[PATCH] ftgmac100: Disable HW checksum generation on AST2500
f.fainelli at gmail.com
Thu Sep 12 04:34:24 AEST 2019
On 9/11/19 11:30 AM, Vijay Khemka wrote:
> On 9/10/19, 4:08 PM, "Linux-aspeed on behalf of Vijay Khemka" <linux-aspeed-bounces+vijaykhemka=fb.com at lists.ozlabs.org on behalf of vijaykhemka at fb.com> wrote:
> On 9/10/19, 3:50 PM, "Linux-aspeed on behalf of Vijay Khemka" <linux-aspeed-bounces+vijaykhemka=fb.com at lists.ozlabs.org on behalf of vijaykhemka at fb.com> wrote:
> On 9/10/19, 3:05 PM, "Florian Fainelli" <f.fainelli at gmail.com> wrote:
> On 9/10/19 2:37 PM, Vijay Khemka wrote:
> > HW checksum generation is not working for AST2500, specially with IPV6
> > over NCSI. All TCP packets with IPv6 get dropped. By disabling this
> > it works perfectly fine with IPV6.
> > Verified with IPV6 enabled and can do ssh.
> How about IPv4, do these packets have problem? If not, can you continue
> advertising NETIF_F_IP_CSUM but take out NETIF_F_IPV6_CSUM?
> I changed code from (netdev->hw_features &= ~NETIF_F_HW_CSUM) to
> (netdev->hw_features &= ~NETIF_F_ IPV6_CSUM). And it is not working.
> Don't know why. IPV4 works without any change but IPv6 needs HW_CSUM
> Now I changed to
> netdev->hw_features &= (~NETIF_F_HW_CSUM) | NETIF_F_IP_CSUM;
> And it works.
> I investigated more on these features and found that we cannot set NETIF_F_IP_CSUM
> While NETIF_F_HW_CSUM is set. So I disabled NETIF_F_HW_CSUM first and enabled
> NETIF_F_IP_CSUM in next statement. And it works fine.
> But as per line 166 in include/linux/skbuff.h,
> * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
> * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
> * checksum offload capability.
> Please suggest which of below 2 I should do. As both works for me.
> 1. Disable completely NETIF_F_HW_CSUM and do nothing. This is original patch.
> 2. Enable NETIF_F_IP_CSUM in addition to 1. I can have v2 if this is accepted.
Sounds like 2 would leave the option of offloading IPv4 checksum
offload, so that would be a better middle group than flat out disable
checksum offload for both IPv4 and IPv6, no?
More information about the Linux-aspeed