On Tue, 29 May 2007 17:10:52 -0700
"Michael Chan" [off-list ref] wrote:

On Wed, 2007-05-30 at 07:36 +1000, Herbert Xu wrote:

quoted

I just checked e1000 and it's correct as it does use the csum_offset
when doing TX offload.  However, you're definitely right that bnx2
seems to be broken.

quoted

A few devices take a offset, starting point, and insertion point. This looks like
the correct model. But no upper layer protocols other than IPV4/IPV6 can do checksum
offload at present, so it seems moot.

I could easily whip up a patch to get GRE to use it for a start :)

quoted

IMHO the correct solution would be to get rid if NETIF_F_HW_SUM and make a new flag
NETIF_F_IPV6_SUM. Devices that can checksum both could do NETIF_F_IPV4_SUM|NETI_F_IPV6_SUM.

We should definitely keep NETIF_F_HW_SUM for sane hardware such as the
e1000.  Unfortunately we may just have to invent IPV6_SUM for the broken
ones.

Ccing Michael to see if the bnx2 chip can actually do offset-based
checksum offload.

bnx2 and tg3 cannot do offset-based checksumming because the hardware
doesn't have room in the buffer descriptors to specify the offsets.  So
regrettably, the NETIF_F_HW_SUM flag has been misused in these drivers.
A new NETIF_F_IPV6_SUM flag will be very useful for us.

Look furthur many drivers are just plain broken and use F_HW_SUM
and can't even do IPV6 properly.  I'll fix

The worst code award goes to: qla3xxx.c
which is broken on IPV6 and goes to trouble of computing all the
offsets and they are already there in skb...



-- 
Stephen Hemminger [off-list ref]