be2net: don't flip hw_features when VXLANs are added/deleted
authorDavide Caratti <dcaratti@redhat.com>
Wed, 3 Oct 2018 13:20:58 +0000 (15:20 +0200)
committerDavid S. Miller <davem@davemloft.net>
Fri, 5 Oct 2018 07:59:21 +0000 (00:59 -0700)
commit2d52527e80c2dc0c5f43f50adf183781262ec565
tree9236400b15184a6d52116da13cdcd0dec1cda418
parent9d2f67e43b73e8af7438be219b66a5de0cfa8bd9
be2net: don't flip hw_features when VXLANs are added/deleted

the be2net implementation of .ndo_tunnel_{add,del}() changes the value of
NETIF_F_GSO_UDP_TUNNEL bit in 'features' and 'hw_features', but it forgets
to call netdev_features_change(). Moreover, ethtool setting for that bit
can potentially be reverted after a tunnel is added or removed.

GSO already does software segmentation when 'hw_enc_features' is 0, even
if VXLAN offload is turned on. In addition, commit 096de2f83ebc ("benet:
stricter vxlan offloading check in be_features_check") avoids hardware
segmentation of non-VXLAN tunneled packets, or VXLAN packets having wrong
destination port. So, it's safe to avoid flipping the above feature on
addition/deletion of VXLAN tunnels.

Fixes: 630f4b70567f ("be2net: Export tunnel offloads only when a VxLAN tunnel is created")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/emulex/benet/be_main.c