[LTP] Cleanup IA-64 related code?

[LTP] Cleanup IA-64 related code?

[Li Wang]

Hi all,

I noticed that the mainline kernel is declaring to dropp support for
Itanium IA-64, actually, the commit[1] has been merged in v6.7-rc2.

I am now thinking should we do the same action for LTP in the
main branch to keep consistent with the mainline kernel?

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=cf8e8658100d4eae80ce9b21f7a81cb024dd5057

[liwang@liwang-workstation ltp]$$ git grep __ia64__
include/lapi/rt_sigaction.h:#if !defined(__ia64__) && !defined(__alpha__)
&& !defined(__hppa__) && !defined(__mips__)
lib/cloner.c:#if defined(__ia64__)
lib/cloner.c: *   2. __ia64__ takes bottom of stack and uses clone2
lib/cloner.c:#if defined(__ia64__)
lib/tst_arch.c:#elif defined(__ia64__)
testcases/kernel/mem/hugetlb/hugemmap/hugemmap05.c:#ifdef __ia64__
testcases/kernel/mem/hugetlb/hugemmap/hugemmap07.c:#elif defined(__ia64__)
testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
|| defined(__powerpc64__) || defined(__ia64__) || \
testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
|| defined(__powerpc64__) || defined(__ia64__) || \
testcases/kernel/mem/shmt/shmt09.c:#ifdef __ia64__
testcases/kernel/mem/shmt/shmt09.c:#if defined (__ia64__) ||
defined(__mips__) || defined(__hppa__) || defined(__arm__) ||
defined(__aarch64__)
testcases/kernel/sched/cfs-scheduler/hackbench.c:#ifndef __ia64__
testcases/kernel/syscalls/mmap/mmap03.c:#if defined(__ia64__) ||
defined(__hppa__) || defined(__mips__)
testcases/kernel/syscalls/mmap/mmap15.c:#ifdef __ia64__
testcases/kernel/syscalls/ptrace/ptrace.h:#ifdef __ia64__ /* what a pos */
testcases/misc/crash/crash02.c:#if defined(__ia64__)
testcases/misc/crash/crash02.c:#endif /* __ia64__ */


-- 
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] Cleanup IA-64 related code?

[Helge Deller]

On 11/20/23 09:11, Li Wang wrote:
> Hi all,
>
> I noticed that the mainline kernel is declaring to dropp support for
> Itanium IA-64, actually, the commit[1] has been merged in v6.7-rc2.
>
> I am now thinking should we do the same action for LTP in the
> main branch to keep consistent with the mainline kernel?

Upstream kernel dropped IA-64 for future releases, but it's still being
used the next few years with existing kernels, machines and distributions,
so I suggest to not *pro-actively* dropping existing code yet.
E.g. it's still supported on debian until end of the current debian releases.

Just my 2c.

Helge


> [1]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=cf8e8658100d4eae80ce9b21f7a81cb024dd5057
>
> [liwang@liwang-workstation ltp]$$ git grep __ia64__
> include/lapi/rt_sigaction.h:#if !defined(__ia64__) && !defined(__alpha__)
> && !defined(__hppa__) && !defined(__mips__)
> lib/cloner.c:#if defined(__ia64__)
> lib/cloner.c: *   2. __ia64__ takes bottom of stack and uses clone2
> lib/cloner.c:#if defined(__ia64__)
> lib/tst_arch.c:#elif defined(__ia64__)
> testcases/kernel/mem/hugetlb/hugemmap/hugemmap05.c:#ifdef __ia64__
> testcases/kernel/mem/hugetlb/hugemmap/hugemmap07.c:#elif defined(__ia64__)
> testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
> || defined(__powerpc64__) || defined(__ia64__) || \
> testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
> || defined(__powerpc64__) || defined(__ia64__) || \
> testcases/kernel/mem/shmt/shmt09.c:#ifdef __ia64__
> testcases/kernel/mem/shmt/shmt09.c:#if defined (__ia64__) ||
> defined(__mips__) || defined(__hppa__) || defined(__arm__) ||
> defined(__aarch64__)
> testcases/kernel/sched/cfs-scheduler/hackbench.c:#ifndef __ia64__
> testcases/kernel/syscalls/mmap/mmap03.c:#if defined(__ia64__) ||
> defined(__hppa__) || defined(__mips__)
> testcases/kernel/syscalls/mmap/mmap15.c:#ifdef __ia64__
> testcases/kernel/syscalls/ptrace/ptrace.h:#ifdef __ia64__ /* what a pos */
> testcases/misc/crash/crash02.c:#if defined(__ia64__)
> testcases/misc/crash/crash02.c:#endif /* __ia64__ */
>
>


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] Cleanup IA-64 related code?

[Petr Vorel]

Hi Helge, Li,

> On 11/20/23 09:11, Li Wang wrote:
> > Hi all,

> > I noticed that the mainline kernel is declaring to dropp support for
> > Itanium IA-64, actually, the commit[1] has been merged in v6.7-rc2.

> > I am now thinking should we do the same action for LTP in the
> > main branch to keep consistent with the mainline kernel?

> Upstream kernel dropped IA-64 for future releases, but it's still being
> used the next few years with existing kernels, machines and distributions,
> so I suggest to not *pro-actively* dropping existing code yet.
> E.g. it's still supported on debian until end of the current debian releases.

> Just my 2c.

I would keep it at least in this upcoming release. I also wonder if anybody
tests IA-64 with new LTP, i.e. how long we should keep it in LTP before
removing.

BTW I removed some IA-64 some years ago during some test rewrite (too long ago
now to search for it). At the time I was not aware that anybody would test IA-64
with LTP.

Kind regards,
Petr

> Helge


> > [1]
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=cf8e8658100d4eae80ce9b21f7a81cb024dd5057

> > [liwang@liwang-workstation ltp]$$ git grep __ia64__
> > include/lapi/rt_sigaction.h:#if !defined(__ia64__) && !defined(__alpha__)
> > && !defined(__hppa__) && !defined(__mips__)
> > lib/cloner.c:#if defined(__ia64__)
> > lib/cloner.c: *   2. __ia64__ takes bottom of stack and uses clone2
> > lib/cloner.c:#if defined(__ia64__)
> > lib/tst_arch.c:#elif defined(__ia64__)
> > testcases/kernel/mem/hugetlb/hugemmap/hugemmap05.c:#ifdef __ia64__
> > testcases/kernel/mem/hugetlb/hugemmap/hugemmap07.c:#elif defined(__ia64__)
> > testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
> > || defined(__powerpc64__) || defined(__ia64__) || \
> > testcases/kernel/mem/hugetlb/hugemmap/hugemmap15.c:#if defined(__powerpc__)
> > || defined(__powerpc64__) || defined(__ia64__) || \
> > testcases/kernel/mem/shmt/shmt09.c:#ifdef __ia64__
> > testcases/kernel/mem/shmt/shmt09.c:#if defined (__ia64__) ||
> > defined(__mips__) || defined(__hppa__) || defined(__arm__) ||
> > defined(__aarch64__)
> > testcases/kernel/sched/cfs-scheduler/hackbench.c:#ifndef __ia64__
> > testcases/kernel/syscalls/mmap/mmap03.c:#if defined(__ia64__) ||
> > defined(__hppa__) || defined(__mips__)
> > testcases/kernel/syscalls/mmap/mmap15.c:#ifdef __ia64__
> > testcases/kernel/syscalls/ptrace/ptrace.h:#ifdef __ia64__ /* what a pos */
> > testcases/misc/crash/crash02.c:#if defined(__ia64__)
> > testcases/misc/crash/crash02.c:#endif /* __ia64__ */

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] Cleanup IA-64 related code?

[Li Wang]

Hi All,

On Wed, Jan 10, 2024 at 1:20 AM Petr Vorel <petr.vorel@gmail.com> wrote:

> Hi Helge, Li,
>
> > On 11/20/23 09:11, Li Wang wrote:
> > > Hi all,
>
> > > I noticed that the mainline kernel is declaring to dropp support for
> > > Itanium IA-64, actually, the commit[1] has been merged in v6.7-rc2.
>
> > > I am now thinking should we do the same action for LTP in the
> > > main branch to keep consistent with the mainline kernel?
>
> > Upstream kernel dropped IA-64 for future releases, but it's still being
> > used the next few years with existing kernels, machines and
> distributions,
> > so I suggest to not *pro-actively* dropping existing code yet.
> > E.g. it's still supported on debian until end of the current debian
> releases.
>
> > Just my 2c.
>
> I would keep it at least in this upcoming release. I also wonder if anybody
> tests IA-64 with new LTP, i.e. how long we should keep it in LTP before
> removing.
>

Good question. I have the same query.

From what I think, we'd better keep the arch support list consistent
with the mainline kernel. If some additional distros (e.g Debian)
hope to extend the life of IA64 support, choosing the old LTP release
package should work.

The latest main branch of LTP maintains appropriate compatibility
and don't want too much, which might be the right way to develop.


-- 
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp