Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1418277 > unrolled thread

Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in function pvclock_read_flags

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in  function pvclock_read_flags Peter Zijlstra <peterz@infradead.org> - 2016-06-09 14:20 +0200
    Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in  function pvclock_read_flags Peter Zijlstra <peterz@infradead.org> - 2016-06-09 14:30 +0200
    Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in  function pvclock_read_flags Paolo Bonzini <pbonzini@redhat.com> - 2016-06-09 14:30 +0200
      Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in  function pvclock_read_flags Peter Zijlstra <peterz@infradead.org> - 2016-06-09 14:40 +0200

#1418277 — Re: [PATCH 3/3 V2] pvclock: Get rid of __pvclock_read_cycles in function pvclock_read_flags

On Sat, May 28, 2016 at 08:27:43PM +0800, Minfei Huang wrote:
> +++ b/arch/x86/kernel/pvclock.c
> @@ -61,11 +61,14 @@ void pvclock_resume(void)
>  u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
>  {
>  	unsigned version;
> -	cycle_t ret;
>  	u8 flags;
>  
>  	do {
> -		version = __pvclock_read_cycles(src, &ret, &flags);
> +		version = src->version;
> +		/* Make the latest version visible */
> +		smp_rmb();
> +
> +		flags = src->flags;

Using a seqcount to load a single byte is insane ;-)

>  		/* Make sure that the version double-check is last. */
>  		smp_rmb();
>  	} while ((src->version & 1) || version != src->version);


What's wrong with:

	u8 flags = READ_ONCE(src->flags);

?

(and have the flags store be done using WRITE_ONCE() of course).

Sure, if your total state is larger than one word you need the
seqcount for integrity, but reading _one_ byte, shees.

[toc] | [next] | [standalone]

#1418280

On Thu, Jun 09, 2016 at 02:16:03PM +0200, Peter Zijlstra wrote:
> What's wrong with:
> 
> 	u8 flags = READ_ONCE(src->flags);
> 
> ?
> 
> (and have the flags store be done using WRITE_ONCE() of course).
> 
> Sure, if your total state is larger than one word you need the
> seqcount for integrity, but reading _one_ byte, shees.

Something like so; although given that I've never seen this code before
I can easily have missed an update site.

diff --git a/arch/x86/kernel/pvclock.c b/arch/x86/kernel/pvclock.c
index 99bfc025111d..7a4bf59aa929 100644
--- a/arch/x86/kernel/pvclock.c
+++ b/arch/x86/kernel/pvclock.c
@@ -60,15 +60,7 @@ void pvclock_resume(void)
 
 u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
 {
-	unsigned version;
-	cycle_t ret;
-	u8 flags;
-
-	do {
-		version = __pvclock_read_cycles(src, &ret, &flags);
-	} while ((src->version & 1) || version != src->version);
-
-	return flags & valid_flags;
+	return READ_ONCE(src->flags) & valid_flags;
 }
 
 cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
@@ -83,7 +75,8 @@ cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src)
 	} while ((src->version & 1) || version != src->version);
 
 	if (unlikely((flags & PVCLOCK_GUEST_STOPPED) != 0)) {
-		src->flags &= ~PVCLOCK_GUEST_STOPPED;
+		flags &= ~PVCLOCK_GUEST_STOPPED;
+		WRITE_ONCE(src->flags, flags);
 		pvclock_touch_watchdogs();
 	}
 
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 1ba3b7d3cae9..54458acce4f5 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -1832,7 +1832,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
 	if (use_master_clock)
 		pvclock_flags |= PVCLOCK_TSC_STABLE_BIT;
 
-	vcpu->hv_clock.flags = pvclock_flags;
+	WRITE_ONCE(vcpu->hv_clock.flags, pvclock_flags);
 
 	trace_kvm_pvclock_update(v->vcpu_id, &vcpu->hv_clock);
 

[toc] | [prev] | [next] | [standalone]

#1418283

On 09/06/2016 14:16, Peter Zijlstra wrote:
> On Sat, May 28, 2016 at 08:27:43PM +0800, Minfei Huang wrote:
>> +++ b/arch/x86/kernel/pvclock.c
>> @@ -61,11 +61,14 @@ void pvclock_resume(void)
>>  u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
>>  {
>>  	unsigned version;
>> -	cycle_t ret;
>>  	u8 flags;
>>  
>>  	do {
>> -		version = __pvclock_read_cycles(src, &ret, &flags);
>> +		version = src->version;
>> +		/* Make the latest version visible */
>> +		smp_rmb();
>> +
>> +		flags = src->flags;
> 
> Using a seqcount to load a single byte is insane ;-)

Only if you know that the writer will not write that byte twice within a
critical section...

Which I guess we do know in this case because the write side is just a
memcpy, but it's still a bit safer when it's not specified by the
pvclock API.  It's not a fast path anyway, it runs literally twice at
startup.

Paolo

[toc] | [prev] | [next] | [standalone]

#1418290

On Thu, Jun 09, 2016 at 02:26:59PM +0200, Paolo Bonzini wrote:
> 
> 
> On 09/06/2016 14:16, Peter Zijlstra wrote:
> > On Sat, May 28, 2016 at 08:27:43PM +0800, Minfei Huang wrote:
> >> +++ b/arch/x86/kernel/pvclock.c
> >> @@ -61,11 +61,14 @@ void pvclock_resume(void)
> >>  u8 pvclock_read_flags(struct pvclock_vcpu_time_info *src)
> >>  {
> >>  	unsigned version;
> >> -	cycle_t ret;
> >>  	u8 flags;
> >>  
> >>  	do {
> >> -		version = __pvclock_read_cycles(src, &ret, &flags);
> >> +		version = src->version;
> >> +		/* Make the latest version visible */
> >> +		smp_rmb();
> >> +
> >> +		flags = src->flags;
> > 
> > Using a seqcount to load a single byte is insane ;-)
> 
> Only if you know that the writer will not write that byte twice within a
> critical section...
> 
> Which I guess we do know in this case because the write side is just a
> memcpy, but it's still a bit safer when it's not specified by the
> pvclock API.  It's not a fast path anyway, it runs literally twice at
> startup.

Fair enough; just thought it was really silly code.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web