From: Jeremy Fitzhardinge Subject: Re: [Xen-devel] Re: [PATCH 09/14] xen/pvticketlock: Xen implementation for PV ticket locks Date: Wed, 17 Nov 2010 01:57:40 -0800 Message-ID: <4CE3A714.9010509@goop.org> References: <4CE39C3C0200007800022AE2@vpn.id2.novell.com> <4CE397E7.2010107@goop.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4CE397E7.2010107@goop.org> Sender: linux-kernel-owner@vger.kernel.org To: Jan Beulich Cc: Jeremy Fitzhardinge , Mathieu Desnoyers , Nick Piggin , Peter Zijlstra , Linux Kernel Mailing List , Srivatsa Vaddagiri , Eric Dumazet , Xen-devel , Avi Kivity , "H. Peter Anvin" , xiyou.wangcong@gmail.com, Linux Virtualization List-Id: xen-devel@lists.xenproject.org On 11/17/2010 12:52 AM, Jeremy Fitzhardinge wrote: > On 11/17/2010 12:11 AM, Jan Beulich wrote: >>>>> On 16.11.10 at 22:08, Jeremy Fitzhardinge wrote: >>> +static void xen_lock_spinning(struct arch_spinlock *lock, unsigned want) >>> { >>> - struct xen_spinlock *xl = (struct xen_spinlock *)lock; >>> - struct xen_spinlock *prev; >>> int irq = __get_cpu_var(lock_kicker_irq); >>> - int ret; >>> + struct xen_lock_waiting *w = &__get_cpu_var(lock_waiting); >>> + int cpu = smp_processor_id(); >>> u64 start; >>> >>> /* If kicker interrupts not initialized yet, just spin */ >>> if (irq == -1) >>> - return 0; >>> + return; >>> >>> start = spin_time_start(); >>> >>> - /* announce we're spinning */ >>> - prev = spinning_lock(xl); >>> + w->want = want; >>> + w->lock = lock; >>> + >>> + /* This uses set_bit, which atomic and therefore a barrier */ >>> + cpumask_set_cpu(cpu, &waiting_cpus); >> Since you don't allow nesting, don't you need to disable >> interrupts before you touch per-CPU state? > Yes, I think you're right - interrupts need to be disabled for the bulk > of this function. Actually, on second thoughts, maybe it doesn't matter so much. The main issue is making sure that the interrupt will make the VCPU drop out of xen_poll_irq() - if it happens before xen_poll_irq(), it should leave the event pending, which will cause the poll to return immediately. I hope. Certainly disabling interrupts for some of the function will make it easier to analyze with respect to interrupt nesting. Another issue may be making sure the writes and reads of "w->want" and "w->lock" are ordered properly to make sure that xen_unlock_kick() never sees an inconsistent view of the (lock,want) tuple. The risk being that xen_unlock_kick() sees a random, spurious (lock,want) pairing and sends the kick event to the wrong VCPU, leaving the deserving one hung. J