RE: why mprotect(PROT_SEM) ?

Hi Uwe

I had a discussion about this w/ Rusty Russel, the implementor of the futex
code in the kernel. Basically it boils down to: it might be needed in some
architectures, so you need to put it for all of them. If I understood
everything correctly, it is only needed for shared futexes [futexes between
clones do not need it, just when shared across different address spaces].

It might go away in the future, maybe yes, maybe not, so we still have to
keep it here. 

I am changing bits in the allocator [well, currently all my cycles are out
for some other thing-blame Intel :)] to acomodate all this, so that at the
end we can have all futexes allocated in the same page; this would take care
of PTH_SEM for that page and at the same time, reduce the pinning of memory
pages transparently for the library developer. I hope ... then mutexes
should be able to benefit directly from it.

Inaky Perez-Gonzalez -- Not speaking for Intel - opinions are my own [or my
fault]