| Message ID | 5f247a2ef20cae297db4d0a130515d0b7a1b8110.1616139307.git.lucien.xin@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [ipsec] esp: delete NETIF_F_SCTP_CRC bit from features for esp offload | expand |
On Fri, Mar 19, 2021 at 03:35:07PM +0800, Xin Long wrote: > Now in esp4/6_gso_segment(), before calling inner proto .gso_segment, > NETIF_F_CSUM_MASK bits are deleted, as HW won't be able to do the > csum for inner proto due to the packet encrypted already. > > So the UDP/TCP packet has to do the checksum on its own .gso_segment. > But SCTP is using CRC checksum, and for that NETIF_F_SCTP_CRC should > be deleted to make SCTP do the csum in own .gso_segment as well. > > In Xiumei's testing with SCTP over IPsec/veth, the packets are kept > dropping due to the wrong CRC checksum. > > Reported-by: Xiumei Mu <xmu@redhat.com> > Fixes: 7862b4058b9f ("esp: Add gso handlers for esp4 and esp6") > Signed-off-by: Xin Long <lucien.xin@gmail.com> Applied, thanks Xin!
diff --git a/net/ipv4/esp4_offload.c b/net/ipv4/esp4_offload.c index 601f5fb..ed3de48 100644 --- a/net/ipv4/esp4_offload.c +++ b/net/ipv4/esp4_offload.c @@ -217,10 +217,12 @@ static struct sk_buff *esp4_gso_segment(struct sk_buff *skb, if ((!(skb->dev->gso_partial_features & NETIF_F_HW_ESP) && !(features & NETIF_F_HW_ESP)) || x->xso.dev != skb->dev) - esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK); + esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK | + NETIF_F_SCTP_CRC); else if (!(features & NETIF_F_HW_ESP_TX_CSUM) && !(skb->dev->gso_partial_features & NETIF_F_HW_ESP_TX_CSUM)) - esp_features = features & ~NETIF_F_CSUM_MASK; + esp_features = features & ~(NETIF_F_CSUM_MASK | + NETIF_F_SCTP_CRC); xo->flags |= XFRM_GSO_SEGMENT; diff --git a/net/ipv6/esp6_offload.c b/net/ipv6/esp6_offload.c index 1ca516f..f35203a 100644 --- a/net/ipv6/esp6_offload.c +++ b/net/ipv6/esp6_offload.c @@ -254,9 +254,11 @@ static struct sk_buff *esp6_gso_segment(struct sk_buff *skb, skb->encap_hdr_csum = 1; if (!(features & NETIF_F_HW_ESP) || x->xso.dev != skb->dev) - esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK); + esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK | + NETIF_F_SCTP_CRC); else if (!(features & NETIF_F_HW_ESP_TX_CSUM)) - esp_features = features & ~NETIF_F_CSUM_MASK; + esp_features = features & ~(NETIF_F_CSUM_MASK | + NETIF_F_SCTP_CRC); xo->flags |= XFRM_GSO_SEGMENT;
Now in esp4/6_gso_segment(), before calling inner proto .gso_segment, NETIF_F_CSUM_MASK bits are deleted, as HW won't be able to do the csum for inner proto due to the packet encrypted already. So the UDP/TCP packet has to do the checksum on its own .gso_segment. But SCTP is using CRC checksum, and for that NETIF_F_SCTP_CRC should be deleted to make SCTP do the csum in own .gso_segment as well. In Xiumei's testing with SCTP over IPsec/veth, the packets are kept dropping due to the wrong CRC checksum. Reported-by: Xiumei Mu <xmu@redhat.com> Fixes: 7862b4058b9f ("esp: Add gso handlers for esp4 and esp6") Signed-off-by: Xin Long <lucien.xin@gmail.com> --- net/ipv4/esp4_offload.c | 6 ++++-- net/ipv6/esp6_offload.c | 6 ++++-- 2 files changed, 8 insertions(+), 4 deletions(-)