* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-02 18:44 [PATCH] x86: vdso: fix pvclock races with task migration Radim Krčmář
@ 2015-04-02 18:59 ` Andy Lutomirski
2015-04-06 20:07 ` Andy Lutomirski
2015-04-06 20:57 ` Marcelo Tosatti
` (2 subsequent siblings)
3 siblings, 1 reply; 9+ messages in thread
From: Andy Lutomirski @ 2015-04-02 18:59 UTC (permalink / raw)
To: Radim Krčmář
Cc: linux-kernel, kvm list, Paolo Bonzini, Marcelo Tosatti, stable
On Thu, Apr 2, 2015 at 11:44 AM, Radim Krčmář <rkrcmar@redhat.com> wrote:
> 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.
Looks good to me.
--Andy
>
> Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"")
> Cc: stable@vger.kernel.org
> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
> ---
> Because it we'll get a complete rewrite, this series does not
> - remove the outdated 'TODO: We can put [...]' comment
> - use a proper encapsulation for the inner do-while loop
> - optimize the outer do-while loop
> (no need to re-read cpu id on version mismatch)
>
> arch/x86/vdso/vclock_gettime.c | 20 ++++++++++++--------
> 1 file changed, 12 insertions(+), 8 deletions(-)
>
> 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));
>
> --
> 2.3.4
>
--
Andy Lutomirski
AMA Capital Management, LLC
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-02 18:59 ` Andy Lutomirski
@ 2015-04-06 20:07 ` Andy Lutomirski
2015-04-06 22:04 ` Paolo Bonzini
0 siblings, 1 reply; 9+ messages in thread
From: Andy Lutomirski @ 2015-04-06 20:07 UTC (permalink / raw)
To: Radim Krčmář
Cc: linux-kernel, kvm list, Paolo Bonzini, Marcelo Tosatti, stable
On 04/02/2015 11:59 AM, Andy Lutomirski wrote:
> On Thu, Apr 2, 2015 at 11:44 AM, Radim Krčmář <rkrcmar@redhat.com> wrote:
>> 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.
>
> Looks good to me.
Just to check: what tree is this intended to go through? I can take it,
but not until the previous patch makes it into Linus' tree or -tip. Or
I can take both patches.
Marcelo, Paolo?
--Andy
>
> --Andy
>
>>
>> Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
>> ---
>> Because it we'll get a complete rewrite, this series does not
>> - remove the outdated 'TODO: We can put [...]' comment
>> - use a proper encapsulation for the inner do-while loop
>> - optimize the outer do-while loop
>> (no need to re-read cpu id on version mismatch)
>>
>> arch/x86/vdso/vclock_gettime.c | 20 ++++++++++++--------
>> 1 file changed, 12 insertions(+), 8 deletions(-)
>>
>> 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));
>>
>> --
>> 2.3.4
>>
>
>
>
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-06 20:07 ` Andy Lutomirski
@ 2015-04-06 22:04 ` Paolo Bonzini
0 siblings, 0 replies; 9+ messages in thread
From: Paolo Bonzini @ 2015-04-06 22:04 UTC (permalink / raw)
To: Andy Lutomirski, Radim Krčmář
Cc: linux-kernel, kvm list, Marcelo Tosatti, stable
On 06/04/2015 22:07, Andy Lutomirski wrote:
> On 04/02/2015 11:59 AM, Andy Lutomirski wrote:
>> On Thu, Apr 2, 2015 at 11:44 AM, Radim Krčmář <rkrcmar@redhat.com> wrote:
>>> 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.
>>
>> Looks good to me.
>
> Just to check: what tree is this intended to go through? I can take it,
> but not until the previous patch makes it into Linus' tree or -tip. Or
> I can take both patches.
I'll take it for 4.1.
Paolo
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-02 18:44 [PATCH] x86: vdso: fix pvclock races with task migration Radim Krčmář
2015-04-02 18:59 ` Andy Lutomirski
@ 2015-04-06 20:57 ` Marcelo Tosatti
2015-04-07 11:11 ` Paolo Bonzini
2015-04-29 12:07 ` Patch "x86: vdso: fix pvclock races with task migration" has been added to the 4.0-stable tree gregkh
3 siblings, 0 replies; 9+ messages in thread
From: Marcelo Tosatti @ 2015-04-06 20:57 UTC (permalink / raw)
To: Radim Krčmář
Cc: linux-kernel, kvm, Paolo Bonzini, Andy Lutomirski, stable
On Thu, Apr 02, 2015 at 08:44:23PM +0200, Radim Krčmář wrote:
> 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.
>
> Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"")
> Cc: stable@vger.kernel.org
> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
> ---
> Because it we'll get a complete rewrite, this series does not
> - remove the outdated 'TODO: We can put [...]' comment
> - use a proper encapsulation for the inner do-while loop
> - optimize the outer do-while loop
> (no need to re-read cpu id on version mismatch)
>
> arch/x86/vdso/vclock_gettime.c | 20 ++++++++++++--------
> 1 file changed, 12 insertions(+), 8 deletions(-)
>
> 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));
>
> --
> 2.3.4
>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-02 18:44 [PATCH] x86: vdso: fix pvclock races with task migration Radim Krčmář
2015-04-02 18:59 ` Andy Lutomirski
2015-04-06 20:57 ` Marcelo Tosatti
@ 2015-04-07 11:11 ` Paolo Bonzini
2015-04-07 12:47 ` Radim Krčmář
2015-04-29 12:07 ` Patch "x86: vdso: fix pvclock races with task migration" has been added to the 4.0-stable tree gregkh
3 siblings, 1 reply; 9+ messages in thread
From: Paolo Bonzini @ 2015-04-07 11:11 UTC (permalink / raw)
To: Radim Krčmář, linux-kernel
Cc: kvm, Marcelo Tosatti, Andy Lutomirski, stable
On 02/04/2015 20:44, Radim Krčmář wrote:
> 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.
>
> Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"")
> Cc: stable@vger.kernel.org
> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Applying this, but removing the "Fixes" tag because a guest patch cannot
fix a host patch (it can work around it or complement it).
Paolo
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH] x86: vdso: fix pvclock races with task migration
2015-04-07 11:11 ` Paolo Bonzini
@ 2015-04-07 12:47 ` Radim Krčmář
2015-04-07 13:57 ` Paolo Bonzini
0 siblings, 1 reply; 9+ messages in thread
From: Radim Krčmář @ 2015-04-07 12:47 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: linux-kernel, kvm, Marcelo Tosatti, Andy Lutomirski, stable
2015-04-07 13:11+0200, Paolo Bonzini:
> On 02/04/2015 20:44, Radim Krčmář wrote:
> > 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.
> >
> > Fixes: 0a4e6be9ca17 ("x86: kvm: Revert "remove sched notifier for cross-cpu migrations"")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
>
> Applying this, but removing the "Fixes" tag because a guest patch cannot
> fix a host patch (it can work around it or complement it).
I think it was correct. Both are guest only, the revert just missed
some races. (0a4e6be9ca17 has misleading commit message ...)
^ permalink raw reply [flat|nested] 9+ messages in thread
* Patch "x86: vdso: fix pvclock races with task migration" has been added to the 4.0-stable tree
2015-04-02 18:44 [PATCH] x86: vdso: fix pvclock races with task migration Radim Krčmář
` (2 preceding siblings ...)
2015-04-07 11:11 ` Paolo Bonzini
@ 2015-04-29 12:07 ` gregkh
3 siblings, 0 replies; 9+ messages in thread
From: gregkh @ 2015-04-29 12:07 UTC (permalink / raw)
To: rkrcmar, gregkh, pbonzini; +Cc: stable, stable-commits
This is a note to let you know that I've just added the patch titled
x86: vdso: fix pvclock races with task migration
to the 4.0-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary
The filename of the patch is:
x86-vdso-fix-pvclock-races-with-task-migration.patch
and it can be found in the queue-4.0 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@vger.kernel.org> know about it.
>From 80f7fdb1c7f0f9266421f823964fd1962681f6ce Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Radim=20Kr=C4=8Dm=C3=A1=C5=99?= <rkrcmar@redhat.com>
Date: Thu, 2 Apr 2015 20:44:23 +0200
Subject: x86: vdso: fix pvclock races with task migration
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
From: =?UTF-8?q?Radim=20Kr=C4=8Dm=C3=A1=C5=99?= <rkrcmar@redhat.com>
commit 80f7fdb1c7f0f9266421f823964fd1962681f6ce upstream.
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.
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
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 <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
arch/x86/vdso/vclock_gettime.c | 20 ++++++++++++--------
1 file changed, 12 insertions(+), 8 deletions(-)
--- 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
* __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));
Patches currently in stable-queue which might be from rkrcmar@redhat.com are
queue-4.0/x86-vdso-fix-pvclock-races-with-task-migration.patch
^ permalink raw reply [flat|nested] 9+ messages in thread