2024-03-11 12:40:52

by Ignat Korchagin

[permalink] [raw]
Subject: [PATCH] net: veth: do not manipulate GRO when using XDP

Commit d3256efd8e8b ("veth: allow enabling NAPI even without XDP") tried to fix
the fact that GRO was not possible without XDP, because veth did not use NAPI
without XDP. However, it also introduced the behaviour that GRO is always
enabled, when XDP is enabled.

While it might be desired for most cases, it is confusing for the user at best
as the GRO flag sudddenly changes, when an XDP program is attached. It also
introduces some complexities in state management as was partially addressed in
commit fe9f801355f0 ("net: veth: clear GRO when clearing XDP even when down").

But the biggest problem is that it is not possible to disable GRO at all, when
an XDP program is attached, which might be needed for some use cases.

Fix this by not touching the GRO flag on XDP enable/disable as the code already
supports switching to NAPI if either GRO or XDP is requested.

Signed-off-by: Ignat Korchagin <[email protected]>
---
drivers/net/veth.c | 18 ------------------
1 file changed, 18 deletions(-)

diff --git a/drivers/net/veth.c b/drivers/net/veth.c
index cd4a6fe458f9..f0b2c4d5fe43 100644
--- a/drivers/net/veth.c
+++ b/drivers/net/veth.c
@@ -1533,8 +1533,6 @@ static netdev_features_t veth_fix_features(struct net_device *dev,
if (peer_priv->_xdp_prog)
features &= ~NETIF_F_GSO_SOFTWARE;
}
- if (priv->_xdp_prog)
- features |= NETIF_F_GRO;

return features;
}
@@ -1638,14 +1636,6 @@ static int veth_xdp_set(struct net_device *dev, struct bpf_prog *prog,
}

if (!old_prog) {
- if (!veth_gro_requested(dev)) {
- /* user-space did not require GRO, but adding
- * XDP is supposed to get GRO working
- */
- dev->features |= NETIF_F_GRO;
- netdev_features_change(dev);
- }
-
peer->hw_features &= ~NETIF_F_GSO_SOFTWARE;
peer->max_mtu = max_mtu;
}
@@ -1661,14 +1651,6 @@ static int veth_xdp_set(struct net_device *dev, struct bpf_prog *prog,
if (dev->flags & IFF_UP)
veth_disable_xdp(dev);

- /* if user-space did not require GRO, since adding XDP
- * enabled it, clear it now
- */
- if (!veth_gro_requested(dev)) {
- dev->features &= ~NETIF_F_GRO;
- netdev_features_change(dev);
- }
-
if (peer) {
peer->hw_features |= NETIF_F_GSO_SOFTWARE;
peer->max_mtu = ETH_MAX_MTU;
--
2.39.2



2024-03-11 20:30:20

by Jakub Kicinski

[permalink] [raw]
Subject: Re: [PATCH] net: veth: do not manipulate GRO when using XDP

On Mon, 11 Mar 2024 12:40:15 +0000 Ignat Korchagin wrote:
> Commit d3256efd8e8b ("veth: allow enabling NAPI even without XDP") tried to fix
> the fact that GRO was not possible without XDP, because veth did not use NAPI
> without XDP. However, it also introduced the behaviour that GRO is always
> enabled, when XDP is enabled.
>
> While it might be desired for most cases, it is confusing for the user at best
> as the GRO flag sudddenly changes, when an XDP program is attached. It also

s/ddd/dd/ spellcheck the whole message, pls

> introduces some complexities in state management as was partially addressed in
> commit fe9f801355f0 ("net: veth: clear GRO when clearing XDP even when down").
>
> But the biggest problem is that it is not possible to disable GRO at all, when
> an XDP program is attached, which might be needed for some use cases.
>
> Fix this by not touching the GRO flag on XDP enable/disable as the code already
> supports switching to NAPI if either GRO or XDP is requested.

Sound legit, AFAIU. But please also adjust
tools/testing/selftests/net/veth.sh because it's failing now.

2024-03-12 09:10:59

by Toke Høiland-Jørgensen

[permalink] [raw]
Subject: Re: [PATCH] net: veth: do not manipulate GRO when using XDP

Ignat Korchagin <[email protected]> writes:

> Commit d3256efd8e8b ("veth: allow enabling NAPI even without XDP") tried to fix
> the fact that GRO was not possible without XDP, because veth did not use NAPI
> without XDP. However, it also introduced the behaviour that GRO is always
> enabled, when XDP is enabled.
>
> While it might be desired for most cases, it is confusing for the user at best
> as the GRO flag sudddenly changes, when an XDP program is attached. It also
> introduces some complexities in state management as was partially addressed in
> commit fe9f801355f0 ("net: veth: clear GRO when clearing XDP even when down").
>
> But the biggest problem is that it is not possible to disable GRO at all, when
> an XDP program is attached, which might be needed for some use cases.
>
> Fix this by not touching the GRO flag on XDP enable/disable as the code already
> supports switching to NAPI if either GRO or XDP is requested.

Sounds reasonable

Reviewed-by: Toke Høiland-Jørgensen <[email protected]>