netfilter: handle ipv6 jumbo packets properly for bridge ovs and tc
Subsystems (5)
v2 cover netfilter: handle ipv6 jumbo packets properly for bridge ovs and tc
Currently pskb_trim_rcsum() is always done on the RX path. However, IPv6 jumbo packets hide the real packet len in the Hop-by-hop option header, which should be parsed before doing the trim.
In ip6_rcv_core() it calls ipv6_parse_hopopts() to handle the Hop-by-hop option header then do pskb_trim_rcsum(). The similar process should also be done properly before pskb_trim_rcsum() on the RX path of bridge and openvswitch and tc.
This patchset improves the function handling the Hop-by-hop option header in bridge, and moves this function into netfilter utils, and then uses it in nf_conntrack_ovs for openvswitch and and tc.
Note that this patch is especially needed after the IPv6 BIG TCP was supported in kernel, which is using IPv6 Jumbo packets, and the last patch adds a big tcp selftest, which also covers it.
v1->v2:
- use the proper cast type in Patch 1, as Simon suggested.
- simplify the return path in Patch 2, as Simon suggested.
- move pkt_len definition into a smaller scope in Patch 3, as Simon suggested.
- move err definition into a smaller scope in Patch 5, as Simon suggested.
Xin Long (6):
netfilter: bridge: call pskb_may_pull in br_nf_check_hbh_len
netfilter: bridge: check len before accessing more nh data
netfilter: bridge: move pskb_trim_rcsum out of br_nf_check_hbh_len
netfilter: move br_nf_check_hbh_len to utils
netfilter: use nf_ip6_check_hbh_len in nf_ct_skb_network_trim
selftests: add a selftest for big tcp
include/linux/netfilter_ipv6.h | 2 + net/bridge/br_netfilter_ipv6.c | 79 ++--------- net/netfilter/nf_conntrack_ovs.c | 11 +- net/netfilter/utils.c | 52 +++++++ tools/testing/selftests/net/Makefile | 1 + tools/testing/selftests/net/big_tcp.sh | 180 +++++++++++++++++++++++++ 6 files changed, 254 insertions(+), 71 deletions(-) create mode 100755 tools/testing/selftests/net/big_tcp.sh
Signature · 1 line
2.39.1
PATCH v2 1/6 netfilter: bridge: call pskb_may_pull in br_nf_check_hbh_len
When checking Hop-by-hop option header, if the option data is in nonlinear area, it should do pskb_may_pull instead of discarding the skb as a bad IPv6 packet.
--- net/bridge/br_netfilter_ipv6.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)
net/bridge/br_netfilter_ipv6.c +9 -5
@@ -45,14 +45,18 @@*/ static int br_nf_check_hbh_len (struct sk_buff *skb) { -unsigned char *raw = (u8 *)(ipv6_hdr (skb) +1 ); +int len, off =sizeof (struct ipv6hdr ); +unsigned char *nh;u32 pkt_len; -const unsigned char *nh =skb_network_header (skb); -int off = raw - nh; -int len = (raw[1 ] +1 ) <<3 ; -if ((raw + len) - skb->data >skb_headlen (skb)) +if (!pskb_may_pull (skb, off +8 )) goto bad; + nh = (unsigned char *)(ipv6_hdr (skb) +1 ); + len = (nh[1 ] +1 ) <<3 ; + +if (!pskb_may_pull (skb, off + len)) + goto bad; + nh =skb_network_header (skb); off +=2 ; len -=2 ;
Signature · 1 line
2.39.1
PATCH v2 2/6 netfilter: bridge: check len before accessing more nh data
In reply to Xin LongIn the while loop of br_nf_check_hbh_len(), similar to ip6_parse_tlv(), before accessing 'nh[off + 1]', it should add a check 'len < 2'; and before parsing IPV6_TLV_JUMBO, it should add a check 'optlen > len', in case of overflows.
--- net/bridge/br_netfilter_ipv6.c | 45 +++++++++++++++------------------- 1 file changed, 20 insertions(+), 25 deletions(-)
net/bridge/br_netfilter_ipv6.c +20 -25
@@ -50,54 +50,49 @@ static int br_nf_check_hbh_len(struct sk_buff *skb)u32 pkt_len;if (!pskb_may_pull (skb, off +8 )) - goto bad; +return -1 ; nh = (unsigned char *)(ipv6_hdr (skb) +1 ); len = (nh[1 ] +1 ) <<3 ;if (!pskb_may_pull (skb, off + len)) - goto bad; +return -1 ; nh =skb_network_header (skb); off +=2 ; len -=2 ; -while (len >0 ) { -int optlen = nh[off +1 ] +2 ; - -switch (nh[off]) { -case IPV6_TLV_PAD1 : - optlen =1 ; -break ; +int optlen; -case IPV6_TLV_PADN : -break ; +if (nh[off] ==IPV6_TLV_PAD1 ) { + off++; + len--; +continue ; + } +if (len <2 ) +return -1 ; + optlen = nh[off +1 ] +2 ; +if (optlen > len) +return -1 ; -case IPV6_TLV_JUMBO : +if (nh[off] ==IPV6_TLV_JUMBO ) {if (nh[off +1 ] !=4 || (off &3 ) !=2 ) - goto bad; +return -1 ; pkt_len =ntohl (*(__be32 *)(nh + off +2 ));if (pkt_len <=IPV6_MAXPLEN ||ipv6_hdr (skb)->payload_len) - goto bad; +return -1 ;if (pkt_len > skb->len -sizeof (struct ipv6hdr )) - goto bad; +return -1 ;if (pskb_trim_rcsum (skb, pkt_len +sizeof (struct ipv6hdr ))) - goto bad; +return -1 ; nh =skb_network_header (skb); -break ; -default : -if (optlen > len) - goto bad; -break ; } off += optlen; len -= optlen; } -if (len ==0 ) -return 0 ; -bad: -return -1 ; + +return len ?-1 :0 ; }int br_validate_ipv6 (struct net *net,struct sk_buff *skb)
Signature · 1 line
2.39.1
PATCH v2 3/6 netfilter: bridge: move pskb_trim_rcsum out of br_nf_check_hbh_len
In reply to Xin Longbr_nf_check_hbh_len() is a function to check the Hop-by-hop option header, and shouldn't do pskb_trim_rcsum() there. This patch is to pass pkt_len out to br_validate_ipv6() and do pskb_trim_rcsum() after calling br_validate_ipv6() instead.
--- net/bridge/br_netfilter_ipv6.c | 33 ++++++++++++++------------------- 1 file changed, 14 insertions(+), 19 deletions(-)
net/bridge/br_netfilter_ipv6.c +14 -19
@@ -43,11 +43,10 @@/* We only check the length. A bridge shouldn't do any hop-by-hop stuff * anyway */ -static int br_nf_check_hbh_len (struct sk_buff *skb) +static int br_nf_check_hbh_len (struct sk_buff *skb,u32 *plen) {int len, off =sizeof (struct ipv6hdr );unsigned char *nh; -u32 pkt_len;if (!pskb_may_pull (skb, off +8 ))return -1 ;
@@ -75,6 +74,8 @@ static int br_nf_check_hbh_len(struct sk_buff *skb)return -1 ;if (nh[off] ==IPV6_TLV_JUMBO ) { +u32 pkt_len; +if (nh[off +1 ] !=4 || (off &3 ) !=2 )return -1 ; pkt_len =ntohl (*(__be32 *)(nh + off +2 ));
@@ -83,10 +84,7 @@ static int br_nf_check_hbh_len(struct sk_buff *skb)return -1 ;if (pkt_len > skb->len -sizeof (struct ipv6hdr ))return -1 ; -if (pskb_trim_rcsum (skb, - pkt_len +sizeof (struct ipv6hdr ))) -return -1 ; - nh =skb_network_header (skb); + *plen = pkt_len; } off += optlen; len -= optlen;
@@ -114,22 +112,19 @@ int br_validate_ipv6(struct net *net, struct sk_buff *skb) goto inhdr_error; pkt_len =ntohs (hdr->payload_len); +if (hdr->nexthdr ==NEXTHDR_HOP &&br_nf_check_hbh_len (skb, &pkt_len)) + goto drop; -if (pkt_len || hdr->nexthdr !=NEXTHDR_HOP ) { -if (pkt_len + ip6h_len > skb->len) { -__IP6_INC_STATS (net, idev, -IPSTATS_MIB_INTRUNCATEDPKTS ); - goto drop; - } -if (pskb_trim_rcsum (skb, pkt_len + ip6h_len)) { -__IP6_INC_STATS (net, idev, -IPSTATS_MIB_INDISCARDS ); - goto drop; - } - hdr =ipv6_hdr (skb); +if (pkt_len + ip6h_len > skb->len) { +__IP6_INC_STATS (net, idev, +IPSTATS_MIB_INTRUNCATEDPKTS ); + goto drop; } -if (hdr->nexthdr ==NEXTHDR_HOP &&br_nf_check_hbh_len (skb)) +if (pskb_trim_rcsum (skb, pkt_len + ip6h_len)) { +__IP6_INC_STATS (net, idev, +IPSTATS_MIB_INDISCARDS ); goto drop; + }memset (IP6CB (skb),0 ,sizeof (struct inet6_skb_parm ));/* No IP options in IPv6 header; however it should be
Signature · 1 line
2.39.1
PATCH v2 5/6 netfilter: use nf_ip6_check_hbh_len in nf_ct_skb_network_trim
In reply to Xin LongFor IPv6 Jumbo packets, the ipv6_hdr(skb)->payload_len is always 0, and its real payload_len ( > 65535) is saved in hbh exthdr. With 0 length for the jumbo packets, all data and exthdr will be trimmed in nf_ct_skb_network_trim().
This patch is to call nf_ip6_check_hbh_len() to get real pkt_len of the IPv6 packet, similar to br_validate_ipv6().
--- net/netfilter/nf_conntrack_ovs.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
net/netfilter/nf_conntrack_ovs.c +9 -2
@@ -6,6 +6,7 @@#include <net/netfilter/ipv6/nf_defrag_ipv6.h> #include <net/ipv6_frag.h> #include <net/ip.h> +#include <linux/netfilter_ipv6.h> /* 'skb' should already be pulled to nh_ofs. */ int nf_ct_helper(struct sk_buff *skb, struct nf_conn *ct,
@@ -120,8 +121,14 @@ int nf_ct_skb_network_trim(struct sk_buff *skb, int family) len =skb_ip_totlen (skb);break ;case NFPROTO_IPV6 : - len =sizeof (struct ipv6hdr ) - +ntohs (ipv6_hdr (skb)->payload_len); + len =ntohs (ipv6_hdr (skb)->payload_len); +if (ipv6_hdr (skb)->nexthdr ==NEXTHDR_HOP ) { +int err =nf_ip6_check_hbh_len (skb, &len); + +if (err) +return err; + } + len +=sizeof (struct ipv6hdr );break ;default : len = skb->len;
Signature · 1 line
2.39.1
PATCH v2 4/6 netfilter: move br_nf_check_hbh_len to utils
In reply to Xin LongRename br_nf_check_hbh_len() to nf_ip6_check_hbh_len() and move it to netfilter utils, so that it can be used by other modules, like ovs and tc.
--- include/linux/netfilter_ipv6.h | 2 ++ net/bridge/br_netfilter_ipv6.c | 55 +--------------------------------- net/netfilter/utils.c | 52 ++++++++++++++++++++++++++++++++ 3 files changed, 55 insertions(+), 54 deletions(-)
include/linux/netfilter_ipv6.h +2 -0
@@ -197,6 +197,8 @@ static inline int nf_cookie_v6_check(const struct ipv6hdr *iph,__sum16 nf_ip6_checksum (struct sk_buff *skb,unsigned int hook,unsigned int dataoff,u_int8_t protocol); +int nf_ip6_check_hbh_len (struct sk_buff *skb,u32 *plen); +int ipv6_netfilter_init (void );void ipv6_netfilter_fini (void );
net/bridge/br_netfilter_ipv6.c +1 -54
67 lines not on this page · raw message
net/netfilter/utils.c +52 -0
55 lines not on this page · raw message
Signature · 1 line
2.39.1
PATCH v2 6/6 selftests: add a selftest for big tcp
In reply to Xin LongThis test runs on the client-router-server topo, and monitors the traffic on the RX devices of router and server while sending BIG TCP packets with netperf from client to server. Meanwhile, it changes 'tso' on the TX devs and 'gro' on the RX devs. Then it checks if any BIG TCP packets appears on the RX devs with 'ip/ip6tables -m length ! --length 0:65535' for each case.
Note that we also add tc action ct in link1 ingress to cover the ipv6 jumbo packets process in nf_ct_skb_network_trim() of nf_conntrack_ovs.
--- tools/testing/selftests/net/Makefile | 1 + tools/testing/selftests/net/big_tcp.sh | 180 +++++++++++++++++++++++++ 2 files changed, 181 insertions(+) create mode 100755 tools/testing/selftests/net/big_tcp.sh
tools/testing/selftests/net/Makefile +1 -0
@@ -48,6 +48,7 @@ TEST_PROGS += l2_tos_ttl_inherit.sh TEST_PROGS += bind_bhash.sh TEST_PROGS += ip_local_port_range.sh TEST_PROGS += rps_default_mask.sh +TEST_PROGS += big_tcp.sh TEST_PROGS_EXTENDED := in_netns.sh setup_loopback.sh setup_veth.sh TEST_PROGS_EXTENDED += toeplitz_client.sh toeplitz.sh TEST_GEN_FILES = socket nettest
tools/testing/selftests/net/big_tcp.sh new file +180 -0
180 lines not on this page · raw message
Signature · 1 line
2.39.1
Re: [PATCHv2 nf-next 6/6] selftests: add a selftest for big tcp
Quoted 10 lines · Xin Long
This test runs on the client-router-server topo, and monitors the traffic on the RX devices of router and server while sending BIG TCP packets with netperf from client to server. Meanwhile, it changes 'tso' on the TX devs and 'gro' on the RX devs. Then it checks if any BIG TCP packets appears on the RX devs with 'ip/ip6tables -m length ! --length 0:65535' for each case.
Note that we also add tc action ct in link1 ingress to cover the ipv6 jumbo packets process in
nf_ct_skb_network_trim()ofnf_conntrack_ovs.
Quoted 7 lines
Signed-off-by Xin Long <[email protected]>Reviewed-by Aaron Conole <[email protected]>--- tools/testing/selftests/net/Makefile | 1 + tools/testing/selftests/net/big_tcp.sh | 180 +++++++++++++++++++++++++ 2 files changed, 181 insertions(+) create mode 100755 tools/testing/selftests/net/big_tcp.sh