Skip to contentSkip to conversation

BIG TCP for UDP tunnels

PATCH Superseded Alice Mikityanska 5 participants 22 messages Jun 11 – Jul 6

Subsystems (25)

v7 of 9 Superseded 2 of 11 applied Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v7 cover BIG TCP for UDP tunnels

From: Alice Mikityanska <[email protected]>

This series is a follow-up to "BIG TCP without HBH in IPv6", and it adds support for BIG TCP IPv4/IPv6 workloads in vxlan and geneve. Now that IPv6 BIG TCP doesn't require stripping the HBH in all various combinations in tunneled traffic, adding BIG TCP becomes feasible.

Patches 01-02 are small fixups to some related code that I'm changing in the series.

Patch 03 adds accessors for the length field in the UDP header, as suggested by Paolo in review. The usage of udp_set_len is then added in the following patches that start using length=0 in BIG TCP UDP packets.

Patches 04-06 close the gaps that prevent BIG TCP packets from going through UDP tunnel code.

Patch 07 validates packets in udp_gro_receive to exclude packets with length=0 from GRO aggregation.

Patch 08 is for proper formatting in tcpdump (set UDP len to 0 rather than a trimmed value on overflow).

Patches 09-10 bump up tso_max_size for VXLAN and GENEVE.

Patch 11 adds selftests.

Thanks all!

v7 changes: Addressed Paolo's comments to properly block malformed packets with UDP length=0 at udp_rcv level.

v6: https://lore.kernel.org/netdev/[email protected]/

v6 changes: Lowered the packets threshold in the selftest to pass upstream CI on debug kernels, also made it configurable.

v5: https://lore.kernel.org/netdev/[email protected]/

v5 changes: Rebased, dropped one of the patches that came in via the net tree, addressed an overflow in nsim_do_psp found by Sashiko.

v4: https://lore.kernel.org/netdev/[email protected]/

v4 changes: Rebased, addressed Sashiko AI review [1] and Willem's comment about netperf flags.

My comments on Sashiko AI review per patch:

01: I'd prefer to keep the cases that I haven't tested outside of the scope of this series, which is for VXLAN/GENEVE tunnels, not for ESP.

02: The patch doesn't have behavioral changes other than fixing the checksum. The final uh->len assignment assigns the actual length, not 64k.

03: The check can't be loosened, because total_len is also assigned to UDP length. BIG TCP works in the mode without GRO hint option.

04: Fixed the fallback value for udplen. BIG TCP is fine, it's the uh->len = 0 case, uh->len can't exceed 64k.

05: In the BIG TCP case, partial GSO splits the SKB in two. If full segmentation is needed, the SKB is split in many MSS-sized SKBs without fragments.

06: Invalid packets from the wire are addressed in 08.

07: __udp_gso_segment only handles UDP GSO packets, which can't be bigger than 64k. Kept udp_set_len_short in nf_nat_mangle_udp_packet, as the function doesn't support BIG TCP packets anyway (see mangle_contents).

08: udp_gro_receive handles packets before aggregation, there are no BIG TCP packets at this point. RFC 768 doesn't say that padded UDP packets are valid. Real jumbograms don't seem to be supported in this path anyway.

09: The packet goes to skb_udp_tunnel_segment, not __udp_gso_segment. __skb_udp_tunnel_segment handles this case.

12: Improved process cleanup by killing everything inside netns before deleting them, and by avoiding killing netserver outside of netns. netperf with -r in TCP_STREAM mode works, and the option has the intended effect, but I replaced it with -m. Added dependency check for tcpdump.

[1]: https://sashiko.dev/#/patchset/20260410150943.993350-1-alice.kernel%40fastmail.im

v3: https://lore.kernel.org/netdev/[email protected]/

v3 changes: Fixed the redirect in the selftest, rebased over my L2TP fix
[2] for the syzbot report [3].

[2]: https://lore.kernel.org/netdev/[email protected]/
[3]: https://lore.kernel.org/netdev/[email protected]/

v2: https://lore.kernel.org/netdev/[email protected]/

v2 changes: Addressed the review comments: added UDP len helpers, consolidated UDP len sanity checks in patch 08 into one, added selftests. Added fixups to related code (patch 01-03).

v1: https://lore.kernel.org/netdev/[email protected]/

Alice Mikityanska (10):
net/sched: act_csum: don't mangle UDP tunnel GSO packets
geneve: Fix off-by-one comparing with GRO_LEGACY_MAX_SIZE
net: Use helpers to get/set UDP len tree-wide
net: Enable BIG TCP with partial GSO
udp: Support BIG TCP GSO packets where they can occur
udp: Support gro_ipv4_max_size > 65536
udp: Validate UDP length in udp_gro_receive
udp: Set length in UDP header to 0 for big GSO packets
vxlan: Enable BIG TCP packets
selftests: net: Add a test for BIG TCP in UDP tunnels

Daniel Borkmann (1):
geneve: Enable BIG TCP packets

 drivers/infiniband/core/lag.c                 |   2 +-
 drivers/infiniband/sw/rxe/rxe_net.c           |   4 +-
 drivers/net/amt.c                             |   6 +-
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   |   2 +-
 drivers/net/ethernet/intel/iavf/iavf_txrx.c   |   2 +-
 drivers/net/ethernet/intel/ice/ice_txrx.c     |   2 +-
 drivers/net/ethernet/intel/idpf/idpf_txrx.c   |   2 +-
 .../marvell/octeontx2/nic/otx2_txrx.c         |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |   4 +-
 .../ethernet/mellanox/mlx5/core/en_selftest.c |   2 +-
 drivers/net/ethernet/sfc/falcon/selftest.c    |   4 +-
 drivers/net/ethernet/sfc/selftest.c           |   4 +-
 drivers/net/ethernet/sfc/siena/selftest.c     |   4 +-
 drivers/net/ethernet/sfc/tc_encap_actions.c   |   2 +-
 .../stmicro/stmmac/stmmac_selftests.c         |   4 +-
 drivers/net/geneve.c                          |   6 +-
 drivers/net/netconsole.c                      |   2 +-
 drivers/net/netdevsim/dev.c                   |   2 +-
 drivers/net/netdevsim/psample.c               |   2 +-
 drivers/net/netdevsim/psp.c                   |   8 +-
 drivers/net/vxlan/vxlan_core.c                |   2 +
 drivers/net/wireguard/receive.c               |   2 +-
 include/linux/udp.h                           |  27 ++++
 include/trace/events/icmp.h                   |   2 +-
 lib/tests/blackhole_dev_kunit.c               |   2 +-
 net/6lowpan/nhc_udp.c                         |  10 +-
 net/core/pktgen.c                             |   4 +-
 net/core/selftests.c                          |   4 +-
 net/core/skbuff.c                             |  10 +-
 net/core/tso.c                                |   3 +-
 net/ipv4/esp4.c                               |   2 +-
 net/ipv4/fou_core.c                           |   2 +-
 net/ipv4/ipconfig.c                           |   6 +-
 net/ipv4/netfilter/nf_nat_snmp_basic_main.c   |   4 +-
 net/ipv4/route.c                              |   2 +-
 net/ipv4/udp.c                                |   7 +-
 net/ipv4/udp_offload.c                        |  49 +++---
 net/ipv4/udp_tunnel_core.c                    |   2 +-
 net/ipv6/esp6.c                               |   5 +-
 net/ipv6/fou6.c                               |   2 +-
 net/ipv6/ip6_udp_tunnel.c                     |   2 +-
 net/ipv6/udp.c                                |   7 +-
 net/ipv6/udp_offload.c                        |   2 +-
 net/l2tp/l2tp_core.c                          |   2 +-
 net/netfilter/ipvs/ip_vs_xmit.c               |   2 +-
 net/netfilter/nf_conntrack_proto_udp.c        |  15 +-
 net/netfilter/nf_log_syslog.c                 |   2 +-
 net/netfilter/nf_nat_helper.c                 |   2 +-
 net/psp/psp_main.c                            |   2 +-
 net/sched/act_csum.c                          |  12 +-
 net/xfrm/xfrm_nat_keepalive.c                 |   2 +-
 tools/testing/selftests/net/Makefile          |   1 +
 .../testing/selftests/net/big_tcp_tunnels.sh  | 151 ++++++++++++++++++
 53 files changed, 310 insertions(+), 105 deletions(-)
 create mode 100755 tools/testing/selftests/net/big_tcp_tunnels.sh
Signature · 1 line
2.54.0
#1 lore raw

PATCH v7 1/11 net/sched: act_csum: don't mangle UDP tunnel GSO packets

From: Alice Mikityanska <[email protected]>

Similar to commit add641e7dee3 ("sched: act_csum: don't mangle TCP and UDP GSO packets"), UDP tunnel GSO packets going through act_csum shouldn't have their checksum calculated at this point, because it will be done after segmentation. Setting the checksum in act_csum modifies skb->ip_summed and prevents inner IP csum offload from kicking in, resulting in a packet with a bad checksum.

Add UDP tunnel GSO packets to the exceptions, and also add UDP GSO (SKB_GSO_UDP_L4), as the same logic as in the commit mentioned above applies to UDP GSO too.

---
 net/sched/act_csum.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
net/sched/act_csum.c +6 -2
index a9e4635d899e..078d3a27130b 100644
@@ -259,7 +259,9 @@ static int tcf_csum_ipv4_udp(struct sk_buff *skb, unsigned int ihl,
 	const struct iphdr *iph;
 	u16 ul;
 
-	if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
+	if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type &
+	    (SKB_GSO_UDP | SKB_GSO_UDP_L4 |
+	     SKB_GSO_UDP_TUNNEL | SKB_GSO_UDP_TUNNEL_CSUM))
 		return 1;
 
 	/*
@@ -315,7 +317,9 @@ static int tcf_csum_ipv6_udp(struct sk_buff *skb, unsigned int ihl,
 	const struct ipv6hdr *ip6h;
 	u16 ul;
 
-	if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
+	if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type &
+	    (SKB_GSO_UDP | SKB_GSO_UDP_L4 |
+	     SKB_GSO_UDP_TUNNEL | SKB_GSO_UDP_TUNNEL_CSUM))
 		return 1;
 
 	/*
Signature · 1 line
2.54.0
#2 lore raw

PATCH v7 2/11 geneve: Fix off-by-one comparing with GRO_LEGACY_MAX_SIZE

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

GRO_LEGACY_MAX_SIZE = 65536; total_len being 65536 is too big to fit into a u16. As can be seen in skb_gro_receive, packets bigger or equal to gro_max_size (or GRO_LEGACY_MAX_SIZE) are dropped with -E2BIG. Apply the same boundary to geneve_post_decap_hint to avoid writing 65536 to a 16-bit iph->tot_len field with an overflow.

Fixes fd0dd796576e ("geneve: use GRO hint option in the RX path")
---
 drivers/net/geneve.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/geneve.c +1 -1
index 23b42466a7c9..9afff7bcaa0b 100644
@@ -604,7 +604,7 @@ static int geneve_post_decap_hint(const struct sock *sk, struct sk_buff *skb,
 	ipv6h = (void *)skb->data + gro_hint->nested_nh_offset;
 	iph = (struct iphdr *)ipv6h;
 	total_len = skb->len - gro_hint->nested_nh_offset;
-	if (total_len > GRO_LEGACY_MAX_SIZE)
+	if (total_len >= GRO_LEGACY_MAX_SIZE)
 		return -E2BIG;
 
 	/*
Signature · 1 line
2.54.0
maintainer
#3 lore raw

Re: [PATCH net-next v7 02/11] geneve: Fix off-by-one comparing with GRO_LEGACY_MAX_SIZE

Quoted 11 lines · Alice Mikityanska

From: Alice Mikityanska <[email protected]>

GRO_LEGACY_MAX_SIZE = 65536; total_len being 65536 is too big to fit into a u16. As can be seen in skb_gro_receive, packets bigger or equal to gro_max_size (or GRO_LEGACY_MAX_SIZE) are dropped with -E2BIG. Apply the same boundary to geneve_post_decap_hint to avoid writing 65536 to a 16-bit iph->tot_len field with an overflow.

Fixes fd0dd796576e ("geneve: use GRO hint option in the RX path")

This patch and the previous one are pretty much uncontroversial. Let me apply them, so that the next iteration will be smaller.

/P

#4 lore raw

PATCH v7 3/11 net: Use helpers to get/set UDP len tree-wide

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

Since BIG TCP for UDP tunnels will start using len=0 in the UDP header as an indicator of a GSO packet bigger than 65535 bytes, this commit introduces the following getter and setters to use tree-wide, in order to explicitly mark places where len=0 may be expected, and handle them properly:

1. udp_set_len() sets uh->len to its real value if it's not bigger than 65535, and to 0 otherwise: to be used in GSO context with aggregated packets.

2. udp_set_len_short() is to be used when the length is known to fit 16 bits. It WARNs when the caller tries to assign a bigger value if CONFIG_DEBUG_NET=y.

3. udp_get_len_short() returns len in host byte order: to be used on the RX side to deal with non-aggregated packets, or to access the raw value of the len field.

4. udp_get_len() decodes uh->len set by udp_set_len(). It checks whether the packet is GSO to guard from malformed packets.

At the moment udp_set_len() is not used, a following commit will start using it after enabling len>65535 for GSO.

Raw uh->len (in network byte order) is still accessed in a few places for checksum calculation purposes, and to decode len=0 in udpv6_rcv for jumbograms. udp_rcv and udpv6_rcv will be addressed by the commit that starts using udp_set_len() to set UDP len=0 for BIG TCP packets in UDP tunnels.

---
 drivers/infiniband/core/lag.c                 |  2 +-
 drivers/infiniband/sw/rxe/rxe_net.c           |  4 +-
 drivers/net/amt.c                             |  6 +--
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   |  2 +-
 drivers/net/ethernet/intel/iavf/iavf_txrx.c   |  2 +-
 drivers/net/ethernet/intel/ice/ice_txrx.c     |  2 +-
 drivers/net/ethernet/intel/idpf/idpf_txrx.c   |  2 +-
 .../marvell/octeontx2/nic/otx2_txrx.c         |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  4 +-
 .../ethernet/mellanox/mlx5/core/en_selftest.c |  2 +-
 drivers/net/ethernet/sfc/falcon/selftest.c    |  4 +-
 drivers/net/ethernet/sfc/selftest.c           |  4 +-
 drivers/net/ethernet/sfc/siena/selftest.c     |  4 +-
 drivers/net/ethernet/sfc/tc_encap_actions.c   |  2 +-
 .../stmicro/stmmac/stmmac_selftests.c         |  4 +-
 drivers/net/geneve.c                          |  2 +-
 drivers/net/netconsole.c                      |  2 +-
 drivers/net/netdevsim/dev.c                   |  2 +-
 drivers/net/netdevsim/psample.c               |  2 +-
 drivers/net/netdevsim/psp.c                   |  8 ++--
 drivers/net/wireguard/receive.c               |  2 +-
 include/linux/udp.h                           | 27 ++++++++++++++
 include/trace/events/icmp.h                   |  2 +-
 lib/tests/blackhole_dev_kunit.c               |  2 +-
 net/6lowpan/nhc_udp.c                         | 10 ++---
 net/core/pktgen.c                             |  4 +-
 net/core/selftests.c                          |  4 +-
 net/core/tso.c                                |  3 +-
 net/ipv4/esp4.c                               |  2 +-
 net/ipv4/fou_core.c                           |  2 +-
 net/ipv4/ipconfig.c                           |  6 +--
 net/ipv4/netfilter/nf_nat_snmp_basic_main.c   |  4 +-
 net/ipv4/route.c                              |  2 +-
 net/ipv4/udp.c                                |  3 +-
 net/ipv4/udp_offload.c                        | 37 +++++++++----------
 net/ipv4/udp_tunnel_core.c                    |  2 +-
 net/ipv6/esp6.c                               |  5 ++-
 net/ipv6/fou6.c                               |  2 +-
 net/ipv6/ip6_udp_tunnel.c                     |  2 +-
 net/ipv6/udp.c                                |  3 +-
 net/ipv6/udp_offload.c                        |  2 +-
 net/l2tp/l2tp_core.c                          |  2 +-
 net/netfilter/ipvs/ip_vs_xmit.c               |  2 +-
 net/netfilter/nf_conntrack_proto_udp.c        | 15 +++++++-
 net/netfilter/nf_log_syslog.c                 |  2 +-
 net/netfilter/nf_nat_helper.c                 |  2 +-
 net/psp/psp_main.c                            |  2 +-
 net/sched/act_csum.c                          |  4 +-
 net/xfrm/xfrm_nat_keepalive.c                 |  2 +-
 49 files changed, 131 insertions(+), 88 deletions(-)
drivers/infiniband/core/lag.c +1 -1
index 8fd80adfe833..00fe241737ff 100644
@@ -36,7 +36,7 @@ static struct sk_buff *rdma_build_skb(struct net_device *netdev,
 	uh->source =
 		htons(rdma_flow_label_to_udp_sport(ah_attr->grh.flow_label));
 	uh->dest = htons(ROCE_V2_UDP_DPORT);
-	uh->len = htons(sizeof(struct udphdr));
+	udp_set_len_short(uh, sizeof(struct udphdr));
 
 	if (is_ipv4) {
 		skb_push(skb, sizeof(struct iphdr));
drivers/infiniband/sw/rxe/rxe_net.c +2 -2
index 082ff387d081..898069deac55 100644
@@ -242,7 +242,7 @@ static int rxe_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
 	pkt->port_num = 1;
 	pkt->hdr = (u8 *)(udph + 1);
 	pkt->mask = RXE_GRH_MASK;
-	pkt->paylen = be16_to_cpu(udph->len) - sizeof(*udph);
+	pkt->paylen = udp_get_len_short(udph) - sizeof(*udph);
 
 	/* remove udp header */
 	skb_pull(skb, sizeof(struct udphdr));
@@ -305,7 +305,7 @@ static void prepare_udp_hdr(struct sk_buff *skb, __be16 src_port,
 
 	udph->dest = dst_port;
 	udph->source = src_port;
-	udph->len = htons(skb->len);
+	udp_set_len_short(udph, skb->len);
 	udph->check = 0;
 }
 
drivers/net/amt.c +3 -3
index 724a8163a514..25f9601512f2 100644
@@ -667,7 +667,7 @@ static void amt_send_discovery(struct amt_dev *amt)
 	udph		= udp_hdr(skb);
 	udph->source	= amt->gw_port;
 	udph->dest	= amt->relay_port;
-	udph->len	= htons(sizeof(*udph) + sizeof(*amtd));
+	udp_set_len_short(udph, sizeof(*udph) + sizeof(*amtd));
 	udph->check	= 0;
 	offset = skb_transport_offset(skb);
 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
@@ -758,7 +758,7 @@ static void amt_send_request(struct amt_dev *amt, bool v6)
 	udph		= udp_hdr(skb);
 	udph->source	= amt->gw_port;
 	udph->dest	= amt->relay_port;
-	udph->len	= htons(sizeof(*amtrh) + sizeof(*udph));
+	udp_set_len_short(udph, sizeof(*amtrh) + sizeof(*udph));
 	udph->check	= 0;
 	offset = skb_transport_offset(skb);
 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
@@ -2608,7 +2608,7 @@ static void amt_send_advertisement(struct amt_dev *amt, __be32 nonce,
 	udph		= udp_hdr(skb);
 	udph->source	= amt->relay_port;
 	udph->dest	= dport;
-	udph->len	= htons(sizeof(*amta) + sizeof(*udph));
+	udp_set_len_short(udph, sizeof(*amta) + sizeof(*udph));
 	udph->check	= 0;
 	offset = skb_transport_offset(skb);
 	skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
drivers/net/ethernet/intel/i40e/i40e_txrx.c +1 -1
index 894f2d06d39d..ef5e657816f0 100644
@@ -3129,7 +3129,7 @@ static int i40e_tso(struct i40e_tx_buffer *first, u8 *hdr_len,
 					 SKB_GSO_UDP_TUNNEL_CSUM)) {
 		if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) &&
 		    (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)) {
-			l4.udp->len = 0;
+			udp_set_len_short(l4.udp, 0);
 
 			/* determine offset of outer transport header */
 			l4_offset = l4.hdr - skb->data;
drivers/net/ethernet/intel/iavf/iavf_txrx.c +1 -1
index 363c42bf3dcf..c30abf17cf5d 100644
@@ -1774,7 +1774,7 @@ static int iavf_tso(struct iavf_tx_buffer *first, u8 *hdr_len,
 					 SKB_GSO_UDP_TUNNEL_CSUM)) {
 		if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) &&
 		    (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)) {
-			l4.udp->len = 0;
+			udp_set_len_short(l4.udp, 0);
 
 			/* determine offset of outer transport header */
 			l4_offset = l4.hdr - skb->data;
drivers/net/ethernet/intel/ice/ice_txrx.c +1 -1
index 4ca1a0602307..fdea2758adf1 100644
@@ -1893,7 +1893,7 @@ int ice_tso(struct ice_tx_buf *first, struct ice_tx_offload_params *off)
 					 SKB_GSO_UDP_TUNNEL_CSUM)) {
 		if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) &&
 		    (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)) {
-			l4.udp->len = 0;
+			udp_set_len_short(l4.udp, 0);
 
 			/* determine offset of outer transport header */
 			l4_start = (u8)(l4.hdr - skb->data);
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index f6b3b15364ff..276296e321ed 100644
@@ -2871,7 +2871,7 @@ int idpf_tso(struct sk_buff *skb, struct idpf_tx_offload_params *off)
 				     (__force __wsum)htonl(paylen));
 		/* compute length of segmentation header */
 		off->tso_hdr_len = sizeof(struct udphdr) + l4_start;
-		l4.udp->len = htons(shinfo->gso_size + sizeof(struct udphdr));
+		udp_set_len_short(l4.udp, shinfo->gso_size + sizeof(struct udphdr));
 		break;
 	default:
 		return -EINVAL;
drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c +1 -1
index 625bb5a05344..8d2d607bc92f 100644
@@ -750,7 +750,7 @@ static void otx2_sqe_add_ext(struct otx2_nic *pfvf, struct otx2_snd_queue *sq,
 				ext->lso_format = pfvf->hw.lso_udpv6_idx;
 			}
 
-			udph->len = htons(sizeof(struct udphdr));
+			udp_set_len_short(udph, sizeof(struct udphdr));
 		}
 	} else if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) {
 		ext->tstmp = 1;
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +2 -2
index 6fbc0441c4b8..04af54b704d8 100644
@@ -1081,7 +1081,7 @@ static void mlx5e_shampo_update_ipv4_udp_hdr(struct mlx5e_rq *rq, struct iphdr *
 	struct udphdr *uh;
 
 	uh = (struct udphdr *)(skb->data + udp_off);
-	uh->len = htons(skb->len - udp_off);
+	udp_set_len_short(uh, skb->len - udp_off);
 
 	if (uh->check)
 		uh->check = ~udp_v4_check(skb->len - udp_off, ipv4->saddr,
@@ -1100,7 +1100,7 @@ static void mlx5e_shampo_update_ipv6_udp_hdr(struct mlx5e_rq *rq, struct ipv6hdr
 	struct udphdr *uh;
 
 	uh = (struct udphdr *)(skb->data + udp_off);
-	uh->len = htons(skb->len - udp_off);
+	udp_set_len_short(uh, skb->len - udp_off);
 
 	if (uh->check)
 		uh->check = ~udp_v6_check(skb->len - udp_off, &ipv6->saddr,
drivers/net/ethernet/mellanox/mlx5/core/en_selftest.c +1 -1
index accc26d1a872..1dcdb86690bb 100644
@@ -113,7 +113,7 @@ static struct sk_buff *mlx5e_test_get_udp_skb(struct mlx5e_priv *priv)
 	/* Fill UDP header */
 	udph->source = htons(9);
 	udph->dest = htons(9); /* Discard Protocol */
-	udph->len = htons(sizeof(struct mlx5ehdr) + sizeof(struct udphdr));
+	udp_set_len_short(udph, sizeof(struct mlx5ehdr) + sizeof(struct udphdr));
 	udph->check = 0;
 
 	/* Fill IP header */
drivers/net/ethernet/sfc/falcon/selftest.c +2 -2
index db4dd7fb77f5..4d29e0baf2eb 100644
@@ -401,8 +401,8 @@ static void ef4_iterate_state(struct ef4_nic *efx)
 
 	/* Initialise udp header */
 	payload->udp.source = 0;
-	payload->udp.len = htons(sizeof(*payload) -
-				 offsetof(struct ef4_loopback_payload, udp));
+	udp_set_len_short(&payload->udp, sizeof(*payload) -
+			  offsetof(struct ef4_loopback_payload, udp));
 	payload->udp.check = 0;	/* checksum ignored */
 
 	/* Fill out payload */
drivers/net/ethernet/sfc/selftest.c +2 -2
index 8ec76329237a..dc716feb79cb 100644
@@ -398,8 +398,8 @@ static void efx_iterate_state(struct efx_nic *efx)
 
 	/* Initialise udp header */
 	payload->udp.source = 0;
-	payload->udp.len = htons(sizeof(*payload) -
-				 offsetof(struct efx_loopback_payload, udp));
+	udp_set_len_short(&payload->udp, sizeof(*payload) -
+			  offsetof(struct efx_loopback_payload, udp));
 	payload->udp.check = 0;	/* checksum ignored */
 
 	/* Fill out payload */
drivers/net/ethernet/sfc/siena/selftest.c +2 -2
index 930643612df5..c74cf5131364 100644
@@ -399,8 +399,8 @@ static void efx_iterate_state(struct efx_nic *efx)
 
 	/* Initialise udp header */
 	payload->udp.source = 0;
-	payload->udp.len = htons(sizeof(*payload) -
-				 offsetof(struct efx_loopback_payload, udp));
+	udp_set_len_short(&payload->udp, sizeof(*payload) -
+			  offsetof(struct efx_loopback_payload, udp));
 	payload->udp.check = 0;	/* checksum ignored */
 
 	/* Fill out payload */
drivers/net/ethernet/sfc/tc_encap_actions.c +1 -1
index db222abef53b..c2ad3a358d20 100644
@@ -311,7 +311,7 @@ static void efx_gen_tun_header_udp(struct efx_tc_encap_action *encap, u8 len)
 	encap->encap_hdr_len += sizeof(*udp);
 
 	udp->dest = key->tp_dst;
-	udp->len = cpu_to_be16(sizeof(*udp) + len);
+	udp_set_len_short(udp, sizeof(*udp) + len);
 }
 
 static void efx_gen_tun_header_vxlan(struct efx_tc_encap_action *encap)
drivers/net/ethernet/stmicro/stmmac/stmmac_selftests.c +2 -2
index a0c75886587c..29e824bd90ca 100644
@@ -154,9 +154,9 @@ static struct sk_buff *stmmac_test_get_udp_skb(struct stmmac_priv *priv,
 	} else {
 		uhdr->source = htons(attr->sport);
 		uhdr->dest = htons(attr->dport);
-		uhdr->len = htons(sizeof(*shdr) + sizeof(*uhdr) + attr->size);
+		udp_set_len_short(uhdr, sizeof(*shdr) + sizeof(*uhdr) + attr->size);
 		if (attr->max_size)
-			uhdr->len = htons(attr->max_size -
+			udp_set_len_short(uhdr, attr->max_size -
 					  (sizeof(*ihdr) + sizeof(*ehdr)));
 		uhdr->check = 0;
 	}
drivers/net/geneve.c +1 -1
index 9afff7bcaa0b..f5ea426c7deb 100644
@@ -631,7 +631,7 @@ static int geneve_post_decap_hint(const struct sock *sk, struct sk_buff *skb,
 
 	/* Adjust the nested UDP header len and checksum. */
 	uh = udp_hdr(skb);
-	uh->len = htons(skb->len - gro_hint->nested_tp_offset);
+	udp_set_len_short(uh, skb->len - gro_hint->nested_tp_offset);
 	if (uh->check) {
 		len = skb->len - gro_hint->nested_tp_offset;
 		skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL_CSUM;
drivers/net/netconsole.c +1 -1
index 606e265cdfd7..b4a9f62cc2f7 100644
@@ -1785,7 +1785,7 @@ static void push_udp(struct netpoll *np, struct sk_buff *skb, int len)
 	udph = udp_hdr(skb);
 	udph->source = htons(np->local_port);
 	udph->dest = htons(np->remote_port);
-	udph->len = htons(udp_len);
+	udp_set_len_short(udph, udp_len);
 
 	netpoll_udp_checksum(np, skb, len);
 }
drivers/net/netdevsim/dev.c +1 -1
index aed9ad5f1b43..f65b4cf4ea39 100644
@@ -845,7 +845,7 @@ static struct sk_buff *nsim_dev_trap_skb_build(void)
 	udph = skb_put_zero(skb, sizeof(struct udphdr) + data_len);
 	get_random_bytes(&udph->source, sizeof(u16));
 	get_random_bytes(&udph->dest, sizeof(u16));
-	udph->len = htons(sizeof(struct udphdr) + data_len);
+	udp_set_len_short(udph, sizeof(struct udphdr) + data_len);
 
 	return skb;
 }
drivers/net/netdevsim/psample.c +1 -1
index 717d157c3ae2..1e71c3da4def 100644
@@ -73,7 +73,7 @@ static struct sk_buff *nsim_dev_psample_skb_build(void)
 	udph = skb_put_zero(skb, sizeof(struct udphdr) + data_len);
 	get_random_bytes(&udph->source, sizeof(u16));
 	get_random_bytes(&udph->dest, sizeof(u16));
-	udph->len = htons(sizeof(struct udphdr) + data_len);
+	udp_set_len_short(udph, sizeof(struct udphdr) + data_len);
 
 	return skb;
 }
drivers/net/netdevsim/psp.c +5 -3
index 59c990fdc79e..d3e36c74be62 100644
@@ -84,6 +84,7 @@ nsim_do_psp(struct sk_buff *skb, struct netdevsim *ns,
 		struct iphdr *iph;
 		struct udphdr *uh;
 		__wsum csum;
+		int udplen;
 
 		/* Do not decapsulate. Receive the skb with the udp and psp
 		 * headers still there as if this is a normal udp packet.
@@ -91,19 +92,20 @@ nsim_do_psp(struct sk_buff *skb, struct netdevsim *ns,
 		 * provide a valid checksum here, so the skb isn't dropped.
 		 */
 		uh = udp_hdr(skb);
+		udplen = ntohs(uh->len) ?: skb->len - skb_transport_offset(skb);
 		csum = skb_checksum(skb, skb_transport_offset(skb),
-				    ntohs(uh->len), 0);
+				    udplen, 0);
 
 		switch (skb->protocol) {
 		case htons(ETH_P_IP):
 			iph = ip_hdr(skb);
-			uh->check = udp_v4_check(ntohs(uh->len), iph->saddr,
+			uh->check = udp_v4_check(udplen, iph->saddr,
 						 iph->daddr, csum);
 			break;
 #if IS_ENABLED(CONFIG_IPV6)
 		case htons(ETH_P_IPV6):
 			ip6h = ipv6_hdr(skb);
-			uh->check = udp_v6_check(ntohs(uh->len), &ip6h->saddr,
+			uh->check = udp_v6_check(udplen, &ip6h->saddr,
 						 &ip6h->daddr, csum);
 			break;
 #endif
drivers/net/wireguard/receive.c +1 -1
index eb8851113654..275fe1bc994c 100644
@@ -62,7 +62,7 @@ static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg)
 		 * to have UDP fields.
 		 */
 		return -EINVAL;
-	data_len = ntohs(udp->len);
+	data_len = udp_get_len_short(udp); /* GRO not expected here. */
 	if (unlikely(data_len < sizeof(struct udphdr) ||
 		     data_len > skb->len - data_offset))
 		/* UDP packet is reporting too small of a size or lying about
include/linux/udp.h +27 -0
index ce56ebcee5cb..998906ec3b32 100644
@@ -23,6 +23,33 @@ static inline struct udphdr *udp_hdr(const struct sk_buff *skb)
 	return (struct udphdr *)skb_transport_header(skb);
 }
 
+static inline unsigned int udp_get_len(const struct sk_buff *skb,
+				       const struct udphdr *uh,
+				       unsigned int dataoff)
+{
+	if (uh->len)
+		return ntohs(uh->len);
+	if (skb_is_gso(skb)) /* BIG TCP */
+		return skb->len - dataoff;
+	return 0;
+}
+
+static inline unsigned int udp_get_len_short(const struct udphdr *uh)
+{
+	return ntohs(uh->len);
+}
+
+static inline void udp_set_len(struct udphdr *uh, unsigned int len)
+{
+	uh->len = len < GRO_LEGACY_MAX_SIZE ? htons(len) : 0;
+}
+
+static inline void udp_set_len_short(struct udphdr *uh, unsigned int len)
+{
+	DEBUG_NET_WARN_ON_ONCE(len >= GRO_LEGACY_MAX_SIZE);
+	uh->len = htons(len);
+}
+
 #define UDP_HTABLE_SIZE_MIN_PERNET	128
 #define UDP_HTABLE_SIZE_MIN		(IS_ENABLED(CONFIG_BASE_SMALL) ? 128 : 256)
 #define UDP_HTABLE_SIZE_MAX		65536
include/trace/events/icmp.h +1 -1
index 31559796949a..09ae115099df 100644
@@ -44,7 +44,7 @@ TRACE_EVENT(icmp_send,
 			} else {
 				__entry->sport = ntohs(uh->source);
 				__entry->dport = ntohs(uh->dest);
-				__entry->ulen = ntohs(uh->len);
+				__entry->ulen = udp_get_len_short(uh);
 			}
 
 			p32 = (__be32 *) __entry->saddr;
lib/tests/blackhole_dev_kunit.c +1 -1
index 06834ab35f43..fa3e0533038d 100644
@@ -46,7 +46,7 @@ static void test_blackholedev(struct kunit *test)
 	uh = (struct udphdr *)skb_push(skb, sizeof(struct udphdr));
 	skb_set_transport_header(skb, 0);
 	uh->source = uh->dest = htons(UDP_PORT);
-	uh->len = htons(data_len);
+	udp_set_len_short(uh, data_len);
 	uh->check = 0;
 	/* (Network) IPv6 */
 	ip6h = (struct ipv6hdr *)skb_push(skb, sizeof(struct ipv6hdr));
net/6lowpan/nhc_udp.c +5 -5
index 0a506c77283d..ed4227e6db74 100644
@@ -88,16 +88,16 @@ static int udp_uncompress(struct sk_buff *skb, size_t needed)
 	switch (lowpan_dev(skb->dev)->lltype) {
 	case LOWPAN_LLTYPE_IEEE802154:
 		if (lowpan_802154_cb(skb)->d_size)
-			uh.len = htons(lowpan_802154_cb(skb)->d_size -
-				       sizeof(struct ipv6hdr));
+			udp_set_len_short(&uh, lowpan_802154_cb(skb)->d_size -
+					  sizeof(struct ipv6hdr));
 		else
-			uh.len = htons(skb->len + sizeof(struct udphdr));
+			udp_set_len_short(&uh, skb->len + sizeof(struct udphdr));
 		break;
 	default:
-		uh.len = htons(skb->len + sizeof(struct udphdr));
+		udp_set_len_short(&uh, skb->len + sizeof(struct udphdr));
 		break;
 	}
-	pr_debug("uncompressed UDP length: src = %d", ntohs(uh.len));
+	pr_debug("uncompressed UDP length: src = %d", udp_get_len_short(&uh));
 
 	/* replace the compressed UDP head by the uncompressed UDP
 	 * header
net/core/pktgen.c +2 -2
index 8e185b318288..5b4dd04d6124 100644
@@ -3005,7 +3005,7 @@ static struct sk_buff *fill_packet_ipv4(struct net_device *odev,
 
 	udph->source = htons(pkt_dev->cur_udp_src);
 	udph->dest = htons(pkt_dev->cur_udp_dst);
-	udph->len = htons(datalen + 8);	/* DATA + udphdr */
+	udp_set_len_short(udph, datalen + 8);	/* DATA + udphdr */
 	udph->check = 0;
 
 	iph->ihl = 5;
@@ -3138,7 +3138,7 @@ static struct sk_buff *fill_packet_ipv6(struct net_device *odev,
 	udplen = datalen + sizeof(struct udphdr);
 	udph->source = htons(pkt_dev->cur_udp_src);
 	udph->dest = htons(pkt_dev->cur_udp_dst);
-	udph->len = htons(udplen);
+	udp_set_len_short(udph, udplen);
 	udph->check = 0;
 
 	*(__be32 *) iph = htonl(0x60000000);	/* Version + flow */
net/core/selftests.c +2 -2
index 0a203d3fb9dc..36b949ae520b 100644
@@ -72,9 +72,9 @@ struct sk_buff *net_test_get_skb(struct net_device *ndev, u8 id,
 	} else {
 		uhdr->source = htons(attr->sport);
 		uhdr->dest = htons(attr->dport);
-		uhdr->len = htons(sizeof(*shdr) + sizeof(*uhdr) + attr->size);
+		udp_set_len_short(uhdr, sizeof(*shdr) + sizeof(*uhdr) + attr->size);
 		if (attr->max_size)
-			uhdr->len = htons(attr->max_size -
+			udp_set_len_short(uhdr, attr->max_size -
 					  (sizeof(*ihdr) + sizeof(*ehdr)));
 		uhdr->check = 0;
 	}
net/core/tso.c +2 -1
index 347b3856ddb9..d2934bcfa795 100644
@@ -39,7 +39,8 @@ void tso_build_hdr(const struct sk_buff *skb, char *hdr, struct tso_t *tso,
 	} else {
 		struct udphdr *uh = (struct udphdr *)hdr;
 
-		uh->len = htons(sizeof(*uh) + size);
+		/* size is after segmentation. */
+		udp_set_len_short(uh, sizeof(*uh) + size);
 	}
 }
 EXPORT_SYMBOL(tso_build_hdr);
net/ipv4/esp4.c +1 -1
index 513c8215c947..d4bce3684b5e 100644
@@ -320,7 +320,7 @@ static struct ip_esp_hdr *esp_output_udp_encap(struct sk_buff *skb,
 	uh = (struct udphdr *)esp->esph;
 	uh->source = sport;
 	uh->dest = dport;
-	uh->len = htons(len);
+	udp_set_len_short(uh, len);
 	uh->check = 0;
 
 	/* For IPv4 ESP with UDP encapsulation, if xo is not null, the skb is in the crypto offload
net/ipv4/fou_core.c +1 -1
index 865bd7205122..a50740d0f288 100644
@@ -1040,7 +1040,7 @@ static void fou_build_udp(struct sk_buff *skb, struct ip_tunnel_encap *e,
 
 	uh->dest = e->dport;
 	uh->source = sport;
-	uh->len = htons(skb->len);
+	udp_set_len_short(uh, skb->len);
 	udp_set_csum(!(e->flags & TUNNEL_ENCAP_FLAG_CSUM), skb,
 		     fl4->saddr, fl4->daddr, skb->len);
 
net/ipv4/ipconfig.c +3 -3
index a35ffedacc7c..155db067eaec 100644
@@ -847,7 +847,7 @@ static void __init ic_bootp_send_if(struct ic_device *d, unsigned long jiffies_d
 	/* Construct UDP header */
 	b->udph.source = htons(68);
 	b->udph.dest = htons(67);
-	b->udph.len = htons(sizeof(struct bootp_pkt) - sizeof(struct iphdr));
+	udp_set_len_short(&b->udph, sizeof(struct bootp_pkt) - sizeof(struct iphdr));
 	/* UDP checksum not calculated -- explicitly allowed in BOOTP RFC */
 
 	/* Construct DHCP/BOOTP header */
@@ -1025,10 +1025,10 @@ static int __init ic_bootp_recv(struct sk_buff *skb, struct net_device *dev, str
 	if (b->udph.source != htons(67) || b->udph.dest != htons(68))
 		goto drop;
 
-	if (ntohs(h->tot_len) < ntohs(b->udph.len) + sizeof(struct iphdr))
+	if (ntohs(h->tot_len) < udp_get_len_short(&b->udph) + sizeof(struct iphdr))
 		goto drop;
 
-	len = ntohs(b->udph.len) - sizeof(struct udphdr);
+	len = udp_get_len_short(&b->udph) - sizeof(struct udphdr);
 	ext_len = len - (sizeof(*b) -
 			 sizeof(struct iphdr) -
 			 sizeof(struct udphdr) -
net/ipv4/netfilter/nf_nat_snmp_basic_main.c +2 -2
index 0ede138dfd29..bd41bdef491c 100644
@@ -127,7 +127,7 @@ static int snmp_translate(struct nf_conn *ct, int dir, struct sk_buff *skb)
 {
 	struct iphdr *iph = ip_hdr(skb);
 	struct udphdr *udph = (struct udphdr *)((__be32 *)iph + iph->ihl);
-	u16 datalen = ntohs(udph->len) - sizeof(struct udphdr);
+	u16 datalen = udp_get_len_short(udph) - sizeof(struct udphdr);
 	char *data = (unsigned char *)udph + sizeof(struct udphdr);
 	struct snmp_ctx ctx;
 	int ret;
@@ -181,7 +181,7 @@ static int help(struct sk_buff *skb, unsigned int protoff,
 	 * enough room for a UDP header.  Just verify the UDP length field so we
 	 * can mess around with the payload.
 	 */
-	if (ntohs(udph->len) != skb->len - (iph->ihl << 2)) {
+	if (udp_get_len_short(udph) != skb->len - (iph->ihl << 2)) {
 		nf_ct_helper_log(skb, ct, "dropping malformed packet\n");
 		return NF_DROP;
 	}
net/ipv4/route.c +1 -1
index 3f3de5164d6e..0825ed983714 100644
@@ -3187,7 +3187,7 @@ static struct sk_buff *inet_rtm_getroute_build_skb(__be32 src, __be32 dst,
 		udph = skb_put_zero(skb, sizeof(struct udphdr));
 		udph->source = sport;
 		udph->dest = dport;
-		udph->len = htons(sizeof(struct udphdr));
+		udp_set_len_short(udph, sizeof(struct udphdr));
 		udph->check = 0;
 		break;
 	}
net/ipv4/udp.c +2 -1
index 70f6cbd4ef73..240a1adab278 100644
@@ -1107,7 +1107,8 @@ static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4,
 	uh = udp_hdr(skb);
 	uh->source = inet_sk(sk)->inet_sport;
 	uh->dest = fl4->fl4_dport;
-	uh->len = htons(len);
+	/* Datagram length checked in udp_sendmsg. */
+	udp_set_len_short(uh, len);
 	uh->check = 0;
 
 	if (cork->gso_size) {
net/ipv4/udp_offload.c +18 -19
index 29651b1a0bc7..493e2b9e16fb 100644
@@ -279,11 +279,11 @@ static struct sk_buff *__skb_udp_tunnel_segment(struct sk_buff *skb,
 		 * segment instead of the entire frame.
 		 */
 		if (gso_partial && skb_is_gso(skb)) {
-			uh->len = htons(skb_shinfo(skb)->gso_size +
-					SKB_GSO_CB(skb)->data_offset +
-					skb->head - (unsigned char *)uh);
+			udp_set_len_short(uh, skb_shinfo(skb)->gso_size +
+					  SKB_GSO_CB(skb)->data_offset +
+					  skb->head - (unsigned char *)uh);
 		} else {
-			uh->len = htons(len);
+			udp_set_len_short(uh, len);
 		}
 
 		if (!need_csum)
@@ -468,7 +468,7 @@ static struct sk_buff *__udp_gso_segment_list(struct sk_buff *skb,
 	if (IS_ERR(skb))
 		return skb;
 
-	udp_hdr(skb)->len = htons(sizeof(struct udphdr) + mss);
+	udp_set_len_short(udp_hdr(skb), sizeof(struct udphdr) + mss);
 
 	if (is_ipv6)
 		return __udpv6_gso_segment_list_csum(skb);
@@ -486,8 +486,8 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
 	unsigned int mss;
 	bool copy_dtor;
 	__sum16 check;
-	__be16 newlen;
 	int ret = 0;
+	u16 newlen;
 
 	mss = skb_shinfo(gso_skb)->gso_size;
 	if (gso_skb->len <= sizeof(*uh) + mss)
@@ -564,8 +564,8 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
 			(skb_shinfo(gso_skb)->tx_flags & SKBTX_ANY_TSTAMP);
 
 	/* compute checksum adjustment based on old length versus new */
-	newlen = htons(sizeof(*uh) + mss);
-	check = csum16_add(csum16_sub(uh->check, uh->len), newlen);
+	newlen = sizeof(*uh) + mss;
+	check = csum16_add(csum16_sub(uh->check, uh->len), htons(newlen));
 
 	for (;;) {
 		if (copy_dtor) {
@@ -577,7 +577,7 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
 		if (!seg->next)
 			break;
 
-		uh->len = newlen;
+		udp_set_len_short(uh, newlen);
 		uh->check = check;
 
 		if (seg->ip_summed == CHECKSUM_PARTIAL)
@@ -594,11 +594,10 @@ struct sk_buff *__udp_gso_segment(struct sk_buff *gso_skb,
 	 * segment may not be full MSS, account for that in the checksum
 	 */
 	if (!skb_is_gso(seg))
-		newlen = htons(skb_tail_pointer(seg) -
-			       skb_transport_header(seg) + seg->data_len);
-	check = csum16_add(csum16_sub(uh->check, uh->len), newlen);
+		newlen = skb_tail_pointer(seg) - skb_transport_header(seg) + seg->data_len;
+	check = csum16_add(csum16_sub(uh->check, uh->len), htons(newlen));
 
-	uh->len = newlen;
+	udp_set_len_short(uh, newlen);
 	uh->check = check;
 
 	if (seg->ip_summed == CHECKSUM_PARTIAL)
@@ -709,7 +708,7 @@ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
 	}
 
 	/* Do not deal with padded or malicious packets, sorry ! */
-	ulen = ntohs(uh->len);
+	ulen = udp_get_len_short(uh);
 	if (ulen <= sizeof(*uh) || ulen != skb_gro_len(skb)) {
 		NAPI_GRO_CB(skb)->flush = 1;
 		return NULL;
@@ -742,7 +741,7 @@ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
 		 * On len mismatch merge the first packet shorter than gso_size,
 		 * otherwise complete the GRO packet.
 		 */
-		if (ulen > ntohs(uh2->len) || flush) {
+		if (ulen > udp_get_len_short(uh2) || flush) {
 			pp = p;
 		} else {
 			if (NAPI_GRO_CB(skb)->is_flist) {
@@ -765,7 +764,7 @@ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
 			}
 		}
 
-		if (ret || ulen != ntohs(uh2->len) ||
+		if (ret || ulen != udp_get_len_short(uh2) ||
 		    NAPI_GRO_CB(p)->count >= UDP_GRO_CNT_MAX)
 			pp = p;
 
@@ -915,12 +914,12 @@ static int udp_gro_complete_segment(struct sk_buff *skb)
 int udp_gro_complete(struct sk_buff *skb, int nhoff,
 		     udp_lookup_t lookup)
 {
-	__be16 newlen = htons(skb->len - nhoff);
 	struct udphdr *uh = (struct udphdr *)(skb->data + nhoff);
+	unsigned int newlen = skb->len - nhoff;
 	struct sock *sk;
 	int err;
 
-	uh->len = newlen;
+	udp_set_len_short(uh, newlen);
 
 	sk = INDIRECT_CALL_INET(lookup, udp6_lib_lookup_skb,
 				udp4_lib_lookup_skb, skb, uh->source, uh->dest);
@@ -957,7 +956,7 @@ INDIRECT_CALLABLE_SCOPE int udp4_gro_complete(struct sk_buff *skb, int nhoff)
 
 	/* do fraglist only if there is no outer UDP encap (or we already processed it) */
 	if (NAPI_GRO_CB(skb)->is_flist && !NAPI_GRO_CB(skb)->encap_mark) {
-		uh->len = htons(skb->len - nhoff);
+		udp_set_len_short(uh, skb->len - nhoff);
 
 		skb_shinfo(skb)->gso_type |= (SKB_GSO_FRAGLIST|SKB_GSO_UDP_L4);
 		skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
net/ipv4/udp_tunnel_core.c +1 -1
index 9ab3728f9630..0fccb38f074d 100644
@@ -178,7 +178,7 @@ void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb
 
 	uh->dest = dst_port;
 	uh->source = src_port;
-	uh->len = htons(skb->len);
+	udp_set_len_short(uh, skb->len);
 
 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 
net/ipv6/esp6.c +3 -2
index 57481e423e59..69ceb36d3734 100644
@@ -227,7 +227,8 @@ static void esp_output_encap_csum(struct sk_buff *skb)
 	if (*skb_mac_header(skb) == IPPROTO_UDP) {
 		struct udphdr *uh = udp_hdr(skb);
 		struct ipv6hdr *ip6h = ipv6_hdr(skb);
-		int len = ntohs(uh->len);
+		/* esp6_output_udp_encap limits len to U16_MAX. */
+		int len = udp_get_len_short(uh);
 		unsigned int offset = skb_transport_offset(skb);
 		__wsum csum = skb_checksum(skb, offset, skb->len - offset, 0);
 
@@ -355,7 +356,7 @@ static struct ip_esp_hdr *esp6_output_udp_encap(struct sk_buff *skb,
 	uh = (struct udphdr *)esp->esph;
 	uh->source = sport;
 	uh->dest = dport;
-	uh->len = htons(len);
+	udp_set_len_short(uh, len);
 	uh->check = 0;
 
 	*skb_mac_header(skb) = IPPROTO_UDP;
net/ipv6/fou6.c +1 -1
index 157765259e2f..588929409241 100644
@@ -30,7 +30,7 @@ static void fou6_build_udp(struct sk_buff *skb, struct ip_tunnel_encap *e,
 
 	uh->dest = e->dport;
 	uh->source = sport;
-	uh->len = htons(skb->len);
+	udp_set_len_short(uh, skb->len);
 	udp6_set_csum(!(e->flags & TUNNEL_ENCAP_FLAG_CSUM6), skb,
 		      &fl6->saddr, &fl6->daddr, skb->len);
 
net/ipv6/ip6_udp_tunnel.c +1 -1
index 9adb5775487f..dcff7fb16ff6 100644
@@ -93,7 +93,7 @@ void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk,
 	uh->dest = dst_port;
 	uh->source = src_port;
 
-	uh->len = htons(skb->len);
+	udp_set_len_short(uh, skb->len);
 
 	skb_dst_set(skb, dst);
 
net/ipv6/udp.c +2 -1
index 15e032194ecc..913a19edfc7c 100644
@@ -1369,7 +1369,8 @@ static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6,
 	uh = udp_hdr(skb);
 	uh->source = fl6->fl6_sport;
 	uh->dest = fl6->fl6_dport;
-	uh->len = htons(len);
+	/* Datagram length checked in udpv6_sendmsg. */
+	udp_set_len_short(uh, len);
 	uh->check = 0;
 
 	if (cork->gso_size) {
net/ipv6/udp_offload.c +1 -1
index 778afc7453ce..c92cf5ee3e6a 100644
@@ -171,7 +171,7 @@ int udp6_gro_complete(struct sk_buff *skb, int nhoff)
 
 	/* do fraglist only if there is no outer UDP encap (or we already processed it) */
 	if (NAPI_GRO_CB(skb)->is_flist && !NAPI_GRO_CB(skb)->encap_mark) {
-		uh->len = htons(skb->len - nhoff);
+		udp_set_len_short(uh, skb->len - nhoff);
 
 		skb_shinfo(skb)->gso_type |= (SKB_GSO_FRAGLIST|SKB_GSO_UDP_L4);
 		skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
net/l2tp/l2tp_core.c +1 -1
index f940914959b1..4712cc41881a 100644
@@ -1296,7 +1296,7 @@ static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, uns
 			ret = NET_XMIT_DROP;
 			goto out_unlock;
 		}
-		uh->len = htons(udp_len);
+		udp_set_len_short(uh, udp_len);
 
 		/* Calculate UDP checksum if configured to do so */
 #if IS_ENABLED(CONFIG_IPV6)
net/netfilter/ipvs/ip_vs_xmit.c +1 -1
index ce542ed4b013..ae3ed2c00ec3 100644
@@ -1100,7 +1100,7 @@ ipvs_gue_encap(struct net *net, struct sk_buff *skb,
 	dport = cp->dest->tun_port;
 	udph->dest = dport;
 	udph->source = sport;
-	udph->len = htons(skb->len);
+	udp_set_len_short(udph, skb->len);
 	udph->check = 0;
 
 	*next_protocol = IPPROTO_UDP;
net/netfilter/nf_conntrack_proto_udp.c +13 -2
index cc9b7e5e1935..8a5675983e7c 100644
@@ -41,11 +41,22 @@ static void udp_error_log(const struct sk_buff *skb,
 	nf_l4proto_log_invalid(skb, state, IPPROTO_UDP, "%s", msg);
 }
 
+static bool udp_validate_len(struct sk_buff *skb,
+			     const struct udphdr *hdr,
+			     unsigned int dataoff)
+{
+	unsigned int udplen = udp_get_len_short(hdr);
+	unsigned int skblen = skb->len - dataoff;
+
+	if (udplen > skblen || udplen < sizeof(*hdr))
+		return false;
+	return true;
+}
+
 static bool udp_error(struct sk_buff *skb,
 		      unsigned int dataoff,
 		      const struct nf_hook_state *state)
 {
-	unsigned int udplen = skb->len - dataoff;
 	const struct udphdr *hdr;
 	struct udphdr _hdr;
 
@@ -57,7 +68,7 @@ static bool udp_error(struct sk_buff *skb,
 	}
 
 	/* Truncated/malformed packets */
-	if (ntohs(hdr->len) > udplen || ntohs(hdr->len) < sizeof(*hdr)) {
+	if (!udp_validate_len(skb, hdr, dataoff)) {
 		udp_error_log(skb, state, "truncated/malformed packet");
 		return true;
 	}
net/netfilter/nf_log_syslog.c +1 -1
index 7a8952b049d1..e98faedd3b95 100644
@@ -301,7 +301,7 @@ nf_log_dump_udp_header(struct nf_log_buf *m,
 
 	/* Max length: 20 "SPT=65535 DPT=65535 " */
 	nf_log_buf_add(m, "SPT=%u DPT=%u LEN=%u ",
-		       ntohs(uh->source), ntohs(uh->dest), ntohs(uh->len));
+		       ntohs(uh->source), ntohs(uh->dest), udp_get_len_short(uh));
 
 out:
 	return 0;
net/netfilter/nf_nat_helper.c +1 -1
index bf591e6af005..3853f41db499 100644
@@ -161,7 +161,7 @@ nf_nat_mangle_udp_packet(struct sk_buff *skb,
 
 	/* update the length of the UDP packet */
 	datalen = skb->len - protoff;
-	udph->len = htons(datalen);
+	udp_set_len_short(udph, datalen);
 
 	/* fix udp checksum if udp checksum was previously calculated */
 	if (!udph->check && skb->ip_summed != CHECKSUM_PARTIAL)
net/psp/psp_main.c +1 -1
index ccbbb2a5fa58..ce7a0ade450f 100644
@@ -211,7 +211,7 @@ static void psp_write_headers(struct net *net, struct sk_buff *skb, __be32 spi,
 		uh->source = udp_flow_src_port(net, skb, 0, 0, false);
 	}
 	uh->check = 0;
-	uh->len = htons(udp_len);
+	udp_set_len_short(uh, udp_len);
 
 	psph->nexthdr = IPPROTO_TCP;
 	psph->hdrlen = PSP_HDRLEN_NOOPT;
net/sched/act_csum.c +2 -2
index 078d3a27130b..5fff52a8ca90 100644
@@ -276,7 +276,7 @@ static int tcf_csum_ipv4_udp(struct sk_buff *skb, unsigned int ihl,
 		return 0;
 
 	iph = ip_hdr(skb);
-	ul = ntohs(udph->len);
+	ul = udp_get_len_short(udph);
 
 	if (udplite || udph->check) {
 
@@ -334,7 +334,7 @@ static int tcf_csum_ipv6_udp(struct sk_buff *skb, unsigned int ihl,
 		return 0;
 
 	ip6h = ipv6_hdr(skb);
-	ul = ntohs(udph->len);
+	ul = udp_get_len_short(udph);
 
 	udph->check = 0;
 
net/xfrm/xfrm_nat_keepalive.c +1 -1
index 458931062a04..906458f3d8c5 100644
@@ -133,7 +133,7 @@ static void nat_keepalive_send(struct nat_keepalive *ka)
 	uh = skb_push(skb, sizeof(*uh));
 	uh->source = ka->encap_sport;
 	uh->dest = ka->encap_dport;
-	uh->len = htons(skb->len);
+	udp_set_len_short(uh, skb->len);
 	uh->check = 0;
 
 	skb->mark = ka->smark;
Signature · 1 line
2.54.0
maintainer Acked-by on v7 3/11
#5 lore raw

Re: [PATCH net-next v7 03/11] net: Use helpers to get/set UDP len tree-wide

drivers/net/wireguard/receive.c static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg) · line 65

Quoted 10 lines · Alice Mikityanska
drivers/net/wireguard/receive.c +1 -1
index eb8851113654..275fe1bc994c 100644
@@ -62,7 +62,7 @@ static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg)
 		 * to have UDP fields.
 		 */
 		return -EINVAL;
-	data_len = ntohs(udp->len);
+	data_len = udp_get_len_short(udp); /* GRO not expected here. */

However, I don't think that comment is necessary and you can remove it if you wind up needing to roll a v+1. But also, I was thinking:

drivers/net/wireguard/receive.c static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg)

Quoted 5 lines
drivers/net/wireguard/receive.c +5 -0
+static inline unsigned int udp_get_len_short(const struct udphdr *uh)
+{
+	return ntohs(uh->len);
+}
+

Do you think it'd be better to call this udp_get_len_nogro(), which would then make its usage rules more explicitly without needing to litter comments?

#6 lore raw

Re: [PATCH net-next v7 03/11] net: Use helpers to get/set UDP len tree-wide

Quoted 15 lines · Jason A. Donenfeld

drivers/net/wireguard/receive.c static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg) · line 65

Quoted 10 lines · Alice Mikityanska
drivers/net/wireguard/receive.c +1 -1
index eb8851113654..275fe1bc994c 100644
@@ -62,7 +62,7 @@ static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg)
               * to have UDP fields.
               */
              return -EINVAL;
-     data_len = ntohs(udp->len);
+     data_len = udp_get_len_short(udp); /* GRO not expected here. */
Quoted 2 lines

However, I don't think that comment is necessary and you can remove it if you wind up needing to roll a v+1.

Thanks! Removing the comment before resubmitting.

Quoted 10 lines

But also, I was thinking:

drivers/net/wireguard/receive.c static int prepare_skb_header(struct sk_buff *skb, struct wg_device *wg)

Quoted 5 lines
drivers/net/wireguard/receive.c +5 -0
+static inline unsigned int udp_get_len_short(const struct udphdr *uh)
+{
+     return ntohs(uh->len);
+}
+

Do you think it'd be better to call this udp_get_len_nogro(), which would then make its usage rules more explicitly without needing to litter comments?

I'll probably keep the original name; "short [packet]" sounds better to me than nogro/nogso, and this naming doesn't exclude the usage of udp_get_len_short to get the raw value of len even in GSO packets.

#7 lore raw

PATCH v7 4/11 net: Enable BIG TCP with partial GSO

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

skb_segment is called for partial GSO, when netif_needs_gso returns true in validate_xmit_skb. Partial GSO is needed, for example, when segmentation of tunneled traffic is offloaded to a NIC that only supports inner checksum offload.

Currently, skb_segment clamps the segment length to 65534 bytes, because gso_size == 65535 is a special value GSO_BY_FRAGS, and we don't want to accidentally assign mss = 65535, as it would fall into the GSO_BY_FRAGS check further in the function.

This implementation, however, artificially blocks len > 65534, which is possible since the introduction of BIG TCP. To allow bigger lengths and avoid resegmentation of BIG TCP packets, store the gso_by_frags flag in the beginning and don't use a special value of mss for this purpose after mss was modified.

---
 drivers/net/netdevsim/psp.c |  2 +-
 net/core/skbuff.c           | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)
drivers/net/netdevsim/psp.c +1 -1
index d3e36c74be62..6b3532b5e360 100644
@@ -92,7 +92,7 @@ nsim_do_psp(struct sk_buff *skb, struct netdevsim *ns,
 		 * provide a valid checksum here, so the skb isn't dropped.
 		 */
 		uh = udp_hdr(skb);
-		udplen = ntohs(uh->len) ?: skb->len - skb_transport_offset(skb);
+		udplen = udp_get_len(skb, uh, skb_transport_offset(skb));
 		csum = skb_checksum(skb, skb_transport_offset(skb),
 				    udplen, 0);
 
net/core/skbuff.c +5 -5
index c64693fcb2d1..5dcee79df8cf 100644
@@ -4773,6 +4773,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	struct sk_buff *tail = NULL;
 	struct sk_buff *list_skb = skb_shinfo(head_skb)->frag_list;
 	unsigned int mss = skb_shinfo(head_skb)->gso_size;
+	bool gso_by_frags = mss == GSO_BY_FRAGS;
 	unsigned int doffset = head_skb->data - skb_mac_header(head_skb);
 	unsigned int offset = doffset;
 	unsigned int tnl_hlen = skb_tnl_header_len(head_skb);
@@ -4788,7 +4789,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	int nfrags, pos;
 
 	if ((skb_shinfo(head_skb)->gso_type & SKB_GSO_DODGY) &&
-	    mss != GSO_BY_FRAGS && mss != skb_headlen(head_skb)) {
+	    !gso_by_frags && mss != skb_headlen(head_skb)) {
 		struct sk_buff *check_skb;
 
 		for (check_skb = list_skb; check_skb; check_skb = check_skb->next) {
@@ -4816,7 +4817,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	sg = !!(features & NETIF_F_SG);
 	csum = !!can_checksum_protocol(features, proto);
 
-	if (sg && csum && (mss != GSO_BY_FRAGS))  {
+	if (sg && csum && !gso_by_frags)  {
 		if (!(features & NETIF_F_GSO_PARTIAL)) {
 			struct sk_buff *iter;
 			unsigned int frag_len;
@@ -4850,9 +4851,8 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 		/* GSO partial only requires that we trim off any excess that
 		 * doesn't fit into an MSS sized block, so take care of that
 		 * now.
-		 * Cap len to not accidentally hit GSO_BY_FRAGS.
 		 */
-		partial_segs = min(len, GSO_BY_FRAGS - 1) / mss;
+		partial_segs = len / mss;
 		if (partial_segs > 1)
 			mss *= partial_segs;
 		else
@@ -4876,7 +4876,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 		int hsize;
 		int size;
 
-		if (unlikely(mss == GSO_BY_FRAGS)) {
+		if (unlikely(gso_by_frags)) {
 			len = list_skb->len;
 		} else {
 			len = head_skb->len - offset;
Signature · 1 line
2.54.0
maintainer
#8 lore raw

Re: [PATCH net-next v7 04/11] net: Enable BIG TCP with partial GSO

Quoted 69 lines · Alice Mikityanska

From: Alice Mikityanska <[email protected]>

skb_segment is called for partial GSO, when netif_needs_gso returns true in validate_xmit_skb. Partial GSO is needed, for example, when segmentation of tunneled traffic is offloaded to a NIC that only supports inner checksum offload.

Currently, skb_segment clamps the segment length to 65534 bytes, because gso_size == 65535 is a special value GSO_BY_FRAGS, and we don't want to accidentally assign mss = 65535, as it would fall into the GSO_BY_FRAGS check further in the function.

This implementation, however, artificially blocks len > 65534, which is possible since the introduction of BIG TCP. To allow bigger lengths and avoid resegmentation of BIG TCP packets, store the gso_by_frags flag in the beginning and don't use a special value of mss for this purpose after mss was modified.

---
 drivers/net/netdevsim/psp.c |  2 +-
 net/core/skbuff.c           | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)
drivers/net/netdevsim/psp.c +1 -1
index d3e36c74be62..6b3532b5e360 100644
@@ -92,7 +92,7 @@ nsim_do_psp(struct sk_buff *skb, struct netdevsim *ns,
 		 * provide a valid checksum here, so the skb isn't dropped.
 		 */
 		uh = udp_hdr(skb);
-		udplen = ntohs(uh->len) ?: skb->len - skb_transport_offset(skb);
+		udplen = udp_get_len(skb, uh, skb_transport_offset(skb));
 		csum = skb_checksum(skb, skb_transport_offset(skb),
 				    udplen, 0);
 
net/core/skbuff.c +3 -2
index c64693fcb2d1..5dcee79df8cf 100644
@@ -4773,6 +4773,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	struct sk_buff *tail = NULL;
 	struct sk_buff *list_skb = skb_shinfo(head_skb)->frag_list;
 	unsigned int mss = skb_shinfo(head_skb)->gso_size;
+	bool gso_by_frags = mss == GSO_BY_FRAGS;
 	unsigned int doffset = head_skb->data - skb_mac_header(head_skb);
 	unsigned int offset = doffset;
 	unsigned int tnl_hlen = skb_tnl_header_len(head_skb);
@@ -4788,7 +4789,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	int nfrags, pos;
 
 	if ((skb_shinfo(head_skb)->gso_type & SKB_GSO_DODGY) &&
-	    mss != GSO_BY_FRAGS && mss != skb_headlen(head_skb)) {
+	    !gso_by_frags && mss != skb_headlen(head_skb)) {
 		struct sk_buff *check_skb;
 
 		for (check_skb = list_skb; check_skb; check_skb = check_skb->next) {
@@ -4816,7 +4817,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	sg = !!(features & NETIF_F_SG);
 	csum = !!can_checksum_protocol(features, proto);
 
-	if (sg && csum && (mss != GSO_BY_FRAGS))  {
+	if (sg && csum && !gso_by_frags)  {
 		if (!(features & NETIF_F_GSO_PARTIAL)) {
 			struct sk_buff *iter;
 			unsigned int frag_len;

net/core/skbuff.c struct sk_buff *skb_segment(struct sk_buff *head_skb, · line 4855

Quoted 7 lines
net/core/skbuff.c +1 -2
@@ -4850,9 +4851,8 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 		/* GSO partial only requires that we trim off any excess that
 		 * doesn't fit into an MSS sized block, so take care of that
 		 * now.
-		 * Cap len to not accidentally hit GSO_BY_FRAGS.
 		 */
-		partial_segs = min(len, GSO_BY_FRAGS - 1) / mss;
+		partial_segs = len / mss;

Sashiko/gemini says the above can lead to hit BUG_ON() later.

I *think* it's not a false positive, as it looks like skb_segment() assumes an skb can hold mss bytes without resorting to frag_list usage, and mss > MAX_SKB_FRAGS * PAGE_SIZE breaks such assumption.

I think handling correctly this case will requires some non trivial surgery to skb_segment: both while (pos < offset + len) { loops must be updated to feed data from frags as needed instead of BUG_ON()/net_warn_ratelimited(skb_shinfo(nskb)->nr_frags >= MAX_SKB_FRAGS);

/P

#9 lore raw

Re: [PATCH net-next v7 04/11] net: Enable BIG TCP with partial GSO

Quoted 80 lines · Paolo Abeni
Quoted 69 lines · Alice Mikityanska

From: Alice Mikityanska <[email protected]>

skb_segment is called for partial GSO, when netif_needs_gso returns true in validate_xmit_skb. Partial GSO is needed, for example, when segmentation of tunneled traffic is offloaded to a NIC that only supports inner checksum offload.

Currently, skb_segment clamps the segment length to 65534 bytes, because gso_size == 65535 is a special value GSO_BY_FRAGS, and we don't want to accidentally assign mss = 65535, as it would fall into the GSO_BY_FRAGS check further in the function.

This implementation, however, artificially blocks len > 65534, which is possible since the introduction of BIG TCP. To allow bigger lengths and avoid resegmentation of BIG TCP packets, store the gso_by_frags flag in the beginning and don't use a special value of mss for this purpose after mss was modified.

---
 drivers/net/netdevsim/psp.c |  2 +-
 net/core/skbuff.c           | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)
drivers/net/netdevsim/psp.c +1 -1
index d3e36c74be62..6b3532b5e360 100644
@@ -92,7 +92,7 @@ nsim_do_psp(struct sk_buff *skb, struct netdevsim *ns,
 		 * provide a valid checksum here, so the skb isn't dropped.
 		 */
 		uh = udp_hdr(skb);
-		udplen = ntohs(uh->len) ?: skb->len - skb_transport_offset(skb);
+		udplen = udp_get_len(skb, uh, skb_transport_offset(skb));
 		csum = skb_checksum(skb, skb_transport_offset(skb),
 				    udplen, 0);
 
net/core/skbuff.c +3 -2
index c64693fcb2d1..5dcee79df8cf 100644
@@ -4773,6 +4773,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	struct sk_buff *tail = NULL;
 	struct sk_buff *list_skb = skb_shinfo(head_skb)->frag_list;
 	unsigned int mss = skb_shinfo(head_skb)->gso_size;
+	bool gso_by_frags = mss == GSO_BY_FRAGS;
 	unsigned int doffset = head_skb->data - skb_mac_header(head_skb);
 	unsigned int offset = doffset;
 	unsigned int tnl_hlen = skb_tnl_header_len(head_skb);
@@ -4788,7 +4789,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	int nfrags, pos;
 
 	if ((skb_shinfo(head_skb)->gso_type & SKB_GSO_DODGY) &&
-	    mss != GSO_BY_FRAGS && mss != skb_headlen(head_skb)) {
+	    !gso_by_frags && mss != skb_headlen(head_skb)) {
 		struct sk_buff *check_skb;
 
 		for (check_skb = list_skb; check_skb; check_skb = check_skb->next) {
@@ -4816,7 +4817,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	sg = !!(features & NETIF_F_SG);
 	csum = !!can_checksum_protocol(features, proto);
 
-	if (sg && csum && (mss != GSO_BY_FRAGS))  {
+	if (sg && csum && !gso_by_frags)  {
 		if (!(features & NETIF_F_GSO_PARTIAL)) {
 			struct sk_buff *iter;
 			unsigned int frag_len;

net/core/skbuff.c struct sk_buff *skb_segment(struct sk_buff *head_skb, · line 4855

Quoted 7 lines
net/core/skbuff.c +1 -2
@@ -4850,9 +4851,8 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 		/* GSO partial only requires that we trim off any excess that
 		 * doesn't fit into an MSS sized block, so take care of that
 		 * now.
-		 * Cap len to not accidentally hit GSO_BY_FRAGS.
 		 */
-		partial_segs = min(len, GSO_BY_FRAGS - 1) / mss;
+		partial_segs = len / mss;

Sashiko/gemini says the above can lead to hit BUG_ON() later.

Quoted 8 lines

I *think* it's not a false positive, as it looks like skb_segment() assumes an skb can hold mss bytes without resorting to frag_list usage, and mss > MAX_SKB_FRAGS * PAGE_SIZE breaks such assumption.

I think handling correctly this case will requires some non trivial surgery to skb_segment: both while (pos < offset + len) { loops must be updated to feed data from frags as needed instead of BUG_ON()/net_warn_ratelimited(skb_shinfo(nskb)->nr_frags >= MAX_SKB_FRAGS);

Thanks, these are some valid points. I analyzed the code in skb_segment better, and from what I can tell, this issue exists before my changes, even without BIG TCP (I have a reproduction). Moreover, I don't think that my changes make the situation worse, because what truly matters is the geometry of the SKB coming into skb_segment, not the sizes of the frags as such.

The root of the issue is that skb_segment attempts to produce SKBs of the same size and without frag_list, but the incoming SKB can have a frag_list and more than 17 frags much smaller than PAGE_SIZE. For example, if the incoming SKB has 30 frags (using frag_list) of 500 bytes each, it's just 15000 bytes (much smaller than 64k or MAX_SKB_FRAGS * PAGE_SIZE), but the partial GSO flow will try to fit almost all of them (mss will be almost 15000) into a single non-frag_list SKB. Since skb_segment just reuses the existing frags without combining them in any way, this will end up with "too many frags".

This means that the assumption that an SKB can hold mss bytes without frag_list is wrong, and skb_segment is already broken as it relies on this assumption.

As for the solution, we can't just break the loop early before we put len bytes into an output SKB, because it will break the guarantee that the output SKBs have the same size. The only way I can imagine is to dry-run the loop over all frags in advance and determine the smallest length that any 17 consecutive frags can take together, and limit len to that value.

Why adding BIG TCP into the picture doesn't make the situation worse, in my opinion? There are two main practical ways to get into skb_segment with BIG TCP:

1. Sending a TCP stream from an application. tcp_sendmsg_locked doesn't build frag_list SKBs, stopping at sysctl_max_skb_frags. If a non- frag_list SKB enters skb_segment, it can handle it even if it's well above 64k. From my observation, 32768-byte high order pages are allocated to back the frags.

2. Receiving a TCP stream with GRO, then forwarding it out of another network interface, e.g., that uses GSO partial. This is indeed broken if GRO resorts to frag_list, but, as shown above, it's broken even without BIG TCP.

Down below, I'll attach a reproducer patch (mostly generated with AI) that adds debug prints to skb_segment and a script that sets up forwarding to a GSO partial VXLAN netdev. Setting net.core.high_order_alloc_disable is not even necessary: I can still observe a 29026-byte SKB with 21 frags and frag_list (most frags are 1448 bytes), it's just these frags are allocated from two order-3 pages.

Here are example logs that show the geometry of an SKB that comes from regular GRO (no BIG TCP) and breaks skb_segment:

skbuff: skb_segment: start head=ffff888101008a00 len=29026 doffset=66 mss=28960 gso_by_frags=0 features=0x0000010e501d0049
skbuff: skb_segment: input head[0] skb=ffff888101008a00 len=29026 data_len=28960 headlen=66 nr_frags=16 frag_list=ffff888104b7e900 head_frag=0 gso_size=1448 gso_segs=20 gso_type=0x801
skbuff: skb_segment: input head[0] frag[0] page=ffffea000420f7c0 head=ffffea000420f7c0 order=0 off=2480 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[1] page=ffffea000420f7c0 head=ffffea000420f7c0 order=0 off=3928 size=168 base_pages=1
skbuff: skb_segment: input head[0] frag[2] page=ffffea00048261c0 head=ffffea00048261c0 order=0 off=0 size=1280 base_pages=1
skbuff: skb_segment: input head[0] frag[3] page=ffffea00048261c0 head=ffffea00048261c0 order=0 off=1280 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[4] page=ffffea00048261c0 head=ffffea00048261c0 order=0 off=2728 size=1368 base_pages=1
skbuff: skb_segment: input head[0] frag[5] page=ffffea0004886400 head=ffffea0004886400 order=0 off=0 size=80 base_pages=1
skbuff: skb_segment: input head[0] frag[6] page=ffffea0004886400 head=ffffea0004886400 order=0 off=80 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[7] page=ffffea0004886400 head=ffffea0004886400 order=0 off=1528 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[8] page=ffffea0004886400 head=ffffea0004886400 order=0 off=2976 size=1120 base_pages=1
skbuff: skb_segment: input head[0] frag[9] page=ffffea00048883c0 head=ffffea00048883c0 order=0 off=0 size=328 base_pages=1
skbuff: skb_segment: input head[0] frag[10] page=ffffea00048883c0 head=ffffea00048883c0 order=0 off=328 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[11] page=ffffea00048883c0 head=ffffea00048883c0 order=0 off=1776 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[12] page=ffffea00048883c0 head=ffffea00048883c0 order=0 off=3224 size=872 base_pages=1
skbuff: skb_segment: input head[0] frag[13] page=ffffea00041bf840 head=ffffea00041bf840 order=0 off=0 size=576 base_pages=1
skbuff: skb_segment: input head[0] frag[14] page=ffffea00041bf840 head=ffffea00041bf840 order=0 off=576 size=1448 base_pages=1
skbuff: skb_segment: input head[0] frag[15] page=ffffea00041bf840 head=ffffea00041bf840 order=0 off=2024 size=1448 base_pages=1
skbuff: skb_segment: input frag_list[0] skb=ffff888104b7e900 len=11584 data_len=11584 headlen=0 nr_frags=11 frag_list=0000000000000000 head_frag=0 gso_size=0 gso_segs=0 gso_type=0x0
skbuff: skb_segment: input frag_list[0] frag[0] page=ffffea00041bf840 head=ffffea00041bf840 order=0 off=3472 size=624 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[1] page=ffffea000413ae40 head=ffffea000413ae40 order=0 off=0 size=824 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[2] page=ffffea000413ae40 head=ffffea000413ae40 order=0 off=824 size=1448 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[3] page=ffffea000413ae40 head=ffffea000413ae40 order=0 off=2272 size=1448 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[4] page=ffffea000413ae40 head=ffffea000413ae40 order=0 off=3720 size=376 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[5] page=ffffea000421d8c0 head=ffffea000421d8c0 order=0 off=0 size=1072 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[6] page=ffffea000421d8c0 head=ffffea000421d8c0 order=0 off=1072 size=1448 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[7] page=ffffea000421d8c0 head=ffffea000421d8c0 order=0 off=2520 size=1448 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[8] page=ffffea000421d8c0 head=ffffea000421d8c0 order=0 off=3968 size=128 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[9] page=ffffea0004826440 head=ffffea0004826440 order=0 off=0 size=1320 base_pages=1
skbuff: skb_segment: input frag_list[0] frag[10] page=ffffea0004826440 head=ffffea0004826440 order=0 off=1320 size=1448 base_pages=1
skbuff: skb_segment: setup after push head=ffff888101008a00 len=29026 payload=28960 offset=66 mss=28960 partial_segs=20 sg=1 csum=1 list_skb=ffff888104b7e900 features=0x0000010e501d0049
skbuff: skb_segment: segment begin offset=66 len=28960 end=29026 pos=66 i=0 nfrags=16 hsize=0 list_skb=ffff888104b7e900 frag_skb=ffff888101008a00
skbuff: skb_segment: alloc nskb=ffff888101cf5100 hsize=0 doffset=66 headroom=190
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=0 source_skb=ffff888101008a00 source=frag[0] page=ffffea000420f7c0 off=2480 size=1448 pos=66 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=1 pos=66 next_pos=1514 next_i=1
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=1 source_skb=ffff888101008a00 source=frag[1] page=ffffea000420f7c0 off=3928 size=168 pos=1514 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=2 pos=1514 next_pos=1682 next_i=2
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=2 source_skb=ffff888101008a00 source=frag[2] page=ffffea00048261c0 off=0 size=1280 pos=1682 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=3 pos=1682 next_pos=2962 next_i=3
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=3 source_skb=ffff888101008a00 source=frag[3] page=ffffea00048261c0 off=1280 size=1448 pos=2962 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=4 pos=2962 next_pos=4410 next_i=4
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=4 source_skb=ffff888101008a00 source=frag[4] page=ffffea00048261c0 off=2728 size=1368 pos=4410 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=5 pos=4410 next_pos=5778 next_i=5
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=5 source_skb=ffff888101008a00 source=frag[5] page=ffffea0004886400 off=0 size=80 pos=5778 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=6 pos=5778 next_pos=5858 next_i=6
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=6 source_skb=ffff888101008a00 source=frag[6] page=ffffea0004886400 off=80 size=1448 pos=5858 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=7 pos=5858 next_pos=7306 next_i=7
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=7 source_skb=ffff888101008a00 source=frag[7] page=ffffea0004886400 off=1528 size=1448 pos=7306 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=8 pos=7306 next_pos=8754 next_i=8
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=8 source_skb=ffff888101008a00 source=frag[8] page=ffffea0004886400 off=2976 size=1120 pos=8754 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=9 pos=8754 next_pos=9874 next_i=9
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=9 source_skb=ffff888101008a00 source=frag[9] page=ffffea00048883c0 off=0 size=328 pos=9874 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=10 pos=9874 next_pos=10202 next_i=10
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=10 source_skb=ffff888101008a00 source=frag[10] page=ffffea00048883c0 off=328 size=1448 pos=10202 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=11 pos=10202 next_pos=11650 next_i=11
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=11 source_skb=ffff888101008a00 source=frag[11] page=ffffea00048883c0 off=1776 size=1448 pos=11650 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=12 pos=11650 next_pos=13098 next_i=12
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=12 source_skb=ffff888101008a00 source=frag[12] page=ffffea00048883c0 off=3224 size=872 pos=13098 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=13 pos=13098 next_pos=13970 next_i=13
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=13 source_skb=ffff888101008a00 source=frag[13] page=ffffea00041bf840 off=0 size=576 pos=13970 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=14 pos=13970 next_pos=14546 next_i=14
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=14 source_skb=ffff888101008a00 source=frag[14] page=ffffea00041bf840 off=576 size=1448 pos=14546 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=15 pos=14546 next_pos=15994 next_i=15
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=15 source_skb=ffff888101008a00 source=frag[15] page=ffffea00041bf840 off=2024 size=1448 pos=15994 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=16 pos=15994 next_pos=17442 next_i=16
skbuff: skb_segment: sg source exhausted pos=17442 target_end=29026 old_frag_skb=ffff888101008a00 old_nfrags=16 next_list=ffff888104b7e900
skbuff: skb_segment: sg entered frag_list skb=ffff888104b7e900 len=11584 headlen=0 nr_frags=11 head_frag=0 start_i=0 next_list=0000000000000000
skbuff: skb_segment: sg add nskb=ffff888101cf5100 out_frag=16 source_skb=ffff888104b7e900 source=frag[0] page=ffffea00041bf840 off=3472 size=624 pos=17442 target=[66,29026)
skbuff: skb_segment: sg consumed source nskb=ffff888101cf5100 out_nr_frags=17 pos=17442 next_pos=18066 next_i=1
skbuff: skb_segment: sg output full nskb=ffff888101cf5100 nr_frags=17 max=17 pos=18066 target_end=29026 source_i=1 source_nfrags=11 source_skb=ffff888104b7e900
skbuff: skb_segment: too many frags: 18066 28960
skbuff: skb_segment: error err=-22 segs=ffff888101cf5100 tail=0000000019d2a099 offset=66 len=28960 pos=18066 i=1 nfrags=11 list_skb=0000000000000000 frag_skb=ffff888104b7e900

Here goes the reproducer patch (applies on top of net-next):

--cut--

net/core/skbuff.c +177 -0
index 18dabb4e9cfa..f23aff92f857 100644
@@ -65,6 +65,7 @@
 #include <linux/kcov.h>
 #include <linux/iov_iter.h>
 #include <linux/crc32.h>
+#include <linux/ratelimit.h>
 
 #include <net/protocol.h>
 #include <net/dst.h>
@@ -4759,6 +4760,77 @@ struct sk_buff *skb_segment_list(struct sk_buff *skb,
 }
 EXPORT_SYMBOL_GPL(skb_segment_list);
 
+static DEFINE_RATELIMIT_STATE(skb_segment_trace_ratelimit, 5 * HZ, 1);
+
+//#define SKB_SEGMENT_TRACE_THRESHOLD GSO_LEGACY_MAX_SIZE
+#define SKB_SEGMENT_TRACE_THRESHOLD 20000
+
+static bool skb_segment_trace(const struct sk_buff *skb,
+			      unsigned int mss)
+{
+	if (!mss || mss == GSO_BY_FRAGS)
+		return false;
+
+	if (mss <= SKB_SEGMENT_TRACE_THRESHOLD && skb->len <= SKB_SEGMENT_TRACE_THRESHOLD)
+		return false;
+
+	return __ratelimit(&skb_segment_trace_ratelimit);
+}
+
+static void skb_segment_dbg_dump_skb(const char *stage, const char *role,
+				     unsigned int idx,
+				     const struct sk_buff *skb)
+{
+	struct skb_shared_info *shinfo = skb_shinfo(skb);
+	unsigned int nr_frags = shinfo->nr_frags;
+	unsigned int i;
+
+	pr_info("skb_segment: %s %s[%u] skb=%px len=%u data_len=%u headlen=%u nr_frags=%u frag_list=%px head_frag=%u gso_size=%u gso_segs=%u gso_type=0x%x\n",
+		stage, role, idx, skb, skb->len, skb->data_len,
+		skb_headlen(skb), nr_frags, shinfo->frag_list,
+		skb->head_frag, shinfo->gso_size, shinfo->gso_segs,
+		shinfo->gso_type);
+
+	for (i = 0; i < nr_frags; i++) {
+		const skb_frag_t *frag = &shinfo->frags[i];
+
+		struct page *page = skb_frag_page(frag);
+		struct page *head = compound_head(page);
+		unsigned int off = skb_frag_off(frag);
+		unsigned int size = skb_frag_size(frag);
+		unsigned int base_pages;
+
+		base_pages = ((off + size - 1) >> PAGE_SHIFT) - (off >> PAGE_SHIFT) + 1;
+
+		pr_info("skb_segment: %s %s[%u] frag[%u] page=%px head=%px order=%u off=%u size=%u base_pages=%u\n",
+			stage, role, idx, i, skb_frag_page(frag),
+			head, compound_order(head),
+			skb_frag_off(frag), skb_frag_size(frag),
+			base_pages);
+	}
+}
+
+static void skb_segment_dbg_dump_tree(const char *stage,
+				      const struct sk_buff *skb)
+{
+	struct sk_buff *iter;
+	unsigned int i = 0;
+
+	skb_segment_dbg_dump_skb(stage, "head", 0, skb);
+	skb_walk_frags(skb, iter)
+		skb_segment_dbg_dump_skb(stage, "frag_list", i++, iter);
+}
+
+static void skb_segment_dbg_dump_list(const char *stage,
+				      const struct sk_buff *skb)
+{
+	const struct sk_buff *iter;
+	unsigned int i = 0;
+
+	for (iter = skb; iter; iter = iter->next)
+		skb_segment_dbg_dump_skb(stage, "out", i++, iter);
+}
+
 /**
  *	skb_segment - Perform protocol segmentation on skb.
  *	@head_skb: buffer to segment
@@ -4775,6 +4847,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	struct sk_buff *tail = NULL;
 	struct sk_buff *list_skb = skb_shinfo(head_skb)->frag_list;
 	unsigned int mss = skb_shinfo(head_skb)->gso_size;
+	bool gso_by_frags = mss == GSO_BY_FRAGS;
 	unsigned int doffset = head_skb->data - skb_mac_header(head_skb);
 	unsigned int offset = doffset;
 	unsigned int tnl_hlen = skb_tnl_header_len(head_skb);
@@ -4784,6 +4857,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	struct sk_buff *frag_skb;
 	skb_frag_t *frag;
 	__be16 proto;
+	bool trace = false;
 	bool csum, sg;
 	int err = -ENOMEM;
 	int i = 0;
@@ -4861,6 +4935,18 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 			partial_segs = 0;
 	}
 
+	trace = skb_segment_trace(head_skb, mss);
+	if (trace) {
+		pr_info("skb_segment: start head=%px len=%u doffset=%u mss=%u gso_by_frags=%u features=%pNF\n",
+			head_skb, head_skb->len, doffset, mss, gso_by_frags,
+			&features);
+		skb_segment_dbg_dump_tree("input", head_skb);
+		pr_info("skb_segment: setup after push head=%px len=%u payload=%u offset=%u mss=%u partial_segs=%u sg=%u csum=%u list_skb=%px features=%pNF\n",
+			head_skb, head_skb->len, head_skb->len - offset,
+			offset, mss, partial_segs, sg, csum, list_skb,
+			&features);
+	}
+
 normal:
 	headroom = skb_headroom(head_skb);
 	pos = skb_headlen(head_skb);
@@ -4888,10 +4974,22 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 
 		hsize = skb_headlen(head_skb) - offset;
 
+		if (trace)
+			pr_info("skb_segment: segment begin offset=%u len=%u end=%u pos=%d i=%d nfrags=%d hsize=%d list_skb=%px frag_skb=%px\n",
+				offset, len, offset + len, pos, i, nfrags,
+				hsize, list_skb, frag_skb);
+
 		if (hsize <= 0 && i >= nfrags && skb_headlen(list_skb) &&
 		    (skb_headlen(list_skb) == len || sg)) {
 			BUG_ON(skb_headlen(list_skb) > len);
 
+			if (trace)
+				pr_info("skb_segment: clone frag_list skb=%px headlen=%u len=%u nr_frags=%u pos=%d target_end=%u sg=%u\n",
+					list_skb, skb_headlen(list_skb),
+					list_skb->len,
+					skb_shinfo(list_skb)->nr_frags,
+					pos, offset + len, sg);
+
 			nskb = skb_clone(list_skb, GFP_ATOMIC);
 			if (unlikely(!nskb))
 				goto err;
@@ -4903,9 +5001,22 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 			pos += skb_headlen(list_skb);
 
 			while (pos < offset + len) {
+				if (trace)
+					pr_info("skb_segment: clone walk pos=%d target_end=%u i=%d nfrags=%d frag_skb=%px\n",
+						pos, offset + len, i, nfrags,
+						frag_skb);
+				if (trace && i >= nfrags)
+					pr_info("skb_segment: clone walk would BUG: pos=%d target_end=%u i=%d nfrags=%d list_skb=%px\n",
+						pos, offset + len, i, nfrags,
+						list_skb);
 				BUG_ON(i >= nfrags);
 
 				size = skb_frag_size(frag);
+				if (trace)
+					pr_info("skb_segment: clone walk frag[%d] page=%px off=%u size=%d pos=%d next=%d target_end=%u\n",
+						i, skb_frag_page(frag),
+						skb_frag_off(frag), size, pos,
+						pos + size, offset + len);
 				if (pos + size > offset + len)
 					break;
 
@@ -4916,6 +5027,10 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 
 			list_skb = list_skb->next;
 
+			if (trace)
+				pr_info("skb_segment: clone walk done nskb=%px pos=%d i=%d next_list=%px\n",
+					nskb, pos, i, list_skb);
+
 			if (unlikely(pskb_trim(nskb, len))) {
 				kfree_skb(nskb);
 				goto err;
@@ -4945,6 +5060,10 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 
 			skb_reserve(nskb, headroom);
 			__skb_put(nskb, doffset);
+
+			if (trace)
+				pr_info("skb_segment: alloc nskb=%px hsize=%d doffset=%u headroom=%u\n",
+					nskb, hsize, doffset, headroom);
 		}
 
 		if (segs)
@@ -4997,6 +5116,10 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 
 		while (pos < offset + len) {
 			if (i >= nfrags) {
+				if (trace)
+					pr_info("skb_segment: sg source exhausted pos=%d target_end=%u old_frag_skb=%px old_nfrags=%d next_list=%px\n",
+						pos, offset + len, frag_skb,
+						nfrags, list_skb);
 				if (skb_orphan_frags(list_skb, GFP_ATOMIC) ||
 				    skb_zerocopy_clone(nskb, list_skb,
 						       GFP_ATOMIC))
@@ -5019,11 +5142,24 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 					frag--;
 				}
 
+				if (trace)
+					pr_info("skb_segment: sg entered frag_list skb=%px len=%u headlen=%u nr_frags=%d head_frag=%u start_i=%d next_list=%px\n",
+						frag_skb, frag_skb->len,
+						skb_headlen(frag_skb), nfrags,
+						frag_skb->head_frag, i,
+						list_skb->next);
+
 				list_skb = list_skb->next;
 			}
 
 			if (unlikely(skb_shinfo(nskb)->nr_frags >=
 				     MAX_SKB_FRAGS)) {
+				if (trace)
+					pr_info("skb_segment: sg output full nskb=%px nr_frags=%u max=%u pos=%d target_end=%u source_i=%d source_nfrags=%d source_skb=%px\n",
+						nskb, skb_shinfo(nskb)->nr_frags,
+						MAX_SKB_FRAGS, pos,
+						offset + len, i, nfrags,
+						frag_skb);
 				net_warn_ratelimited(
 					"skb_segment: too many frags: %u %u\n",
 					pos, mss);
@@ -5035,18 +5171,46 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 			__skb_frag_ref(nskb_frag);
 			size = skb_frag_size(nskb_frag);
 
+			if (trace)
+				pr_info("skb_segment: sg add nskb=%px out_frag=%u source_skb=%px source=%s[%d] page=%px off=%u size=%d pos=%d target=[%u,%u)\n",
+					nskb, skb_shinfo(nskb)->nr_frags,
+					frag_skb, i < 0 ? "head_frag" : "frag",
+					i, skb_frag_page(nskb_frag),
+					skb_frag_off(nskb_frag), size, pos,
+					offset, offset + len);
+
 			if (pos < offset) {
+				if (trace)
+					pr_info("skb_segment: sg trim front nskb=%px out_frag=%u trim=%u old_off=%u old_size=%d\n",
+						nskb, skb_shinfo(nskb)->nr_frags,
+						offset - pos,
+						skb_frag_off(nskb_frag), size);
 				skb_frag_off_add(nskb_frag, offset - pos);
 				skb_frag_size_sub(nskb_frag, offset - pos);
+				if (trace)
+					pr_info("skb_segment: sg trim front done nskb=%px out_frag=%u new_off=%u new_size=%u\n",
+						nskb, skb_shinfo(nskb)->nr_frags,
+						skb_frag_off(nskb_frag),
+						skb_frag_size(nskb_frag));
 			}
 
 			skb_shinfo(nskb)->nr_frags++;
 
 			if (pos + size <= offset + len) {
+				if (trace)
+					pr_info("skb_segment: sg consumed source nskb=%px out_nr_frags=%u pos=%d next_pos=%d next_i=%d\n",
+						nskb, skb_shinfo(nskb)->nr_frags,
+						pos, pos + size, i + 1);
 				i++;
 				frag++;
 				pos += size;
 			} else {
+				if (trace)
+					pr_info("skb_segment: sg trim tail nskb=%px out_frag=%u trim=%u pos=%d size=%d target_end=%u\n",
+						nskb,
+						skb_shinfo(nskb)->nr_frags - 1,
+						pos + size - (offset + len),
+						pos, size, offset + len);
 				skb_frag_size_sub(nskb_frag, pos + size - (offset + len));
 				goto skip_fraglist;
 			}
@@ -5059,6 +5223,12 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 		nskb->len += nskb->data_len;
 		nskb->truesize += nskb->data_len;
 
+		if (trace)
+			pr_info("skb_segment: segment built nskb=%px len=%u data_len=%u nr_frags=%u hsize=%d offset=%u seg_len=%u pos=%d i=%d next_offset=%u\n",
+				nskb, nskb->len, nskb->data_len,
+				skb_shinfo(nskb)->nr_frags, hsize, offset,
+				len, pos, i, offset + len);
+
 perform_csum_check:
 		if (!csum) {
 			if (skb_has_shared_frag(nskb) &&
@@ -5106,6 +5276,9 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 			skb_shinfo(tail)->gso_segs = DIV_ROUND_UP(tail->len - doffset, gso_size);
 	}
 
+	if (trace)
+		skb_segment_dbg_dump_list("output", segs);
+
 	/* Following permits correct backpressure, for protocols
 	 * using skb_set_owner_w().
 	 * Idea is to tranfert ownership from head_skb to last segment.
@@ -5118,6 +5291,10 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 	return segs;
 
 err:
+	if (trace)
+		pr_info("skb_segment: error err=%d segs=%px tail=%p offset=%u len=%u pos=%d i=%d nfrags=%d list_skb=%px frag_skb=%px\n",
+			err, segs, tail, offset, len, pos, i, nfrags,
+			list_skb, frag_skb);
 	kfree_skb_list(segs);
 	return ERR_PTR(err);
 }
tools/testing/selftests/net/big_tcp_repro.sh new file +226 -0

226 lines not on this page · raw message

--cut--

Thanks,
Alice

#10 lore raw

PATCH v7 5/11 udp: Support BIG TCP GSO packets where they can occur

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

Wherever a GSO packet can occur, and its length is used to fill the UDP header, use udp_set_len that assigns 0 if the length doesn't fit 16 bits, so that the packet can be properly parsed and segmented later, instead of having truncated length.

Use udp_get_len in udp_validate_len to treat BIG TCP packets as valid.

---
 net/ipv4/fou_core.c                    | 2 +-
 net/ipv6/fou6.c                        | 2 +-
 net/netfilter/ipvs/ip_vs_xmit.c        | 2 +-
 net/netfilter/nf_conntrack_proto_udp.c | 2 +-
 net/psp/psp_main.c                     | 2 +-
 5 files changed, 5 insertions(+), 5 deletions(-)
net/ipv4/fou_core.c +1 -1
index a50740d0f288..aef3ce1dec7a 100644
@@ -1040,7 +1040,7 @@ static void fou_build_udp(struct sk_buff *skb, struct ip_tunnel_encap *e,
 
 	uh->dest = e->dport;
 	uh->source = sport;
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);
 	udp_set_csum(!(e->flags & TUNNEL_ENCAP_FLAG_CSUM), skb,
 		     fl4->saddr, fl4->daddr, skb->len);
 
net/ipv6/fou6.c +1 -1
index 588929409241..4b659ca60ba9 100644
@@ -30,7 +30,7 @@ static void fou6_build_udp(struct sk_buff *skb, struct ip_tunnel_encap *e,
 
 	uh->dest = e->dport;
 	uh->source = sport;
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);
 	udp6_set_csum(!(e->flags & TUNNEL_ENCAP_FLAG_CSUM6), skb,
 		      &fl6->saddr, &fl6->daddr, skb->len);
 
net/netfilter/ipvs/ip_vs_xmit.c +1 -1
index ae3ed2c00ec3..c51ebd83a476 100644
@@ -1100,7 +1100,7 @@ ipvs_gue_encap(struct net *net, struct sk_buff *skb,
 	dport = cp->dest->tun_port;
 	udph->dest = dport;
 	udph->source = sport;
-	udp_set_len_short(udph, skb->len);
+	udp_set_len(udph, skb->len);
 	udph->check = 0;
 
 	*next_protocol = IPPROTO_UDP;
net/netfilter/nf_conntrack_proto_udp.c +1 -1
index 8a5675983e7c..a7edfefd6fd1 100644
@@ -45,7 +45,7 @@ static bool udp_validate_len(struct sk_buff *skb,
 			     const struct udphdr *hdr,
 			     unsigned int dataoff)
 {
-	unsigned int udplen = udp_get_len_short(hdr);
+	unsigned int udplen = udp_get_len(skb, hdr, dataoff);
 	unsigned int skblen = skb->len - dataoff;
 
 	if (udplen > skblen || udplen < sizeof(*hdr))
net/psp/psp_main.c +1 -1
index ce7a0ade450f..0475e680651d 100644
@@ -211,7 +211,7 @@ static void psp_write_headers(struct net *net, struct sk_buff *skb, __be32 spi,
 		uh->source = udp_flow_src_port(net, skb, 0, 0, false);
 	}
 	uh->check = 0;
-	udp_set_len_short(uh, udp_len);
+	udp_set_len(uh, udp_len);
 
 	psph->nexthdr = IPPROTO_TCP;
 	psph->hdrlen = PSP_HDRLEN_NOOPT;
Signature · 1 line
2.54.0
#11 lore raw

PATCH v7 6/11 udp: Support gro_ipv4_max_size > 65536

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

Currently, gro_max_size and gro_ipv4_max_size can be set to values bigger than 65536, and GRO will happily aggregate UDP to the configured size (for example, with TCP traffic in VXLAN tunnels). However, udp_gro_complete uses the 16-bit length field in the UDP header to store the length of the aggregated packet. It leads to the packet truncation later in udp_rcv.

Fix this by storing 0 to the UDP length field and by restoring the real length from skb->len in udp_rcv.

Additionally, restrict handling uh->len=0 in udpv6_rcv to BIG TCP and jumbograms only by using the udp_get_len helper.

---
 net/ipv4/udp.c         | 4 ++--
 net/ipv4/udp_offload.c | 4 ++--
 net/ipv6/udp.c         | 4 ++--
 net/ipv6/udp_offload.c | 2 +-
 4 files changed, 7 insertions(+), 7 deletions(-)
net/ipv4/udp.c +2 -2
index 240a1adab278..3713ae77606f 100644
@@ -2591,8 +2591,8 @@ int udp_rcv(struct sk_buff *skb)
 	struct rtable *rt = skb_rtable(skb);
 	struct net *net = dev_net(skb->dev);
 	struct sock *sk = NULL;
-	unsigned short ulen;
 	__be32 saddr, daddr;
+	unsigned int ulen;
 	struct udphdr *uh;
 	bool refcounted;
 	int drop_reason;
@@ -2606,7 +2606,7 @@ int udp_rcv(struct sk_buff *skb)
 		goto drop;		/* No space for header. */
 
 	uh   = udp_hdr(skb);
-	ulen = ntohs(uh->len);
+	ulen = udp_get_len(skb, uh, 0);
 	saddr = ip_hdr(skb)->saddr;
 	daddr = ip_hdr(skb)->daddr;
 
net/ipv4/udp_offload.c +2 -2
index 493e2b9e16fb..4f9a3922937c 100644
@@ -919,7 +919,7 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff,
 	struct sock *sk;
 	int err;
 
-	udp_set_len_short(uh, newlen);
+	udp_set_len(uh, newlen);
 
 	sk = INDIRECT_CALL_INET(lookup, udp6_lib_lookup_skb,
 				udp4_lib_lookup_skb, skb, uh->source, uh->dest);
@@ -956,7 +956,7 @@ INDIRECT_CALLABLE_SCOPE int udp4_gro_complete(struct sk_buff *skb, int nhoff)
 
 	/* do fraglist only if there is no outer UDP encap (or we already processed it) */
 	if (NAPI_GRO_CB(skb)->is_flist && !NAPI_GRO_CB(skb)->encap_mark) {
-		udp_set_len_short(uh, skb->len - nhoff);
+		udp_set_len(uh, skb->len - nhoff);
 
 		skb_shinfo(skb)->gso_type |= (SKB_GSO_FRAGLIST|SKB_GSO_UDP_L4);
 		skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
net/ipv6/udp.c +2 -2
index 913a19edfc7c..e434d52cc049 100644
@@ -1081,12 +1081,12 @@ INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb)
 	daddr = &ipv6_hdr(skb)->daddr;
 	uh = udp_hdr(skb);
 
-	ulen = ntohs(uh->len);
+	ulen = udp_get_len(skb, uh, 0);
 	if (ulen > skb->len)
 		goto short_packet;
 
 	/* Check for jumbo payload */
-	if (ulen == 0)
+	if (ulen == 0 && inet6_is_jumbogram(skb))
 		ulen = skb->len;
 
 	if (ulen < sizeof(*uh))
net/ipv6/udp_offload.c +1 -1
index c92cf5ee3e6a..7370bcb80332 100644
@@ -171,7 +171,7 @@ int udp6_gro_complete(struct sk_buff *skb, int nhoff)
 
 	/* do fraglist only if there is no outer UDP encap (or we already processed it) */
 	if (NAPI_GRO_CB(skb)->is_flist && !NAPI_GRO_CB(skb)->encap_mark) {
-		udp_set_len_short(uh, skb->len - nhoff);
+		udp_set_len(uh, skb->len - nhoff);
 
 		skb_shinfo(skb)->gso_type |= (SKB_GSO_FRAGLIST|SKB_GSO_UDP_L4);
 		skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
Signature · 1 line
2.54.0
maintainer
#12 lore raw

Re: [PATCH net-next v7 06/11] udp: Support gro_ipv4_max_size > 65536

Quoted 5 lines · Alice Mikityanska

From: Alice Mikityanska <[email protected]>

Currently, gro_max_size and gro_ipv4_max_size can be set to values bigger than 65536, and GRO will happily aggregate UDP to the configured size (for example, with TCP traffic in VXLAN tunnels).

Sashiko gemini says IPv4 will not be happy, suspecting iph_totlen() will not work properly, but that is wrong, because skb_is_gso_tcp() will still be true for the relevant GSO packets.

It would be nice to explicitly mentioning such fact here.

/P

#13 lore raw

PATCH v7 7/11 udp: Validate UDP length in udp_gro_receive

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

In the previous commit we started using uh->len = 0 as a marker of a GRO packet bigger than 65536 bytes. Filter out malformed packets coming from the wire with len=0 at udp_gro_receive to exclude them from GRO.

Note that a similar check was present in udp_gro_receive_segment, but not in the UDP socket gro_receive flow. By adding an early check to udp_gro_receive, the check in udp_gro_receive_segment can be dropped.

---
 net/ipv4/udp_offload.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)
net/ipv4/udp_offload.c +7 -5
index 4f9a3922937c..8f77c8788f6d 100644
@@ -707,12 +707,8 @@ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
 		return NULL;
 	}
 
-	/* Do not deal with padded or malicious packets, sorry ! */
 	ulen = udp_get_len_short(uh);
-	if (ulen <= sizeof(*uh) || ulen != skb_gro_len(skb)) {
-		NAPI_GRO_CB(skb)->flush = 1;
-		return NULL;
-	}
+
 	/* pull encapsulating udp header */
 	skb_gro_pull(skb, sizeof(struct udphdr));
 
@@ -782,8 +778,14 @@ struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
 	struct sk_buff *p;
 	struct udphdr *uh2;
 	unsigned int off = skb_gro_offset(skb);
+	unsigned int ulen;
 	int flush = 1;
 
+	/* Do not deal with padded or malicious packets, sorry! */
+	ulen = udp_get_len_short(uh);
+	if (ulen <= sizeof(*uh) || ulen != skb_gro_len(skb))
+		goto out;
+
 	/* We can do L4 aggregation only if the packet can't land in a tunnel
 	 * otherwise we could corrupt the inner stream. Detecting such packets
 	 * cannot be foolproof and the aggregation might still happen in some
Signature · 1 line
2.54.0
#14 lore raw

PATCH v7 8/11 udp: Set length in UDP header to 0 for big GSO packets

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

skb->len may be bigger than 65535 in UDP-based tunnels that have BIG TCP enabled. If GSO aggregates packets that large, set the length in the UDP header to 0, so that tcpdump can print such packets properly (treating them as RFC 2675 jumbograms). Later in the pipeline, __udp_gso_segment will set uh->len to the size of individual packets.

---
 net/ipv4/udp_tunnel_core.c | 2 +-
 net/ipv6/ip6_udp_tunnel.c  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
net/ipv4/udp_tunnel_core.c +1 -1
index 0fccb38f074d..a128fe85620d 100644
@@ -178,7 +178,7 @@ void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb
 
 	uh->dest = dst_port;
 	uh->source = src_port;
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);
 
 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 
net/ipv6/ip6_udp_tunnel.c +1 -1
index dcff7fb16ff6..32525a051a6f 100644
@@ -93,7 +93,7 @@ void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk,
 	uh->dest = dst_port;
 	uh->source = src_port;
 
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);
 
 	skb_dst_set(skb, dst);
 
Signature · 1 line
2.54.0
maintainer
#15 lore raw

Re: [PATCH net-next v7 08/11] udp: Set length in UDP header to 0 for big GSO packets

net/ipv6/ip6_udp_tunnel.c void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk, · line 96

Quoted 10 lines · Alice Mikityanska
net/ipv6/ip6_udp_tunnel.c +1 -1
index dcff7fb16ff6..32525a051a6f 100644
@@ -93,7 +93,7 @@ void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk,
 	uh->dest = dst_port;
 	uh->source = src_port;
 
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);

Both Sashikos noted the above breaks GSO csum, as the following udp_set_csum() will use skb->len to compute the csum partial.

I think it would be useful to consolidate udp_set_len() and udp_set_csum() in a single helper - say udp_set_csum_len():

void udp_set_csum_len(bool nocheck, struct sk_buff *skb,

		      __be32 saddr, __be32 daddr, int len)
{
	struct udphdr *uh = udp_hdr(skb);

	
	uh->len = htons(len);
	if (nocheck) {
		uh->check = 0;
	} else if (skb_is_gso(skb)) {
		uh->len = len < GRO_LEGACY_MAX_SIZE ? htons(len) : 0;
		uh->check = ~udp_v4_check(len, saddr, daddr, 0);
	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
		uh->check = 0;
		uh->check = udp_v4_check(len, saddr, daddr,
	// ...

so that csum and len are always consistent, the 'set' logic is symmetric with udp_get_len(), and no duplicate checks are needed.

/P

#16 lore raw

Re: [PATCH net-next v7 08/11] udp: Set length in UDP header to 0 for big GSO packets

Quoted 9 lines · Paolo Abeni
Quoted 8 lines · Alice Mikityanska
net/ipv6/ip6_udp_tunnel.c +0 -0
index dcff7fb16ff6..32525a051a6f 100644
@@ -93,7 +93,7 @@ void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk,
 	uh->dest = dst_port;
 	uh->source = src_port;
 
Quoted 5 lines

net/ipv6/ip6_udp_tunnel.c void udp_tunnel6_xmit_skb(struct dst_entry *dst, struct sock *sk, · line 96

Quoted 2 lines
net/ipv6/ip6_udp_tunnel.c +1 -1
-	udp_set_len_short(uh, skb->len);
+	udp_set_len(uh, skb->len);

Both Sashikos noted the above breaks GSO csum, as the following udp_set_csum() will use skb->len to compute the csum partial.

I think the existing convention is to actually use skb->len even when it's bigger than 64k. For example, see [1]. In order to keep this working, I kept using skb->len for the pseudoheader checksum. I think I made some fixups in one of previous iterations, because some places missed using skb->len.

Regarding Sashiko's comment, I've already replied to it when it commented the same thing on v3. These packets don't do to __udp_gso_segment, they go to skb_udp_tunnel_segment, which calls __skb_udp_tunnel_segment and adjusts the checksum with skb->len [1].

Please correct me if I'm missing some cases, but it looks correct to me as is.

[1]: https://elixir.bootlin.com/linux/v7.1.2/source/net/ipv4/udp_offload.c#L192-L202

Quoted 24 lines

I think it would be useful to consolidate udp_set_len() and udp_set_csum() in a single helper - say udp_set_csum_len():

void udp_set_csum_len(bool nocheck, struct sk_buff *skb,

		      __be32 saddr, __be32 daddr, int len)
{
	struct udphdr *uh = udp_hdr(skb);

	
	uh->len = htons(len);
	if (nocheck) {
		uh->check = 0;
	} else if (skb_is_gso(skb)) {
		uh->len = len < GRO_LEGACY_MAX_SIZE ? htons(len) : 0;
		uh->check = ~udp_v4_check(len, saddr, daddr, 0);
	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
		uh->check = 0;
		uh->check = udp_v4_check(len, saddr, daddr,
	// ...

so that csum and len are always consistent, the 'set' logic is symmetric with udp_get_len(), and no duplicate checks are needed.

/P

#17 lore raw

PATCH v7 9/11 vxlan: Enable BIG TCP packets

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

In Cilium we do support BIG TCP, but so far the latter has only been enabled for direct routing use-cases. A lot of users rely on Cilium with vxlan/geneve tunneling though. The underlying kernel infra for tunneling has not been supporting BIG TCP up to this point.

Given we do now, bump tso_max_size for vxlan netdevs up to GSO_MAX_SIZE to allow the admin to use BIG TCP with vxlan tunnels.

BIG TCP on vxlan disabled:

Standard MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    30.00    34440.00

8k MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    262144  32768  32768    30.00    55684.26

BIG TCP on vxlan enabled:

Standard MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    30.00    39564.78

8k MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    262144  32768  32768    30.00    61466.47

When tunnel offloads are not enabled/exposed and we fully need to rely on SW-based segmentation on transmit (e.g. in case of Azure) then the more aggressive batching also has a visible effect. Below example was on the same setup as with above benchmarks but with HW support disabled:

  # ethtool -k enp10s0f0np0 | grep udp
  tx-udp_tnl-segmentation: off
  tx-udp_tnl-csum-segmentation: off
  tx-udp-segmentation: off
  rx-udp_tunnel-port-offload: off
  rx-udp-gro-forwarding: off

Before:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    60.00    21820.82

After:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    60.00    29390.78

Example receive side:

swapper 0 [002] 4712.645070: net:netif_receive_skb: dev=enp10s0f0np0 skbaddr=0xffff8f3b086e0200 len=129542

        ffffffff8cfe3aaa __netif_receive_skb_core.constprop.0+0x6ca ([kernel.kallsyms])
        ffffffff8cfe3aaa __netif_receive_skb_core.constprop.0+0x6ca ([kernel.kallsyms])
        ffffffff8cfe47dd __netif_receive_skb_list_core+0xed ([kernel.kallsyms])
        ffffffff8cfe4e52 netif_receive_skb_list_internal+0x1d2 ([kernel.kallsyms])
        ffffffff8d0210d8 gro_complete.constprop.0+0x108 ([kernel.kallsyms])
        ffffffff8d021724 dev_gro_receive+0x4e4 ([kernel.kallsyms])
        ffffffff8d021a99 gro_receive_skb+0x89 ([kernel.kallsyms])
        ffffffffc06edb71 mlx5e_handle_rx_cqe_mpwrq+0x131 ([kernel.kallsyms])
        ffffffffc06ee38a mlx5e_poll_rx_cq+0x9a ([kernel.kallsyms])
        ffffffffc06ef2c7 mlx5e_napi_poll+0x107 ([kernel.kallsyms])
        ffffffff8cfe586d __napi_poll+0x2d ([kernel.kallsyms])
        ffffffff8cfe5f8d net_rx_action+0x20d ([kernel.kallsyms])
        ffffffff8c35d252 handle_softirqs+0xe2 ([kernel.kallsyms])
        ffffffff8c35d556 __irq_exit_rcu+0xd6 ([kernel.kallsyms])
        ffffffff8c35d81e irq_exit_rcu+0xe ([kernel.kallsyms])
        ffffffff8d2602b8 common_interrupt+0x98 ([kernel.kallsyms])
        ffffffff8c000da7 asm_common_interrupt+0x27 ([kernel.kallsyms])
        ffffffff8d2645c5 cpuidle_enter_state+0xd5 ([kernel.kallsyms])
        ffffffff8cf6358e cpuidle_enter+0x2e ([kernel.kallsyms])
        ffffffff8c3ba932 call_cpuidle+0x22 ([kernel.kallsyms])
        ffffffff8c3bfb5e do_idle+0x1ce ([kernel.kallsyms])
        ffffffff8c3bfd79 cpu_startup_entry+0x29 ([kernel.kallsyms])
        ffffffff8c30a6c2 start_secondary+0x112 ([kernel.kallsyms])
        ffffffff8c2c142d common_startup_64+0x13e ([kernel.kallsyms])

Example transmit side:

swapper 0 [005] 4768.021375: net:net_dev_xmit: dev=enp10s0f0np0 skbaddr=0xffff8af32ebe1200 len=129556 rc=0

        ffffffffa75e19c3 dev_hard_start_xmit+0x173 ([kernel.kallsyms])
        ffffffffa75e19c3 dev_hard_start_xmit+0x173 ([kernel.kallsyms])
        ffffffffa7653823 sch_direct_xmit+0x143 ([kernel.kallsyms])
        ffffffffa75e2780 __dev_queue_xmit+0xc70 ([kernel.kallsyms])
        ffffffffa76a1205 ip_finish_output2+0x265 ([kernel.kallsyms])
        ffffffffa76a1577 __ip_finish_output+0x87 ([kernel.kallsyms])
        ffffffffa76a165b ip_finish_output+0x2b ([kernel.kallsyms])
        ffffffffa76a179e ip_output+0x5e ([kernel.kallsyms])
        ffffffffa76a19d5 ip_local_out+0x35 ([kernel.kallsyms])
        ffffffffa770d0e5 iptunnel_xmit+0x185 ([kernel.kallsyms])
        ffffffffc179634e nf_nat_used_tuple_new.cold+0x1129 ([kernel.kallsyms])
        ffffffffc17a7301 vxlan_xmit_one+0xc21 ([kernel.kallsyms])
        ffffffffc17a80a2 vxlan_xmit+0x4a2 ([kernel.kallsyms])
        ffffffffa75e18af dev_hard_start_xmit+0x5f ([kernel.kallsyms])
        ffffffffa75e1d3f __dev_queue_xmit+0x22f ([kernel.kallsyms])
        ffffffffa76a1205 ip_finish_output2+0x265 ([kernel.kallsyms])
        ffffffffa76a1577 __ip_finish_output+0x87 ([kernel.kallsyms])
        ffffffffa76a165b ip_finish_output+0x2b ([kernel.kallsyms])
        ffffffffa76a179e ip_output+0x5e ([kernel.kallsyms])
        ffffffffa76a1de2 __ip_queue_xmit+0x1b2 ([kernel.kallsyms])
        ffffffffa76a2135 ip_queue_xmit+0x15 ([kernel.kallsyms])
        ffffffffa76c70a2 __tcp_transmit_skb+0x522 ([kernel.kallsyms])
        ffffffffa76c931a tcp_write_xmit+0x65a ([kernel.kallsyms])
        ffffffffa76cb42e tcp_tsq_write+0x5e ([kernel.kallsyms])
        ffffffffa76cb7ef tcp_tasklet_func+0x10f ([kernel.kallsyms])
        ffffffffa695d9f7 tasklet_action_common+0x107 ([kernel.kallsyms])
        ffffffffa695db99 tasklet_action+0x29 ([kernel.kallsyms])
        ffffffffa695d252 handle_softirqs+0xe2 ([kernel.kallsyms])
        ffffffffa695d556 __irq_exit_rcu+0xd6 ([kernel.kallsyms])
        ffffffffa695d81e irq_exit_rcu+0xe ([kernel.kallsyms])
        ffffffffa78602b8 common_interrupt+0x98 ([kernel.kallsyms])
        ffffffffa6600da7 asm_common_interrupt+0x27 ([kernel.kallsyms])
        ffffffffa78645c5 cpuidle_enter_state+0xd5 ([kernel.kallsyms])
        ffffffffa756358e cpuidle_enter+0x2e ([kernel.kallsyms])
        ffffffffa69ba932 call_cpuidle+0x22 ([kernel.kallsyms])
        ffffffffa69bfb5e do_idle+0x1ce ([kernel.kallsyms])
        ffffffffa69bfd79 cpu_startup_entry+0x29 ([kernel.kallsyms])
        ffffffffa690a6c2 start_secondary+0x112 ([kernel.kallsyms])
        ffffffffa68c142d common_startup_64+0x13e ([kernel.kallsyms])
---
 drivers/net/vxlan/vxlan_core.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/net/vxlan/vxlan_core.c +2 -0
index 67c367cc5662..3a32f067e701 100644
@@ -3370,6 +3370,8 @@ static void vxlan_setup(struct net_device *dev)
 	dev->mangleid_features = NETIF_F_GSO_PARTIAL;
 
 	netif_keep_dst(dev);
+	netif_set_tso_max_size(dev, GSO_MAX_SIZE);
+
 	dev->priv_flags |= IFF_NO_QUEUE;
 	dev->change_proto_down = true;
 	dev->lltx = true;
Signature · 1 line
2.54.0
#18 lore raw

PATCH v7 10/11 geneve: Enable BIG TCP packets

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

From: Daniel Borkmann <[email protected]>

In Cilium we do support BIG TCP, but so far the latter has only been enabled for direct routing use-cases. A lot of users rely on Cilium with vxlan/geneve tunneling though. The underlying kernel infra for tunneling has not been supporting BIG TCP up to this point.

Given we do now, bump tso_max_size for geneve netdevs up to GSO_MAX_SIZE to allow the admin to use BIG TCP with geneve tunnels.

BIG TCP on geneve disabled:

Standard MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    30.00    37391.34

8k MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    262144  32768  32768    60.00    58030.19

BIG TCP on geneve enabled:

Standard MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    131072  16384  16384    30.00    40891.57

8k MTU:

    # netperf -H 10.1.0.2 -t TCP_STREAM -l60
    MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.1.0.2 () port 0 AF_INET : demo
    Recv   Send    Send
    Socket Socket  Message  Elapsed
    Size   Size    Size     Time     Throughput
    bytes  bytes   bytes    secs.    10^6bits/sec
    262144  32768  32768    60.00    61458.39

Example receive side:

swapper 0 [008] 3682.509996: net:netif_receive_skb: dev=geneve0 skbaddr=0xffff8f3b0a781800 len=129492

        ffffffff8cfe3aaa __netif_receive_skb_core.constprop.0+0x6ca ([kernel.kallsyms])
        ffffffff8cfe3aaa __netif_receive_skb_core.constprop.0+0x6ca ([kernel.kallsyms])
        ffffffff8cfe47dd __netif_receive_skb_list_core+0xed ([kernel.kallsyms])
        ffffffff8cfe4e52 netif_receive_skb_list_internal+0x1d2 ([kernel.kallsyms])
        ffffffff8cfe573c napi_complete_done+0x7c ([kernel.kallsyms])
        ffffffff8d046c23 gro_cell_poll+0x83 ([kernel.kallsyms])
        ffffffff8cfe586d __napi_poll+0x2d ([kernel.kallsyms])
        ffffffff8cfe5f8d net_rx_action+0x20d ([kernel.kallsyms])
        ffffffff8c35d252 handle_softirqs+0xe2 ([kernel.kallsyms])
        ffffffff8c35d556 __irq_exit_rcu+0xd6 ([kernel.kallsyms])
        ffffffff8c35d81e irq_exit_rcu+0xe ([kernel.kallsyms])
        ffffffff8d2602b8 common_interrupt+0x98 ([kernel.kallsyms])
        ffffffff8c000da7 asm_common_interrupt+0x27 ([kernel.kallsyms])
        ffffffff8d2645c5 cpuidle_enter_state+0xd5 ([kernel.kallsyms])
        ffffffff8cf6358e cpuidle_enter+0x2e ([kernel.kallsyms])
        ffffffff8c3ba932 call_cpuidle+0x22 ([kernel.kallsyms])
        ffffffff8c3bfb5e do_idle+0x1ce ([kernel.kallsyms])
        ffffffff8c3bfd79 cpu_startup_entry+0x29 ([kernel.kallsyms])
        ffffffff8c30a6c2 start_secondary+0x112 ([kernel.kallsyms])
        ffffffff8c2c142d common_startup_64+0x13e ([kernel.kallsyms])

Example transmit side:

swapper 0 [002] 3403.688687: net:net_dev_xmit: dev=enp10s0f0np0 skbaddr=0xffff8af31d104ae8 len=129556 rc=0

        ffffffffa75e19c3 dev_hard_start_xmit+0x173 ([kernel.kallsyms])
        ffffffffa75e19c3 dev_hard_start_xmit+0x173 ([kernel.kallsyms])
        ffffffffa7653823 sch_direct_xmit+0x143 ([kernel.kallsyms])
        ffffffffa75e2780 __dev_queue_xmit+0xc70 ([kernel.kallsyms])
        ffffffffa76a1205 ip_finish_output2+0x265 ([kernel.kallsyms])
        ffffffffa76a1577 __ip_finish_output+0x87 ([kernel.kallsyms])
        ffffffffa76a165b ip_finish_output+0x2b ([kernel.kallsyms])
        ffffffffa76a179e ip_output+0x5e ([kernel.kallsyms])
        ffffffffa76a19d5 ip_local_out+0x35 ([kernel.kallsyms])
        ffffffffa770d0e5 iptunnel_xmit+0x185 ([kernel.kallsyms])
        ffffffffc179634e nf_nat_used_tuple_new.cold+0x1129 ([kernel.kallsyms])
        ffffffffc179d3e0 geneve_xmit+0x920 ([kernel.kallsyms])
        ffffffffa75e18af dev_hard_start_xmit+0x5f ([kernel.kallsyms])
        ffffffffa75e1d3f __dev_queue_xmit+0x22f ([kernel.kallsyms])
        ffffffffa76a1205 ip_finish_output2+0x265 ([kernel.kallsyms])
        ffffffffa76a1577 __ip_finish_output+0x87 ([kernel.kallsyms])
        ffffffffa76a165b ip_finish_output+0x2b ([kernel.kallsyms])
        ffffffffa76a179e ip_output+0x5e ([kernel.kallsyms])
        ffffffffa76a1de2 __ip_queue_xmit+0x1b2 ([kernel.kallsyms])
        ffffffffa76a2135 ip_queue_xmit+0x15 ([kernel.kallsyms])
        ffffffffa76c70a2 __tcp_transmit_skb+0x522 ([kernel.kallsyms])
        ffffffffa76c931a tcp_write_xmit+0x65a ([kernel.kallsyms])
        ffffffffa76ca3b9 __tcp_push_pending_frames+0x39 ([kernel.kallsyms])
        ffffffffa76c1fb6 tcp_rcv_established+0x276 ([kernel.kallsyms])
        ffffffffa76d3957 tcp_v4_do_rcv+0x157 ([kernel.kallsyms])
        ffffffffa76d6053 tcp_v4_rcv+0x1243 ([kernel.kallsyms])
        ffffffffa769b8ea ip_protocol_deliver_rcu+0x2a ([kernel.kallsyms])
        ffffffffa769bab7 ip_local_deliver_finish+0x77 ([kernel.kallsyms])
        ffffffffa769bb4d ip_local_deliver+0x6d ([kernel.kallsyms])
        ffffffffa769abe7 ip_sublist_rcv_finish+0x37 ([kernel.kallsyms])
        ffffffffa769b713 ip_sublist_rcv+0x173 ([kernel.kallsyms])
        ffffffffa769bde2 ip_list_rcv+0x102 ([kernel.kallsyms])
        ffffffffa75e4868 __netif_receive_skb_list_core+0x178 ([kernel.kallsyms])
        ffffffffa75e4e52 netif_receive_skb_list_internal+0x1d2 ([kernel.kallsyms])
        ffffffffa75e573c napi_complete_done+0x7c ([kernel.kallsyms])
        ffffffffa7646c23 gro_cell_poll+0x83 ([kernel.kallsyms])
        ffffffffa75e586d __napi_poll+0x2d ([kernel.kallsyms])
        ffffffffa75e5f8d net_rx_action+0x20d ([kernel.kallsyms])
        ffffffffa695d252 handle_softirqs+0xe2 ([kernel.kallsyms])
        ffffffffa695d556 __irq_exit_rcu+0xd6 ([kernel.kallsyms])
        ffffffffa695d81e irq_exit_rcu+0xe ([kernel.kallsyms])
        ffffffffa78602b8 common_interrupt+0x98 ([kernel.kallsyms])
        ffffffffa6600da7 asm_common_interrupt+0x27 ([kernel.kallsyms])
        ffffffffa78645c5 cpuidle_enter_state+0xd5 ([kernel.kallsyms])
        ffffffffa756358e cpuidle_enter+0x2e ([kernel.kallsyms])
        ffffffffa69ba932 call_cpuidle+0x22 ([kernel.kallsyms])
        ffffffffa69bfb5e do_idle+0x1ce ([kernel.kallsyms])
        ffffffffa69bfd79 cpu_startup_entry+0x29 ([kernel.kallsyms])
        ffffffffa690a6c2 start_secondary+0x112 ([kernel.kallsyms])
        ffffffffa68c142d common_startup_64+0x13e ([kernel.kallsyms])
---
 drivers/net/geneve.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/net/geneve.c +2 -0
index f5ea426c7deb..52c5453ac7d8 100644
@@ -1731,6 +1731,8 @@ static void geneve_setup(struct net_device *dev)
 	dev->max_mtu = IP_MAX_MTU - GENEVE_BASE_HLEN - dev->hard_header_len;
 
 	netif_keep_dst(dev);
+	netif_set_tso_max_size(dev, GSO_MAX_SIZE);
+
 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
 	dev->lltx = true;
Signature · 1 line
2.54.0
#19 lore raw

PATCH v7 11/11 selftests: net: Add a test for BIG TCP in UDP tunnels

In reply to Alice Mikityanska

From: Alice Mikityanska <[email protected]>

The test sets up VXLAN and GENEVE tunnels over IPv4 and IPv6 and runs IPv4 and IPv6 traffic through them with BIG TCP enabled. It checks that a non-negligible amount of big aggregated packets are seen in tcpdump.

---
 tools/testing/selftests/net/Makefile          |   1 +
 .../testing/selftests/net/big_tcp_tunnels.sh  | 151 ++++++++++++++++++
 2 files changed, 152 insertions(+)
 create mode 100755 tools/testing/selftests/net/big_tcp_tunnels.sh
tools/testing/selftests/net/Makefile +1 -0
index 6a190a525a39..9b97aedd8058 100644
@@ -13,6 +13,7 @@ TEST_PROGS := \
 	arp_ndisc_untracked_subnets.sh \
 	bareudp.sh \
 	big_tcp.sh \
+	big_tcp_tunnels.sh \
 	bind_bhash.sh \
 	bpf_offload.py \
 	bridge_stp_mode.sh \
tools/testing/selftests/net/big_tcp_tunnels.sh new file +151 -0

151 lines not on this page · raw message

Signature · 1 line
2.54.0
maintainer
#20 lore raw

Re: [PATCH net-next v7 11/11] selftests: net: Add a test for BIG TCP in UDP tunnels

tools/testing/selftests/net/big_tcp_tunnels.sh line 151

Quoted 11 lines · Alice Mikityanska
tools/testing/selftests/net/big_tcp_tunnels.sh new file +11 -0
+trap cleanup EXIT
+setup
+for tunnel in vxlan geneve; do
+	for tun_family in 4 6; do
+		for traffic_family in 4 6; do
+			setup_tunnel "$tunnel" "$tun_family" || exit "$?"
+			do_test "$traffic_family" || exit "$?"
+			cleanup_tunnel
+		done
+	done
+done

Sashiko noted that the above does not validate the GRO/GSO csum code path. It could be useful to add another variant, with veth csum offload disabled.

/P

#21 lore raw
In reply to Alice Mikityanska
This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:

Quoted 8 lines · On Thu, 11 Jun 2026 21:29:44 +0200 you wrote:

From: Alice Mikityanska <[email protected]>

This series is a follow-up to "BIG TCP without HBH in IPv6", and it adds support for BIG TCP IPv4/IPv6 workloads in vxlan and geneve. Now that IPv6 BIG TCP doesn't require stripping the HBH in all various combinations in tunneled traffic, adding BIG TCP becomes feasible.

[...]

Here is the summary with links:
- [net-next,v7,01/11] net/sched: act_csum: don't mangle UDP tunnel GSO packets

    https://git.kernel.org/netdev/net-next/c/9bcb30b389ec
  - [net-next,v7,02/11] geneve: Fix off-by-one comparing with GRO_LEGACY_MAX_SIZE
    https://git.kernel.org/netdev/net-next/c/2319688890d9

- [net-next,v7,03/11] net: Use helpers to get/set UDP len tree-wide (no matching commit)
- [net-next,v7,04/11] net: Enable BIG TCP with partial GSO
(no matching commit)
- [net-next,v7,05/11] udp: Support BIG TCP GSO packets where they can occur (no matching commit)
- [net-next,v7,06/11] udp: Support gro_ipv4_max_size > 65536 (no matching commit)
- [net-next,v7,07/11] udp: Validate UDP length in udp_gro_receive (no matching commit)
- [net-next,v7,08/11] udp: Set length in UDP header to 0 for big GSO packets (no matching commit)
- [net-next,v7,09/11] vxlan: Enable BIG TCP packets
(no matching commit)
- [net-next,v7,10/11] geneve: Enable BIG TCP packets
(no matching commit)
- [net-next,v7,11/11] selftests: net: Add a test for BIG TCP in UDP tunnels (no matching commit)

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html