From 9f9818f8c1cf44055634297247620be4755e7af2 Mon Sep 17 00:00:00 2001 From: Jesse Gross Date: Sat, 19 Mar 2016 09:32:02 -0700 Subject: tunnels: Remove encapsulation offloads on decap. commit a09a4c8dd1ec7f830e1fb9e59eb72bddc965d168 upstream. If a packet is either locally encapsulated or processed through GRO it is marked with the offloads that it requires. However, when it is decapsulated these tunnel offload indications are not removed. This means that if we receive an encapsulated TCP packet, aggregate it with GRO, decapsulate, and retransmit the resulting frame on a NIC that does not support encapsulation, we won't be able to take advantage of hardware offloads even though it is just a simple TCP packet at this point. This fixes the problem by stripping off encapsulation offload indications when packets are decapsulated. The performance impacts of this bug are significant. In a test where a Geneve encapsulated TCP stream is sent to a hypervisor, GRO'ed, decapsulated, and bridged to a VM performance is improved by 60% (5Gbps->8Gbps) as a result of avoiding unnecessary segmentation at the VM tap interface. Reported-by: Ramu Ramamurthy Fixes: 68c33163 ("v4 GRE: Add TCP segmentation offload for GRE") Signed-off-by: Jesse Gross Signed-off-by: David S. Miller (backported from commit a09a4c8dd1ec7f830e1fb9e59eb72bddc965d168) [adapt iptunnel_pull_header arguments, avoid 7f290c9] Signed-off-by: Stefan Bader Signed-off-by: Juerg Haefliger Signed-off-by: Greg Kroah-Hartman --- include/net/ip_tunnels.h | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'include/net') diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h index af40bc586a1b..86a7bdd61d1a 100644 --- a/include/net/ip_tunnels.h +++ b/include/net/ip_tunnels.h @@ -283,6 +283,22 @@ struct metadata_dst *iptunnel_metadata_reply(struct metadata_dst *md, struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb, bool gre_csum, int gso_type_mask); +static inline int iptunnel_pull_offloads(struct sk_buff *skb) +{ + if (skb_is_gso(skb)) { + int err; + + err = skb_unclone(skb, GFP_ATOMIC); + if (unlikely(err)) + return err; + skb_shinfo(skb)->gso_type &= ~(NETIF_F_GSO_ENCAP_ALL >> + NETIF_F_GSO_SHIFT); + } + + skb->encapsulation = 0; + return 0; +} + static inline void iptunnel_xmit_stats(int err, struct net_device_stats *err_stats, struct pcpu_sw_netstats __percpu *stats) -- cgit v1.2.3 From d2e01b15657c394085fd810f4da6b5ef6574e14b Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 15 Sep 2016 08:48:46 -0700 Subject: net: avoid sk_forward_alloc overflows [ Upstream commit 20c64d5cd5a2bdcdc8982a06cb05e5e1bd851a3d ] A malicious TCP receiver, sending SACK, can force the sender to split skbs in write queue and increase its memory usage. Then, when socket is closed and its write queue purged, we might overflow sk_forward_alloc (It becomes negative) sk_mem_reclaim() does nothing in this case, and more than 2GB are leaked from TCP perspective (tcp_memory_allocated is not changed) Then warnings trigger from inet_sock_destruct() and sk_stream_kill_queues() seeing a not zero sk_forward_alloc All TCP stack can be stuck because TCP is under memory pressure. A simple fix is to preemptively reclaim from sk_mem_uncharge(). This makes sure a socket wont have more than 2 MB forward allocated, after burst and idle period. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller Signed-off-by: Greg Kroah-Hartman --- include/net/sock.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'include/net') diff --git a/include/net/sock.h b/include/net/sock.h index 14d3c0734007..3d5ff7436f41 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1425,6 +1425,16 @@ static inline void sk_mem_uncharge(struct sock *sk, int size) if (!sk_has_account(sk)) return; sk->sk_forward_alloc += size; + + /* Avoid a possible overflow. + * TCP send queues can make this happen, if sk_mem_reclaim() + * is not called and more than 2 GBytes are released at once. + * + * If we reach 2 MBytes, reclaim 1 MBytes right now, there is + * no need to hold that much forward allocation anyway. + */ + if (unlikely(sk->sk_forward_alloc >= 1 << 21)) + __sk_mem_reclaim(sk, 1 << 20); } static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) -- cgit v1.2.3 From 9edbf4a0b60b62a1fb5f57248f6c9b9ffb30c328 Mon Sep 17 00:00:00 2001 From: Shmulik Ladkani Date: Thu, 29 Sep 2016 12:10:40 +0300 Subject: net/sched: act_vlan: Push skb->data to mac_header prior calling skb_vlan_*() functions [ Upstream commit f39acc84aad10710e89835c60d3b6694c43a8dd9 ] Generic skb_vlan_push/skb_vlan_pop functions don't properly handle the case where the input skb data pointer does not point at the mac header: - They're doing push/pop, but fail to properly unwind data back to its original location. For example, in the skb_vlan_push case, any subsequent 'skb_push(skb, skb->mac_len)' calls make the skb->data point 4 bytes BEFORE start of frame, leading to bogus frames that may be transmitted. - They update rcsum per the added/removed 4 bytes tag. Alas if data is originally after the vlan/eth headers, then these bytes were already pulled out of the csum. OTOH calling skb_vlan_push/skb_vlan_pop with skb->data at mac_header present no issues. act_vlan is the only caller to skb_vlan_*() that has skb->data pointing at network header (upon ingress). Other calles (ovs, bpf) already adjust skb->data at mac_header. This patch fixes act_vlan to point to the mac_header prior calling skb_vlan_*() functions, as other callers do. Signed-off-by: Shmulik Ladkani Cc: Daniel Borkmann Cc: Pravin Shelar Cc: Jiri Pirko Signed-off-by: David S. Miller Signed-off-by: Greg Kroah-Hartman --- include/net/sch_generic.h | 9 +++++++++ net/sched/act_vlan.c | 9 +++++++++ 2 files changed, 18 insertions(+) (limited to 'include/net') diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h index 86df0835f6b5..e5bba897d206 100644 --- a/include/net/sch_generic.h +++ b/include/net/sch_generic.h @@ -408,6 +408,15 @@ bool tcf_destroy(struct tcf_proto *tp, bool force); void tcf_destroy_chain(struct tcf_proto __rcu **fl); int skb_do_redirect(struct sk_buff *); +static inline bool skb_at_tc_ingress(const struct sk_buff *skb) +{ +#ifdef CONFIG_NET_CLS_ACT + return G_TC_AT(skb->tc_verd) & AT_INGRESS; +#else + return false; +#endif +} + /* Reset all TX qdiscs greater then index of a device. */ static inline void qdisc_reset_all_tx_gt(struct net_device *dev, unsigned int i) { diff --git a/net/sched/act_vlan.c b/net/sched/act_vlan.c index 796785e0bf96..d7edba4536bd 100644 --- a/net/sched/act_vlan.c +++ b/net/sched/act_vlan.c @@ -33,6 +33,12 @@ static int tcf_vlan(struct sk_buff *skb, const struct tc_action *a, bstats_update(&v->tcf_bstats, skb); action = v->tcf_action; + /* Ensure 'data' points at mac_header prior calling vlan manipulating + * functions. + */ + if (skb_at_tc_ingress(skb)) + skb_push_rcsum(skb, skb->mac_len); + switch (v->tcfv_action) { case TCA_VLAN_ACT_POP: err = skb_vlan_pop(skb); @@ -54,6 +60,9 @@ drop: action = TC_ACT_SHOT; v->tcf_qstats.drops++; unlock: + if (skb_at_tc_ingress(skb)) + skb_pull_rcsum(skb, skb->mac_len); + spin_unlock(&v->tcf_lock); return action; } -- cgit v1.2.3 From d46c76765da696502837d823227d4c32c28d8c05 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 23 Oct 2016 18:03:06 -0700 Subject: udp: fix IP_CHECKSUM handling [ Upstream commit 10df8e6152c6c400a563a673e9956320bfce1871 ] First bug was added in commit ad6f939ab193 ("ip: Add offset parameter to ip_cmsg_recv") : Tom missed that ipv4 udp messages could be received on AF_INET6 socket. ip_cmsg_recv(msg, skb) should have been replaced by ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr)); Then commit e6afc8ace6dd ("udp: remove headers from UDP packets before queueing") forgot to adjust the offsets now UDP headers are pulled before skb are put in receive queue. Fixes: ad6f939ab193 ("ip: Add offset parameter to ip_cmsg_recv") Fixes: e6afc8ace6dd ("udp: remove headers from UDP packets before queueing") Signed-off-by: Eric Dumazet Cc: Sam Kumar Cc: Willem de Bruijn Tested-by: Willem de Bruijn Signed-off-by: David S. Miller Signed-off-by: Greg Kroah-Hartman --- include/net/ip.h | 4 ++-- net/ipv4/ip_sockglue.c | 10 ++++++---- net/ipv4/udp.c | 2 +- net/ipv6/udp.c | 3 ++- 4 files changed, 11 insertions(+), 8 deletions(-) (limited to 'include/net') diff --git a/include/net/ip.h b/include/net/ip.h index 1a98f1ca1638..b450d8653b30 100644 --- a/include/net/ip.h +++ b/include/net/ip.h @@ -553,7 +553,7 @@ int ip_options_rcv_srr(struct sk_buff *skb); */ void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb); -void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, int offset); +void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, int tlen, int offset); int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc, bool allow_ipv6); int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval, @@ -575,7 +575,7 @@ void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport, static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb) { - ip_cmsg_recv_offset(msg, skb, 0); + ip_cmsg_recv_offset(msg, skb, 0, 0); } bool icmp_global_allow(void); diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c index a50124260f5a..9ce202549e7a 100644 --- a/net/ipv4/ip_sockglue.c +++ b/net/ipv4/ip_sockglue.c @@ -98,7 +98,7 @@ static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) } static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, - int offset) + int tlen, int offset) { __wsum csum = skb->csum; @@ -106,7 +106,9 @@ static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, return; if (offset != 0) - csum = csum_sub(csum, csum_partial(skb->data, offset, 0)); + csum = csum_sub(csum, + csum_partial(skb->data + tlen, + offset, 0)); put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum); } @@ -152,7 +154,7 @@ static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) } void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, - int offset) + int tlen, int offset) { struct inet_sock *inet = inet_sk(skb->sk); unsigned int flags = inet->cmsg_flags; @@ -215,7 +217,7 @@ void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, } if (flags & IP_CMSG_CHECKSUM) - ip_cmsg_recv_checksum(msg, skb, offset); + ip_cmsg_recv_checksum(msg, skb, tlen, offset); } EXPORT_SYMBOL(ip_cmsg_recv_offset); diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c index 0b1ea5abcc04..e9513e397c4f 100644 --- a/net/ipv4/udp.c +++ b/net/ipv4/udp.c @@ -1342,7 +1342,7 @@ try_again: *addr_len = sizeof(*sin); } if (inet->cmsg_flags) - ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr)); + ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr), off); err = copied; if (flags & MSG_TRUNC) diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c index e6092bd72ee2..dfa85e7264df 100644 --- a/net/ipv6/udp.c +++ b/net/ipv6/udp.c @@ -498,7 +498,8 @@ try_again: if (is_udp4) { if (inet->cmsg_flags) - ip_cmsg_recv(msg, skb); + ip_cmsg_recv_offset(msg, skb, + sizeof(struct udphdr), off); } else { if (np->rxopt.all) ip6_datagram_recv_specific_ctl(sk, msg, skb); -- cgit v1.2.3