query for idea about a kexec regression

[Dave Young]

Hi

While I testing efi runtime patches for kexec kernel, there's
2 machines kexec fails on them. It can be reproduced with
non-efi kernel, so it's not related to the patchset.

Problem is kexec kernel will immediatly reboot after below
kernel message:
smpboot: Booting Node   0, Processors  #   1

The latter CPUs were not printed.

nr_cpus=1 works, so there might be some bug on smp code.

Any idea or suspection acout some patches? This is not 100%
reproducible, to me it's about 30% thus it's hard to bisect.

Thanks
Dave