[PATCH net-next] net: nexthop: reduce rcu synchronizations when replacing resilient groups

From: Nikolay Aleksandrov <redacted>

We can optimize resilient nexthop group replaces by reducing the number of
synchronize_net calls. After commit 1005f19b9357 ("net: nexthop: release
IPv6 per-cpu dsts when replacing a nexthop group") we always do a
synchronize_net because we must ensure no new dsts can be created for the
replaced group's removed nexthops, but we already did that when replacing
resilient groups, so if we always call synchronize_net after any group
type replacement we'll take care of both cases and reduce synchronize_net
calls for resilient groups.

Suggested-by: Ido Schimmel <redacted>
Signed-off-by: Nikolay Aleksandrov <redacted>
---
 net/ipv4/nexthop.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/net/ipv4/nexthop.c b/net/ipv4/nexthop.c
index 5dbd4b5505eb..1319d093cdda 100644
--- a/net/ipv4/nexthop.c
+++ b/net/ipv4/nexthop.c

@@ -1918,9 +1918,6 @@ static void nh_rt_cache_flush(struct net *net, struct nexthop *nh,
 	if (!replaced_nh->is_group)
 		return;
 
-	/* new dsts must use only the new nexthop group */
-	synchronize_net();
-
 	nhg = rtnl_dereference(replaced_nh->nh_grp);
 	for (i = 0; i < nhg->num_nh; i++) {
 		struct nh_grp_entry *nhge = &nhg->nh_entries[i];

@@ -2002,9 +1999,10 @@ static int replace_nexthop_grp(struct net *net, struct nexthop *old,
 
 	rcu_assign_pointer(old->nh_grp, newg);
 
+	/* Make sure concurrent readers are not using 'oldg' anymore. */
+	synchronize_net();
+
 	if (newg->resilient) {
-		/* Make sure concurrent readers are not using 'oldg' anymore. */
-		synchronize_net();
 		rcu_assign_pointer(oldg->res_table, tmp_table);
 		rcu_assign_pointer(oldg->spare->res_table, tmp_table);
 	}

-- 
2.31.1

On Mon, Nov 29, 2021 at 02:09:24PM +0200, Nikolay Aleksandrov wrote:

From: Nikolay Aleksandrov <redacted>

We can optimize resilient nexthop group replaces by reducing the number of
synchronize_net calls. After commit 1005f19b9357 ("net: nexthop: release
IPv6 per-cpu dsts when replacing a nexthop group") we always do a
synchronize_net because we must ensure no new dsts can be created for the
replaced group's removed nexthops, but we already did that when replacing
resilient groups, so if we always call synchronize_net after any group
type replacement we'll take care of both cases and reduce synchronize_net
calls for resilient groups.

Suggested-by: Ido Schimmel <redacted>
Signed-off-by: Nikolay Aleksandrov <redacted>

I ran fib_nexthops.sh that used to trigger the bug fixed by commit
563f23b00253 ("nexthop: Fix division by zero while replacing a resilient
group") and it looks good.

Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>

Thanks!

On 11/29/21 5:09 AM, Nikolay Aleksandrov wrote:

From: Nikolay Aleksandrov <redacted>

We can optimize resilient nexthop group replaces by reducing the number of
synchronize_net calls. After commit 1005f19b9357 ("net: nexthop: release
IPv6 per-cpu dsts when replacing a nexthop group") we always do a
synchronize_net because we must ensure no new dsts can be created for the
replaced group's removed nexthops, but we already did that when replacing
resilient groups, so if we always call synchronize_net after any group
type replacement we'll take care of both cases and reduce synchronize_net
calls for resilient groups.

Suggested-by: Ido Schimmel <redacted>
Signed-off-by: Nikolay Aleksandrov <redacted>
---
 net/ipv4/nexthop.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

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

Hello:

This patch was applied to netdev/net-next.git (master)
by David S. Miller [off-list ref]:

On Mon, 29 Nov 2021 14:09:24 +0200 you wrote:

From: Nikolay Aleksandrov <redacted>

We can optimize resilient nexthop group replaces by reducing the number of
synchronize_net calls. After commit 1005f19b9357 ("net: nexthop: release
IPv6 per-cpu dsts when replacing a nexthop group") we always do a
synchronize_net because we must ensure no new dsts can be created for the
replaced group's removed nexthops, but we already did that when replacing
resilient groups, so if we always call synchronize_net after any group
type replacement we'll take care of both cases and reduce synchronize_net
calls for resilient groups.

[...]

Here is the summary with links:
  - [net-next] net: nexthop: reduce rcu synchronizations when replacing resilient groups
    https://git.kernel.org/netdev/net-next/c/7709efa62c4f

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html