[PATCH v2] tracing/ipv4/ipv6: Use static array for name field in fib*_lookup_table event

[PATCH v2] tracing/ipv4/ipv6: Use static array for name field in fib*_lookup_table event

[Steven Rostedt]

From: "Steven Rostedt (Google)" <rostedt@goodmis.org>

The fib_lookup_table and fib6_lookup_table events declare name as a
dynamic_array, but also give it a fixed size, which defeats the purpose of
the dynamic array, especially since the dynamic array also includes meta
data in the event to specify its size.

Since the size of the name is at most 16 bytes (defined by IFNAMSIZ),
it is not worth spending the effort to determine the size of the string.

Just use a fixed size array and copy into it. This will save 4 bytes that
are used for the meta data that saves the size and position of a dynamic
array, and even slightly speed up the event processing.

Cc: David Ahern <dsahern@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
---
Changes since v1: https://lkml.kernel.org/r/20220703102359.30f12e39@rorschach.local.home
 - Just use a fixed size array instead of calculating the
   size needed for the dynamic allocation.

 include/trace/events/fib.h  | 6 +++---
 include/trace/events/fib6.h | 8 ++++----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/include/trace/events/fib.h b/include/trace/events/fib.h
index 6f2a4dc35e37..c2300c407f58 100644
--- a/include/trace/events/fib.h
+++ b/include/trace/events/fib.h
@@ -32,7 +32,7 @@ TRACE_EVENT(fib_table_lookup,
 		__array(	__u8,	gw6,	16	)
 		__field(	u16,	sport		)
 		__field(	u16,	dport		)
-		__dynamic_array(char,  name,   IFNAMSIZ )
+		__array(char,  name,   IFNAMSIZ )
 	),
 
 	TP_fast_assign(
@@ -66,7 +66,7 @@ TRACE_EVENT(fib_table_lookup,
 		}
 
 		dev = nhc ? nhc->nhc_dev : NULL;
-		__assign_str(name, dev ? dev->name : "-");
+		strlcpy(__entry->name, dev ? dev->name : "-", IFNAMSIZ);
 
 		if (nhc) {
 			if (nhc->nhc_gw_family == AF_INET) {
@@ -95,7 +95,7 @@ TRACE_EVENT(fib_table_lookup,
 		  __entry->tb_id, __entry->oif, __entry->iif, __entry->proto,
 		  __entry->src, __entry->sport, __entry->dst, __entry->dport,
 		  __entry->tos, __entry->scope, __entry->flags,
-		  __get_str(name), __entry->gw4, __entry->gw6, __entry->err)
+		  __entry->name, __entry->gw4, __entry->gw6, __entry->err)
 );
 #endif /* _TRACE_FIB_H */
 
diff --git a/include/trace/events/fib6.h b/include/trace/events/fib6.h
index c6abdcc77c12..6e821eb79450 100644
--- a/include/trace/events/fib6.h
+++ b/include/trace/events/fib6.h
@@ -31,7 +31,7 @@ TRACE_EVENT(fib6_table_lookup,
 		__field(        u16,	dport		)
 		__field(        u8,	proto		)
 		__field(        u8,	rt_type		)
-		__dynamic_array(	char,	name,	IFNAMSIZ )
+		__array(		char,	name,	IFNAMSIZ )
 		__array(		__u8,	gw,	16	 )
 	),
 
@@ -63,9 +63,9 @@ TRACE_EVENT(fib6_table_lookup,
 		}
 
 		if (res->nh && res->nh->fib_nh_dev) {
-			__assign_str(name, res->nh->fib_nh_dev);
+			strlcpy(__entry->name, res->nh->fib_nh_dev->name, IFNAMSIZ);
 		} else {
-			__assign_str(name, "-");
+			strcpy(__entry->name, "-");
 		}
 		if (res->f6i == net->ipv6.fib6_null_entry) {
 			struct in6_addr in6_zero = {};
@@ -83,7 +83,7 @@ TRACE_EVENT(fib6_table_lookup,
 		  __entry->tb_id, __entry->oif, __entry->iif, __entry->proto,
 		  __entry->src, __entry->sport, __entry->dst, __entry->dport,
 		  __entry->tos, __entry->scope, __entry->flags,
-		  __get_str(name), __entry->gw, __entry->err)
+		  __entry->name, __entry->gw, __entry->err)
 );
 
 #endif /* _TRACE_FIB6_H */
-- 
2.35.1

Re: [PATCH v2] tracing/ipv4/ipv6: Use static array for name field in fib*_lookup_table event

[David Ahern]

On 7/4/22 7:14 AM, Steven Rostedt wrote:
> From: "Steven Rostedt (Google)" <rostedt@goodmis.org>
> 
> The fib_lookup_table and fib6_lookup_table events declare name as a
> dynamic_array, but also give it a fixed size, which defeats the purpose of
> the dynamic array, especially since the dynamic array also includes meta
> data in the event to specify its size.
> 
> Since the size of the name is at most 16 bytes (defined by IFNAMSIZ),
> it is not worth spending the effort to determine the size of the string.
> 
> Just use a fixed size array and copy into it. This will save 4 bytes that
> are used for the meta data that saves the size and position of a dynamic
> array, and even slightly speed up the event processing.
> 
> Cc: David Ahern <dsahern@gmail.com>
> Cc: David S. Miller <davem@davemloft.net>
> Cc: netdev@vger.kernel.org
> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
> ---
> Changes since v1: https://lkml.kernel.org/r/20220703102359.30f12e39@rorschach.local.home
>  - Just use a fixed size array instead of calculating the
>    size needed for the dynamic allocation.
> 
>  include/trace/events/fib.h  | 6 +++---
>  include/trace/events/fib6.h | 8 ++++----
>  2 files changed, 7 insertions(+), 7 deletions(-)
> 

Reviewed-by: David Ahern <dsahern@kernel.org>

Re: [PATCH v2] tracing/ipv4/ipv6: Use static array for name field in fib*_lookup_table event

[Steven Rostedt]

On Mon, 4 Jul 2022 13:09:01 -0600
David Ahern <dsahern@gmail.com> wrote:

> >  include/trace/events/fib.h  | 6 +++---
> >  include/trace/events/fib6.h | 8 ++++----
> >  2 files changed, 7 insertions(+), 7 deletions(-)
> >   
> 
> Reviewed-by: David Ahern <dsahern@kernel.org>
> 

Is everyone OK if I take this through my tree then?

I'm fine if it goes through net-next too.

-- Steve

Re: [PATCH v2] tracing/ipv4/ipv6: Use static array for name field in fib*_lookup_table event

[Jakub Kicinski]

On Tue, 5 Jul 2022 10:00:41 -0400 Steven Rostedt wrote:
> On Mon, 4 Jul 2022 13:09:01 -0600
> David Ahern <dsahern@gmail.com> wrote:
> 
> > >  include/trace/events/fib.h  | 6 +++---
> > >  include/trace/events/fib6.h | 8 ++++----
> > >  2 files changed, 7 insertions(+), 7 deletions(-)
> > >     
> > 
> > Reviewed-by: David Ahern <dsahern@kernel.org>
> >   
> 
> Is everyone OK if I take this through my tree then?
> 
> I'm fine if it goes through net-next too.

Slight risk of conflicts there if someone decides to add another FIB
trace point, but seems unlikely so feel free to take it:

Acked-by: Jakub Kicinski <kuba@kernel.org>