Hi Jakub,
On 9/30/26 03:09, Jakub Kicinski wrote:
On Tue, 29 Sep 2026 10:57:10 +0200 Maxime Chevallier wrote:
quoted
quoted
quoted
@@ -943,6 +949,9 @@ static int stmmac_get_rxfh(struct net_device *dev,
struct stmmac_priv *priv = netdev_priv(dev);
int i;
+ if (!stmmac_rss_is_supported(priv))
+ return -EOPNOTSUPP;
+
[Severity: Medium]
Does this -EOPNOTSUPP break namespace-wide ETHTOOL_MSG_RSS_GET dumps?
This was reported by Nicolai already, and other drivers report -EOPNOTSUPP
here.
If we address that, I'd say we do that in the RSS ethnl code.
Let's get this in the same series, cause I'm not 100% convinced.
My instinct is that returning 0 size for indir table and key
would be correct and less hairy for the core to handle.
All pieces of RSS are optional so returning 0 size for everything
should be equivalent to having no RSS.
Ok no problem, I'll also give it a test on mvpp2 as well then.
Thanks,
Maxime