[PATCH 4/5] s390: Move files out of s390-64 folders

[Adhemerval Zanella Netto]

On 16/04/26 11:52, Stefan Liebler wrote:
> On 4/15/26 21:39, Adhemerval Zanella Netto wrote:
>>
>>
>> On 09/04/26 05:47, Stefan Liebler wrote:
>>> All the files in subfolders s390/s390-64 in sysdeps directory are moved
>>> up to the s390/ ones.  If necessary the files were merged with the existing
>>> ones.
>>>
>>> sysdeps/s390/preconfigure.ac was updated to reflect the removal of s390-64
>>> subdirectory.
>>
>> Patch looks good, just some question below if we still need some configure
>> and build directives.
>>
> SNIP...
>>> diff --git a/sysdeps/s390/Makefile b/sysdeps/s390/Makefile
>>> index 985b4f25ee..29d11dfdab 100644
>>> --- a/sysdeps/s390/Makefile
>>> +++ b/sysdeps/s390/Makefile
>>> @@ -1,3 +1,7 @@
>>> +ifeq ($(subdir),gmon)
>>> +sysdep_routines += s390x-mcount
>>> +endif
>>> +
>>>  ifeq ($(subdir),iconvdata)
>>>  ISO-8859-1_CP037_Z900-routines := iso-8859-1_cp037_z900
>>>  ISO-8859-1_CP037_Z900-map := gconv.map
>>> @@ -43,6 +47,91 @@ $(modpfx)gconv-modules-s390.conf: ../sysdeps/s390/gconv-modules-s390.conf \
>>>  endif
>>>  
>>>  ifeq ($(subdir),elf)
>>> +CFLAGS-rtld.c += -Wno-uninitialized -Wno-unused
>>> +CFLAGS-dl-load.c += -Wno-unused
>>> +CFLAGS-dl-reloc.c += -Wno-unused
>>
>> Do we still need these?
> Good question. It seems it originally comes from sysdeps/i386/Makefile
> where it is still available.I've also found:
> sysdeps/m68k/Makefile:44:CFLAGS-rtld.c += -Wno-uninitialized -Wno-unused
> 
> At least with a single build of gcc 15, I don't need it (also not on
> 31bit). I have to dig deeper in the history why it was added in the past
> and do some more tests.
> 
> As this is a change to s390-64, I would prefer to investigate first and
> create a patch independent of the s390-32 removal.
> 
> SNIP...
>>> diff --git a/sysdeps/s390/configure.ac b/sysdeps/s390/configure.ac
>>> index 82e1a4445f..3c384f9670 100644
>>> --- a/sysdeps/s390/configure.ac
>>> +++ b/sysdeps/s390/configure.ac
>>> @@ -133,5 +133,57 @@ then
>>>    AC_DEFINE(HAVE_S390_MIN_ARCH13_ZARCH_ASM_SUPPORT)
>>>  fi
>>>  
>>> -test -n "$critic_missing" && AC_MSG_ERROR([
>>> -*** $critic_missing])
>>> +# Minimal checking for static PIE support in ld.
>>> +# Compare to ld testcase/bugzilla:
>>> +# <binutils-source>/ld/testsuite/ld-elf/pr22263-1.rd
>>> +AC_CACHE_CHECK([for s390-specific static PIE requirements], \
>>> +[libc_cv_s390x_staticpie_req_runtime], [dnl
>>> +  cat > conftest1.c <<EOF
>>> +__thread int * foo;
>>> +
>>> +void
>>> +bar (void)
>>> +{
>>> +  *foo = 1;
>>> +}
>>> +
>>> +void bar2 (void);
>>> +void* bar2_ifunc (void) __asm__ ("bar2");
>>> +__asm__ (".type bar2, %gnu_indirect_function");
>>> +void* bar2_ifunc (void) { return 0; }
>>> +EOF
>>> +  cat > conftest2.c <<EOF
>>> +extern __thread int *foo;
>>> +extern void bar (void);
>>> +extern void bar2 (void);
>>> +static int x;
>>> +
>>> +int
>>> +main ()
>>> +{
>>> +  foo = &x;
>>> +  bar2 ();
>>> +  return 0;
>>> +}
>>> +EOF
>>> +  libc_cv_s390x_staticpie_req_runtime=no
>>> +  # Check if the static linker does not generate dynamic TLS relocs in PIE
>>> +  # (binutils PR ld/22263), if it accepts --no-dynamic-linker
>>> +  # (by using -static-pie), and if it adds a DT_JMPREL pointing to .rela.iplt
>>> +  # with static pie.
>>> +  if AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS $LDFLAGS -fPIE -c conftest1.c -o conftest1.o]) \
>>> +     && AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS $LDFLAGS -fPIE -c conftest2.c -o conftest2.o]) \
>>> +     && AC_TRY_COMMAND([${CC-cc} $CFLAGS $CPPFLAGS $LDFLAGS -static-pie -nostartfiles -nostdlib -fPIE -o conftest conftest1.o conftest2.o]) \
>>> +     && AC_TRY_COMMAND([! LC_ALL=C $READELF -Wr conftest | grep R_390_TLS_TPOFF] >&AS_MESSAGE_LOG_FD) \
>>> +     && AC_TRY_COMMAND([LC_ALL=C $READELF -Wd conftest | grep JMPREL >&AS_MESSAGE_LOG_FD])
>>> +  then
>>> +    libc_cv_s390x_staticpie_req_runtime=yes
>>> +  fi
>>> +  rm -rf conftest.*])
>>> +
>>> +if test $libc_cv_s390x_staticpie_req_runtime = yes; then
>>> +   # Some kernels might fail with /proc/sys/kernel/randomize_va_space set to 0
>>> +   # due not enough space for a brk call.  However, there is no reliable way to
>>> +   # test it.
>>> +   AC_DEFINE(SUPPORT_STATIC_PIE)
>>> +fi
>>
>> Now that we require binutils 2.39, do we still need these check for
>> static PIE?
>>
> For binutils-bfd all required patches are included in binutils 2.39.
> 
> From my tests in the past:
> lld had also generated TPOFF relocations, which was fixed with llvm 19
> and I think it was backported to llvm 18 release branch:
> "[lld/ELF] Avoid unnecessary TPOFF relocations in GOT for -pie (#81739)"
> https://github.com/llvm/llvm-project/commit/6f907733e65d24edad65f763fb14402464bd578b
> 
> But I have to dig deeper in recent lld versions. It seems as static-pie
> is disabled. At least when building with llvm 21, config.h claims:
> /* Define if static PIE is supported.  */
> /* #undef SUPPORT_STATIC_PIE */
> 
> For gold, there is the bug:
> "Bug 22221 - add --no-dynamic-linker option"
> https://sourceware.org/bugzilla/show_bug.cgi?id=22221
> 
> Thus I would leave the test as is for now.
> 
> 
> 
> 
> 
> Thanks for the two points. I have to check further. But as mentioned I
> would prefer to leave it as is for the s390-32 removal and do some
> follow-up work.
> Is this okay for you?

Sound good, thanks.