[PATCH net v2] net: don't touch dev->stats in BPF redirect paths

[PATCH net v2] net: don't touch dev->stats in BPF redirect paths

[Jakub Kicinski]

Gal reports that BPF redirect increments dev->stats.tx_errors
on failure. This is not correct, most modern drivers completely
ignore dev->stats so these drops will be invisible to the user.
Core code should use the dedicated core stats which are folded
into device stats in dev_get_stats().

Note that we're switching from tx_errors to tx_dropped.
Core only has tx_dropped, hence presumably users already expect
that counter to increment for "stack" Tx issues.

Acked-by: Martin KaFai Lau <martin.lau@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Reported-by: Gal Pressman <gal@nvidia.com>
Link: https://lore.kernel.org/c5df3b60-246a-4030-9c9a-0a35cd1ca924@nvidia.com
Fixes: b4ab31414970 ("bpf: Add redirect_neigh helper as redirect drop-in")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
v2:
 - add note on tx_dropped to the commit msg
v1: https://lore.kernel.org/20260128161030.39870-1-kuba@kernel.org

CC: ast@kernel.org
CC: eddyz87@gmail.com
CC: song@kernel.org
CC: yonghong.song@linux.dev
CC: john.fastabend@gmail.com
CC: kpsingh@kernel.org
CC: sdf@fomichev.me
CC: haoluo@google.com
CC: jolsa@kernel.org
CC: dsahern@gmail.com
CC: bpf@vger.kernel.org
---
 net/core/filter.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/net/core/filter.c b/net/core/filter.c
index 616e0520a0bb..2c21735798a5 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2289,12 +2289,12 @@ static int __bpf_redirect_neigh_v6(struct sk_buff *skb, struct net_device *dev,
 
 	err = bpf_out_neigh_v6(net, skb, dev, nh);
 	if (unlikely(net_xmit_eval(err)))
-		DEV_STATS_INC(dev, tx_errors);
+		dev_core_stats_tx_dropped_inc(dev);
 	else
 		ret = NET_XMIT_SUCCESS;
 	goto out_xmit;
 out_drop:
-	DEV_STATS_INC(dev, tx_errors);
+	dev_core_stats_tx_dropped_inc(dev);
 	kfree_skb(skb);
 out_xmit:
 	return ret;
@@ -2396,12 +2396,12 @@ static int __bpf_redirect_neigh_v4(struct sk_buff *skb, struct net_device *dev,
 
 	err = bpf_out_neigh_v4(net, skb, dev, nh);
 	if (unlikely(net_xmit_eval(err)))
-		DEV_STATS_INC(dev, tx_errors);
+		dev_core_stats_tx_dropped_inc(dev);
 	else
 		ret = NET_XMIT_SUCCESS;
 	goto out_xmit;
 out_drop:
-	DEV_STATS_INC(dev, tx_errors);
+	dev_core_stats_tx_dropped_inc(dev);
 	kfree_skb(skb);
 out_xmit:
 	return ret;
-- 
2.52.0

Re: [PATCH net v2] net: don't touch dev->stats in BPF redirect paths

[Eric Dumazet]

On Fri, Jan 30, 2026 at 4:38 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Gal reports that BPF redirect increments dev->stats.tx_errors
> on failure. This is not correct, most modern drivers completely
> ignore dev->stats so these drops will be invisible to the user.
> Core code should use the dedicated core stats which are folded
> into device stats in dev_get_stats().
>
> Note that we're switching from tx_errors to tx_dropped.
> Core only has tx_dropped, hence presumably users already expect
> that counter to increment for "stack" Tx issues.

Reviewed-by: Eric Dumazet <edumazet@google.com>

Re: [PATCH net v2] net: don't touch dev->stats in BPF redirect paths

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Thu, 29 Jan 2026 19:38:27 -0800 you wrote:
> Gal reports that BPF redirect increments dev->stats.tx_errors
> on failure. This is not correct, most modern drivers completely
> ignore dev->stats so these drops will be invisible to the user.
> Core code should use the dedicated core stats which are folded
> into device stats in dev_get_stats().
> 
> Note that we're switching from tx_errors to tx_dropped.
> Core only has tx_dropped, hence presumably users already expect
> that counter to increment for "stack" Tx issues.
> 
> [...]

Here is the summary with links:
  - [net,v2] net: don't touch dev->stats in BPF redirect paths
    https://git.kernel.org/netdev/net/c/fdf3f6800be3

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