[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <14385191E87B904DBD836449AA30269DABBFEC@MORGANITE.micrel.com>
Date: Mon, 11 Apr 2011 11:21:24 -0700
From: "Ha, Tristram" <Tristram.Ha@...rel.Com>
To: Michał Mirosław <mirq-linux@...e.qmqm.pl>
Cc: <netdev@...r.kernel.org>, "David S. Miller" <davem@...emloft.net>
Subject: RE: [PATCH] net: ksz884x: convert to hw_features
Michał Mirosław wrote:
> This also fixes possible race when changing receive checksum state and removes
> IPV6_CSUM_GEN_HACK as it's always set.
>
> BTW, The claim about fake IPV6 checksum looks dubious. If that were true, then there's a problem
> in networking core and should be fixed there and not in random drivers.
>
> BTW#2, there's no MAINTAINERS entry for this driver. I assume this driver is supported by Micrel?
>
> Signed-off-by: Michał Mirosław <mirq-linux@...e.qmqm.pl>
> ---
Micrel KSZ884X PCI Ethernet Controller does not support generating checksums on IPv6 packets. But if the driver advertises IPv6 checksum generation support and uses the skb_copy_and_csum_dev function to consolidate the socket buffer, the network performance is actually better than the normal method. Common sense seems to dictate that using the normal method is faster than the scatter/gather then consolidation method, but actually this method is faster. Indeed that is another network driver in the kernel that advertises IPv4 checksum generation support but does not do anything special in hardware and just uses the skb_copy_and_csum_dev function. If this method is no longer faster or workable, then KSZ884X driver should not advertise IPv6 checksum generation.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists