Re: [PATCH v2] linux-user: Fix strace output for s390x mmap()

[Philippe Mathieu-Daudé]

On 21/11/24 10:26, iii wrote:
> On 2024-11-21 10:00, Philippe Mathieu-Daudé wrote:
> > On 20/11/24 22:26, Ilya Leoshkevich wrote:
> > > print_mmap() assumes that mmap() receives arguments via memory if
> > > mmap2() is present. s390x (as opposed to s390) does not fit this
> > > pattern: it does not have mmap2(), but mmap() still receives arguments
> > > via memory.
> > >
> > > Fix by sharing the detection logic between syscall.c and strace.c.
> > >
> > > Cc: qemu-stable@nongnu.org
> > > Fixes: d971040c2d16 ("linux-user: Fix strace output for old_mmap")
> > > Suggested-by: Richard Henderson <richard.henderson@linaro.org>
> > > Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
> > > ---
> > >
> > > v1: 
> > > 20241119211138.148806-1-iii@linux.ibm.com/">https://lore.kernel.org/qemu-devel/20241119211138.148806-1-iii@linux.ibm.com/
> > > v1 -> v2: Share the detection logic between syscall.c and strace.c
> > >            (Richard).
> > >
> > >   linux-user/strace.c       | 2 +-
> > >   linux-user/syscall.c      | 5 +----
> > >   linux-user/syscall_defs.h | 7 +++++++
> > >   3 files changed, 9 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/linux-user/strace.c b/linux-user/strace.c
> > > index b70eadc19ef..9c55f39b095 100644
> > > --- a/linux-user/strace.c
> > > +++ b/linux-user/strace.c
> > > @@ -3971,7 +3971,7 @@ print_mmap(CPUArchState *cpu_env, const struct 
> > > syscallname *name,
> > >   {
> > >       return print_mmap_both(cpu_env, name, arg0, arg1, arg2, arg3,
> > >                              arg4, arg5,
> > > -#if defined(TARGET_NR_mmap2)
> > > +#ifdef TARGET_ARCH_WANT_SYS_OLD_MMAP
> >
> > We still want to print for mmap2, so:
> >
> >   #if defined(TARGET_NR_mmap2) || defined(TARGET_ARCH_WANT_SYS_OLD_MMAP)
> >
> > >                               true
> > >   #else
> > >                               false
>
> mmap2() has its own flow from what I can see:
>
> print_mmap2()
>    print_mmap_both(..., is_old_mmap=false)
>
> It should not call print_mmap(), which I'm changing here.
>
> [...]

If so, better to clean that in a previous patch.

Cc'ing Helge since I'm a bit confused by commit d971040c2d intent
("linux-user: Fix strace output for old_mmap").