Re: [PATCH 01/20] asm-generic/mmiowb: Add generic implementation of mmiowb() tracking
From: Peter Zijlstra
Date: Thu Mar 07 2019 - 04:14:20 EST
On Thu, Mar 07, 2019 at 11:47:53AM +1100, Michael Ellerman wrote:
> The mutex unlock fast path is just:
> if (atomic_long_cmpxchg_release(&lock->owner, curr, 0UL) == curr)
> return true;
> And because it's the "release" variant we just use lwsync, which doesn't
> order MMIO. If it was just atomic_long_cmpxchg() that would work because
> we use sync for those.
> __up_write() uses atomic_long_sub_return_release(), so same story.
As does spin_unlock() of course, which is a great segway into...
my RCsc desires :-)
If all your unlocks were to have SYNC, your locks would, aside from
ordering MMIO, also be RCsc, Win-Win :-)
There is, of course, that pesky little performance detail that keeps
getting in the way.