[PATCH iwl-net] ice: keep the RSS hash function on indirection-only updates

[PATCH iwl-net] ice: keep the RSS hash function on indirection-only updates

[Maciej Fijalkowski]

The ethtool core passes RXH_XFRM_NO_CHANGE (0xff) in input_xfrm when a
request does not touch the input transformation, e.g. when it only
rewrites the RSS indirection table. ice_set_rxfh() tests input_xfrm only
for RXH_XFRM_SYM_XOR, which 0xff has set, so such an update switches the
VSI to symmetric Toeplitz hashing behind the user's back.

Once symmetric-xor is reported, the core requires all hashed flow types
to be symmetric for any following RSS set. The GTP flow types hash on
the TEID, so every later netlink RSS update fails with "hash field config
is not symmetric", including the one that would put the indirection
table back.

Keep the current hash function when input_xfrm is RXH_XFRM_NO_CHANGE.
ice_set_rss_hfunc() then returns early and leaves the VSI as it is.

This has been encountered while trying to run xskxceiver within drv-net
tests where one of the setup steps was queue pull out of RSS indirection
table.

Fixes: 352e9bf23813 ("ice: enable symmetric-xor RSS for Toeplitz hash function")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_ethtool.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_ethtool.c b/drivers/net/ethernet/intel/ice/ice_ethtool.c
index dffa213734f1..fd2f3cb9faba 100644
--- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
+++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
@@ -3704,7 +3704,9 @@ ice_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
 	}
 
 	/* Update the VSI's hash function */
-	if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
+	if (rxfh->input_xfrm == RXH_XFRM_NO_CHANGE)
+		hfunc = vsi->rss_hfunc;
+	else if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
 		hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ;
 
 	err = ice_set_rss_hfunc(vsi, hfunc);
-- 
2.43.0

Re: [PATCH iwl-net] ice: keep the RSS hash function on indirection-only updates

[Jakub Kicinski]

On Wed,  7 Oct 2026 22:03:11 +0200 Maciej Fijalkowski wrote:
> The ethtool core passes RXH_XFRM_NO_CHANGE (0xff) in input_xfrm when a
> request does not touch the input transformation, e.g. when it only
> rewrites the RSS indirection table. ice_set_rxfh() tests input_xfrm only
> for RXH_XFRM_SYM_XOR, which 0xff has set, so such an update switches the
> VSI to symmetric Toeplitz hashing behind the user's back.
> 
> Once symmetric-xor is reported, the core requires all hashed flow types
> to be symmetric for any following RSS set. The GTP flow types hash on
> the TEID, so every later netlink RSS update fails with "hash field config
> is not symmetric", including the one that would put the indirection
> table back.
> 
> Keep the current hash function when input_xfrm is RXH_XFRM_NO_CHANGE.
> ice_set_rss_hfunc() then returns early and leaves the VSI as it is.
> 
> This has been encountered while trying to run xskxceiver within drv-net
> tests where one of the setup steps was queue pull out of RSS indirection
> table.
> 
> Fixes: 352e9bf23813 ("ice: enable symmetric-xor RSS for Toeplitz hash function")
> Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>

Hm, not seeing it on the list now but I thought I submitted this 
a while back:

--- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
+++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
@@ -3681,10 +3681,10 @@ ice_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
             struct netlink_ext_ack *extack)
 {
        struct ice_netdev_priv *np = netdev_priv(netdev);
-       u8 hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
        struct ice_vsi *vsi = np->vsi;
        struct ice_pf *pf = vsi->back;
        struct device *dev;
+       u8 hfunc;
        int err;
 
        dev = ice_pf_to_dev(pf);
@@ -3703,9 +3703,12 @@ ice_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
                return -EOPNOTSUPP;
        }
 
-       /* Update the VSI's hash function */
-       if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
+       if (rxfh->input_xfrm == RXH_XFRM_NO_CHANGE)
+               hfunc = vsi->rss_hfunc;
+       else if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
                hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ;
+       else /* input_xfrm == 0; core rejects any other value */
+               hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
 
        err = ice_set_rss_hfunc(vsi, hfunc);
        if (err)

Could you double check if we need the trailing else as well? 
Not 100% sure now if I added it just in case or it mattered :S

Re: [PATCH iwl-net] ice: keep the RSS hash function on indirection-only updates

[Maciej Fijalkowski]

On Thu, Oct 08, 2026 at 08:34:12AM -0700, Jakub Kicinski wrote:
> On Wed,  7 Oct 2026 22:03:11 +0200 Maciej Fijalkowski wrote:
> > The ethtool core passes RXH_XFRM_NO_CHANGE (0xff) in input_xfrm when a
> > request does not touch the input transformation, e.g. when it only
> > rewrites the RSS indirection table. ice_set_rxfh() tests input_xfrm only
> > for RXH_XFRM_SYM_XOR, which 0xff has set, so such an update switches the
> > VSI to symmetric Toeplitz hashing behind the user's back.
> > 
> > Once symmetric-xor is reported, the core requires all hashed flow types
> > to be symmetric for any following RSS set. The GTP flow types hash on
> > the TEID, so every later netlink RSS update fails with "hash field config
> > is not symmetric", including the one that would put the indirection
> > table back.
> > 
> > Keep the current hash function when input_xfrm is RXH_XFRM_NO_CHANGE.
> > ice_set_rss_hfunc() then returns early and leaves the VSI as it is.
> > 
> > This has been encountered while trying to run xskxceiver within drv-net
> > tests where one of the setup steps was queue pull out of RSS indirection
> > table.
> > 
> > Fixes: 352e9bf23813 ("ice: enable symmetric-xor RSS for Toeplitz hash function")
> > Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> 
> Hm, not seeing it on the list now but I thought I submitted this 
> a while back:

Hi! This comes from your response where you asked Aleksandr to pull out
this fix out of -next changes and send to -net separately:

https://lore.kernel.org/netdev/20260626152636.1e48330f@kernel.org/

> 
> --- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
> +++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
> @@ -3681,10 +3681,10 @@ ice_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
>              struct netlink_ext_ack *extack)
>  {
>         struct ice_netdev_priv *np = netdev_priv(netdev);
> -       u8 hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
>         struct ice_vsi *vsi = np->vsi;
>         struct ice_pf *pf = vsi->back;
>         struct device *dev;
> +       u8 hfunc;
>         int err;
>  
>         dev = ice_pf_to_dev(pf);
> @@ -3703,9 +3703,12 @@ ice_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
>                 return -EOPNOTSUPP;
>         }
>  
> -       /* Update the VSI's hash function */
> -       if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
> +       if (rxfh->input_xfrm == RXH_XFRM_NO_CHANGE)
> +               hfunc = vsi->rss_hfunc;
> +       else if (rxfh->input_xfrm & RXH_XFRM_SYM_XOR)
>                 hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ;
> +       else /* input_xfrm == 0; core rejects any other value */
> +               hfunc = ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ;
>  
>         err = ice_set_rss_hfunc(vsi, hfunc);
>         if (err)
> 
> Could you double check if we need the trailing else as well? 
> Not 100% sure now if I added it just in case or it mattered :S

else part seems to be just a move from hfunc being initialized to
ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ.

Re: [PATCH iwl-net] ice: keep the RSS hash function on indirection-only updates

[Jakub Kicinski]

On Thu, 8 Oct 2026 20:38:19 +0200 Maciej Fijalkowski wrote:
> > Could you double check if we need the trailing else as well? 
> > Not 100% sure now if I added it just in case or it mattered :S  
> 
> else part seems to be just a move from hfunc being initialized to
> ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ.

of course, lack of caffeine