Re: [RFC][PATCH] mips: Fix arch_spin_unlock()

[Ralf Baechle]

On Tue, Feb 02, 2016 at 12:19:04AM -0800, Linus Torvalds wrote:

> Memory ordering is confusing enough as it is. We should not make
> people worry more than they already have to. Strong rules are good.

Confusing and the resulting bugs can be very hard to debug.

One of the problems I've experienced is that Linux does support liberal
memory ordering, even as extreme as the Alpha.  And every once in a
while a hardware guy is asking me if Linux their preferred variant of
weak ordering and answering honestly I have to say yes.  Even advising
against it in strong words against it, guess what will happen - another
weakly ordered core implementation.

To this point we've only fixed theoretical memory ordering issues on MIPS
but it's only a matter of time until we get bitten by a bug that does
actual damage.

SGI and a few others have managed to build large systems with significant
memory latencies yet managed to get decent performance with strong
ordering.

  Ralf