On Fri, 20 Mar 2015, David Ahern wrote:
Here's another data point: If I disable NUMA I don't see the problem.
Performance drops, but no NULL pointer splats which would have been panics.
The 128 cpu ldom with NUMA enabled shows the problem every single time I do a
kernel compile (-j 128). With NUMA disabled I have done 3 allyesconfig
compiles without hitting the problem. I'll put the compiles into a loop while
I head out for dinner.
It might be helpful to enable CONFIG_DEBUG_SLAB if you're reproducing it.