From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from out1-smtp.messagingengine.com ([66.111.4.25]:41850 "EHLO out1-smtp.messagingengine.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1422883AbbD2Lzf (ORCPT ); Wed, 29 Apr 2015 07:55:35 -0400 Received: from compute1.internal (compute1.nyi.internal [10.202.2.41]) by mailout.nyi.internal (Postfix) with ESMTP id A07A42089A for ; Wed, 29 Apr 2015 07:55:33 -0400 (EDT) Subject: FAILED: patch "[PATCH] x86: vdso: fix pvclock races with task migration" failed to apply to 3.19-stable tree To: rkrcmar@redhat.com, pbonzini@redhat.com Cc: From: Date: Wed, 29 Apr 2015 13:55:31 +0200 Message-ID: <14303085315171@kroah.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Sender: stable-owner@vger.kernel.org List-ID: The patch below does not apply to the 3.19-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to . thanks, greg k-h ------------------ original commit in Linus's tree ------------------ >>From 80f7fdb1c7f0f9266421f823964fd1962681f6ce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Radim=20Kr=C4=8Dm=C3=A1=C5=99?= Date: Thu, 2 Apr 2015 20:44:23 +0200 Subject: [PATCH] x86: vdso: fix pvclock races with task migration MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If we were migrated right after __getcpu, but before reading the migration_count, we wouldn't notice that we read TSC of a different VCPU, nor that KVM's bug made pvti invalid, as only migration_count on source VCPU is increased. Change vdso instead of updating migration_count on destination. Cc: stable@vger.kernel.org Signed-off-by: Radim Krčmář Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"") Message-Id: <1428000263-11892-1-git-send-email-rkrcmar@redhat.com> Signed-off-by: Paolo Bonzini diff --git a/arch/x86/vdso/vclock_gettime.c b/arch/x86/vdso/vclock_gettime.c index 30933760ee5f..40d2473836c9 100644 --- a/arch/x86/vdso/vclock_gettime.c +++ b/arch/x86/vdso/vclock_gettime.c @@ -99,21 +99,25 @@ static notrace cycle_t vread_pvclock(int *mode) * __getcpu() calls (Gleb). */ - pvti = get_pvti(cpu); + /* Make sure migrate_count will change if we leave the VCPU. */ + do { + pvti = get_pvti(cpu); + migrate_count = pvti->migrate_count; - migrate_count = pvti->migrate_count; + cpu1 = cpu; + cpu = __getcpu() & VGETCPU_CPU_MASK; + } while (unlikely(cpu != cpu1)); version = __pvclock_read_cycles(&pvti->pvti, &ret, &flags); /* * Test we're still on the cpu as well as the version. - * We could have been migrated just after the first - * vgetcpu but before fetching the version, so we - * wouldn't notice a version change. + * - We must read TSC of pvti's VCPU. + * - KVM doesn't follow the versioning protocol, so data could + * change before version if we left the VCPU. */ - cpu1 = __getcpu() & VGETCPU_CPU_MASK; - } while (unlikely(cpu != cpu1 || - (pvti->pvti.version & 1) || + smp_rmb(); + } while (unlikely((pvti->pvti.version & 1) || pvti->pvti.version != version || pvti->migrate_count != migrate_count));