Re: [RFC patch 2/5] trace event skb fix unassigned field

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Frederic Weisbecker]

On Tue, Jan 04, 2011 at 06:46:06PM -0500, nhorman@tuxdriver.com wrote:
> Acked- by: Neil Horman <nhorman@tuxdriver.com>
> 
> 
> Sent from my Verizon Wireless Phone
> 
> ----- Reply message -----
> From: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>
> Date: Tue, Jan 4, 2011 6:16 pm
> Subject: [RFC patch 2/5] trace event skb fix unassigned field
> To: "LKML" <linux-kernel@vger.kernel.org>
> Cc: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frederic Weisbecker" <fweisbec@gmail.com>, "Ingo Molnar" <mingo@elte.hu>, "Neil Horman" <nhorman@tuxdriver.com>, "Thomas Gleixner" <tglx@linutronix.de>
> 
> 
> The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
> leaving its trace output as garbage. Assign the value to 0 when skb is NULL
> instead.

Hm, if the skb is already null, we probably shouldn't send any trace.

What about using TP_CONDITION() ?

> 
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> CC: Steven Rostedt <rostedt@goodmis.org>
> CC: Frederic Weisbecker <fweisbec@gmail.com>
> CC: Ingo Molnar <mingo@elte.hu>
> CC: Neil Horman <nhorman@tuxdriver.com>
> CC: Thomas Gleixner <tglx@linutronix.de>
> ---
>  include/trace/events/skb.h |    4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> Index: linux-2.6-lttng/include/trace/events/skb.h
> ===================================================================
> --- linux-2.6-lttng.orig/include/trace/events/skb.h
> +++ linux-2.6-lttng/include/trace/events/skb.h
> @@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
>  
>  	TP_fast_assign(
>  		__entry->skbaddr = skb;
> -		if (skb) {
> -			__entry->protocol = ntohs(skb->protocol);
> -		}
> +		__entry->protocol = skb ? ntohs(skb->protocol) : 0;
>  		__entry->location = location;
>  	),
>  
> 
> 

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Mathieu Desnoyers]

* Frederic Weisbecker (fweisbec@gmail.com) wrote:
> On Tue, Jan 04, 2011 at 06:46:06PM -0500, nhorman@tuxdriver.com wrote:
> > Acked- by: Neil Horman <nhorman@tuxdriver.com>
> > 
> > 
> > Sent from my Verizon Wireless Phone
> > 
> > ----- Reply message -----
> > From: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>
> > Date: Tue, Jan 4, 2011 6:16 pm
> > Subject: [RFC patch 2/5] trace event skb fix unassigned field
> > To: "LKML" <linux-kernel@vger.kernel.org>
> > Cc: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frederic Weisbecker" <fweisbec@gmail.com>, "Ingo Molnar" <mingo@elte.hu>, "Neil Horman" <nhorman@tuxdriver.com>, "Thomas Gleixner" <tglx@linutronix.de>
> > 
> > 
> > The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
> > leaving its trace output as garbage. Assign the value to 0 when skb is NULL
> > instead.
> 
> Hm, if the skb is already null, we probably shouldn't send any trace.
> 
> What about using TP_CONDITION() ?

Hrm, let's see. It's been introduced by commit
5cb3d1d9d34ac04bcaa2034139345b2a5fea54c1
by Zhaolei.

Event at the time of that commit, the only caller looked like:

void kfree_skb(struct sk_buff *skb)
{
        if (unlikely(!skb))
                return;
        if (likely(atomic_read(&skb->users) == 1))
                smp_rmb();
        else if (likely(!atomic_dec_and_test(&skb->users)))
                return;
        trace_kfree_skb(skb, __builtin_return_address(0));
        __kfree_skb(skb);
}
EXPORT_SYMBOL(kfree_skb);

So it already checks for a null pointer before calling the tracepoint. This
leads me to wonder why why this check was added in the first place ?

Mathieu

> 
> > 
> > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> > CC: Steven Rostedt <rostedt@goodmis.org>
> > CC: Frederic Weisbecker <fweisbec@gmail.com>
> > CC: Ingo Molnar <mingo@elte.hu>
> > CC: Neil Horman <nhorman@tuxdriver.com>
> > CC: Thomas Gleixner <tglx@linutronix.de>
> > ---
> >  include/trace/events/skb.h |    4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> > 
> > Index: linux-2.6-lttng/include/trace/events/skb.h
> > ===================================================================
> > --- linux-2.6-lttng.orig/include/trace/events/skb.h
> > +++ linux-2.6-lttng/include/trace/events/skb.h
> > @@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
> >  
> >  	TP_fast_assign(
> >  		__entry->skbaddr = skb;
> > -		if (skb) {
> > -			__entry->protocol = ntohs(skb->protocol);
> > -		}
> > +		__entry->protocol = skb ? ntohs(skb->protocol) : 0;
> >  		__entry->location = location;
> >  	),
> >  
> > 
> > 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Frederic Weisbecker]

On Tue, Jan 04, 2011 at 07:40:38PM -0500, Mathieu Desnoyers wrote:
> * Frederic Weisbecker (fweisbec@gmail.com) wrote:
> > On Tue, Jan 04, 2011 at 06:46:06PM -0500, nhorman@tuxdriver.com wrote:
> > > Acked- by: Neil Horman <nhorman@tuxdriver.com>
> > > 
> > > 
> > > Sent from my Verizon Wireless Phone
> > > 
> > > ----- Reply message -----
> > > From: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>
> > > Date: Tue, Jan 4, 2011 6:16 pm
> > > Subject: [RFC patch 2/5] trace event skb fix unassigned field
> > > To: "LKML" <linux-kernel@vger.kernel.org>
> > > Cc: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frederic Weisbecker" <fweisbec@gmail.com>, "Ingo Molnar" <mingo@elte.hu>, "Neil Horman" <nhorman@tuxdriver.com>, "Thomas Gleixner" <tglx@linutronix.de>
> > > 
> > > 
> > > The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
> > > leaving its trace output as garbage. Assign the value to 0 when skb is NULL
> > > instead.
> > 
> > Hm, if the skb is already null, we probably shouldn't send any trace.
> > 
> > What about using TP_CONDITION() ?
> 
> Hrm, let's see. It's been introduced by commit
> 5cb3d1d9d34ac04bcaa2034139345b2a5fea54c1
> by Zhaolei.
> 
> Event at the time of that commit, the only caller looked like:
> 
> void kfree_skb(struct sk_buff *skb)
> {
>         if (unlikely(!skb))
>                 return;
>         if (likely(atomic_read(&skb->users) == 1))
>                 smp_rmb();
>         else if (likely(!atomic_dec_and_test(&skb->users)))
>                 return;
>         trace_kfree_skb(skb, __builtin_return_address(0));
>         __kfree_skb(skb);
> }
> EXPORT_SYMBOL(kfree_skb);
> 
> So it already checks for a null pointer before calling the tracepoint. This
> leads me to wonder why why this check was added in the first place ?

Likely for no strong reasons :)

So I guess we can remove the check from the tracepoint?

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Mathieu Desnoyers]

* Frederic Weisbecker (fweisbec@gmail.com) wrote:
> On Tue, Jan 04, 2011 at 07:40:38PM -0500, Mathieu Desnoyers wrote:
> > * Frederic Weisbecker (fweisbec@gmail.com) wrote:
> > > On Tue, Jan 04, 2011 at 06:46:06PM -0500, nhorman@tuxdriver.com wrote:
> > > > Acked- by: Neil Horman <nhorman@tuxdriver.com>
> > > > 
> > > > 
> > > > Sent from my Verizon Wireless Phone
> > > > 
> > > > ----- Reply message -----
> > > > From: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>
> > > > Date: Tue, Jan 4, 2011 6:16 pm
> > > > Subject: [RFC patch 2/5] trace event skb fix unassigned field
> > > > To: "LKML" <linux-kernel@vger.kernel.org>
> > > > Cc: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frederic Weisbecker" <fweisbec@gmail.com>, "Ingo Molnar" <mingo@elte.hu>, "Neil Horman" <nhorman@tuxdriver.com>, "Thomas Gleixner" <tglx@linutronix.de>
> > > > 
> > > > 
> > > > The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
> > > > leaving its trace output as garbage. Assign the value to 0 when skb is NULL
> > > > instead.
> > > 
> > > Hm, if the skb is already null, we probably shouldn't send any trace.
> > > 
> > > What about using TP_CONDITION() ?
> > 
> > Hrm, let's see. It's been introduced by commit
> > 5cb3d1d9d34ac04bcaa2034139345b2a5fea54c1
> > by Zhaolei.
> > 
> > Event at the time of that commit, the only caller looked like:
> > 
> > void kfree_skb(struct sk_buff *skb)
> > {
> >         if (unlikely(!skb))
> >                 return;
> >         if (likely(atomic_read(&skb->users) == 1))
> >                 smp_rmb();
> >         else if (likely(!atomic_dec_and_test(&skb->users)))
> >                 return;
> >         trace_kfree_skb(skb, __builtin_return_address(0));
> >         __kfree_skb(skb);
> > }
> > EXPORT_SYMBOL(kfree_skb);
> > 
> > So it already checks for a null pointer before calling the tracepoint. This
> > leads me to wonder why why this check was added in the first place ?
> 
> Likely for no strong reasons :)
> 
> So I guess we can remove the check from the tracepoint?

Yep, leading to this patch instead:

trace event skb remove duplicate null-pointer check

The check for NULL skb in the kfree_skb trace event is a duplicate from the
check already done in its only caller, kfree_skb(). Remove this duplicate check.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Neil Horman <nhorman@tuxdriver.com>
---
 include/trace/events/skb.h |    4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

Index: linux-2.6-lttng/include/trace/events/skb.h
===================================================================
--- linux-2.6-lttng.orig/include/trace/events/skb.h
+++ linux-2.6-lttng/include/trace/events/skb.h
@@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
 
 	TP_fast_assign(
 		__entry->skbaddr = skb;
-		if (skb) {
-			__entry->protocol = ntohs(skb->protocol);
-		}
+		__entry->protocol = ntohs(skb->protocol);
 		__entry->location = location;
 	),
 


-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Neil Horman]

On Tue, Jan 04, 2011 at 08:21:23PM -0500, Mathieu Desnoyers wrote:
> * Frederic Weisbecker (fweisbec@gmail.com) wrote:
> > On Tue, Jan 04, 2011 at 07:40:38PM -0500, Mathieu Desnoyers wrote:
> > > * Frederic Weisbecker (fweisbec@gmail.com) wrote:
> > > > On Tue, Jan 04, 2011 at 06:46:06PM -0500, nhorman@tuxdriver.com wrote:
> > > > > Acked- by: Neil Horman <nhorman@tuxdriver.com>
> > > > > 
> > > > > 
> > > > > Sent from my Verizon Wireless Phone
> > > > > 
> > > > > ----- Reply message -----
> > > > > From: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>
> > > > > Date: Tue, Jan 4, 2011 6:16 pm
> > > > > Subject: [RFC patch 2/5] trace event skb fix unassigned field
> > > > > To: "LKML" <linux-kernel@vger.kernel.org>
> > > > > Cc: "Mathieu Desnoyers" <mathieu.desnoyers@efficios.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frederic Weisbecker" <fweisbec@gmail.com>, "Ingo Molnar" <mingo@elte.hu>, "Neil Horman" <nhorman@tuxdriver.com>, "Thomas Gleixner" <tglx@linutronix.de>
> > > > > 
> > > > > 
> > > > > The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
> > > > > leaving its trace output as garbage. Assign the value to 0 when skb is NULL
> > > > > instead.
> > > > 
> > > > Hm, if the skb is already null, we probably shouldn't send any trace.
> > > > 
> > > > What about using TP_CONDITION() ?
> > > 
> > > Hrm, let's see. It's been introduced by commit
> > > 5cb3d1d9d34ac04bcaa2034139345b2a5fea54c1
> > > by Zhaolei.
> > > 
> > > Event at the time of that commit, the only caller looked like:
> > > 
> > > void kfree_skb(struct sk_buff *skb)
> > > {
> > >         if (unlikely(!skb))
> > >                 return;
> > >         if (likely(atomic_read(&skb->users) == 1))
> > >                 smp_rmb();
> > >         else if (likely(!atomic_dec_and_test(&skb->users)))
> > >                 return;
> > >         trace_kfree_skb(skb, __builtin_return_address(0));
> > >         __kfree_skb(skb);
> > > }
> > > EXPORT_SYMBOL(kfree_skb);
> > > 
> > > So it already checks for a null pointer before calling the tracepoint. This
> > > leads me to wonder why why this check was added in the first place ?
> > 
> > Likely for no strong reasons :)
> > 
> > So I guess we can remove the check from the tracepoint?
> 
> Yep, leading to this patch instead:
> 
> trace event skb remove duplicate null-pointer check
> 
> The check for NULL skb in the kfree_skb trace event is a duplicate from the
> check already done in its only caller, kfree_skb(). Remove this duplicate check.
> 
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> CC: Steven Rostedt <rostedt@goodmis.org>
> CC: Frederic Weisbecker <fweisbec@gmail.com>
> CC: Ingo Molnar <mingo@elte.hu>
> CC: Thomas Gleixner <tglx@linutronix.de>
> CC: Neil Horman <nhorman@tuxdriver.com>
> ---
>  include/trace/events/skb.h |    4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> Index: linux-2.6-lttng/include/trace/events/skb.h
> ===================================================================
> --- linux-2.6-lttng.orig/include/trace/events/skb.h
> +++ linux-2.6-lttng/include/trace/events/skb.h
> @@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
>  
>  	TP_fast_assign(
>  		__entry->skbaddr = skb;
> -		if (skb) {
> -			__entry->protocol = ntohs(skb->protocol);
> -		}
> +		__entry->protocol = ntohs(skb->protocol);
>  		__entry->location = location;
>  	),
>  
> 
> 
Acked-by: Neil Horman <nhorman@tuxdriver.com>

> -- 
> Mathieu Desnoyers
> Operating System Efficiency R&D Consultant
> EfficiOS Inc.
> http://www.efficios.com
> 

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Mathieu Desnoyers]

* Neil Horman (nhorman@tuxdriver.com) wrote:
[...]
> > trace event skb remove duplicate null-pointer check
> > 
> > The check for NULL skb in the kfree_skb trace event is a duplicate from the
> > check already done in its only caller, kfree_skb(). Remove this duplicate check.
> > 
> > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> > CC: Steven Rostedt <rostedt@goodmis.org>
> > CC: Frederic Weisbecker <fweisbec@gmail.com>
> > CC: Ingo Molnar <mingo@elte.hu>
> > CC: Thomas Gleixner <tglx@linutronix.de>
> > CC: Neil Horman <nhorman@tuxdriver.com>
> > ---
> >  include/trace/events/skb.h |    4 +---
> >  1 file changed, 1 insertion(+), 3 deletions(-)
> > 
> > Index: linux-2.6-lttng/include/trace/events/skb.h
> > ===================================================================
> > --- linux-2.6-lttng.orig/include/trace/events/skb.h
> > +++ linux-2.6-lttng/include/trace/events/skb.h
> > @@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
> >  
> >  	TP_fast_assign(
> >  		__entry->skbaddr = skb;
> > -		if (skb) {
> > -			__entry->protocol = ntohs(skb->protocol);
> > -		}
> > +		__entry->protocol = ntohs(skb->protocol);
> >  		__entry->location = location;
> >  	),
> >  
> > 
> > 
> Acked-by: Neil Horman <nhorman@tuxdriver.com>

Thanks!

Mathieu

> 
> > -- 
> > Mathieu Desnoyers
> > Operating System Efficiency R&D Consultant
> > EfficiOS Inc.
> > http://www.efficios.com
> > 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Re: [RFC patch 2/5] trace event skb fix unassigned field

[Steven Rostedt]

On Wed, 2011-01-05 at 06:59 -0500, Neil Horman wrote:

> > Yep, leading to this patch instead:
> > 
> > trace event skb remove duplicate null-pointer check
> > 
> > The check for NULL skb in the kfree_skb trace event is a duplicate from the
> > check already done in its only caller, kfree_skb(). Remove this duplicate check.
> > 

> Acked-by: Neil Horman <nhorman@tuxdriver.com>

Mathieu, can you repost this as a non-RFC patch, and include Neil's
acked-by.

Thanks,

-- Steve

[RFC patch 0/5] Trace event fixes and cleanups

[Mathieu Desnoyers]

Hi,

I noticed a 2 bugs in TRACE_EVENT() users (unassigned fields), along with 3
inconsistencies in the Ftrace comments and macros, which should be considered as
cleanups.

I made sure to only include trivial patches that do not change anything
ABI-wise.

(tested against -tip)

Comments are welcome,

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

[RFC patch 2/5] trace event skb fix unassigned field

[Mathieu Desnoyers]

[-- Attachment #1: trace-event-skb-fix-unassigned-field.patch --]
[-- Type: text/plain, Size: 985 bytes --]

The field "protocol" in event kfree_skb is left unassigned if skb is NULL,
leaving its trace output as garbage. Assign the value to 0 when skb is NULL
instead.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Neil Horman <nhorman@tuxdriver.com>
CC: Thomas Gleixner <tglx@linutronix.de>
---
 include/trace/events/skb.h |    4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

Index: linux-2.6-lttng/include/trace/events/skb.h
===================================================================
--- linux-2.6-lttng.orig/include/trace/events/skb.h
+++ linux-2.6-lttng/include/trace/events/skb.h
@@ -25,9 +25,7 @@ TRACE_EVENT(kfree_skb,
 
 	TP_fast_assign(
 		__entry->skbaddr = skb;
-		if (skb) {
-			__entry->protocol = ntohs(skb->protocol);
-		}
+		__entry->protocol = skb ? ntohs(skb->protocol) : 0;
 		__entry->location = location;
 	),