From 69fec325a64383667b8a35df5d48d6ce52fb2782 Mon Sep 17 00:00:00 2001 From: Xin Long Date: Sun, 18 Nov 2018 16:14:47 +0800 Subject: Revert "sctp: remove sctp_transport_pmtu_check" This reverts commit 22d7be267eaa8114dcc28d66c1c347f667d7878a. The dst's mtu in transport can be updated by a non sctp place like in xfrm where the MTU information didn't get synced between asoc, transport and dst, so it is still needed to do the pmtu check in sctp_packet_config. Acked-by: Neil Horman Signed-off-by: David S. Miller --- include/net/sctp/sctp.h | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'include') diff --git a/include/net/sctp/sctp.h b/include/net/sctp/sctp.h index 8c2caa370e0f..ab9242e51d9e 100644 --- a/include/net/sctp/sctp.h +++ b/include/net/sctp/sctp.h @@ -608,4 +608,16 @@ static inline __u32 sctp_dst_mtu(const struct dst_entry *dst) SCTP_DEFAULT_MINSEGMENT)); } +static inline bool sctp_transport_pmtu_check(struct sctp_transport *t) +{ + __u32 pmtu = sctp_dst_mtu(t->dst); + + if (t->pathmtu == pmtu) + return true; + + t->pathmtu = pmtu; + + return false; +} + #endif /* __net_sctp_h__ */ -- cgit v1.2.3 From 86de5921a3d5dd246df661e09bdd0a6131b39ae3 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 20 Nov 2018 05:53:59 -0800 Subject: tcp: defer SACK compression after DupThresh Jean-Louis reported a TCP regression and bisected to recent SACK compression. After a loss episode (receiver not able to keep up and dropping packets because its backlog is full), linux TCP stack is sending a single SACK (DUPACK). Sender waits a full RTO timer before recovering losses. While RFC 6675 says in section 5, "Algorithm Details", (2) If DupAcks < DupThresh but IsLost (HighACK + 1) returns true -- indicating at least three segments have arrived above the current cumulative acknowledgment point, which is taken to indicate loss -- go to step (4). ... (4) Invoke fast retransmit and enter loss recovery as follows: there are old TCP stacks not implementing this strategy, and still counting the dupacks before starting fast retransmit. While these stacks probably perform poorly when receivers implement LRO/GRO, we should be a little more gentle to them. This patch makes sure we do not enable SACK compression unless 3 dupacks have been sent since last rcv_nxt update. Ideally we should even rearm the timer to send one or two more DUPACK if no more packets are coming, but that will be work aiming for linux-4.21. Many thanks to Jean-Louis for bisecting the issue, providing packet captures and testing this patch. Fixes: 5d9f4262b7ea ("tcp: add SACK compression") Reported-by: Jean-Louis Dupond Tested-by: Jean-Louis Dupond Signed-off-by: Eric Dumazet Acked-by: Neal Cardwell Signed-off-by: David S. Miller --- include/linux/tcp.h | 1 + net/ipv4/tcp_input.c | 14 ++++++++++++-- net/ipv4/tcp_output.c | 6 +++--- net/ipv4/tcp_timer.c | 2 +- 4 files changed, 17 insertions(+), 6 deletions(-) (limited to 'include') diff --git a/include/linux/tcp.h b/include/linux/tcp.h index 8ed77bb4ed86..a9b0280687d5 100644 --- a/include/linux/tcp.h +++ b/include/linux/tcp.h @@ -196,6 +196,7 @@ struct tcp_sock { u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */ u32 lsndtime; /* timestamp of last sent data packet (for restart window) */ u32 last_oow_ack_time; /* timestamp of last out-of-window ACK */ + u32 compressed_ack_rcv_nxt; u32 tsoffset; /* timestamp offset */ diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c index e695584bb33f..1e37c1388189 100644 --- a/net/ipv4/tcp_input.c +++ b/net/ipv4/tcp_input.c @@ -4268,7 +4268,7 @@ static void tcp_sack_new_ofo_skb(struct sock *sk, u32 seq, u32 end_seq) * If the sack array is full, forget about the last one. */ if (this_sack >= TCP_NUM_SACKS) { - if (tp->compressed_ack) + if (tp->compressed_ack > TCP_FASTRETRANS_THRESH) tcp_send_ack(sk); this_sack--; tp->rx_opt.num_sacks--; @@ -5189,7 +5189,17 @@ send_now: if (!tcp_is_sack(tp) || tp->compressed_ack >= sock_net(sk)->ipv4.sysctl_tcp_comp_sack_nr) goto send_now; - tp->compressed_ack++; + + if (tp->compressed_ack_rcv_nxt != tp->rcv_nxt) { + tp->compressed_ack_rcv_nxt = tp->rcv_nxt; + if (tp->compressed_ack > TCP_FASTRETRANS_THRESH) + NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPACKCOMPRESSED, + tp->compressed_ack - TCP_FASTRETRANS_THRESH); + tp->compressed_ack = 0; + } + + if (++tp->compressed_ack <= TCP_FASTRETRANS_THRESH) + goto send_now; if (hrtimer_is_queued(&tp->compressed_ack_timer)) return; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 9c34b97d365d..3f510cad0b3e 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -180,10 +180,10 @@ static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts, { struct tcp_sock *tp = tcp_sk(sk); - if (unlikely(tp->compressed_ack)) { + if (unlikely(tp->compressed_ack > TCP_FASTRETRANS_THRESH)) { NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPACKCOMPRESSED, - tp->compressed_ack); - tp->compressed_ack = 0; + tp->compressed_ack - TCP_FASTRETRANS_THRESH); + tp->compressed_ack = TCP_FASTRETRANS_THRESH; if (hrtimer_try_to_cancel(&tp->compressed_ack_timer) == 1) __sock_put(sk); } diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 676020663ce8..5f8b6d3cd855 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -740,7 +740,7 @@ static enum hrtimer_restart tcp_compressed_ack_kick(struct hrtimer *timer) bh_lock_sock(sk); if (!sock_owned_by_user(sk)) { - if (tp->compressed_ack) + if (tp->compressed_ack > TCP_FASTRETRANS_THRESH) tcp_send_ack(sk); } else { if (!test_and_set_bit(TCP_DELACK_TIMER_DEFERRED, -- cgit v1.2.3 From 0211dda68a4f6531923a2f72d8e8959207f59fba Mon Sep 17 00:00:00 2001 From: Tal Gilboa Date: Wed, 21 Nov 2018 16:28:23 +0200 Subject: net/dim: Update DIM start sample after each DIM iteration On every iteration of net_dim, the algorithm may choose to check for the system state by comparing current data sample with previous data sample. After each of these comparison, regardless of the action taken, the sample used as baseline is needed to be updated. This patch fixes a bug that causes DIM to take wrong decisions, due to never updating the baseline sample for comparison between iterations. This way, DIM always compares current sample with zeros. Although this is a functional fix, it also improves and stabilizes performance as the algorithm works properly now. Performance: Tested single UDP TX stream with pktgen: samples/pktgen/pktgen_sample03_burst_single_flow.sh -i p4p2 -d 1.1.1.1 -m 24:8a:07:88:26:8b -f 3 -b 128 ConnectX-5 100GbE packet rate improved from 15-19Mpps to 19-20Mpps. Also, toggling between profiles is less frequent with the fix. Fixes: 8115b750dbcb ("net/dim: use struct net_dim_sample as arg to net_dim") Signed-off-by: Tal Gilboa Reviewed-by: Tariq Toukan Signed-off-by: David S. Miller --- include/linux/net_dim.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include') diff --git a/include/linux/net_dim.h b/include/linux/net_dim.h index c79e859408e6..fd458389f7d1 100644 --- a/include/linux/net_dim.h +++ b/include/linux/net_dim.h @@ -406,6 +406,8 @@ static inline void net_dim(struct net_dim *dim, } /* fall through */ case NET_DIM_START_MEASURE: + net_dim_sample(end_sample.event_ctr, end_sample.pkt_ctr, end_sample.byte_ctr, + &dim->start_sample); dim->state = NET_DIM_MEASURE_IN_PROGRESS; break; case NET_DIM_APPLY_NEW_PROFILE: -- cgit v1.2.3 From 5cd8d46ea1562be80063f53c7c6a5f40224de623 Mon Sep 17 00:00:00 2001 From: Willem de Bruijn Date: Tue, 20 Nov 2018 13:00:18 -0500 Subject: packet: copy user buffers before orphan or clone tpacket_snd sends packets with user pages linked into skb frags. It notifies that pages can be reused when the skb is released by setting skb->destructor to tpacket_destruct_skb. This can cause data corruption if the skb is orphaned (e.g., on transmit through veth) or cloned (e.g., on mirror to another psock). Create a kernel-private copy of data in these cases, same as tun/tap zerocopy transmission. Reuse that infrastructure: mark the skb as SKBTX_ZEROCOPY_FRAG, which will trigger copy in skb_orphan_frags(_rx). Unlike other zerocopy packets, do not set shinfo destructor_arg to struct ubuf_info. tpacket_destruct_skb already uses that ptr to notify when the original skb is released and a timestamp is recorded. Do not change this timestamp behavior. The ubuf_info->callback is not needed anyway, as no zerocopy notification is expected. Mark destructor_arg as not-a-uarg by setting the lower bit to 1. The resulting value is not a valid ubuf_info pointer, nor a valid tpacket_snd frame address. Add skb_zcopy_.._nouarg helpers for this. The fix relies on features introduced in commit 52267790ef52 ("sock: add MSG_ZEROCOPY"), so can be backported as is only to 4.14. Tested with from `./in_netns.sh ./txring_overwrite` from http://github.com/wdebruij/kerneltools/tests Fixes: 69e3c75f4d54 ("net: TX_RING and packet mmap") Reported-by: Anand H. Krishnan Signed-off-by: Willem de Bruijn Signed-off-by: David S. Miller --- include/linux/skbuff.h | 18 +++++++++++++++++- net/packet/af_packet.c | 4 ++-- 2 files changed, 19 insertions(+), 3 deletions(-) (limited to 'include') diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index 0ba687454267..0d1b2c3f127b 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -1326,6 +1326,22 @@ static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg) } } +static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) +{ + skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); + skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG; +} + +static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) +{ + return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; +} + +static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) +{ + return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); +} + /* Release a reference on a zerocopy structure */ static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy) { @@ -1335,7 +1351,7 @@ static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy) if (uarg->callback == sock_zerocopy_callback) { uarg->zerocopy = uarg->zerocopy && zerocopy; sock_zerocopy_put(uarg); - } else { + } else if (!skb_zcopy_is_nouarg(skb)) { uarg->callback(uarg, zerocopy); } diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c index ec3095f13aae..a74650e98f42 100644 --- a/net/packet/af_packet.c +++ b/net/packet/af_packet.c @@ -2394,7 +2394,7 @@ static void tpacket_destruct_skb(struct sk_buff *skb) void *ph; __u32 ts; - ph = skb_shinfo(skb)->destructor_arg; + ph = skb_zcopy_get_nouarg(skb); packet_dec_pending(&po->tx_ring); ts = __packet_set_timestamp(po, ph, skb); @@ -2461,7 +2461,7 @@ static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, skb->mark = po->sk.sk_mark; skb->tstamp = sockc->transmit_time; sock_tx_timestamp(&po->sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags); - skb_shinfo(skb)->destructor_arg = ph.raw; + skb_zcopy_set_nouarg(skb, ph.raw); skb_reserve(skb, hlen); skb_reset_network_header(skb); -- cgit v1.2.3