> Yup.
>
> Adding 7 extra void*'s seems to affect only dbench, which
> whose degration with the nsproxy falls outside the noise.
> The odd thing isn't so much the degradation, but the widely
> scattered values, compared to without nsproxy.
Well dbench is filesystem heavy so it may have some need to
actually be using the filesystem namespace. Although I can't
think of why it would. Do you wan to track down that performance
regression or do you want to give up on that space optimization for
now?