Add support for UDP Segmentation Offloading in software (SW USO). This
is helpful for testing when real hardware is not available. A test which
uses this codepath will be added in a following commit.
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Joe Damato <joe@dama.to>
---
v5:
- Added Pavan's Reviewed-by. No functional changes.
v4:
- Added parentheses around the gso_type check for clarity. No functional
change.
rfcv2:
- new in rfcv2
drivers/net/netdevsim/netdev.c | 100 ++++++++++++++++++++++++++++++++-
1 file changed, 99 insertions(+), 1 deletion(-)
diff --git a/drivers/net/netdevsim/netdev.c b/drivers/net/netdevsim/netdev.c
index c71b8d116f18..f228bcf3d190 100644
--- a/drivers/net/netdevsim/netdev.c
+++ b/drivers/net/netdevsim/netdev.c
@@ -30,6 +30,7 @@
#include <net/rtnetlink.h>
#include <net/udp_tunnel.h>
#include <net/busy_poll.h>
+#include <net/tso.h>
#include "netdevsim.h"
@@ -120,6 +121,98 @@ static int nsim_forward_skb(struct net_device *tx_dev,
return nsim_napi_rx(tx_dev, rx_dev, rq, skb);
}
+static netdev_tx_t nsim_uso_segment_xmit(struct net_device *dev,
+ struct sk_buff *skb)
+{
+ unsigned int hdr_len, mss, total_payload, num_segs;
+ struct netdevsim *ns = netdev_priv(dev);
+ struct net_device *peer_dev;
+ unsigned int total_len = 0;
+ struct netdevsim *peer_ns;
+ struct nsim_rq *rq;
+ struct tso_t tso;
+ int i, rxq;
+
+ hdr_len = tso_start(skb, &tso);
+ mss = skb_shinfo(skb)->gso_size;
+ total_payload = skb->len - hdr_len;
+ num_segs = DIV_ROUND_UP(total_payload, mss);
+
+ udp_hdr(skb)->check = 0;
+ if (!tso.ipv6)
+ ip_hdr(skb)->check = 0;
+
+ rcu_read_lock();
+ peer_ns = rcu_dereference(ns->peer);
+ if (!peer_ns)
+ goto out_drop_free;
+
+ peer_dev = peer_ns->netdev;
+ rxq = skb_get_queue_mapping(skb);
+ if (rxq >= peer_dev->num_rx_queues)
+ rxq = rxq % peer_dev->num_rx_queues;
+ rq = peer_ns->rq[rxq];
+
+ for (i = 0; i < num_segs; i++) {
+ unsigned int seg_payload = min_t(unsigned int, mss,
+ total_payload);
+ bool last = (i == num_segs - 1);
+ unsigned int seg_remaining;
+ struct sk_buff *seg;
+
+ seg = alloc_skb(hdr_len + seg_payload, GFP_ATOMIC);
+ if (!seg)
+ break;
+
+ seg->dev = dev;
+
+ tso_build_hdr(skb, skb_put(seg, hdr_len), &tso,
+ seg_payload, last);
+
+ if (!tso.ipv6) {
+ unsigned int nh_off = skb_network_offset(skb);
+ struct iphdr *iph;
+
+ iph = (struct iphdr *)(seg->data + nh_off);
+ iph->check = ip_fast_csum(iph, iph->ihl);
+ }
+
+ seg_remaining = seg_payload;
+ while (seg_remaining > 0) {
+ unsigned int chunk = min_t(unsigned int, tso.size,
+ seg_remaining);
+
+ memcpy(skb_put(seg, chunk), tso.data, chunk);
+ tso_build_data(skb, &tso, chunk);
+ seg_remaining -= chunk;
+ }
+
+ total_payload -= seg_payload;
+
+ seg->ip_summed = CHECKSUM_UNNECESSARY;
+
+ if (nsim_forward_skb(dev, peer_dev, seg, rq, NULL) == NET_RX_DROP)
+ continue;
+
+ total_len += hdr_len + seg_payload;
+ }
+
+ if (!hrtimer_active(&rq->napi_timer))
+ hrtimer_start(&rq->napi_timer, us_to_ktime(5),
+ HRTIMER_MODE_REL);
+
+ rcu_read_unlock();
+ dev_kfree_skb(skb);
+ dev_dstats_tx_add(dev, total_len);
+ return NETDEV_TX_OK;
+
+out_drop_free:
+ dev_kfree_skb(skb);
+ rcu_read_unlock();
+ dev_dstats_tx_dropped(dev);
+ return NETDEV_TX_OK;
+}
+
static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct netdevsim *ns = netdev_priv(dev);
@@ -132,6 +225,10 @@ static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
int rxq;
int dr;
+ if (skb_is_gso(skb) &&
+ skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4)
+ return nsim_uso_segment_xmit(dev, skb);
+
rcu_read_lock();
if (!nsim_ipsec_tx(ns, skb))
goto out_drop_any;
@@ -938,7 +1035,8 @@ static void nsim_setup(struct net_device *dev)
NETIF_F_HW_CSUM |
NETIF_F_LRO |
NETIF_F_TSO |
- NETIF_F_LOOPBACK;
+ NETIF_F_LOOPBACK |
+ NETIF_F_GSO_UDP_L4;
dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS;
dev->max_mtu = ETH_MAX_MTU;
dev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_HW_OFFLOAD;
--
2.52.0