From 89340d0935c9296c7b8222b6eab30e67cb57ab82 Mon Sep 17 00:00:00 2001 From: Wanpeng Li Date: Mon, 9 Sep 2019 09:40:28 +0800 Subject: Revert "locking/pvqspinlock: Don't wait if vCPU is preempted" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This patch reverts commit 75437bb304b20 (locking/pvqspinlock: Don't wait if vCPU is preempted). A large performance regression was caused by this commit. on over-subscription scenarios. The test was run on a Xeon Skylake box, 2 sockets, 40 cores, 80 threads, with three VMs of 80 vCPUs each. The score of ebizzy -M is reduced from 13000-14000 records/s to 1700-1800 records/s: Host Guest score vanilla w/o kvm optimizations upstream 1700-1800 records/s vanilla w/o kvm optimizations revert 13000-14000 records/s vanilla w/ kvm optimizations upstream 4500-5000 records/s vanilla w/ kvm optimizations revert 14000-15500 records/s Exit from aggressive wait-early mechanism can result in premature yield and extra scheduling latency. Actually, only 6% of wait_early events are caused by vcpu_is_preempted() being true. However, when one vCPU voluntarily releases its vCPU, all the subsequently waiters in the queue will do the same and the cascading effect leads to bad performance. kvm optimizations: [1] commit d73eb57b80b (KVM: Boost vCPUs that are delivering interrupts) [2] commit 266e85a5ec9 (KVM: X86: Boost queue head vCPU to mitigate lock waiter preemption) Tested-by: loobinliu@tencent.com Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Waiman Long Cc: Paolo Bonzini Cc: Radim Krčmář Cc: loobinliu@tencent.com Cc: stable@vger.kernel.org Fixes: 75437bb304b20 (locking/pvqspinlock: Don't wait if vCPU is preempted) Signed-off-by: Wanpeng Li Signed-off-by: Paolo Bonzini --- kernel/locking/qspinlock_paravirt.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel') diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h index 89bab079e7a4..e84d21aa0722 100644 --- a/kernel/locking/qspinlock_paravirt.h +++ b/kernel/locking/qspinlock_paravirt.h @@ -269,7 +269,7 @@ pv_wait_early(struct pv_node *prev, int loop) if ((loop & PV_PREV_CHECK_MASK) != 0) return false; - return READ_ONCE(prev->state) != vcpu_running || vcpu_is_preempted(prev->cpu); + return READ_ONCE(prev->state) != vcpu_running; } /* -- cgit v1.2.3