On 10/18/19, 5:03 PM, "Benjamin Herrenschmidt" <[email protected]>
wrote:
On Fri, 2019-10-18 at 22:50 +0000, Vijay Khemka wrote:
> I don't have much understanding of IP Stack but I went through code
details and
> you are right and found that it should fallback to SW calculation for
IPV6 but it doesn't
> happen because ftgmac100_hard_start_xmit checks for CHECKSUM_PARTIAL
before
> setting HW checksum and calling ftgmac100_prep_tx_csum function. And in
my
> understanding, this value is set CHECKSUM_PARTIAL in IP stack. I looked
up IP stack for
> IPV6, file net/ipv6/ip6_output.c, function __ip6_append_data: here it
sets
> CHECKSUM_PARTIAL only for UDP packets not for TCP packets. Please look at
line
> number 1880. This could be an issue we are seeing here as why
> ftgmac100_prep_tx_csum is not getting triggered for IPV6 with TCP. Please
correct
> me if my understanding is wrong.
>
Not entirely sure. tcp_v6_send_response() in tcp_ipv6.c does set
CHECKSUM_PARTIAL as well. I don't really know how things are being
handled in that part of the network stack though.
From a driver perspective, if the value of ip_summed is not
CHECKSUM_PARTIAL it means we should not have to calculate any checksum.
At least that's my understanding here.
You may need to add some traces to the driver to see what you get in
there, what protocol indication etc... and analyze the corresponding
packets with something like tcpdump or wireshark on the other end.
Thanks Ben,
I will try to add some trace and test whatever possible and test it. As we
don't have tcpdump into our image and I have limited understanding of
networking stack so if you get some time to verify ipv6, it will be really
helpful.
Cheers,
Ben.
