On Mon, Feb 26, 2018 at 09:00:43AM -0800, Linus Torvalds wrote: > On Mon, Feb 26, 2018 at 8:24 AM, Will Deacon <[email protected]> wrote: > > > > Strictly speaking, that's not what we've got implemented on arm64: only > > the read part of the RmW has Acquire semantics, but there is a total > > order on the lock/unlock operations for the lock. > > Hmm. > > I thought we had exactly that bug on some architecture with the queued > spinlocks, and people decided it was wrong.
So ARM64 and Power have the acquire-on-load only thing, but qspinlock has it per construction on anything that allowes reordering stores. Given that unlock/lock are ordered, which covers about 99% of the users out there, and fixing the issue would make things significantly slower on the weak architectures we let it be. But yes, its a pesky detail.
