[PATCH v3 net] udp: gso: do not drop small packets when PMTU reduces

[Yan Zhai <yan@cloudflare.com>]

Commit 4094871db1d6 ("udp: only do GSO if # of segs > 1") avoided GSO
for small packets. But the kernel currently dismisses GSO requests only
after checking MTU/PMTU on gso_size. This means any packets, regardless
of their payload sizes, could be dropped when PMTU becomes smaller than
requested gso_size. We encountered this issue in production and it
caused a reliability problem that new QUIC connection cannot be
established before PMTU cache expired, while non GSO sockets still
worked fine at the same time.

Ideally, do not check any GSO related constraints when payload size is
smaller than requested gso_size, and return EMSGSIZE instead of EINVAL
on MTU/PMTU check failure to be more specific on the error cause.

Fixes: 4094871db1d6 ("udp: only do GSO if # of segs > 1")
Signed-off-by: Yan Zhai <yan@cloudflare.com>
Suggested-by: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
---
v2->v3: simplify the code; adding two test cases
v1->v2: add a missing MTU check when fall back to no GSO mode; Fixed up
commit message to be more precise.

v2: https://lore.kernel.org/netdev/Z5swit7ykNRbJFMS@debian.debian/T/#u
v1: https://lore.kernel.org/all/Z5cgWh%2F6bRQm9vVU@debian.debian/
---
 net/ipv4/udp.c                       |  4 ++--
 net/ipv6/udp.c                       |  4 ++--
 tools/testing/selftests/net/udpgso.c | 26 ++++++++++++++++++++++++++
 3 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index c472c9a57cf6..a9bb9ce5438e 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -1141,9 +1141,9 @@ static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4,
 		const int hlen = skb_network_header_len(skb) +
 				 sizeof(struct udphdr);
 
-		if (hlen + cork->gso_size > cork->fragsize) {
+		if (hlen + min(datalen, cork->gso_size) > cork->fragsize) {
 			kfree_skb(skb);
-			return -EINVAL;
+			return -EMSGSIZE;
 		}
 		if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) {
 			kfree_skb(skb);
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index 6671daa67f4f..c6ea438b5c75 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -1389,9 +1389,9 @@ static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6,
 		const int hlen = skb_network_header_len(skb) +
 				 sizeof(struct udphdr);
 
-		if (hlen + cork->gso_size > cork->fragsize) {
+		if (hlen + min(datalen, cork->gso_size) > cork->fragsize) {
 			kfree_skb(skb);
-			return -EINVAL;
+			return -EMSGSIZE;
 		}
 		if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) {
 			kfree_skb(skb);
diff --git a/tools/testing/selftests/net/udpgso.c b/tools/testing/selftests/net/udpgso.c
index 3f2fca02fec5..36ff28af4b19 100644
--- a/tools/testing/selftests/net/udpgso.c
+++ b/tools/testing/selftests/net/udpgso.c
@@ -102,6 +102,19 @@ struct testcase testcases_v4[] = {
 		.gso_len = CONST_MSS_V4,
 		.r_num_mss = 1,
 	},
+	{
+		/* datalen <= MSS < gso_len: will fall back to no GSO */
+		.tlen = CONST_MSS_V4,
+		.gso_len = CONST_MSS_V4 + 1,
+		.r_num_mss = 0,
+		.r_len_last = CONST_MSS_V4,
+	},
+	{
+		/* MSS < datalen < gso_len: fail */
+		.tlen = CONST_MSS_V4 + 1,
+		.gso_len = CONST_MSS_V4 + 2,
+		.tfail = true,
+	},
 	{
 		/* send a single MSS + 1B */
 		.tlen = CONST_MSS_V4 + 1,
@@ -205,6 +218,19 @@ struct testcase testcases_v6[] = {
 		.gso_len = CONST_MSS_V6,
 		.r_num_mss = 1,
 	},
+	{
+		/* datalen <= MSS < gso_len: will fall back to no GSO */
+		.tlen = CONST_MSS_V6,
+		.gso_len = CONST_MSS_V6 + 1,
+		.r_num_mss = 0,
+		.r_len_last = CONST_MSS_V6,
+	},
+	{
+		/* MSS < datalen < gso_len: fail */
+		.tlen = CONST_MSS_V6 + 1,
+		.gso_len = CONST_MSS_V6 + 2,
+		.tfail = true
+	},
 	{
 		/* send a single MSS + 1B */
 		.tlen = CONST_MSS_V6 + 1,
-- 
2.30.2