Re: ld.lld: error: vmlinux.a(af_vsock.o at 1186296) <inline asm>:5:8: unpredictable STXR instruction, status is also a source

[Nathan Chancellor]

On Fri, Jul 10, 2026 at 04:22:20PM +0200, Stefano Garzarella wrote:
> On Mon, 6 Jul 2026 at 10:44, kernel test robot <lkp@xxxxxxxxx> wrote:
> >
> > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> > head:   8cdeaa50eae8dad34885515f62559ee83e7e8dda
> > commit: 102eab95f025b4d3f3a6c0a858400aca2af2fe52 vsock: lock down child_ns_mode as write-once
> > date:   4 months ago
> > config: arm64-randconfig-001-20260706 (https://download.01.org/0day-ci/archive/20260706/202607061644.jD93AjWc-lkp@xxxxxxxxx/config)
> > compiler: clang version 17.0.6 (https://github.com/llvm/llvm-project 6009708b4367171ccdbf4b5905cb6a803753fe18)
> > reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260706/202607061644.jD93AjWc-lkp@xxxxxxxxx/reproduce)
> >
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > the same patch/commit), kindly add following tags
> > | Fixes: 102eab95f025 ("vsock: lock down child_ns_mode as write-once")
> > | Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > | Closes: https://lore.kernel.org/oe-kbuild-all/202607061644.jD93AjWc-lkp@xxxxxxxxx/
> >
> > All errors (new ones prefixed by >>):
> >
> > >> ld.lld: error: vmlinux.a(af_vsock.o at 1186296) <inline asm>:5:8: unpredictable STXR instruction, status is also a source
> >    stlxr   w8, w8, [x21]
> >    ^
> >
> 
> Could this be related to https://github.com/llvm/llvm-project/issues/106380 ?
> That issue seems fixed in llvm >= 20

Indeed, this is resolved by

  https://github.com/llvm/llvm-project/commit/7183771834d9035ffbedd8f1ff9233b16722b986

in LLVM 20.1.0, I confirmed with a reverse bisect. It looks like
this issue triggers on undefined values being passed to inline asm but I
don't see where there is an undefined value in the blamed commit but I
did not look too hard. Presumably this is around the try_cmpxchg() in
vsock_net_set_child_mode(). I am not sure this will be easily to
workaround so we might just have to leave it broken.

-- 
Cheers,
Nathan