[PATCH 1/2] net: ipv4: route.c: add likely() statements

Add likely() statements in ipv4_confirm_neigh() for 'rt->rt_gw_family
== AF_INET'.

Signed-off-by: Yejune Deng <redacted>
---
 net/ipv4/route.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index fa68c2612252..5762d9bc671c 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c

@@ -440,7 +440,7 @@ static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr)
 	struct net_device *dev = dst->dev;
 	const __be32 *pkey = daddr;
 
-	if (rt->rt_gw_family == AF_INET) {
+	if (likely(rt->rt_gw_family == AF_INET)) {
 		pkey = (const __be32 *)&rt->rt_gw4;
 	} else if (rt->rt_gw_family == AF_INET6) {
 		return __ipv6_confirm_neigh_stub(dev, &rt->rt_gw6);

-- 
2.29.0

On Wed, Mar 24, 2021 at 11:09:22AM +0800, Yejune Deng wrote:

Add likely() statements in ipv4_confirm_neigh() for 'rt->rt_gw_family
== AF_INET'.

Why? Such macros are beneficial in only specific cases, most of the time,
likely/unlikely is cargo cult.

My reasons are as following: ipv4_confirm_neigh() belongs to
ipv4_dst_ops that family is AF_INET, and ipv4_neigh_lookup() is also
added likely() when rt->rt_gw_family is equal to AF_INET.

On Wed, Mar 24, 2021 at 6:34 PM Leon Romanovsky [off-list ref] wrote:

On Wed, Mar 24, 2021 at 11:09:22AM +0800, Yejune Deng wrote:

quoted

Add likely() statements in ipv4_confirm_neigh() for 'rt->rt_gw_family
== AF_INET'.

Why? Such macros are beneficial in only specific cases, most of the time,
likely/unlikely is cargo cult.

quoted

Signed-off-by: Yejune Deng <redacted>
---
 net/ipv4/route.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index fa68c2612252..5762d9bc671c 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c

@@ -440,7 +440,7 @@ static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr)
      struct net_device *dev = dst->dev;
      const __be32 *pkey = daddr;

-     if (rt->rt_gw_family == AF_INET) {
+     if (likely(rt->rt_gw_family == AF_INET)) {
              pkey = (const __be32 *)&rt->rt_gw4;
      } else if (rt->rt_gw_family == AF_INET6) {
              return __ipv6_confirm_neigh_stub(dev, &rt->rt_gw6);
--

2.29.0

On Wed, Mar 24, 2021 at 07:01:19PM +0800, Yejune Deng wrote:

My reasons are as following: ipv4_confirm_neigh() belongs to
ipv4_dst_ops that family is AF_INET, and ipv4_neigh_lookup() is also
added likely() when rt->rt_gw_family is equal to AF_INET.

It is part of that cargo cult. Please support your claim with
performance numbers when this likely/unlikely will give any difference.

Thanks

On Wed, Mar 24, 2021 at 6:34 PM Leon Romanovsky [off-list ref] wrote:

quoted

On Wed, Mar 24, 2021 at 11:09:22AM +0800, Yejune Deng wrote:

quoted

Add likely() statements in ipv4_confirm_neigh() for 'rt->rt_gw_family
== AF_INET'.

Why? Such macros are beneficial in only specific cases, most of the time,
likely/unlikely is cargo cult.

quoted

Signed-off-by: Yejune Deng <redacted>
---
 net/ipv4/route.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index fa68c2612252..5762d9bc671c 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c

@@ -440,7 +440,7 @@ static void ipv4_confirm_neigh(const struct dst_entry *dst, const void *daddr)
      struct net_device *dev = dst->dev;
      const __be32 *pkey = daddr;

-     if (rt->rt_gw_family == AF_INET) {
+     if (likely(rt->rt_gw_family == AF_INET)) {
              pkey = (const __be32 *)&rt->rt_gw4;
      } else if (rt->rt_gw_family == AF_INET6) {
              return __ipv6_confirm_neigh_stub(dev, &rt->rt_gw6);
--

2.29.0