[PATCH net-next] igc: avoid kernel warning when changing RX ring parameters

[PATCH net-next] igc: avoid kernel warning when changing RX ring parameters

[Corinna Vinschen]

Calling ethtool changing the RX ring parameters like this:

  $ ethtool -G eth0 rx 1024

triggers the "Missing unregister, handled but fix driver" warning in
xdp_rxq_info_reg().

igc_ethtool_set_ringparam() copies the igc_ring structure but neglects to
reset the xdp_rxq_info member before calling igc_setup_rx_resources().
This in turn calls xdp_rxq_info_reg() with an already registered xdp_rxq_info.

This fix initializes the xdp_rxq_info member prior to calling
igc_setup_rx_resources(), exactly like igb.

Fixes: 73f1071c1d29 ("igc: Add support for XDP_TX action")
Signed-off-by: Corinna Vinschen <vinschen@redhat.com>
---
 drivers/net/ethernet/intel/igc/igc_ethtool.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/ethernet/intel/igc/igc_ethtool.c b/drivers/net/ethernet/intel/igc/igc_ethtool.c
index 8cc077b712ad..93839106504d 100644
--- a/drivers/net/ethernet/intel/igc/igc_ethtool.c
+++ b/drivers/net/ethernet/intel/igc/igc_ethtool.c
@@ -671,6 +671,10 @@ igc_ethtool_set_ringparam(struct net_device *netdev,
 			memcpy(&temp_ring[i], adapter->rx_ring[i],
 			       sizeof(struct igc_ring));
 
+			/* Clear copied XDP RX-queue info */
+			memset(&temp_ring[i].xdp_rxq, 0,
+			       sizeof(temp_ring[i].xdp_rxq));
+
 			temp_ring[i].count = new_rx_count;
 			err = igc_setup_rx_resources(&temp_ring[i]);
 			if (err) {
-- 
2.27.0

Re: [PATCH net-next] igc: avoid kernel warning when changing RX ring parameters

[Vinicius Costa Gomes]

Corinna Vinschen <vinschen@redhat.com> writes:

> Calling ethtool changing the RX ring parameters like this:
>
>   $ ethtool -G eth0 rx 1024
>
> triggers the "Missing unregister, handled but fix driver" warning in
> xdp_rxq_info_reg().
>
> igc_ethtool_set_ringparam() copies the igc_ring structure but neglects to
> reset the xdp_rxq_info member before calling igc_setup_rx_resources().
> This in turn calls xdp_rxq_info_reg() with an already registered xdp_rxq_info.
>
> This fix initializes the xdp_rxq_info member prior to calling
> igc_setup_rx_resources(), exactly like igb.
>
> Fixes: 73f1071c1d29 ("igc: Add support for XDP_TX action")
> Signed-off-by: Corinna Vinschen <vinschen@redhat.com>
> ---
>  drivers/net/ethernet/intel/igc/igc_ethtool.c | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/drivers/net/ethernet/intel/igc/igc_ethtool.c b/drivers/net/ethernet/intel/igc/igc_ethtool.c
> index 8cc077b712ad..93839106504d 100644
> --- a/drivers/net/ethernet/intel/igc/igc_ethtool.c
> +++ b/drivers/net/ethernet/intel/igc/igc_ethtool.c
> @@ -671,6 +671,10 @@ igc_ethtool_set_ringparam(struct net_device *netdev,
>  			memcpy(&temp_ring[i], adapter->rx_ring[i],
>  			       sizeof(struct igc_ring));
>  
> +			/* Clear copied XDP RX-queue info */
> +			memset(&temp_ring[i].xdp_rxq, 0,
> +			       sizeof(temp_ring[i].xdp_rxq));
> +

Reaching "inside" xdp_rxq to reset it doesn't feel right in this
context, even if it's going to work fine (for now).

I think that the suggestion that Alexander gave in that other thread is
the best approach. And thanks for noticing that igb '_set_ringparam()'
has the same underlying problem and also needs to be fixed.


Cheers,
-- 
Vinicius

Re: [Intel-wired-lan] [PATCH net-next] igc: avoid kernel warning when changing RX ring parameters

[Corinna Vinschen]

On Jan 13 16:37, Vinicius Costa Gomes wrote:
> Corinna Vinschen <vinschen@redhat.com> writes:
> 
> > Calling ethtool changing the RX ring parameters like this:
> >
> >   $ ethtool -G eth0 rx 1024
> >
> > triggers the "Missing unregister, handled but fix driver" warning in
> > xdp_rxq_info_reg().
> >[...]
> 
> Reaching "inside" xdp_rxq to reset it doesn't feel right in this
> context, even if it's going to work fine (for now).
> 
> I think that the suggestion that Alexander gave in that other thread is
> the best approach. And thanks for noticing that igb '_set_ringparam()'
> has the same underlying problem and also needs to be fixed.

Yeah, it didn't sit overly well with me either, but I thought if it's
good for igb...

Either way, the better approach should be sth. like this in both,
ig[bc]_setup_rx_resources:

	if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
		xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
	res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, ndev, index,
			       rx_ring->q_vector->napi.napi_id);

And while at it, wouldn't it make sense to move the xdp_rxq_info_reg
call in igc_setup_rx_resources down to be the last action, so the error
path doesn't have to call xdp_rxq_info_unreg, just like in
igb_setup_rx_resources?


Thanks,
Corinna