Re: [musl] Re: add loongarch64 port v3

[Arnd Bergmann <arnd@kernel.org>]

On Wed, May 11, 2022 at 5:57 PM Rich Felker <dalias@libc.org> wrote:
>
> On Mon, May 09, 2022 at 09:34:59AM +0800, 王洪亮 wrote:
> >
> > 在 2022/5/5 上午11:21, 王洪亮 写道:
> > >Hi,
> > >
> > >I published loongarch64 port v3 in
> > >https://github.com/loongson/musl/tree/loongarch-v2.0.
> > >
> > >fixed two problem:
> > >
> > >1.use __NR_clone3 in  __clone().
> > >
> > >2.remove __NR_fstat and __NR_newfstatat.
> > >
> > >please code review if there are any other problems ? thanks.
> > >
> > >
> > >Hongliang Wang.
> > >
> > Hi,
> >
> > this port has any other problems need to fix?
>
> I'm at least waiting for confirmation from the kernel side that
> everything is okay with syscall usage to move forward. I'll try to
> look over the rest again soon to check that there are no remaining
> issues.

Since musl is using clone3 internally and no longer exports the
__NR_fstat/newfstatat macros, that part is fine. The clone()
one still isn't correct I think, as the mainline kernel does not yet
support the emulation with a zero-length stack length. Once
we add that part, musl should be fine, but then the __NR_clone
definition is mismatched, as the kernel won't provide that call
to user applications any more.

      Arnd