_HAS_SYSCALL_WRAPPER
> #define __SYSCALL(nr, entry) [nr] = entry,
> #else
> /*
>  * Coerce syscall handlers with arbitrary parameters to common type
>  * requires cast to void* to avoid -Wcast-function-type.
>  */
> #define __SYSCALL(nr, entry) [nr] = (void *) entry,
> #endif

I think this is not the complete picture though. The definition of the 
syscall table did not change but the actual wrappers (or syscall functions 
naming before) did change a couple of times even on powerpc if I am 
reading the git history right. ARCH_HAS_SYSCALL_WRAPPER also changed how 
syscall parameters are consumed (from registers to stack frame (pt_regs)). 
It is not particularly relevant for getpid() which is SYSCALL_DEFINE0() 
but I wanted to point that out.

Miroslav