netdev - Re: [PATCH net-next 6/6] eth: sfc: falcon: migrate to new RXFH callbacks

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6425933b-3b17-4509-86be-be4a75f12e17@gmail.com>
Date: Fri, 13 Jun 2025 14:44:40 +0100
From: Edward Cree <ecree.xilinx@...il.com>
To: Jakub Kicinski <kuba@...nel.org>, davem@...emloft.net
Cc: netdev@...r.kernel.org, edumazet@...gle.com, pabeni@...hat.com,
 andrew+netdev@...n.ch, horms@...nel.org, bharat@...lsio.com,
 benve@...co.com, satishkh@...co.com, claudiu.manoil@....com,
 vladimir.oltean@....com, wei.fang@....com, xiaoning.wang@....com,
 anthony.l.nguyen@...el.com, przemyslaw.kitszel@...el.com,
 bryan.whitehead@...rochip.com, rosenp@...il.com, imx@...ts.linux.dev
Subject: Re: [PATCH net-next 6/6] eth: sfc: falcon: migrate to new RXFH
 callbacks

On 13/06/2025 01:54, Jakub Kicinski wrote:
> Migrate to new callbacks added by commit 9bb00786fc61 ("net: ethtool:
> add dedicated callbacks for getting and setting rxfh fields").
> This driver's RXFH config is read only / fixed so the conversion
> is purely factoring out the handling into a helper.
> 
> Signed-off-by: Jakub Kicinski <kuba@...nel.org>
> ---
>  drivers/net/ethernet/sfc/falcon/ethtool.c | 51 +++++++++++++----------
>  1 file changed, 28 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/net/ethernet/sfc/falcon/ethtool.c b/drivers/net/ethernet/sfc/falcon/ethtool.c
> index 04766448a545..6685e71ab13f 100644
> --- a/drivers/net/ethernet/sfc/falcon/ethtool.c
> +++ b/drivers/net/ethernet/sfc/falcon/ethtool.c
> @@ -943,6 +943,33 @@ static int ef4_ethtool_get_class_rule(struct ef4_nic *efx,
>  	return rc;
>  }
>  
> +static int
> +ef4_ethtool_get_rxfh_fields(struct net_device *net_dev,
> +			    struct ethtool_rxfh_fields *info)
> +{
> +	struct ef4_nic *efx = netdev_priv(net_dev);
> +	unsigned int min_revision = 0;
> +
> +	info->data = 0;
> +	switch (info->flow_type) {
> +	case TCP_V4_FLOW:
> +		info->data |= RXH_L4_B_0_1 | RXH_L4_B_2_3;
> +		fallthrough;
> +	case UDP_V4_FLOW:
> +	case SCTP_V4_FLOW:
> +	case AH_ESP_V4_FLOW:
> +	case IPV4_FLOW:
> +		info->data |= RXH_IP_SRC | RXH_IP_DST;
> +		min_revision = EF4_REV_FALCON_B0;
> +		break;
> +	default:
> +		break;
> +	}
> +	if (ef4_nic_rev(efx) < min_revision)
> +		info->data = 0;
> +	return 0;

So granted that you're only moving code, but looking at this it doesn't
 actually make sense, since every path that sets info->data to nonzero
 also sets min_revision, so why not just do the ef4_nic_rev() check at
 the start?  Answer, from git log spelunking, is that when this code was
 shared with Siena, EFX_REV_SIENA_A0 supported IPv6 here.

Have a
Reviewed-By: Edward Cree <ecree.xilinx@...il.com>
... but this patch could be followed-up with a simplification to put
	if (ef4_nic_rev(efx) < EF4_REV_FALCON_B0)
		return 0;
 before the switch and get rid of min_revision.
Falcon is long since end-of-life, so I don't have any NICs and can't run
 any tests, which maybe means the smart thing to do is just to leave well
 alone and not touch this code beyond your factoring.

*twitches with barely-suppressed urge to fix it anyway*
-ed

PS: I spent about two hours reading device documentation from 2008
 because I thought it said Falcon did 4-tuple hashing on UDP too.  For
 the record: the 'Falcon hash' was broken (in some unspecified way), so
 falcon_init_rx_cfg() selects the Toeplitz hash which does indeed only
 consume port numbers on these devices if protocol is TCP.  And I will
 never get that time back :/