Extend bpf_skb_adjust_room growth to mark inner MAC header so that L2 encapsulation can be used for tc tunnels.
Patch #1 extends the existing test_tc_tunnel to support UDP encapsulation; later we want to be able to test MPLS over UDP and MPLS over GRE encapsulation.
Patch #2 adds the BPF_F_ADJ_ROOM_ENCAP_L2(len) macro, which allows specification of inner mac length. Other approaches were explored prior to taking this approach. Specifically, I tried automatically computing the inner mac length on the basis of the specified flags (so inner maclen for GRE/IPv4 encap is the len_diff specified to bpf_skb_adjust_room minus GRE + IPv4 header length for example). Problem with this is that we don't know for sure what form of GRE/UDP header we have; is it a full GRE header, or is it a FOU UDP header or generic UDP encap header? My fear here was we'd end up with an explosion of flags. The other approach tried was to support inner L2 header marking as a separate room adjustment, i.e. adjust for L3/L4 encap, then call bpf_skb_adjust_room for L2 encap. This can be made to work but because it imposed an order on operations, felt a bit clunky.
Patch #3 syncs tools/ bpf.h.
Patch #4 extends the tests again to support MPLSoverGRE, MPLSoverUDP, and transparent ethernet bridging (TEB) where the inner L2 header is an ethernet header. Testing of BPF encap against tunnels is done for cases where configuration of such tunnels is possible (MPLSoverGRE[6], MPLSoverUDP, gre[6]tap), and skipped otherwise. Testing of BPF encap/decap is always carried out.
Changes since v2: - updated tools/testing/selftest/bpf/config with FOU/MPLS CONFIG variables (patches 1, 4) - reduced noise in patch 1 by avoiding unnecessary movement of code - eliminated inner_mac variable in bpf_skb_net_grow (patch 2)
Changes since v1: - fixed formatting of commit references. - BPF_F_ADJ_ROOM_FIXED_GSO flag enabled on all variants (patch 1) - fixed fou6 options for UDP encap; checksum errors observed were due to the fact fou6 tunnel was not set up with correct ipproto options (41 -6). 0 checksums work fine (patch 1) - added definitions for mask and shift used in setting L2 length (patch 2) - allow udp encap with fixed GSO (patch 2) - changed "elen" to "l2_len" to be more descriptive (patch 4)
Alan Maguire (4): selftests_bpf: extend test_tc_tunnel for UDP encap bpf: add layer 2 encap support to bpf_skb_adjust_room bpf: sync bpf.h to tools/ for BPF_F_ADJ_ROOM_ENCAP_L2 selftests_bpf: add L2 encap to test_tc_tunnel
include/uapi/linux/bpf.h | 10 + net/core/filter.c | 12 +- tools/include/uapi/linux/bpf.h | 10 + tools/testing/selftests/bpf/config | 8 + tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 321 +++++++++++++++++---- tools/testing/selftests/bpf/test_tc_tunnel.sh | 136 +++++++-- 6 files changed, 417 insertions(+), 80 deletions(-)
commit 868d523535c2 ("bpf: add bpf_skb_adjust_room encap flags") introduced support to bpf_skb_adjust_room for GSO-friendly GRE and UDP encapsulation and later introduced associated test_tc_tunnel tests. Here those tests are extended to cover UDP encapsulation also.
Signed-off-by: Alan Maguire alan.maguire@oracle.com --- tools/testing/selftests/bpf/config | 4 + tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 140 ++++++++++++++------- tools/testing/selftests/bpf/test_tc_tunnel.sh | 47 ++++++- 3 files changed, 143 insertions(+), 48 deletions(-)
diff --git a/tools/testing/selftests/bpf/config b/tools/testing/selftests/bpf/config index a42f4fc..8e749c1 100644 --- a/tools/testing/selftests/bpf/config +++ b/tools/testing/selftests/bpf/config @@ -25,3 +25,7 @@ CONFIG_XDP_SOCKETS=y CONFIG_FTRACE_SYSCALLS=y CONFIG_IPV6_TUNNEL=y CONFIG_IPV6_GRE=y +CONFIG_NET_FOU=m +CONFIG_NET_FOU_IP_TUNNELS=y +CONFIG_IPV6_FOU=m +CONFIG_IPV6_FOU_TUNNEL=m diff --git a/tools/testing/selftests/bpf/progs/test_tc_tunnel.c b/tools/testing/selftests/bpf/progs/test_tc_tunnel.c index f541c2d..33b338e 100644 --- a/tools/testing/selftests/bpf/progs/test_tc_tunnel.c +++ b/tools/testing/selftests/bpf/progs/test_tc_tunnel.c @@ -12,6 +12,7 @@ #include <linux/ip.h> #include <linux/ipv6.h> #include <linux/tcp.h> +#include <linux/udp.h> #include <linux/pkt_cls.h> #include <linux/types.h>
@@ -20,16 +21,27 @@
static const int cfg_port = 8000;
-struct grev4hdr { - struct iphdr ip; +static const int cfg_udp_src = 20000; +static const int cfg_udp_dst = 5555; + +struct gre_hdr { __be16 flags; __be16 protocol; } __attribute__((packed));
-struct grev6hdr { +union l4hdr { + struct udphdr udp; + struct gre_hdr gre; +}; + +struct v4hdr { + struct iphdr ip; + union l4hdr l4hdr; +} __attribute__((packed)); + +struct v6hdr { struct ipv6hdr ip; - __be16 flags; - __be16 protocol; + union l4hdr l4hdr; } __attribute__((packed));
static __always_inline void set_ipv4_csum(struct iphdr *iph) @@ -47,10 +59,10 @@ static __always_inline void set_ipv4_csum(struct iphdr *iph) iph->check = ~((csum & 0xffff) + (csum >> 16)); }
-static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre) +static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto) { - struct grev4hdr h_outer; struct iphdr iph_inner; + struct v4hdr h_outer; struct tcphdr tcph; __u64 flags; int olen; @@ -70,12 +82,29 @@ 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;
+ olen = sizeof(h_outer.ip); + flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4; - if (with_gre) { + switch (encap_proto) { + case IPPROTO_GRE: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE; - olen = sizeof(h_outer); - } else { - olen = sizeof(h_outer.ip); + olen += sizeof(h_outer.l4hdr.gre); + h_outer.l4hdr.gre.protocol = bpf_htons(ETH_P_IP); + h_outer.l4hdr.gre.flags = 0; + break; + case IPPROTO_UDP: + flags |= BPF_F_ADJ_ROOM_ENCAP_L4_UDP; + olen += sizeof(h_outer.l4hdr.udp); + h_outer.l4hdr.udp.source = __bpf_constant_htons(cfg_udp_src); + h_outer.l4hdr.udp.dest = __bpf_constant_htons(cfg_udp_dst); + h_outer.l4hdr.udp.check = 0; + h_outer.l4hdr.udp.len = bpf_htons(bpf_ntohs(iph_inner.tot_len) + + sizeof(h_outer.l4hdr.udp)); + break; + case IPPROTO_IPIP: + break; + default: + return TC_ACT_OK; }
/* add room between mac and network header */ @@ -85,16 +114,10 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre) /* prepare new outer network header */ 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; - } + bpf_ntohs(h_outer.ip.tot_len)); + h_outer.ip.protocol = encap_proto;
- set_ipv4_csum((void *)&h_outer.ip); + set_ipv4_csum(&h_outer.ip);
/* store new outer network header */ if (bpf_skb_store_bytes(skb, ETH_HLEN, &h_outer, olen, @@ -104,11 +127,12 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, bool with_gre) return TC_ACT_OK; }
-static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre) +static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto) { struct ipv6hdr iph_inner; - struct grev6hdr h_outer; + struct v6hdr h_outer; struct tcphdr tcph; + __u16 tot_len; __u64 flags; int olen;
@@ -124,15 +148,32 @@ 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;
+ olen = sizeof(h_outer.ip); + flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV6; - if (with_gre) { + switch (encap_proto) { + case IPPROTO_GRE: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE; - olen = sizeof(h_outer); - } else { - olen = sizeof(h_outer.ip); + olen += sizeof(h_outer.l4hdr.gre); + h_outer.l4hdr.gre.protocol = bpf_htons(ETH_P_IPV6); + h_outer.l4hdr.gre.flags = 0; + break; + case IPPROTO_UDP: + flags |= BPF_F_ADJ_ROOM_ENCAP_L4_UDP; + olen += sizeof(h_outer.l4hdr.udp); + h_outer.l4hdr.udp.source = __bpf_constant_htons(cfg_udp_src); + h_outer.l4hdr.udp.dest = __bpf_constant_htons(cfg_udp_dst); + tot_len = bpf_ntohs(iph_inner.payload_len) + sizeof(iph_inner) + + sizeof(h_outer.l4hdr.udp); + h_outer.l4hdr.udp.check = 0; + h_outer.l4hdr.udp.len = bpf_htons(tot_len); + break; + case IPPROTO_IPV6: + break; + default: + return TC_ACT_OK; }
- /* add room between mac and network header */ if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags)) return TC_ACT_SHOT; @@ -141,13 +182,8 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre) 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; - } + + h_outer.ip.nexthdr = encap_proto;
/* store new outer network header */ if (bpf_skb_store_bytes(skb, ETH_HLEN, &h_outer, olen, @@ -161,7 +197,7 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, bool with_gre) int __encap_ipip(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) - return encap_ipv4(skb, false); + return encap_ipv4(skb, IPPROTO_IPIP); else return TC_ACT_OK; } @@ -170,7 +206,16 @@ int __encap_ipip(struct __sk_buff *skb) int __encap_gre(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) - return encap_ipv4(skb, true); + return encap_ipv4(skb, IPPROTO_GRE); + else + return TC_ACT_OK; +} + +SEC("encap_udp") +int __encap_udp(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) + return encap_ipv4(skb, IPPROTO_UDP); else return TC_ACT_OK; } @@ -179,7 +224,7 @@ int __encap_gre(struct __sk_buff *skb) int __encap_ip6tnl(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) - return encap_ipv6(skb, false); + return encap_ipv6(skb, IPPROTO_IPV6); else return TC_ACT_OK; } @@ -188,23 +233,34 @@ int __encap_ip6tnl(struct __sk_buff *skb) int __encap_ip6gre(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) - return encap_ipv6(skb, true); + return encap_ipv6(skb, IPPROTO_GRE); + else + return TC_ACT_OK; +} + +SEC("encap_ip6udp") +int __encap_ip6udp(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) + return encap_ipv6(skb, IPPROTO_UDP); else return TC_ACT_OK; }
static int decap_internal(struct __sk_buff *skb, int off, int len, char proto) { - char buf[sizeof(struct grev6hdr)]; - int olen; + char buf[sizeof(struct v6hdr)]; + int olen = len;
switch (proto) { case IPPROTO_IPIP: case IPPROTO_IPV6: - olen = len; break; case IPPROTO_GRE: - olen = len + 4 /* gre hdr */; + olen += sizeof(struct gre_hdr); + break; + case IPPROTO_UDP: + olen += sizeof(struct udphdr); break; default: return TC_ACT_OK; diff --git a/tools/testing/selftests/bpf/test_tc_tunnel.sh b/tools/testing/selftests/bpf/test_tc_tunnel.sh index c805adb..64f3091 100755 --- a/tools/testing/selftests/bpf/test_tc_tunnel.sh +++ b/tools/testing/selftests/bpf/test_tc_tunnel.sh @@ -15,6 +15,9 @@ readonly ns2_v4=192.168.1.2 readonly ns1_v6=fd::1 readonly ns2_v6=fd::2
+# Must match port used by bpf program +readonly udpport=5555 + readonly infile="$(mktemp)" readonly outfile="$(mktemp)"
@@ -38,8 +41,8 @@ setup() { # 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 + ip -netns "${ns1}" -4 route add "${ns2_v4}" mtu 1472 dev veth1 + ip -netns "${ns1}" -6 route add "${ns2_v6}" mtu 1452 dev veth1
sleep 1
@@ -103,6 +106,18 @@ if [[ "$#" -eq "0" ]]; then echo "ip6 gre gso" $0 ipv6 ip6gre 2000
+ echo "ip udp" + $0 ipv4 udp 100 + + echo "ip6 udp" + $0 ipv6 ip6udp 100 + + echo "ip udp gso" + $0 ipv4 udp 2000 + + echo "ip6 udp gso" + $0 ipv6 ip6udp 2000 + echo "OK. All tests passed" exit 0 fi @@ -117,12 +132,20 @@ case "$1" in "ipv4") readonly addr1="${ns1_v4}" readonly addr2="${ns2_v4}" - readonly netcat_opt=-4 + readonly ipproto=4 + readonly netcat_opt=-${ipproto} + readonly foumod=fou + readonly foutype=ipip + readonly fouproto=4 ;; "ipv6") readonly addr1="${ns1_v6}" readonly addr2="${ns2_v6}" - readonly netcat_opt=-6 + readonly ipproto=6 + readonly netcat_opt=-${ipproto} + readonly foumod=fou6 + readonly foutype=ip6tnl + readonly fouproto="41 -6" ;; *) echo "unknown arg: $1" @@ -155,11 +178,23 @@ echo "test bpf encap without decap (expect failure)" server_listen ! client_connect
+if [[ "$tuntype" =~ "udp" ]]; then + # Set up fou tunnel. + ttype="${foutype}" + targs="encap fou encap-sport auto encap-dport $udpport" + # fou may be a module; allow this to fail. + modprobe "${foumod}" ||true + ip netns exec "${ns2}" ip fou add port 5555 ipproto ${fouproto} +else + ttype=$tuntype + targs="" +fi + # serverside, insert decap module # server is still running # client can connect again -ip netns exec "${ns2}" ip link add dev testtun0 type "${tuntype}" \ - remote "${addr1}" local "${addr2}" +ip netns exec "${ns2}" ip link add name testtun0 type "${ttype}" \ + remote "${addr1}" local "${addr2}" $targs # Because packets are decapped by the tunnel they arrive on testtun0 from # the IP stack perspective. Ensure reverse path filtering is disabled # otherwise we drop the TCP SYN as arriving on testtun0 instead of the
commit 868d523535c2 ("bpf: add bpf_skb_adjust_room encap flags") introduced support to bpf_skb_adjust_room for GSO-friendly GRE and UDP encapsulation.
For GSO to work for skbs, the inner headers (mac and network) need to be marked. For L3 encapsulation using bpf_skb_adjust_room, the mac and network headers are identical. Here we provide a way of specifying the inner mac header length for cases where L2 encap is desired. Such an approach can support encapsulated ethernet headers, MPLS headers etc. For example to convert from a packet of form [eth][ip][tcp] to [eth][ip][udp][inner mac][ip][tcp], something like the following could be done:
headroom = sizeof(iph) + sizeof(struct udphdr) + inner_maclen;
ret = bpf_skb_adjust_room(skb, headroom, BPF_ADJ_ROOM_MAC, BPF_F_ADJ_ROOM_ENCAP_L4_UDP | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | BPF_F_ADJ_ROOM_ENCAP_L2(inner_maclen));
Signed-off-by: Alan Maguire alan.maguire@oracle.com --- include/uapi/linux/bpf.h | 10 ++++++++++ net/core/filter.c | 12 ++++++++---- 2 files changed, 18 insertions(+), 4 deletions(-)
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 8370245..912391c 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -1500,6 +1500,10 @@ struct bpf_stack_build_id { * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **: * Use with ENCAP_L3 flags to further specify the tunnel type. * + * * **BPF_F_ADJ_ROOM_ENCAP_L2(len) **: + * Use with ENCAP_L3/L4 flags to further specify the tunnel + * type; **len** is the length of the inner MAC header. + * * 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 @@ -2641,10 +2645,16 @@ enum bpf_func_id { /* BPF_FUNC_skb_adjust_room flags. */ #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
+#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff +#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 + #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_L4_GRE (1ULL << 3) #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) +#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ + BPF_ADJ_ROOM_ENCAP_L2_MASK) \ + << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
/* Mode for BPF_FUNC_skb_adjust_room helper. */ enum bpf_adj_room_mode { diff --git a/net/core/filter.c b/net/core/filter.c index 22eb2ed..a1654ef62 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -2969,11 +2969,14 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb) #define BPF_F_ADJ_ROOM_MASK (BPF_F_ADJ_ROOM_FIXED_GSO | \ BPF_F_ADJ_ROOM_ENCAP_L3_MASK | \ BPF_F_ADJ_ROOM_ENCAP_L4_GRE | \ - BPF_F_ADJ_ROOM_ENCAP_L4_UDP) + BPF_F_ADJ_ROOM_ENCAP_L4_UDP | \ + BPF_F_ADJ_ROOM_ENCAP_L2( \ + BPF_ADJ_ROOM_ENCAP_L2_MASK))
static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, u64 flags) { + u8 inner_mac_len = flags >> BPF_ADJ_ROOM_ENCAP_L2_SHIFT; bool encap = flags & BPF_F_ADJ_ROOM_ENCAP_L3_MASK; u16 mac_len = 0, inner_net = 0, inner_trans = 0; unsigned int gso_type = SKB_GSO_DODGY; @@ -3008,6 +3011,8 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff,
mac_len = skb->network_header - skb->mac_header; inner_net = skb->network_header; + if (inner_mac_len > len_diff) + return -EINVAL; inner_trans = skb->transport_header; }
@@ -3016,8 +3021,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, return ret;
if (encap) { - /* inner mac == inner_net on l3 encap */ - skb->inner_mac_header = inner_net; + skb->inner_mac_header = inner_net - inner_mac_len; skb->inner_network_header = inner_net; skb->inner_transport_header = inner_trans; skb_set_inner_protocol(skb, skb->protocol); @@ -3031,7 +3035,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, gso_type |= SKB_GSO_GRE; else if (flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV6) gso_type |= SKB_GSO_IPXIP6; - else + else if (flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV4) gso_type |= SKB_GSO_IPXIP4;
if (flags & BPF_F_ADJ_ROOM_ENCAP_L4_GRE ||
On 04/09/2019 04:06 PM, Alan Maguire wrote:
commit 868d523535c2 ("bpf: add bpf_skb_adjust_room encap flags") introduced support to bpf_skb_adjust_room for GSO-friendly GRE and UDP encapsulation.
For GSO to work for skbs, the inner headers (mac and network) need to be marked. For L3 encapsulation using bpf_skb_adjust_room, the mac and network headers are identical. Here we provide a way of specifying the inner mac header length for cases where L2 encap is desired. Such an approach can support encapsulated ethernet headers, MPLS headers etc. For example to convert from a packet of form [eth][ip][tcp] to [eth][ip][udp][inner mac][ip][tcp], something like the following could be done:
headroom = sizeof(iph) + sizeof(struct udphdr) + inner_maclen;
ret = bpf_skb_adjust_room(skb, headroom, BPF_ADJ_ROOM_MAC, BPF_F_ADJ_ROOM_ENCAP_L4_UDP | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | BPF_F_ADJ_ROOM_ENCAP_L2(inner_maclen));
Signed-off-by: Alan Maguire alan.maguire@oracle.com
include/uapi/linux/bpf.h | 10 ++++++++++ net/core/filter.c | 12 ++++++++---- 2 files changed, 18 insertions(+), 4 deletions(-)
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 8370245..912391c 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -1500,6 +1500,10 @@ struct bpf_stack_build_id {
* **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
Use with ENCAP_L3 flags to further specify the tunnel type.
* **BPF_F_ADJ_ROOM_ENCAP_L2(len) **:
Use with ENCAP_L3/L4 flags to further specify the tunnel
type; **len** is the length of the inner MAC header.
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
@@ -2641,10 +2645,16 @@ enum bpf_func_id { /* BPF_FUNC_skb_adjust_room flags. */ #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) +#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff +#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
Would appreciate a small whitespace cleanup in that tab after the define is replaced with a normal space like in the rest of the file.
#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_L4_GRE (1ULL << 3) #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) +#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
BPF_ADJ_ROOM_ENCAP_L2_MASK) \
<< BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Ditto here.
/* Mode for BPF_FUNC_skb_adjust_room helper. */ enum bpf_adj_room_mode { diff --git a/net/core/filter.c b/net/core/filter.c index 22eb2ed..a1654ef62 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -2969,11 +2969,14 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb) #define BPF_F_ADJ_ROOM_MASK (BPF_F_ADJ_ROOM_FIXED_GSO | \ BPF_F_ADJ_ROOM_ENCAP_L3_MASK | \ BPF_F_ADJ_ROOM_ENCAP_L4_GRE | \
BPF_F_ADJ_ROOM_ENCAP_L4_UDP)
BPF_F_ADJ_ROOM_ENCAP_L4_UDP | \
BPF_F_ADJ_ROOM_ENCAP_L2( \
BPF_ADJ_ROOM_ENCAP_L2_MASK))
static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, u64 flags) {
One thing that it seems missing in this but also in Willem's earlier set is that we reject these flags in bpf_skb_net_shrink(). Both are called from bpf_skb_adjust_room(), so if we don't support decap yet we should reject such flags so we can extend in future. Could you or Willem send a follow-up? Thanks!
- u8 inner_mac_len = flags >> BPF_ADJ_ROOM_ENCAP_L2_SHIFT; bool encap = flags & BPF_F_ADJ_ROOM_ENCAP_L3_MASK; u16 mac_len = 0, inner_net = 0, inner_trans = 0; unsigned int gso_type = SKB_GSO_DODGY;
@@ -3008,6 +3011,8 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, mac_len = skb->network_header - skb->mac_header; inner_net = skb->network_header;
if (inner_mac_len > len_diff)
inner_trans = skb->transport_header; }return -EINVAL;
@@ -3016,8 +3021,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, return ret; if (encap) {
/* inner mac == inner_net on l3 encap */
skb->inner_mac_header = inner_net;
skb->inner_network_header = inner_net; skb->inner_transport_header = inner_trans; skb_set_inner_protocol(skb, skb->protocol);skb->inner_mac_header = inner_net - inner_mac_len;
@@ -3031,7 +3035,7 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, gso_type |= SKB_GSO_GRE; else if (flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV6) gso_type |= SKB_GSO_IPXIP6;
else
else if (flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV4) gso_type |= SKB_GSO_IPXIP4;
if (flags & BPF_F_ADJ_ROOM_ENCAP_L4_GRE ||
On Thu, Apr 11, 2019 at 5:09 PM Daniel Borkmann daniel@iogearbox.net wrote:
On 04/09/2019 04:06 PM, Alan Maguire wrote:
commit 868d523535c2 ("bpf: add bpf_skb_adjust_room encap flags") introduced support to bpf_skb_adjust_room for GSO-friendly GRE and UDP encapsulation.
For GSO to work for skbs, the inner headers (mac and network) need to be marked. For L3 encapsulation using bpf_skb_adjust_room, the mac and network headers are identical. Here we provide a way of specifying the inner mac header length for cases where L2 encap is desired. Such an approach can support encapsulated ethernet headers, MPLS headers etc. For example to convert from a packet of form [eth][ip][tcp] to [eth][ip][udp][inner mac][ip][tcp], something like the following could be done:
headroom = sizeof(iph) + sizeof(struct udphdr) + inner_maclen; ret = bpf_skb_adjust_room(skb, headroom, BPF_ADJ_ROOM_MAC, BPF_F_ADJ_ROOM_ENCAP_L4_UDP | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 | BPF_F_ADJ_ROOM_ENCAP_L2(inner_maclen));
Signed-off-by: Alan Maguire alan.maguire@oracle.com
include/uapi/linux/bpf.h | 10 ++++++++++ net/core/filter.c | 12 ++++++++---- 2 files changed, 18 insertions(+), 4 deletions(-)
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 8370245..912391c 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -1500,6 +1500,10 @@ struct bpf_stack_build_id {
* **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **:
Use with ENCAP_L3 flags to further specify the tunnel type.
* **BPF_F_ADJ_ROOM_ENCAP_L2(len) **:
Use with ENCAP_L3/L4 flags to further specify the tunnel
type; **len** is the length of the inner MAC header.
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
@@ -2641,10 +2645,16 @@ enum bpf_func_id { /* BPF_FUNC_skb_adjust_room flags. */ #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
+#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff +#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56
Would appreciate a small whitespace cleanup in that tab after the define is replaced with a normal space like in the rest of the file.
#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_L4_GRE (1ULL << 3) #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) +#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \
BPF_ADJ_ROOM_ENCAP_L2_MASK) \
<< BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
Ditto here.
/* Mode for BPF_FUNC_skb_adjust_room helper. */ enum bpf_adj_room_mode { diff --git a/net/core/filter.c b/net/core/filter.c index 22eb2ed..a1654ef62 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -2969,11 +2969,14 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb) #define BPF_F_ADJ_ROOM_MASK (BPF_F_ADJ_ROOM_FIXED_GSO | \ BPF_F_ADJ_ROOM_ENCAP_L3_MASK | \ BPF_F_ADJ_ROOM_ENCAP_L4_GRE | \
BPF_F_ADJ_ROOM_ENCAP_L4_UDP)
BPF_F_ADJ_ROOM_ENCAP_L4_UDP | \
BPF_F_ADJ_ROOM_ENCAP_L2( \
BPF_ADJ_ROOM_ENCAP_L2_MASK))
static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, u64 flags) {
One thing that it seems missing in this but also in Willem's earlier set is that we reject these flags in bpf_skb_net_shrink(). Both are called from bpf_skb_adjust_room(), so if we don't support decap yet we should reject such flags so we can extend in future. Could you or Willem send a follow-up?
Indeed. Thanks for bringing that up, Daniel. I only thought of that after submitting the original patchset, too. Will send a patch.
On 04/11/2019 11:18 PM, Willem de Bruijn wrote: [...]
diff --git a/net/core/filter.c b/net/core/filter.c index 22eb2ed..a1654ef62 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -2969,11 +2969,14 @@ static u32 bpf_skb_net_base_len(const struct sk_buff *skb) #define BPF_F_ADJ_ROOM_MASK (BPF_F_ADJ_ROOM_FIXED_GSO | \ BPF_F_ADJ_ROOM_ENCAP_L3_MASK | \ BPF_F_ADJ_ROOM_ENCAP_L4_GRE | \
BPF_F_ADJ_ROOM_ENCAP_L4_UDP)
BPF_F_ADJ_ROOM_ENCAP_L4_UDP | \
BPF_F_ADJ_ROOM_ENCAP_L2( \
BPF_ADJ_ROOM_ENCAP_L2_MASK))
static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff, u64 flags) {
One thing that it seems missing in this but also in Willem's earlier set is that we reject these flags in bpf_skb_net_shrink(). Both are called from bpf_skb_adjust_room(), so if we don't support decap yet we should reject such flags so we can extend in future. Could you or Willem send a follow-up?
Indeed. Thanks for bringing that up, Daniel. I only thought of that after submitting the original patchset, too. Will send a patch.
Great, thanks, appreciated! :)
Sync include/uapi/linux/bpf.h with tools/ equivalent to add BPF_F_ADJ_ROOM_ENCAP_L2(len) macro.
Signed-off-by: Alan Maguire alan.maguire@oracle.com --- tools/include/uapi/linux/bpf.h | 10 ++++++++++ 1 file changed, 10 insertions(+)
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 8370245..912391c 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -1500,6 +1500,10 @@ struct bpf_stack_build_id { * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **: * Use with ENCAP_L3 flags to further specify the tunnel type. * + * * **BPF_F_ADJ_ROOM_ENCAP_L2(len) **: + * Use with ENCAP_L3/L4 flags to further specify the tunnel + * type; **len** is the length of the inner MAC header. + * * 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 @@ -2641,10 +2645,16 @@ enum bpf_func_id { /* BPF_FUNC_skb_adjust_room flags. */ #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0)
+#define BPF_ADJ_ROOM_ENCAP_L2_MASK 0xff +#define BPF_ADJ_ROOM_ENCAP_L2_SHIFT 56 + #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_L4_GRE (1ULL << 3) #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) +#define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ + BPF_ADJ_ROOM_ENCAP_L2_MASK) \ + << BPF_ADJ_ROOM_ENCAP_L2_SHIFT)
/* Mode for BPF_FUNC_skb_adjust_room helper. */ enum bpf_adj_room_mode {
Update test_tc_tunnel to verify adding inner L2 header encapsulation (an MPLS label or ethernet header) works.
Signed-off-by: Alan Maguire alan.maguire@oracle.com --- tools/testing/selftests/bpf/config | 4 + tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 219 ++++++++++++++++++--- tools/testing/selftests/bpf/test_tc_tunnel.sh | 113 ++++++++--- 3 files changed, 277 insertions(+), 59 deletions(-)
diff --git a/tools/testing/selftests/bpf/config b/tools/testing/selftests/bpf/config index 8e749c1..8c97647 100644 --- a/tools/testing/selftests/bpf/config +++ b/tools/testing/selftests/bpf/config @@ -29,3 +29,7 @@ CONFIG_NET_FOU=m CONFIG_NET_FOU_IP_TUNNELS=y CONFIG_IPV6_FOU=m CONFIG_IPV6_FOU_TUNNEL=m +CONFIG_MPLS=y +CONFIG_NET_MPLS_GSO=m +CONFIG_MPLS_ROUTING=m +CONFIG_MPLS_IPTUNNEL=m diff --git a/tools/testing/selftests/bpf/progs/test_tc_tunnel.c b/tools/testing/selftests/bpf/progs/test_tc_tunnel.c index 33b338e..bcb00d7 100644 --- a/tools/testing/selftests/bpf/progs/test_tc_tunnel.c +++ b/tools/testing/selftests/bpf/progs/test_tc_tunnel.c @@ -11,6 +11,7 @@ #include <linux/in.h> #include <linux/ip.h> #include <linux/ipv6.h> +#include <linux/mpls.h> #include <linux/tcp.h> #include <linux/udp.h> #include <linux/pkt_cls.h> @@ -22,7 +23,14 @@ static const int cfg_port = 8000;
static const int cfg_udp_src = 20000; -static const int cfg_udp_dst = 5555; + +#define UDP_PORT 5555 +#define MPLS_OVER_UDP_PORT 6635 +#define ETH_OVER_UDP_PORT 7777 + +/* MPLS label 1000 with S bit (last label) set and ttl of 255. */ +static const __u32 mpls_label = __bpf_constant_htonl(1000 << 12 | + MPLS_LS_S_MASK | 0xff);
struct gre_hdr { __be16 flags; @@ -37,11 +45,13 @@ struct gre_hdr { struct v4hdr { struct iphdr ip; union l4hdr l4hdr; + __u8 pad[16]; /* enough space for L2 header */ } __attribute__((packed));
struct v6hdr { struct ipv6hdr ip; union l4hdr l4hdr; + __u8 pad[16]; /* enough space for L2 header */ } __attribute__((packed));
static __always_inline void set_ipv4_csum(struct iphdr *iph) @@ -59,13 +69,15 @@ static __always_inline void set_ipv4_csum(struct iphdr *iph) iph->check = ~((csum & 0xffff) + (csum >> 16)); }
-static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto) +static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto, + __u16 l2_proto) { + __u16 udp_dst = UDP_PORT; struct iphdr iph_inner; struct v4hdr h_outer; struct tcphdr tcph; + int olen, l2_len; __u64 flags; - int olen;
if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner, sizeof(iph_inner)) < 0) @@ -83,23 +95,38 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK;
olen = sizeof(h_outer.ip); + l2_len = 0;
flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV4; + + switch (l2_proto) { + case ETH_P_MPLS_UC: + l2_len = sizeof(mpls_label); + udp_dst = MPLS_OVER_UDP_PORT; + break; + case ETH_P_TEB: + l2_len = ETH_HLEN; + udp_dst = ETH_OVER_UDP_PORT; + break; + } + flags |= BPF_F_ADJ_ROOM_ENCAP_L2(l2_len); + switch (encap_proto) { case IPPROTO_GRE: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE; olen += sizeof(h_outer.l4hdr.gre); - h_outer.l4hdr.gre.protocol = bpf_htons(ETH_P_IP); + h_outer.l4hdr.gre.protocol = bpf_htons(l2_proto); h_outer.l4hdr.gre.flags = 0; break; case IPPROTO_UDP: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_UDP; olen += sizeof(h_outer.l4hdr.udp); h_outer.l4hdr.udp.source = __bpf_constant_htons(cfg_udp_src); - h_outer.l4hdr.udp.dest = __bpf_constant_htons(cfg_udp_dst); + h_outer.l4hdr.udp.dest = bpf_htons(udp_dst); h_outer.l4hdr.udp.check = 0; h_outer.l4hdr.udp.len = bpf_htons(bpf_ntohs(iph_inner.tot_len) + - sizeof(h_outer.l4hdr.udp)); + sizeof(h_outer.l4hdr.udp) + + l2_len); break; case IPPROTO_IPIP: break; @@ -107,6 +134,19 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK; }
+ /* add L2 encap (if specified) */ + switch (l2_proto) { + case ETH_P_MPLS_UC: + *((__u32 *)((__u8 *)&h_outer + olen)) = mpls_label; + break; + case ETH_P_TEB: + if (bpf_skb_load_bytes(skb, 0, (__u8 *)&h_outer + olen, + ETH_HLEN)) + return TC_ACT_SHOT; + break; + } + olen += l2_len; + /* add room between mac and network header */ if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags)) return TC_ACT_SHOT; @@ -127,14 +167,16 @@ static __always_inline int encap_ipv4(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK; }
-static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto) +static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto, + __u16 l2_proto) { + __u16 udp_dst = UDP_PORT; struct ipv6hdr iph_inner; struct v6hdr h_outer; struct tcphdr tcph; + int olen, l2_len; __u16 tot_len; __u64 flags; - int olen;
if (bpf_skb_load_bytes(skb, ETH_HLEN, &iph_inner, sizeof(iph_inner)) < 0) @@ -149,20 +191,34 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK;
olen = sizeof(h_outer.ip); + l2_len = 0;
flags = BPF_F_ADJ_ROOM_FIXED_GSO | BPF_F_ADJ_ROOM_ENCAP_L3_IPV6; + + switch (l2_proto) { + case ETH_P_MPLS_UC: + l2_len = sizeof(mpls_label); + udp_dst = MPLS_OVER_UDP_PORT; + break; + case ETH_P_TEB: + l2_len = ETH_HLEN; + udp_dst = ETH_OVER_UDP_PORT; + break; + } + flags |= BPF_F_ADJ_ROOM_ENCAP_L2(l2_len); + switch (encap_proto) { case IPPROTO_GRE: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_GRE; olen += sizeof(h_outer.l4hdr.gre); - h_outer.l4hdr.gre.protocol = bpf_htons(ETH_P_IPV6); + h_outer.l4hdr.gre.protocol = bpf_htons(l2_proto); h_outer.l4hdr.gre.flags = 0; break; case IPPROTO_UDP: flags |= BPF_F_ADJ_ROOM_ENCAP_L4_UDP; olen += sizeof(h_outer.l4hdr.udp); h_outer.l4hdr.udp.source = __bpf_constant_htons(cfg_udp_src); - h_outer.l4hdr.udp.dest = __bpf_constant_htons(cfg_udp_dst); + h_outer.l4hdr.udp.dest = bpf_htons(udp_dst); tot_len = bpf_ntohs(iph_inner.payload_len) + sizeof(iph_inner) + sizeof(h_outer.l4hdr.udp); h_outer.l4hdr.udp.check = 0; @@ -174,6 +230,19 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK; }
+ /* add L2 encap (if specified) */ + switch (l2_proto) { + case ETH_P_MPLS_UC: + *((__u32 *)((__u8 *)&h_outer + olen)) = mpls_label; + break; + case ETH_P_TEB: + if (bpf_skb_load_bytes(skb, 0, (__u8 *)&h_outer + olen, + ETH_HLEN)) + return TC_ACT_SHOT; + break; + } + olen += l2_len; + /* add room between mac and network header */ if (bpf_skb_adjust_room(skb, olen, BPF_ADJ_ROOM_MAC, flags)) return TC_ACT_SHOT; @@ -193,56 +262,128 @@ static __always_inline int encap_ipv6(struct __sk_buff *skb, __u8 encap_proto) return TC_ACT_OK; }
-SEC("encap_ipip") -int __encap_ipip(struct __sk_buff *skb) +SEC("encap_ipip_none") +int __encap_ipip_none(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) + return encap_ipv4(skb, IPPROTO_IPIP, ETH_P_IP); + else + return TC_ACT_OK; +} + +SEC("encap_gre_none") +int __encap_gre_none(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) + return encap_ipv4(skb, IPPROTO_GRE, ETH_P_IP); + else + return TC_ACT_OK; +} + +SEC("encap_gre_mpls") +int __encap_gre_mpls(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) + return encap_ipv4(skb, IPPROTO_GRE, ETH_P_MPLS_UC); + else + return TC_ACT_OK; +} + +SEC("encap_gre_eth") +int __encap_gre_eth(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) - return encap_ipv4(skb, IPPROTO_IPIP); + return encap_ipv4(skb, IPPROTO_GRE, ETH_P_TEB); else return TC_ACT_OK; }
-SEC("encap_gre") -int __encap_gre(struct __sk_buff *skb) +SEC("encap_udp_none") +int __encap_udp_none(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) - return encap_ipv4(skb, IPPROTO_GRE); + return encap_ipv4(skb, IPPROTO_UDP, ETH_P_IP); else return TC_ACT_OK; }
-SEC("encap_udp") -int __encap_udp(struct __sk_buff *skb) +SEC("encap_udp_mpls") +int __encap_udp_mpls(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) - return encap_ipv4(skb, IPPROTO_UDP); + return encap_ipv4(skb, IPPROTO_UDP, ETH_P_MPLS_UC); + else + return TC_ACT_OK; +} + +SEC("encap_udp_eth") +int __encap_udp_eth(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IP)) + return encap_ipv4(skb, IPPROTO_UDP, ETH_P_TEB); + else + return TC_ACT_OK; +} + +SEC("encap_ip6tnl_none") +int __encap_ip6tnl_none(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) + return encap_ipv6(skb, IPPROTO_IPV6, ETH_P_IPV6); + else + return TC_ACT_OK; +} + +SEC("encap_ip6gre_none") +int __encap_ip6gre_none(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) + return encap_ipv6(skb, IPPROTO_GRE, ETH_P_IPV6); + else + return TC_ACT_OK; +} + +SEC("encap_ip6gre_mpls") +int __encap_ip6gre_mpls(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) + return encap_ipv6(skb, IPPROTO_GRE, ETH_P_MPLS_UC); + else + return TC_ACT_OK; +} + +SEC("encap_ip6gre_eth") +int __encap_ip6gre_eth(struct __sk_buff *skb) +{ + if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) + return encap_ipv6(skb, IPPROTO_GRE, ETH_P_TEB); else return TC_ACT_OK; }
-SEC("encap_ip6tnl") -int __encap_ip6tnl(struct __sk_buff *skb) +SEC("encap_ip6udp_none") +int __encap_ip6udp_none(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) - return encap_ipv6(skb, IPPROTO_IPV6); + return encap_ipv6(skb, IPPROTO_UDP, ETH_P_IPV6); else return TC_ACT_OK; }
-SEC("encap_ip6gre") -int __encap_ip6gre(struct __sk_buff *skb) +SEC("encap_ip6udp_mpls") +int __encap_ip6udp_mpls(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) - return encap_ipv6(skb, IPPROTO_GRE); + return encap_ipv6(skb, IPPROTO_UDP, ETH_P_MPLS_UC); else return TC_ACT_OK; }
-SEC("encap_ip6udp") -int __encap_ip6udp(struct __sk_buff *skb) +SEC("encap_ip6udp_eth") +int __encap_ip6udp_eth(struct __sk_buff *skb) { if (skb->protocol == __bpf_constant_htons(ETH_P_IPV6)) - return encap_ipv6(skb, IPPROTO_UDP); + return encap_ipv6(skb, IPPROTO_UDP, ETH_P_TEB); else return TC_ACT_OK; } @@ -250,6 +391,8 @@ int __encap_ip6udp(struct __sk_buff *skb) static int decap_internal(struct __sk_buff *skb, int off, int len, char proto) { char buf[sizeof(struct v6hdr)]; + struct gre_hdr greh; + struct udphdr udph; int olen = len;
switch (proto) { @@ -258,9 +401,29 @@ static int decap_internal(struct __sk_buff *skb, int off, int len, char proto) break; case IPPROTO_GRE: olen += sizeof(struct gre_hdr); + if (bpf_skb_load_bytes(skb, off + len, &greh, sizeof(greh)) < 0) + return TC_ACT_OK; + switch (bpf_ntohs(greh.protocol)) { + case ETH_P_MPLS_UC: + olen += sizeof(mpls_label); + break; + case ETH_P_TEB: + olen += ETH_HLEN; + break; + } break; case IPPROTO_UDP: olen += sizeof(struct udphdr); + if (bpf_skb_load_bytes(skb, off + len, &udph, sizeof(udph)) < 0) + return TC_ACT_OK; + switch (bpf_ntohs(udph.dest)) { + case MPLS_OVER_UDP_PORT: + olen += sizeof(mpls_label); + break; + case ETH_OVER_UDP_PORT: + olen += ETH_HLEN; + break; + } break; default: return TC_ACT_OK; diff --git a/tools/testing/selftests/bpf/test_tc_tunnel.sh b/tools/testing/selftests/bpf/test_tc_tunnel.sh index 64f3091..d4d8d5d 100755 --- a/tools/testing/selftests/bpf/test_tc_tunnel.sh +++ b/tools/testing/selftests/bpf/test_tc_tunnel.sh @@ -17,6 +17,9 @@ readonly ns2_v6=fd::2
# Must match port used by bpf program readonly udpport=5555 +# MPLSoverUDP +readonly mplsudpport=6635 +readonly mplsproto=137
readonly infile="$(mktemp)" readonly outfile="$(mktemp)" @@ -41,8 +44,8 @@ setup() { # clamp route to reserve room for tunnel headers ip -netns "${ns1}" -4 route flush table main ip -netns "${ns1}" -6 route flush table main - ip -netns "${ns1}" -4 route add "${ns2_v4}" mtu 1472 dev veth1 - ip -netns "${ns1}" -6 route add "${ns2_v6}" mtu 1452 dev veth1 + ip -netns "${ns1}" -4 route add "${ns2_v4}" mtu 1458 dev veth1 + ip -netns "${ns1}" -6 route add "${ns2_v6}" mtu 1438 dev veth1
sleep 1
@@ -89,42 +92,44 @@ set -e # no arguments: automated test, run all if [[ "$#" -eq "0" ]]; then echo "ipip" - $0 ipv4 ipip 100 + $0 ipv4 ipip none 100
echo "ip6ip6" - $0 ipv6 ip6tnl 100 + $0 ipv6 ip6tnl none 100
- echo "ip gre" - $0 ipv4 gre 100 + for mac in none mpls eth ; do + echo "ip gre $mac" + $0 ipv4 gre $mac 100
- echo "ip6 gre" - $0 ipv6 ip6gre 100 + echo "ip6 gre $mac" + $0 ipv6 ip6gre $mac 100
- echo "ip gre gso" - $0 ipv4 gre 2000 + echo "ip gre $mac gso" + $0 ipv4 gre $mac 2000
- echo "ip6 gre gso" - $0 ipv6 ip6gre 2000 + echo "ip6 gre $mac gso" + $0 ipv6 ip6gre $mac 2000
- echo "ip udp" - $0 ipv4 udp 100 + echo "ip udp $mac" + $0 ipv4 udp $mac 100
- echo "ip6 udp" - $0 ipv6 ip6udp 100 + echo "ip6 udp $mac" + $0 ipv6 ip6udp $mac 100
- echo "ip udp gso" - $0 ipv4 udp 2000 + echo "ip udp $mac gso" + $0 ipv4 udp $mac 2000
- echo "ip6 udp gso" - $0 ipv6 ip6udp 2000 + echo "ip6 udp $mac gso" + $0 ipv6 ip6udp $mac 2000 + done
echo "OK. All tests passed" exit 0 fi
-if [[ "$#" -ne "3" ]]; then +if [[ "$#" -ne "4" ]]; then echo "Usage: $0" - echo " or: $0 <ipv4|ipv6> <tuntype> <data_len>" + echo " or: $0 <ipv4|ipv6> <tuntype> <none|mpls|eth> <data_len>" exit 1 fi
@@ -137,6 +142,8 @@ case "$1" in readonly foumod=fou readonly foutype=ipip readonly fouproto=4 + readonly fouproto_mpls=${mplsproto} + readonly gretaptype=gretap ;; "ipv6") readonly addr1="${ns1_v6}" @@ -146,6 +153,8 @@ case "$1" in readonly foumod=fou6 readonly foutype=ip6tnl readonly fouproto="41 -6" + readonly fouproto_mpls="${mplsproto} -6" + readonly gretaptype=ip6gretap ;; *) echo "unknown arg: $1" @@ -154,9 +163,10 @@ case "$1" in esac
readonly tuntype=$2 -readonly datalen=$3 +readonly mac=$3 +readonly datalen=$4
-echo "encap ${addr1} to ${addr2}, type ${tuntype}, len ${datalen}" +echo "encap ${addr1} to ${addr2}, type ${tuntype}, mac ${mac} len ${datalen}"
trap cleanup EXIT
@@ -173,7 +183,7 @@ verify_data ip netns exec "${ns1}" tc qdisc add dev veth1 clsact ip netns exec "${ns1}" tc filter add dev veth1 egress \ bpf direct-action object-file ./test_tc_tunnel.o \ - section "encap_${tuntype}" + section "encap_${tuntype}_${mac}" echo "test bpf encap without decap (expect failure)" server_listen ! client_connect @@ -184,7 +194,18 @@ if [[ "$tuntype" =~ "udp" ]]; then targs="encap fou encap-sport auto encap-dport $udpport" # fou may be a module; allow this to fail. modprobe "${foumod}" ||true - ip netns exec "${ns2}" ip fou add port 5555 ipproto ${fouproto} + if [[ "$mac" == "mpls" ]]; then + dport=${mplsudpport} + dproto=${fouproto_mpls} + tmode="mode any ttl 255" + else + dport=${udpport} + dproto=${fouproto} + fi + ip netns exec "${ns2}" ip fou add port $dport ipproto ${dproto} + targs="encap fou encap-sport auto encap-dport $dport" +elif [[ "$tuntype" =~ "gre" && "$mac" == "eth" ]]; then + ttype=$gretaptype else ttype=$tuntype targs="" @@ -194,7 +215,31 @@ fi # server is still running # client can connect again ip netns exec "${ns2}" ip link add name testtun0 type "${ttype}" \ - remote "${addr1}" local "${addr2}" $targs + ${tmode} remote "${addr1}" local "${addr2}" $targs + +expect_tun_fail=0 + +if [[ "$tuntype" == "ip6udp" && "$mac" == "mpls" ]]; then + # No support for MPLS IPv6 fou tunnel; expect failure. + expect_tun_fail=1 +elif [[ "$tuntype" =~ "udp" && "$mac" == "eth" ]]; then + # No support for TEB fou tunnel; expect failure. + expect_tun_fail=1 +elif [[ "$tuntype" =~ "gre" && "$mac" == "eth" ]]; then + # Share ethernet address between tunnel/veth2 so L2 decap works. + ethaddr=$(ip netns exec "${ns2}" ip link show veth2 | \ + awk '/ether/ { print $2 }') + ip netns exec "${ns2}" ip link set testtun0 address $ethaddr +elif [[ "$mac" == "mpls" ]]; then + modprobe mpls_iptunnel ||true + modprobe mpls_gso ||true + ip netns exec "${ns2}" sysctl -qw net.mpls.platform_labels=65536 + ip netns exec "${ns2}" ip -f mpls route add 1000 dev lo + ip netns exec "${ns2}" ip link set lo up + ip netns exec "${ns2}" sysctl -qw net.mpls.conf.testtun0.input=1 + ip netns exec "${ns2}" sysctl -qw net.ipv4.conf.lo.rp_filter=0 +fi + # Because packets are decapped by the tunnel they arrive on testtun0 from # the IP stack perspective. Ensure reverse path filtering is disabled # otherwise we drop the TCP SYN as arriving on testtun0 instead of the @@ -204,16 +249,22 @@ ip netns exec "${ns2}" sysctl -qw net.ipv4.conf.all.rp_filter=0 # selected as the max of the "all" and device-specific values. ip netns exec "${ns2}" sysctl -qw net.ipv4.conf.testtun0.rp_filter=0 ip netns exec "${ns2}" ip link set dev testtun0 up -echo "test bpf encap with tunnel device decap" -client_connect -verify_data +if [[ "$expect_tun_fail" == 1 ]]; then + # This tunnel mode is not supported, so we expect failure. + echo "test bpf encap with tunnel device decap (expect failure)" + ! client_connect +else + echo "test bpf encap with tunnel device decap" + client_connect + verify_data + server_listen +fi
# serverside, use BPF for decap ip netns exec "${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 verify_data
On Tue, Apr 9, 2019 at 10:08 AM Alan Maguire alan.maguire@oracle.com wrote:
Extend bpf_skb_adjust_room growth to mark inner MAC header so that L2 encapsulation can be used for tc tunnels.
Patch #1 extends the existing test_tc_tunnel to support UDP encapsulation; later we want to be able to test MPLS over UDP and MPLS over GRE encapsulation.
Patch #2 adds the BPF_F_ADJ_ROOM_ENCAP_L2(len) macro, which allows specification of inner mac length. Other approaches were explored prior to taking this approach. Specifically, I tried automatically computing the inner mac length on the basis of the specified flags (so inner maclen for GRE/IPv4 encap is the len_diff specified to bpf_skb_adjust_room minus GRE + IPv4 header length for example). Problem with this is that we don't know for sure what form of GRE/UDP header we have; is it a full GRE header, or is it a FOU UDP header or generic UDP encap header? My fear here was we'd end up with an explosion of flags. The other approach tried was to support inner L2 header marking as a separate room adjustment, i.e. adjust for L3/L4 encap, then call bpf_skb_adjust_room for L2 encap. This can be made to work but because it imposed an order on operations, felt a bit clunky.
Patch #3 syncs tools/ bpf.h.
Patch #4 extends the tests again to support MPLSoverGRE, MPLSoverUDP, and transparent ethernet bridging (TEB) where the inner L2 header is an ethernet header. Testing of BPF encap against tunnels is done for cases where configuration of such tunnels is possible (MPLSoverGRE[6], MPLSoverUDP, gre[6]tap), and skipped otherwise. Testing of BPF encap/decap is always carried out.
Changes since v2:
- updated tools/testing/selftest/bpf/config with FOU/MPLS CONFIG variables (patches 1, 4)
- reduced noise in patch 1 by avoiding unnecessary movement of code
- eliminated inner_mac variable in bpf_skb_net_grow (patch 2)
Changes since v1:
- fixed formatting of commit references.
- BPF_F_ADJ_ROOM_FIXED_GSO flag enabled on all variants (patch 1)
- fixed fou6 options for UDP encap; checksum errors observed were due to the fact fou6 tunnel was not set up with correct ipproto options (41 -6). 0 checksums work fine (patch 1)
- added definitions for mask and shift used in setting L2 length (patch 2)
- allow udp encap with fixed GSO (patch 2)
- changed "elen" to "l2_len" to be more descriptive (patch 4)
Alan Maguire (4): selftests_bpf: extend test_tc_tunnel for UDP encap bpf: add layer 2 encap support to bpf_skb_adjust_room bpf: sync bpf.h to tools/ for BPF_F_ADJ_ROOM_ENCAP_L2 selftests_bpf: add L2 encap to test_tc_tunnel
include/uapi/linux/bpf.h | 10 + net/core/filter.c | 12 +- tools/include/uapi/linux/bpf.h | 10 + tools/testing/selftests/bpf/config | 8 + tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 321 +++++++++++++++++---- tools/testing/selftests/bpf/test_tc_tunnel.sh | 136 +++++++-- 6 files changed, 417 insertions(+), 80 deletions(-)
For the series:
Acked-by: Willem de Bruijn willemb@google.com
Thanks Alan.
(quick response due to iterative review: checked out both v2 and v3 series from patchwork (a super useful feature!) and did a git diff)
On 04/09/2019 04:34 PM, Willem de Bruijn wrote:
On Tue, Apr 9, 2019 at 10:08 AM Alan Maguire alan.maguire@oracle.com wrote:
Extend bpf_skb_adjust_room growth to mark inner MAC header so that L2 encapsulation can be used for tc tunnels.
Patch #1 extends the existing test_tc_tunnel to support UDP encapsulation; later we want to be able to test MPLS over UDP and MPLS over GRE encapsulation.
Patch #2 adds the BPF_F_ADJ_ROOM_ENCAP_L2(len) macro, which allows specification of inner mac length. Other approaches were explored prior to taking this approach. Specifically, I tried automatically computing the inner mac length on the basis of the specified flags (so inner maclen for GRE/IPv4 encap is the len_diff specified to bpf_skb_adjust_room minus GRE + IPv4 header length for example). Problem with this is that we don't know for sure what form of GRE/UDP header we have; is it a full GRE header, or is it a FOU UDP header or generic UDP encap header? My fear here was we'd end up with an explosion of flags. The other approach tried was to support inner L2 header marking as a separate room adjustment, i.e. adjust for L3/L4 encap, then call bpf_skb_adjust_room for L2 encap. This can be made to work but because it imposed an order on operations, felt a bit clunky.
Patch #3 syncs tools/ bpf.h.
Patch #4 extends the tests again to support MPLSoverGRE, MPLSoverUDP, and transparent ethernet bridging (TEB) where the inner L2 header is an ethernet header. Testing of BPF encap against tunnels is done for cases where configuration of such tunnels is possible (MPLSoverGRE[6], MPLSoverUDP, gre[6]tap), and skipped otherwise. Testing of BPF encap/decap is always carried out.
Changes since v2:
- updated tools/testing/selftest/bpf/config with FOU/MPLS CONFIG variables (patches 1, 4)
- reduced noise in patch 1 by avoiding unnecessary movement of code
- eliminated inner_mac variable in bpf_skb_net_grow (patch 2)
Changes since v1:
- fixed formatting of commit references.
- BPF_F_ADJ_ROOM_FIXED_GSO flag enabled on all variants (patch 1)
- fixed fou6 options for UDP encap; checksum errors observed were due to the fact fou6 tunnel was not set up with correct ipproto options (41 -6). 0 checksums work fine (patch 1)
- added definitions for mask and shift used in setting L2 length (patch 2)
- allow udp encap with fixed GSO (patch 2)
- changed "elen" to "l2_len" to be more descriptive (patch 4)
Alan Maguire (4): selftests_bpf: extend test_tc_tunnel for UDP encap bpf: add layer 2 encap support to bpf_skb_adjust_room bpf: sync bpf.h to tools/ for BPF_F_ADJ_ROOM_ENCAP_L2 selftests_bpf: add L2 encap to test_tc_tunnel
include/uapi/linux/bpf.h | 10 + net/core/filter.c | 12 +- tools/include/uapi/linux/bpf.h | 10 + tools/testing/selftests/bpf/config | 8 + tools/testing/selftests/bpf/progs/test_tc_tunnel.c | 321 +++++++++++++++++---- tools/testing/selftests/bpf/test_tc_tunnel.sh | 136 +++++++-- 6 files changed, 417 insertions(+), 80 deletions(-)
For the series:
Acked-by: Willem de Bruijn willemb@google.com
Thanks Alan.
(quick response due to iterative review: checked out both v2 and v3 series from patchwork (a super useful feature!) and did a git diff)
Applied, thanks! Small follow-up request though, will comment in patch 2.
linux-kselftest-mirror@lists.linaro.org