detach-sigkill-race failures with and without utrace patches

[Oleg Nesterov]

On 11/27, CAI Qian wrote:
>
> I have been consistently reproduced detach-sigkill-race failures
> on an x86_64 F12 box on the both patched and unpatched kernels.
>
> # uname -r
> 2.6.32-0.53.rc8.496.fc13.x86_64
>
> Are those expected?

This can be explained by clone() bug too, so far I think this
failure has nothing to do with ptrace.

When I reproduced this failure, the subthread was killed by
SIGILL in _dl_x86_64_restore_sse(), like it happens with the
test-case I sent in another thread.

Oleg.