Skip to contentSkip to conversation

bpf tc tunneling

PATCH Superseded Willem de Bruijn 2 participants 18 messages Mar 22, 2019

Subsystems (6)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover bpf tc tunneling

From: Willem de Bruijn <[email protected]>

BPF allows for dynamic tunneling, choosing the tunnel destination and features on-demand. Extend bpf_skb_adjust_room to allow for efficient tunneling at the TC hooks.

Most features are required for large packets with GSO, as these will be modified after this patch.

Patch 1
is a performance optimization, avoiding an unnecessary unclone for the TCP hot path.

Patches 2..6
introduce a regression test. These can be squashed, but the code is arguably more readable when gradually expanding the feature set.

Patch 7
is a performance optimization, avoid copying network headers that are going to be overwritten. This also simplifies the bpf program.

Patch 8
reenables bpf_skb_adjust_room for UDP packets.

Patch 9
configures skb tunneling metadata analogous to tunnel devices.

Patches 10..13
expand the regression test to make use of the new features and enable the GSO testcases.

Changes v1->v2
- move BPF_F_ADJ_ROOM_MASK out of uapi as it can be expanded
- document new flags
- in tests replace netcat -q flag with coreutils timeout:
the -q flag is not supported in all netcat versions

Willem de Bruijn (13):
bpf: in bpf_skb_adjust_room avoid copy in tx fast path selftests/bpf: bpf tunnel encap test
selftests/bpf: expand bpf tunnel test with decap selftests/bpf: expand bpf tunnel test to ipv6 selftests/bpf: extend bpf tunnel test with gre selftests/bpf: extend bpf tunnel test with tso
bpf: add bpf_skb_adjust_room mode BPF_ADJ_ROOM_MAC
bpf: add bpf_skb_adjust_room flag BPF_F_ADJ_ROOM_FIXED_GSO
bpf: add bpf_skb_adjust_room encap flags
bpf: Sync bpf.h to tools
selftests/bpf: convert bpf tunnel test to BPF_ADJ_ROOM_MAC selftests/bpf: convert bpf tunnel test to BPF_F_ADJ_ROOM_FIXED_GSO selftests/bpf: convert bpf tunnel test to encap modes

 include/uapi/linux/bpf.h                      |  32 ++-
 net/core/filter.c                             | 129 +++++++--
 tools/include/uapi/linux/bpf.h                |  32 ++-
 tools/testing/selftests/bpf/Makefile          |   3 +-
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/progs/test_tc_tunnel.c      | 261 ++++++++++++++++++
 tools/testing/selftests/bpf/test_tc_tunnel.sh | 178 ++++++++++++
 7 files changed, 601 insertions(+), 36 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/progs/test_tc_tunnel.c
 create mode 100755 tools/testing/selftests/bpf/test_tc_tunnel.sh
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#1 lore raw

PATCH v2 1/13 bpf: in bpf_skb_adjust_room avoid copy in tx fast path

From: Willem de Bruijn <[email protected]>

bpf_skb_adjust_room calls skb_cow on grow.

This expensive operation can be avoided in the fast path when the only other clone has released the header. This is the common case for TCP, where one headerless clone is kept on the retransmit queue.

It is safe to do so even when touching the gso fields in skb_shinfo. Regular tunnel encap with iptunnel_handle_offloads takes the same optimization.

The tcp stack unclones in the unlikely case that it accesses these fields through headerless clones packets on the retransmit queue (see __tcp_retransmit_skb).

If any other clones are present, e.g., from packet sockets, skb_cow_head returns the same value as skb_cow().

---
 net/core/filter.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/core/filter.c +1 -1
index d2511fe46db3..d21e1acdde29 100644
@@ -2971,7 +2971,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 len_diff)
 	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
 		return -ENOTSUPP;
 
-	ret = skb_cow(skb, len_diff);
+	ret = skb_cow_head(skb, len_diff);
 	if (unlikely(ret < 0))
 		return ret;
 
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#2 lore raw

PATCH v2 2/13 selftests/bpf: bpf tunnel encap test

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Validate basic tunnel encapsulation using ipip.

Set up two namespaces connected by veth. Connect a client and server. Do this with and without bpf encap.

---
 tools/testing/selftests/bpf/Makefile          |  3 +-
 .../selftests/bpf/progs/test_tc_tunnel.c      | 83 +++++++++++++++++++
 tools/testing/selftests/bpf/test_tc_tunnel.sh | 75 +++++++++++++++++
 3 files changed, 160 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/bpf/progs/test_tc_tunnel.c
 create mode 100755 tools/testing/selftests/bpf/test_tc_tunnel.sh
tools/testing/selftests/bpf/Makefile +2 -1
index edd59707cb1f..cdcc54ddf4b9 100644
@@ -52,7 +52,8 @@ TEST_PROGS := test_kmod.sh \
 	test_flow_dissector.sh \
 	test_xdp_vlan.sh \
 	test_lwt_ip_encap.sh \
-	test_tcp_check_syncookie.sh
+	test_tcp_check_syncookie.sh \
+	test_tc_tunnel.sh
 
 TEST_PROGS_EXTENDED := with_addr.sh \
 	with_tunnels.sh \
tools/testing/selftests/bpf/progs/test_tc_tunnel.c new file +83 -0

83 lines not on this page · raw message

tools/testing/selftests/bpf/test_tc_tunnel.sh new file +75 -0

75 lines not on this page · raw message

Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#3 lore raw

PATCH v2 3/13 selftests/bpf: expand bpf tunnel test with decap

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

The bpf tunnel test encapsulates using bpf, then decapsulates using a standard tunnel device to verify correctness.

Once encap is verified, also test decap, by replacing the tunnel device on decap with another bpf program.

---
 .../selftests/bpf/progs/test_tc_tunnel.c      | 31 +++++++++++++++++++
 tools/testing/selftests/bpf/test_tc_tunnel.sh |  9 ++++++
 2 files changed, 40 insertions(+)
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +31 -0
index 8223e4347be8..25db148635ab 100644
@@ -80,4 +80,35 @@ int encap_f(struct __sk_buff *skb)
 	return TC_ACT_OK;
 }
 
+SEC("decap")
+int decap_f(struct __sk_buff *skb)
+{
+	struct iphdr iph_outer, iph_inner;
+
+	if (skb->protocol != __bpf_constant_htons(ETH_P_IP))
+		return TC_ACT_OK;
+
+	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_outer,
+			       sizeof(iph_outer)) < 0)
+		return TC_ACT_OK;
+
+	if (iph_outer.ihl != 5 || iph_outer.protocol != IPPROTO_IPIP)
+		return TC_ACT_OK;
+
+	if (bpf_skb_load_bytes(skb, ETH_HLEN + sizeof(iph_outer),
+			       &iph_inner, sizeof(iph_inner)) < 0)
+		return TC_ACT_OK;
+
+	if (bpf_skb_adjust_room(skb, -(int)sizeof(iph_outer),
+				BPF_ADJ_ROOM_NET, 0))
+		return TC_ACT_SHOT;
+
+	/* bpf_skb_adjust_room has moved outer over inner header: restore */
+	if (bpf_skb_store_bytes(skb, ETH_HLEN, &iph_inner, sizeof(iph_inner),
+				BPF_F_INVALIDATE_HASH) < 0)
+		return TC_ACT_SHOT;
+
+	return TC_ACT_OK;
+}
+
 char __license[] SEC("license") = "GPL";
tools/testing/selftests/bpf/test_tc_tunnel.sh +9 -0
index 6ebb288a3afc..91151d91e5a1 100755
@@ -72,4 +72,13 @@ ip netns exec "${ns2}" ip link set dev testtun0 up
 echo "test bpf encap with tunnel device decap"
 client_connect
 
+# serverside, use BPF for decap
+ip netns exec "${ns2}" ip link del dev testtun0
+ip netns exec "${ns2}" tc qdisc add dev veth2 clsact
+ip netns exec "${ns2}" tc filter add dev veth2 ingress \
+	bpf direct-action object-file ./test_tc_tunnel.o section decap
+server_listen
+echo "test bpf encap with bpf decap"
+client_connect
+
 echo OK
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#4 lore raw

PATCH v2 4/13 selftests/bpf: expand bpf tunnel test to ipv6

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

The test only uses ipv4 so far, expand to ipv6. This is mostly a boilerplate near copy of the ipv4 path.

---
 tools/testing/selftests/bpf/config            |   2 +
 .../selftests/bpf/progs/test_tc_tunnel.c      | 116 +++++++++++++++---
 tools/testing/selftests/bpf/test_tc_tunnel.sh |  53 +++++++-
 3 files changed, 149 insertions(+), 22 deletions(-)
tools/testing/selftests/bpf/config +2 -0
index 37f947ec44ed..a42f4fc4dc11 100644
@@ -23,3 +23,5 @@ CONFIG_LWTUNNEL=y
 CONFIG_BPF_STREAM_PARSER=y
 CONFIG_XDP_SOCKETS=y
 CONFIG_FTRACE_SYSCALLS=y
+CONFIG_IPV6_TUNNEL=y
+CONFIG_IPV6_GRE=y
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +98 -18

154 lines not on this page · raw message

tools/testing/selftests/bpf/test_tc_tunnel.sh +49 -4

85 lines not on this page · raw message

Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#5 lore raw

PATCH v2 5/13 selftests/bpf: extend bpf tunnel test with gre

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

GRE is a commonly used protocol. Add GRE cases for both IPv4 and IPv6.

It also inserts different sized headers, which can expose some unexpected edge cases.

---
 .../selftests/bpf/progs/test_tc_tunnel.c      | 148 +++++++++++++-----
 tools/testing/selftests/bpf/test_tc_tunnel.sh |  21 ++-
 2 files changed, 123 insertions(+), 46 deletions(-)
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +109 -39
index 591f540ce513..900c5653105f 100644
@@ -2,6 +2,9 @@
 
 /* In-place tunneling */
 
+#include <stdbool.h>
+#include <string.h>
+
 #include <linux/stddef.h>
 #include <linux/bpf.h>
 #include <linux/if_ether.h>
@@ -17,6 +20,18 @@
 
 static const int cfg_port = 8000;
 
+struct grev4hdr {
+	struct iphdr ip;
+	__be16 flags;
+	__be16 protocol;
+} __attribute__((packed));
+
+struct grev6hdr {
+	struct ipv6hdr ip;
+	__be16 flags;
+	__be16 protocol;
+} __attribute__((packed));
+
 static __always_inline void set_ipv4_csum(struct iphdr *iph)
 {
 	__u16 *iph16 = (__u16 *)iph;
@@ -32,10 +47,12 @@ static __always_inline void set_ipv4_csum(struct iphdr *iph)
 	iph->check = ~((csum & 0xffff) + (csum >> 16));
 }
 
-static int encap_ipv4(struct __sk_buff *skb)
+static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 {
-	struct iphdr iph_outer, iph_inner;
+	struct grev4hdr h_outer;
+	struct iphdr iph_inner;
 	struct tcphdr tcph;
+	int olen;
 
 	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
 			       sizeof(iph_inner)) < 0)
@@ -52,24 +69,33 @@ static int encap_ipv4(struct __sk_buff *skb)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
+	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
+
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, sizeof(iph_outer), BPF_ADJ_ROOM_NET, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_NET, 0))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
-	iph_outer = iph_inner;
-	iph_outer.protocol = IPPROTO_IPIP;
-	iph_outer.tot_len = bpf_htons(sizeof(iph_outer) +
-				      bpf_htons(iph_outer.tot_len));
-	set_ipv4_csum(&iph_outer);
+	h_outer.ip = iph_inner;
+	h_outer.ip.tot_len = bpf_htons(olen +
+				      bpf_htons(h_outer.ip.tot_len));
+	if (with_gre) {
+		h_outer.ip.protocol = IPPROTO_GRE;
+		h_outer.protocol = bpf_htons(ETH_P_IP);
+		h_outer.flags = 0;
+	} else {
+		h_outer.ip.protocol = IPPROTO_IPIP;
+	}
+
+	set_ipv4_csum((void *)&h_outer.ip);
 
 	/* store new outer network header */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN, &iph_outer, sizeof(iph_outer),
+	if (bpf_skb_store_bytes(skb, ETH_HLEN, &h_outer, olen,
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
 
 	/* bpf_skb_adjust_room has moved header to start of room: restore */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN + sizeof(iph_outer),
+	if (bpf_skb_store_bytes(skb, ETH_HLEN + olen,
 				&iph_inner, sizeof(iph_inner),
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
@@ -77,10 +103,12 @@ static int encap_ipv4(struct __sk_buff *skb)
 	return TC_ACT_OK;
 }
 
-static int encap_ipv6(struct __sk_buff *skb)
+static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 {
-	struct ipv6hdr iph_outer, iph_inner;
+	struct ipv6hdr iph_inner;
+	struct grev6hdr h_outer;
 	struct tcphdr tcph;
+	int olen;
 
 	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
 			       sizeof(iph_inner)) < 0)
@@ -94,23 +122,31 @@ static int encap_ipv6(struct __sk_buff *skb)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
+	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
+
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, sizeof(iph_outer), BPF_ADJ_ROOM_NET, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_NET, 0))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
-	iph_outer = iph_inner;
-	iph_outer.nexthdr = IPPROTO_IPV6;
-	iph_outer.payload_len = bpf_htons(sizeof(iph_outer) +
-					  bpf_ntohs(iph_outer.payload_len));
+	h_outer.ip = iph_inner;
+	h_outer.ip.payload_len = bpf_htons(olen +
+					   bpf_ntohs(h_outer.ip.payload_len));
+	if (with_gre) {
+		h_outer.ip.nexthdr = IPPROTO_GRE;
+		h_outer.protocol = bpf_htons(ETH_P_IPV6);
+		h_outer.flags = 0;
+	} else {
+		h_outer.ip.nexthdr = IPPROTO_IPV6;
+	}
 
 	/* store new outer network header */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN, &iph_outer, sizeof(iph_outer),
+	if (bpf_skb_store_bytes(skb, ETH_HLEN, &h_outer, olen,
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
 
 	/* bpf_skb_adjust_room has moved header to start of room: restore */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN + sizeof(iph_outer),
+	if (bpf_skb_store_bytes(skb, ETH_HLEN + olen,
 				&iph_inner, sizeof(iph_inner),
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
@@ -118,28 +154,63 @@ static int encap_ipv6(struct __sk_buff *skb)
 	return TC_ACT_OK;
 }
 
-SEC("encap")
-int encap_f(struct __sk_buff *skb)
+SEC("encap_ipip")
+int __encap_ipip(struct __sk_buff *skb)
 {
-	switch (skb->protocol) {
-	case __bpf_constant_htons(ETH_P_IP):
-		return encap_ipv4(skb);
-	case __bpf_constant_htons(ETH_P_IPV6):
-		return encap_ipv6(skb);
-	default:
-		/* does not match, ignore */
+	if (skb->protocol == __bpf_constant_htons(ETH_P_IP))
+		return encap_ipv4(skb, false);
+	else
 		return TC_ACT_OK;
-	}
 }
 
-static int decap_internal(struct __sk_buff *skb, int off, int len)
+SEC("encap_gre")
+int __encap_gre(struct __sk_buff *skb)
 {
-	char buf[sizeof(struct ipv6hdr)];
+	if (skb->protocol == __bpf_constant_htons(ETH_P_IP))
+		return encap_ipv4(skb, true);
+	else
+		return TC_ACT_OK;
+}
 
-	if (bpf_skb_load_bytes(skb, off + len, &buf, len) < 0)
+SEC("encap_ip6tnl")
+int __encap_ip6tnl(struct __sk_buff *skb)
+{
+	if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6))
+		return encap_ipv6(skb, false);
+	else
+		return TC_ACT_OK;
+}
+
+SEC("encap_ip6gre")
+int __encap_ip6gre(struct __sk_buff *skb)
+{
+	if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6))
+		return encap_ipv6(skb, true);
+	else
 		return TC_ACT_OK;
+}
 
-	if (bpf_skb_adjust_room(skb, -len, BPF_ADJ_ROOM_NET, 0))
+static int decap_internal(struct __sk_buff *skb, int off, int len, char proto)
+{
+	char buf[sizeof(struct grev6hdr)];
+	int olen;
+
+	switch (proto) {
+	case IPPROTO_IPIP:
+	case IPPROTO_IPV6:
+		olen = len;
+		break;
+	case IPPROTO_GRE:
+		olen = len + 4 /* gre hdr */;
+		break;
+	default:
+		return TC_ACT_OK;
+	}
+
+	if (bpf_skb_load_bytes(skb, off + olen, &buf, olen) < 0)
+		return TC_ACT_OK;
+
+	if (bpf_skb_adjust_room(skb, -olen, BPF_ADJ_ROOM_NET, 0))
 		return TC_ACT_SHOT;
 
 	/* bpf_skb_adjust_room has moved outer over inner header: restore */
@@ -157,10 +228,11 @@ static int decap_ipv4(struct __sk_buff *skb)
 			       sizeof(iph_outer)) < 0)
 		return TC_ACT_OK;
 
-	if (iph_outer.ihl != 5 || iph_outer.protocol != IPPROTO_IPIP)
+	if (iph_outer.ihl != 5)
 		return TC_ACT_OK;
 
-	return decap_internal(skb, ETH_HLEN, sizeof(iph_outer));
+	return decap_internal(skb, ETH_HLEN, sizeof(iph_outer),
+			      iph_outer.protocol);
 }
 
 static int decap_ipv6(struct __sk_buff *skb)
@@ -171,10 +243,8 @@ static int decap_ipv6(struct __sk_buff *skb)
 			       sizeof(iph_outer)) < 0)
 		return TC_ACT_OK;
 
-	if (iph_outer.nexthdr != IPPROTO_IPV6)
-		return TC_ACT_OK;
-
-	return decap_internal(skb, ETH_HLEN, sizeof(iph_outer));
+	return decap_internal(skb, ETH_HLEN, sizeof(iph_outer),
+			      iph_outer.nexthdr);
 }
 
 SEC("decap")
tools/testing/selftests/bpf/test_tc_tunnel.sh +14 -7
index 7b1758f3006b..c78922048610 100755
@@ -54,30 +54,36 @@ set -e
 # no arguments: automated test, run all
 if [[ "$#" -eq "0" ]]; then
 	echo "ipip"
-	$0 ipv4
+	$0 ipv4 ipip
 
 	echo "ip6ip6"
-	$0 ipv6
+	$0 ipv6 ip6tnl
+
+	echo "ip gre"
+	$0 ipv4 gre
+
+	echo "ip6 gre"
+	$0 ipv6 ip6gre
 
 	echo "OK. All tests passed"
 	exit 0
 fi
 
-if [[ "$#" -ne "1" ]]; then
+if [[ "$#" -ne "2" ]]; then
 	echo "Usage: $0"
-	echo "   or: $0 <ipv4|ipv6>"
+	echo "   or: $0 <ipv4|ipv6> <tuntype>"
 	exit 1
 fi
 
 case "$1" in
 "ipv4")
-	readonly tuntype=ipip
+	readonly tuntype=$2
 	readonly addr1="${ns1_v4}"
 	readonly addr2="${ns2_v4}"
 	readonly netcat_opt=-4
 	;;
 "ipv6")
-	readonly tuntype=ip6tnl
+	readonly tuntype=$2
 	readonly addr1="${ns1_v6}"
 	readonly addr2="${ns2_v6}"
 	readonly netcat_opt=-6
@@ -103,7 +109,8 @@ client_connect
 # client can no longer connect
 ip netns exec "${ns1}" tc qdisc add dev veth1 clsact
 ip netns exec "${ns1}" tc filter add dev veth1 egress \
-	bpf direct-action object-file ./test_tc_tunnel.o section encap
+	bpf direct-action object-file ./test_tc_tunnel.o \
+	section "encap_${tuntype}"
 echo "test bpf encap without decap (expect failure)"
 server_listen
 ! client_connect
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#6 lore raw

PATCH v2 6/13 selftests/bpf: extend bpf tunnel test with tso

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Segmentation offload takes a longer path. Verify that the feature works with large packets.

The test succeeds if not setting dodgy in bpf_skb_adjust_room, as veth TSO is permissive.

If not setting SKB_GSO_DODGY, this enables tunneled TSO offload on supporting NICs.

The feature sets SKB_GSO_DODGY because the caller is untrusted. As a result the packets traverse through the gso stack at least up to TCP. And fail the gso_type validation, such as the skb->encapsulation check in gre_gso_segment and the gso_type checks introduced in commit 418e897e0716 ("gso: validate gso_type on ipip style tunnel").

This will be addressed in a follow-on feature patch. In the meantime, disable the new gso tests.

Changes v1->v2:
- not all netcat versions support flag '-q', use timeout instead

---
 tools/testing/selftests/bpf/test_tc_tunnel.sh | 60 +++++++++++++++----
 1 file changed, 49 insertions(+), 11 deletions(-)
tools/testing/selftests/bpf/test_tc_tunnel.sh +49 -11
index c78922048610..9e18754f2354 100755
@@ -15,6 +15,8 @@ readonly ns2_v4=192.168.1.2
 readonly ns1_v6=fd::1
 readonly ns2_v6=fd::2
 
+readonly infile="$(mktemp)"
+readonly outfile="$(mktemp)"
 
 setup() {
 	ip netns add "${ns1}"
@@ -23,6 +25,8 @@ setup() {
 	ip link add dev veth1 mtu 1500 netns "${ns1}" type veth \
 	      peer name veth2 mtu 1500 netns "${ns2}"
 
+	ip netns exec "${ns1}" ethtool -K veth1 tso off
+
 	ip -netns "${ns1}" link set veth1 up
 	ip -netns "${ns2}" link set veth2 up
 
@@ -32,58 +36,86 @@ setup() {
 	ip -netns "${ns2}" -6 addr add "${ns2_v6}/64" dev veth2 nodad
 
 	sleep 1
+
+	dd if=/dev/urandom of="${infile}" bs="${datalen}" count=1 status=none
 }
 
 cleanup() {
 	ip netns del "${ns2}"
 	ip netns del "${ns1}"
+
+	if [[ -f "${outfile}" ]]; then
+		rm "${outfile}"
+	fi
+	if [[ -f "${infile}" ]]; then
+		rm "${infile}"
+	fi
 }
 
 server_listen() {
-	ip netns exec "${ns2}" nc "${netcat_opt}" -l -p "${port}" &
+	ip netns exec "${ns2}" nc "${netcat_opt}" -l -p "${port}" > "${outfile}" &
+	server_pid=$!
 	sleep 0.2
 }
 
 client_connect() {
-	ip netns exec "${ns1}" nc "${netcat_opt}" -z -w 1 "${addr2}" "${port}"
+	ip netns exec "${ns1}" timeout 2 nc "${netcat_opt}" -w 1 "${addr2}" "${port}" < "${infile}"
 	echo $?
 }
 
+verify_data() {
+	wait "${server_pid}"
+	# sha1sum returns two fields [sha1] [filepath]
+	# convert to bash array and access first elem
+	insum=($(sha1sum ${infile}))
+	outsum=($(sha1sum ${outfile}))
+	if [[ "${insum[0]}" != "${outsum[0]}" ]]; then
+		echo "data mismatch"
+		exit 1
+	fi
+}
+
 set -e
 
 # no arguments: automated test, run all
 if [[ "$#" -eq "0" ]]; then
 	echo "ipip"
-	$0 ipv4 ipip
+	$0 ipv4 ipip 100
 
 	echo "ip6ip6"
-	$0 ipv6 ip6tnl
+	$0 ipv6 ip6tnl 100
 
 	echo "ip gre"
-	$0 ipv4 gre
+	$0 ipv4 gre 100
 
 	echo "ip6 gre"
-	$0 ipv6 ip6gre
+	$0 ipv6 ip6gre 100
+
+	# disabled until passes SKB_GSO_DODGY checks
+	# echo "ip gre gso"
+	# $0 ipv4 gre 2000
+
+	# disabled until passes SKB_GSO_DODGY checks
+	# echo "ip6 gre gso"
+	# $0 ipv6 ip6gre 2000
 
 	echo "OK. All tests passed"
 	exit 0
 fi
 
-if [[ "$#" -ne "2" ]]; then
+if [[ "$#" -ne "3" ]]; then
 	echo "Usage: $0"
-	echo "   or: $0 <ipv4|ipv6> <tuntype>"
+	echo "   or: $0 <ipv4|ipv6> <tuntype> <data_len>"
 	exit 1
 fi
 
 case "$1" in
 "ipv4")
-	readonly tuntype=$2
 	readonly addr1="${ns1_v4}"
 	readonly addr2="${ns2_v4}"
 	readonly netcat_opt=-4
 	;;
 "ipv6")
-	readonly tuntype=$2
 	readonly addr1="${ns1_v6}"
 	readonly addr2="${ns2_v6}"
 	readonly netcat_opt=-6
@@ -94,7 +126,10 @@ case "$1" in
 	;;
 esac
 
-echo "encap ${addr1} to ${addr2}, type ${tuntype}"
+readonly tuntype=$2
+readonly datalen=$3
+
+echo "encap ${addr1} to ${addr2}, type ${tuntype}, len ${datalen}"
 
 trap cleanup EXIT
 
@@ -104,6 +139,7 @@ setup
 echo "test basic connectivity"
 server_listen
 client_connect
+verify_data
 
 # clientside, insert bpf program to encap all TCP to port ${port}
 # client can no longer connect
@@ -123,6 +159,7 @@ ip netns exec "${ns2}" ip link add dev testtun0 type "${tuntype}" \
 ip netns exec "${ns2}" ip link set dev testtun0 up
 echo "test bpf encap with tunnel device decap"
 client_connect
+verify_data
 
 # serverside, use BPF for decap
 ip netns exec "${ns2}" ip link del dev testtun0
@@ -132,5 +169,6 @@ ip netns exec "${ns2}" tc filter add dev veth2 ingress \
 server_listen
 echo "test bpf encap with bpf decap"
 client_connect
+verify_data
 
 echo OK
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#7 lore raw

PATCH v2 7/13 bpf: add bpf_skb_adjust_room mode BPF_ADJ_ROOM_MAC

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

bpf_skb_adjust_room net allows inserting room in an skb.

Existing mode BPF_ADJ_ROOM_NET inserts room after the network header by pulling the skb, moving the network header forward and zeroing the new space.

Add new mode BPF_ADJUST_ROOM_MAC that inserts room after the mac header. This allows inserting tunnel headers in front of the network header without having to recreate the network header in the original space, avoiding two copies.

---
 include/uapi/linux/bpf.h |  6 +++++-
 net/core/filter.c        | 38 ++++++++++++++++++++------------------
 2 files changed, 25 insertions(+), 19 deletions(-)
include/uapi/linux/bpf.h +5 -1
index 3c04410137d9..7c8fd0647070 100644
@@ -1478,7 +1478,10 @@ union bpf_attr {
  * 		Grow or shrink the room for data in the packet associated to
  * 		*skb* by *len_diff*, and according to the selected *mode*.
  *
- * 		There is a single supported mode at this time:
+ *		There are two supported modes at this time:
+ *
+ *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
+ *		  (room space is added or removed below the layer 2 header).
  *
  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  * 		  (room space is added or removed below the layer 3 header).
@@ -2627,6 +2630,7 @@ enum bpf_func_id {
 /* Mode for BPF_FUNC_skb_adjust_room helper. */
 enum bpf_adj_room_mode {
 	BPF_ADJ_ROOM_NET,
+	BPF_ADJ_ROOM_MAC,
 };
 
 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
net/core/filter.c +20 -18
index d21e1acdde29..e7b7720b18e9 100644
@@ -2963,9 +2963,8 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb)
 	}
 }
 
-static int bpf_skb_net_grow(struct sk_buff *skb, u32 len_diff)
+static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff)
 {
-	u32 off = skb_mac_header_len(skb) + bpf_skb_net_base_len(skb);
 	int ret;
 
 	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
@@ -2992,9 +2991,8 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 len_diff)
 	return 0;
 }
 
-static int bpf_skb_net_shrink(struct sk_buff *skb, u32 len_diff)
+static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff)
 {
-	u32 off = skb_mac_header_len(skb) + bpf_skb_net_base_len(skb);
 	int ret;
 
 	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
@@ -3027,7 +3025,8 @@ static u32 __bpf_skb_max_len(const struct sk_buff *skb)
 			  SKB_MAX_ALLOC;
 }
 
-static int bpf_skb_adjust_net(struct sk_buff *skb, s32 len_diff)
+BPF_CALL_4(bpf_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
+	   u32, mode, u64, flags)
 {
 	bool trans_same = skb->transport_header == skb->network_header;
 	u32 len_cur, len_diff_abs = abs(len_diff);
@@ -3035,14 +3034,28 @@ static int bpf_skb_adjust_net(struct sk_buff *skb, s32 len_diff)
 	u32 len_max = __bpf_skb_max_len(skb);
 	__be16 proto = skb->protocol;
 	bool shrink = len_diff < 0;
+	u32 off;
 	int ret;
 
+	if (unlikely(flags))
+		return -EINVAL;
 	if (unlikely(len_diff_abs > 0xfffU))
 		return -EFAULT;
 	if (unlikely(proto != htons(ETH_P_IP) &&
 		     proto != htons(ETH_P_IPV6)))
 		return -ENOTSUPP;
 
+	off = skb_mac_header_len(skb);
+	switch (mode) {
+	case BPF_ADJ_ROOM_NET:
+		off += bpf_skb_net_base_len(skb);
+		break;
+	case BPF_ADJ_ROOM_MAC:
+		break;
+	default:
+		return -ENOTSUPP;
+	}
+
 	len_cur = skb->len - skb_network_offset(skb);
 	if (skb_transport_header_was_set(skb) && !trans_same)
 		len_cur = skb_network_header_len(skb);
@@ -3052,24 +3065,13 @@ static int bpf_skb_adjust_net(struct sk_buff *skb, s32 len_diff)
 			 !skb_is_gso(skb))))
 		return -ENOTSUPP;
 
-	ret = shrink ? bpf_skb_net_shrink(skb, len_diff_abs) :
-		       bpf_skb_net_grow(skb, len_diff_abs);
+	ret = shrink ? bpf_skb_net_shrink(skb, off, len_diff_abs) :
+		       bpf_skb_net_grow(skb, off, len_diff_abs);
 
 	bpf_compute_data_pointers(skb);
 	return ret;
 }
 
-BPF_CALL_4(bpf_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
-	   u32, mode, u64, flags)
-{
-	if (unlikely(flags))
-		return -EINVAL;
-	if (likely(mode == BPF_ADJ_ROOM_NET))
-		return bpf_skb_adjust_net(skb, len_diff);
-
-	return -ENOTSUPP;
-}
-
 static const struct bpf_func_proto bpf_skb_adjust_room_proto = {
 	.func		= bpf_skb_adjust_room,
 	.gpl_only	= false,
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#8 lore raw

PATCH v2 8/13 bpf: add bpf_skb_adjust_room flag BPF_F_ADJ_ROOM_FIXED_GSO

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

bpf_skb_adjust_room adjusts gso_size of gso packets to account for the pushed or popped header room.

This is not allowed with UDP, where gso_size delineates datagrams. Add an option to avoid these updates and allow this call for datagrams.

It can also be used with TCP, when MSS is known to allow headroom, e.g., through MSS clamping or route MTU.

Changes v1->v2:
- document flag BPF_F_ADJ_ROOM_FIXED_GSO
- do not expose BPF_F_ADJ_ROOM_MASK through uapi, as it may change.

---
 include/uapi/linux/bpf.h |  9 +++++++--
 net/core/filter.c        | 38 +++++++++++++++++++++++++++-----------
 2 files changed, 34 insertions(+), 13 deletions(-)
include/uapi/linux/bpf.h +7 -2
index 7c8fd0647070..4f157d0ec571 100644
@@ -1486,8 +1486,10 @@ union bpf_attr {
  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  * 		  (room space is added or removed below the layer 3 header).
  *
- * 		All values for *flags* are reserved for future usage, and must
- * 		be left at zero.
+ *		There is one supported flag at this time:
+ *
+ *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
+ *		  Adjusting mss in this way is not allowed for datagrams.
  *
  * 		A call to this helper is susceptible to change the underlaying
  * 		packet buffer. Therefore, at load time, all checks on pointers
@@ -2627,6 +2629,9 @@ enum bpf_func_id {
 /* Current network namespace */
 #define BPF_F_CURRENT_NETNS		(-1L)
 
+/* BPF_FUNC_skb_adjust_room flags. */
+#define BPF_F_ADJ_ROOM_FIXED_GSO	(1ULL << 0)
+
 /* Mode for BPF_FUNC_skb_adjust_room helper. */
 enum bpf_adj_room_mode {
 	BPF_ADJ_ROOM_NET,
net/core/filter.c +27 -11
index e7b7720b18e9..393d1e4903b5 100644
@@ -2963,12 +2963,17 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb)
 	}
 }
 
-static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff)
+static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff,
+			    u64 flags)
 {
 	int ret;
 
-	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
-		return -ENOTSUPP;
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb)) {
+		/* udp gso_size delineates datagrams, only allow if fixed */
+		if (!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ||
+		    !(flags & BPF_F_ADJ_ROOM_FIXED_GSO))
+			return -ENOTSUPP;
+	}
 
 	ret = skb_cow_head(skb, len_diff);
 	if (unlikely(ret < 0))
@@ -2982,7 +2987,9 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff)
 		struct skb_shared_info *shinfo = skb_shinfo(skb);
 
 		/* Due to header grow, MSS needs to be downgraded. */
-		skb_decrease_gso_size(shinfo, len_diff);
+		if (!(flags & BPF_F_ADJ_ROOM_FIXED_GSO))
+			skb_decrease_gso_size(shinfo, len_diff);
+
 		/* Header must be checked, and gso_segs recomputed. */
 		shinfo->gso_type |= SKB_GSO_DODGY;
 		shinfo->gso_segs = 0;
@@ -2991,12 +2998,17 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff)
 	return 0;
 }
 
-static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff)
+static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff,
+			      u64 flags)
 {
 	int ret;
 
-	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb))
-		return -ENOTSUPP;
+	if (skb_is_gso(skb) && !skb_is_gso_tcp(skb)) {
+		/* udp gso_size delineates datagrams, only allow if fixed */
+		if (!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ||
+		    !(flags & BPF_F_ADJ_ROOM_FIXED_GSO))
+			return -ENOTSUPP;
+	}
 
 	ret = skb_unclone(skb, GFP_ATOMIC);
 	if (unlikely(ret < 0))
@@ -3010,7 +3022,9 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff)
 		struct skb_shared_info *shinfo = skb_shinfo(skb);
 
 		/* Due to header shrink, MSS can be upgraded. */
-		skb_increase_gso_size(shinfo, len_diff);
+		if (!(flags & BPF_F_ADJ_ROOM_FIXED_GSO))
+			skb_increase_gso_size(shinfo, len_diff);
+
 		/* Header must be checked, and gso_segs recomputed. */
 		shinfo->gso_type |= SKB_GSO_DODGY;
 		shinfo->gso_segs = 0;
@@ -3025,6 +3039,8 @@ static u32 __bpf_skb_max_len(const struct sk_buff *skb)
 			  SKB_MAX_ALLOC;
 }
 
+#define BPF_F_ADJ_ROOM_MASK		(BPF_F_ADJ_ROOM_FIXED_GSO)
+
 BPF_CALL_4(bpf_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
 	   u32, mode, u64, flags)
 {
@@ -3037,7 +3053,7 @@ BPF_CALL_4(bpf_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
 	u32 off;
 	int ret;
 
-	if (unlikely(flags))
+	if (unlikely(flags & ~BPF_F_ADJ_ROOM_MASK))
 		return -EINVAL;
 	if (unlikely(len_diff_abs > 0xfffU))
 		return -EFAULT;
@@ -3065,8 +3081,8 @@ BPF_CALL_4(bpf_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
 			 !skb_is_gso(skb))))
 		return -ENOTSUPP;
 
-	ret = shrink ? bpf_skb_net_shrink(skb, off, len_diff_abs) :
-		       bpf_skb_net_grow(skb, off, len_diff_abs);
+	ret = shrink ? bpf_skb_net_shrink(skb, off, len_diff_abs, flags) :
+		       bpf_skb_net_grow(skb, off, len_diff_abs, flags);
 
 	bpf_compute_data_pointers(skb);
 	return ret;
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#9 lore raw

PATCH v2 9/13 bpf: add bpf_skb_adjust_room encap flags

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

When pushing tunnel headers, annotate skbs in the same way as tunnel devices.

For GSO packets, the network stack requires certain fields set to segment packets with tunnel headers. gro_gse_segment depends on transport and inner mac header, for instance.

Add an option to pass this information.

Remove the restriction on len_diff to network header length, which is too short, e.g., for GRE protocols.

Changes v1->v2:
- document new flags
- BPF_F_ADJ_ROOM_MASK moved

---
 include/uapi/linux/bpf.h | 19 +++++++++++-
 net/core/filter.c        | 63 ++++++++++++++++++++++++++++++++++++----
 2 files changed, 76 insertions(+), 6 deletions(-)
include/uapi/linux/bpf.h +18 -1
index 4f157d0ec571..f770f0de5b9c 100644
@@ -1486,11 +1486,20 @@ union bpf_attr {
  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  * 		  (room space is added or removed below the layer 3 header).
  *
- *		There is one supported flag at this time:
+ *		The following flags are supported at this time:
  *
  *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
  *		  Adjusting mss in this way is not allowed for datagrams.
  *
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **:
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **:
+ *		  Any new space is reserved to hold a tunnel header.
+ *		  Configure skb offsets and other fields accordingly.
+ *
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **:
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
+ *		  Use with ENCAP_L3 flags to further specify the tunnel type.
+ *
  * 		A call to this helper is susceptible to change the underlaying
  * 		packet buffer. Therefore, at load time, all checks on pointers
  * 		previously done by the verifier are invalidated and must be
@@ -2632,6 +2641,14 @@ enum bpf_func_id {
 /* BPF_FUNC_skb_adjust_room flags. */
 #define BPF_F_ADJ_ROOM_FIXED_GSO	(1ULL << 0)
 
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	(1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	(1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK	(BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+					 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)
+
+#define BPF_F_ADJ_ROOM_ENCAP_L4_GRE	(1ULL << 3)
+#define BPF_F_ADJ_ROOM_ENCAP_L4_UDP	(1ULL << 4)
+
 /* Mode for BPF_FUNC_skb_adjust_room helper. */
 enum bpf_adj_room_mode {
 	BPF_ADJ_ROOM_NET,
net/core/filter.c +58 -5

101 lines not on this page · raw message

Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#10 lore raw

Re: [PATCH bpf-next v2 09/13] bpf: add bpf_skb_adjust_room encap flags

include/uapi/linux/bpf.h enum bpf_func_id { · line 2647

Quoted 5 lines · Willem de Bruijn
include/uapi/linux/bpf.h +4 -0
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4   (1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6   (1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK   (BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+                                        BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)

mask is still in uapi...

#11 lore raw

Re: [PATCH bpf-next v2 09/13] bpf: add bpf_skb_adjust_room encap flags

Quoted 10 lines · Alexei Starovoitov

include/uapi/linux/bpf.h enum bpf_func_id { · line 2647

Quoted 5 lines · Willem de Bruijn
include/uapi/linux/bpf.h +4 -0
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4   (1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6   (1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK   (BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+                                        BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)

mask is still in uapi...

That's only the L3 MASK, which captures ipv4 or ipv6. I don't see that being expanded.

BPF_F_ADJ_ROOM_MASK was moved to net/core/filter.c

#12 lore raw

Re: [PATCH bpf-next v2 09/13] bpf: add bpf_skb_adjust_room encap flags

Quoted 14 lines · Willem de Bruijn
Quoted 10 lines · Alexei Starovoitov

include/uapi/linux/bpf.h enum bpf_func_id { · line 2647

Quoted 5 lines · Willem de Bruijn
include/uapi/linux/bpf.h +4 -0
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4   (1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6   (1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK   (BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+                                        BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)

mask is still in uapi...

Quoted 2 lines

That's only the L3 MASK, which captures ipv4 or ipv6. I don't see that being expanded.

and what's the use of it for user space?
packet is either ipv4 or ipv6.
passing two flags will be rejected:
if (flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 &&
+ flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)
+ return -EINVAL;

#13 lore raw

Re: [PATCH bpf-next v2 09/13] bpf: add bpf_skb_adjust_room encap flags

Quoted 17 lines · Alexei Starovoitov
Quoted 14 lines · Willem de Bruijn
Quoted 10 lines · Alexei Starovoitov

include/uapi/linux/bpf.h enum bpf_func_id { · line 2647

Quoted 5 lines · Willem de Bruijn
include/uapi/linux/bpf.h +4 -0
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4   (1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6   (1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK   (BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+                                        BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)

mask is still in uapi...

Quoted 4 lines
Quoted 2 lines

That's only the L3 MASK, which captures ipv4 or ipv6. I don't see that being expanded.

and what's the use of it for user space?

I see. Indeed, it has none there. Will remove. Apologies for the extra round of revision as a result of this.

#14 lore raw

PATCH v2 10/13 bpf: Sync bpf.h to tools

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Sync include/uapi/linux/bpf.h with tools/

Changes v1->v2:
- BPF_F_ADJ_ROOM_MASK moved, no longer in this commit

---
 tools/include/uapi/linux/bpf.h | 32 +++++++++++++++++++++++++++++---
 1 file changed, 29 insertions(+), 3 deletions(-)
tools/include/uapi/linux/bpf.h +29 -3
index 3c04410137d9..f770f0de5b9c 100644
@@ -1478,13 +1478,27 @@ union bpf_attr {
  * 		Grow or shrink the room for data in the packet associated to
  * 		*skb* by *len_diff*, and according to the selected *mode*.
  *
- * 		There is a single supported mode at this time:
+ *		There are two supported modes at this time:
+ *
+ *		* **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer
+ *		  (room space is added or removed below the layer 2 header).
  *
  * 		* **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  * 		  (room space is added or removed below the layer 3 header).
  *
- * 		All values for *flags* are reserved for future usage, and must
- * 		be left at zero.
+ *		The following flags are supported at this time:
+ *
+ *		* **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size.
+ *		  Adjusting mss in this way is not allowed for datagrams.
+ *
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **:
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **:
+ *		  Any new space is reserved to hold a tunnel header.
+ *		  Configure skb offsets and other fields accordingly.
+ *
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **:
+ *		* **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
+ *		  Use with ENCAP_L3 flags to further specify the tunnel type.
  *
  * 		A call to this helper is susceptible to change the underlaying
  * 		packet buffer. Therefore, at load time, all checks on pointers
@@ -2624,9 +2638,21 @@ enum bpf_func_id {
 /* Current network namespace */
 #define BPF_F_CURRENT_NETNS		(-1L)
 
+/* BPF_FUNC_skb_adjust_room flags. */
+#define BPF_F_ADJ_ROOM_FIXED_GSO	(1ULL << 0)
+
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4	(1ULL << 1)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6	(1ULL << 2)
+#define BPF_F_ADJ_ROOM_ENCAP_L3_MASK	(BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | \
+					 BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)
+
+#define BPF_F_ADJ_ROOM_ENCAP_L4_GRE	(1ULL << 3)
+#define BPF_F_ADJ_ROOM_ENCAP_L4_UDP	(1ULL << 4)
+
 /* Mode for BPF_FUNC_skb_adjust_room helper. */
 enum bpf_adj_room_mode {
 	BPF_ADJ_ROOM_NET,
+	BPF_ADJ_ROOM_MAC,
 };
 
 /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#15 lore raw

PATCH v2 11/13 selftests/bpf: convert bpf tunnel test to BPF_ADJ_ROOM_MAC

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Avoid moving the network layer header when prefixing tunnel headers.

This avoids an explicit call to bpf_skb_store_bytes and an implicit move of the network header bytes in bpf_skb_adjust_room.

---
 .../selftests/bpf/progs/test_tc_tunnel.c      | 25 +++----------------
 1 file changed, 3 insertions(+), 22 deletions(-)
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +3 -22
index 900c5653105f..f6a16fd23dbd 100644
@@ -72,7 +72,7 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
 
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_NET, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, 0))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
@@ -94,12 +94,6 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
 
-	/* bpf_skb_adjust_room has moved header to start of room: restore */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN + olen,
-				&iph_inner, sizeof(iph_inner),
-				BPF_F_INVALIDATE_HASH) < 0)
-		return TC_ACT_SHOT;
-
 	return TC_ACT_OK;
 }
 
@@ -125,7 +119,7 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
 
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_NET, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, 0))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
@@ -145,12 +139,6 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 				BPF_F_INVALIDATE_HASH) < 0)
 		return TC_ACT_SHOT;
 
-	/* bpf_skb_adjust_room has moved header to start of room: restore */
-	if (bpf_skb_store_bytes(skb, ETH_HLEN + olen,
-				&iph_inner, sizeof(iph_inner),
-				BPF_F_INVALIDATE_HASH) < 0)
-		return TC_ACT_SHOT;
-
 	return TC_ACT_OK;
 }
 
@@ -207,14 +195,7 @@ static int decap_internal(struct __sk_buff *skb, int off, int len, char proto)
 		return TC_ACT_OK;
 	}
 
-	if (bpf_skb_load_bytes(skb, off + olen, &buf, olen) < 0)
-		return TC_ACT_OK;
-
-	if (bpf_skb_adjust_room(skb, -olen, BPF_ADJ_ROOM_NET, 0))
-		return TC_ACT_SHOT;
-
-	/* bpf_skb_adjust_room has moved outer over inner header: restore */
-	if (bpf_skb_store_bytes(skb, off, buf, len, BPF_F_INVALIDATE_HASH) < 0)
+	if (bpf_skb_adjust_room(skb, -olen, BPF_ADJ_ROOM_MAC, 0))
 		return TC_ACT_SHOT;
 
 	return TC_ACT_OK;
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#16 lore raw

PATCH v2 12/13 selftests/bpf: convert bpf tunnel test to BPF_F_ADJ_ROOM_FIXED_GSO

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Lower route MTU to ensure packets fit in device MTU after encap, then skip the gso_size changes.

---
 tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 11 ++++++++---
 tools/testing/selftests/bpf/test_tc_tunnel.sh      |  6 ++++++
 2 files changed, 14 insertions(+), 3 deletions(-)
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +8 -3
index f6a16fd23dbd..3b79dffb8103 100644
@@ -52,6 +52,7 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 	struct grev4hdr h_outer;
 	struct iphdr iph_inner;
 	struct tcphdr tcph;
+	__u64 flags;
 	int olen;
 
 	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
@@ -69,10 +70,11 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
+	flags = BPF_F_ADJ_ROOM_FIXED_GSO;
 	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
 
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
@@ -102,6 +104,7 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 	struct ipv6hdr iph_inner;
 	struct grev6hdr h_outer;
 	struct tcphdr tcph;
+	__u64 flags;
 	int olen;
 
 	if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner,
@@ -116,10 +119,11 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
+	flags = BPF_F_ADJ_ROOM_FIXED_GSO;
 	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
 
 	/* add room between mac and network header */
-	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, 0))
+	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags))
 		return TC_ACT_SHOT;
 
 	/* prepare new outer network header */
@@ -195,7 +199,8 @@ static int decap_internal(struct __sk_buff *skb, int off, int len, char proto)
 		return TC_ACT_OK;
 	}
 
-	if (bpf_skb_adjust_room(skb, -olen, BPF_ADJ_ROOM_MAC, 0))
+	if (bpf_skb_adjust_room(skb, -olen, BPF_ADJ_ROOM_MAC,
+				BPF_F_ADJ_ROOM_FIXED_GSO))
 		return TC_ACT_SHOT;
 
 	return TC_ACT_OK;
tools/testing/selftests/bpf/test_tc_tunnel.sh +6 -0
index 9e18754f2354..cda5317790d2 100755
@@ -35,6 +35,12 @@ setup() {
 	ip -netns "${ns1}" -6 addr add "${ns1_v6}/64" dev veth1 nodad
 	ip -netns "${ns2}" -6 addr add "${ns2_v6}/64" dev veth2 nodad
 
+	# clamp route to reserve room for tunnel headers
+	ip -netns "${ns1}" -4 route flush table main
+	ip -netns "${ns1}" -6 route flush table main
+	ip -netns "${ns1}" -4 route add "${ns2_v4}" mtu 1476 dev veth1
+	ip -netns "${ns1}" -6 route add "${ns2_v6}" mtu 1456 dev veth1
+
 	sleep 1
 
 	dd if=/dev/urandom of="${infile}" bs="${datalen}" count=1 status=none
Signature · 1 line
2.21.0.392.gf8f6787159e-goog
#17 lore raw

PATCH v2 13/13 selftests/bpf: convert bpf tunnel test to encap modes

In reply to Willem de Bruijn

From: Willem de Bruijn <[email protected]>

Make the tests correctly annotate skbs with tunnel metadata.

This makes the gso tests succeed. Enable them.

---
 .../selftests/bpf/progs/test_tc_tunnel.c      | 19 +++++++++++++++----
 tools/testing/selftests/bpf/test_tc_tunnel.sh | 10 ++++------
 2 files changed, 19 insertions(+), 10 deletions(-)
tools/testing/selftests/bpf/progs/test_tc_tunnel.c +15 -4
index 3b79dffb8103..f541c2de947d 100644
@@ -70,8 +70,13 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
-	flags = BPF_F_ADJ_ROOM_FIXED_GSO;
-	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
+	flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4;
+	if (with_gre) {
+		flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE;
+		olen = sizeof(h_outer);
+	} else {
+		olen = sizeof(h_outer.ip);
+	}
 
 	/* add room between mac and network header */
 	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags))
@@ -119,8 +124,14 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre)
 	if (tcph.dest != __bpf_constant_htons(cfg_port))
 		return TC_ACT_OK;
 
-	flags = BPF_F_ADJ_ROOM_FIXED_GSO;
-	olen = with_gre ? sizeof(h_outer) : sizeof(h_outer.ip);
+	flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV6;
+	if (with_gre) {
+		flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE;
+		olen = sizeof(h_outer);
+	} else {
+		olen = sizeof(h_outer.ip);
+	}
+
 
 	/* add room between mac and network header */
 	if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags))
tools/testing/selftests/bpf/test_tc_tunnel.sh +4 -6
index cda5317790d2..dcf320626931 100755
@@ -97,13 +97,11 @@ if [[ "$#" -eq "0" ]]; then
 	echo "ip6 gre"
 	$0 ipv6 ip6gre 100
 
-	# disabled until passes SKB_GSO_DODGY checks
-	# echo "ip gre gso"
-	# $0 ipv4 gre 2000
+	echo "ip gre gso"
+	$0 ipv4 gre 2000
 
-	# disabled until passes SKB_GSO_DODGY checks
-	# echo "ip6 gre gso"
-	# $0 ipv6 ip6gre 2000
+	echo "ip6 gre gso"
+	$0 ipv6 ip6gre 2000
 
 	echo "OK. All tests passed"
 	exit 0
Signature · 1 line
2.21.0.392.gf8f6787159e-goog