Re: [RFC][PATCH] locking: Generic ticket-lock

From: Will Deacon <will@kernel.org>
To: Catalin Marinas <catalin.marinas@arm.com>
Cc: "Peter Zijlstra" <peterz@infradead.org>,
	"Stafford Horne" <shorne@gmail.com>,
	"Guo Ren" <guoren@kernel.org>,
	"Christoph Müllner" <christophm30@gmail.com>,
	"Palmer Dabbelt" <palmer@dabbelt.com>,
	"Anup Patel" <anup@brainfault.org>,
	linux-riscv <linux-riscv@lists.infradead.org>,
	"Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
	"Guo Ren" <guoren@linux.alibaba.com>,
	"Arnd Bergmann" <arnd@arndb.de>,
	jonas@southpole.se, stefan.kristiansson@saunalahti.fi
Subject: Re: [RFC][PATCH] locking: Generic ticket-lock
Date: Thu, 15 Apr 2021 10:22:12 +0100	[thread overview]
Message-ID: <20210415092212.GA26151@willie-the-truck> (raw)
In-Reply-To: <20210415090215.GA1015@arm.com>

On Thu, Apr 15, 2021 at 10:02:18AM +0100, Catalin Marinas wrote:
> (fixed Will's email address)
> 
> On Thu, Apr 15, 2021 at 10:09:54AM +0200, Peter Zijlstra wrote:
> > On Thu, Apr 15, 2021 at 05:47:34AM +0900, Stafford Horne wrote:
> > > > How's this then? Compile tested only on openrisc/simple_smp_defconfig.
> > > 
> > > I did my testing with this FPGA build SoC:
> > > 
> > >  https://github.com/stffrdhrn/de0_nano-multicore
> > > 
> > > Note, the CPU timer sync logic uses mb() and is a bit flaky.  So missing mb()
> > > might be a reason.  I thought we had defined mb() and l.msync, but it seems to
> > > have gotten lost.
> > > 
> > > With that said I could test out this ticket-lock implementation.  How would I
> > > tell if its better than qspinlock?
> > 
> > Mostly if it isn't worse, it's better for being *much* simpler. As you
> > can see, the guts of ticket is like 16 lines of C (lock+unlock) and you
> > only need the behaviour of atomic_fetch_add() to reason about behaviour
> > of the whole thing. qspinlock OTOH is mind bending painful to reason
> > about.
> > 
> > There are some spinlock tests in locktorture; but back when I had a
> > userspace copy of the lot and would measure min,avg,max acquire times
> > under various contention loads (making sure to only run a single task
> > per CPU etc.. to avoid lock holder preemption and other such 'fun'
> > things).
> > 
> > It took us a fair amount of work to get qspinlock to compete with ticket
> > for low contention cases (by far the most common in the kernel), and it
> > took a fairly large amount of CPUs for qspinlock to really win from
> > ticket on the contended case. Your hardware may vary. In particular the
> > access to the external cacheline (for queueing, see the queue: label in
> > queued_spin_lock_slowpath) is a pain-point and the relative cost of
> > cacheline misses for your arch determines where (and if) low contention
> > behaviour is competitive.
> > 
> > Also, less variance (the reason for the min/max measure) is better.
> > Large variance is typically a sign of fwd progress trouble.
> 
> IIRC, one issue we had with ticket spinlocks on arm64 was on big.LITTLE
> systems where the little CPUs were always last to get a ticket when
> racing with the big cores. That was with load/store exclusives (LR/SC
> style) and would have probably got better with atomics but we moved to
> qspinlocks eventually (the Juno board didn't have atomics).
> 
> (leaving the rest of the text below for Will's convenience)

Yes, I think it was this thread:

https://lore.kernel.org/lkml/alpine.DEB.2.20.1707261548560.2186@nanos

but I don't think you can really fix such hardware by changing the locking
algorithm (although my proposed cpu_relax() hack was worryingly effective).

Will