Linux TCP stack painfully segments all TSO/GSO packets before retransmits.

This was fine back in the days when TSO/GSO were emerging, with their
bugs, but we believe the dark age is over.

Keeping big packets in write queues, but also in stack traversal
has a lot of benefits.
 - Less memory overhead, because write queues have less skbs
 - Less cpu overhead at ACK processing.
 - Better SACK processing, as lot of studies mentioned how
   awful linux was at this ;)
 - Less cpu overhead to send the rtx packets
   (IP stack traversal, netfilter traversal, drivers...)
 - Better latencies in presence of losses.
 - Smaller spikes in fq like packet schedulers, as retransmits
   are not constrained by TCP Small Queues.

1 % packet losses are common today, and at 100Gbit speeds, this
translates to ~80,000 losses per second.
Losses are often correlated, and we see many retransmit events
leading to 1-MSS train of packets, at the time hosts are already
under stress.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After receiving sacks, tcp_shifted_skb() will collapse
skbs if possible.  tx_flags and tskey also have to be
merged.

This patch reuses the tcp_skb_collapse_tstamp() to handle
them.

BPF Output Before:
~~~~~
<no-output-due-to-missing-tstamp-event>

BPF Output After:
~~~~~
<...>-2024  [007] d.s.    88.644374: : ee_data:14599

Packetdrill Script:
~~~~~
+0 `sysctl -q -w net.ipv4.tcp_min_tso_segs=10`
+0 `sysctl -q -w net.ipv4.tcp_no_metrics_save=1`
+0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0

0.100 < S 0:0(0) win 32792 <mss 1460,sackOK,nop,nop,nop,wscale 7>
0.100 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 7>
0.200 < . 1:1(0) ack 1 win 257
0.200 accept(3, ..., ...) = 4
+0 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0

0.200 write(4, ..., 1460) = 1460
+0 setsockopt(4, SOL_SOCKET, 37, [2688], 4) = 0
0.200 write(4, ..., 13140) = 13140

0.200 > P. 1:1461(1460) ack 1
0.200 > . 1461:8761(7300) ack 1
0.200 > P. 8761:14601(5840) ack 1

0.300 < . 1:1(0) ack 1 win 257 <sack 1461:14601,nop,nop>
0.300 > P. 1:1461(1460) ack 1
0.400 < . 1:1(0) ack 14601 win 257

0.400 close(4) = 0
0.400 > F. 14601:14601(0) ack 1
0.500 < F. 1:1(0) ack 14602 win 257
0.500 > . 14602:14602(0) ack 2
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: NSoheil Hassas Yeganeh <soheil@google.com>
Tested-by: NSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Assuming SOF_TIMESTAMPING_TX_ACK is on. When dup acks are received,
it could incorrectly think that a skb has already
been acked and queue a SCM_TSTAMP_ACK cmsg to the
sk->sk_error_queue.

In tcp_ack_tstamp(), it checks
'between(shinfo->tskey, prior_snd_una, tcp_sk(sk)->snd_una - 1)'.
If prior_snd_una == tcp_sk(sk)->snd_una like the following packetdrill
script, between() returns true but the tskey is actually not acked.
e.g. try between(3, 2, 1).

The fix is to replace between() with one before() and one !before().
By doing this, the -1 offset on the tcp_sk(sk)->snd_una can also be
removed.

A packetdrill script is used to reproduce the dup ack scenario.
Due to the lacking cmsg support in packetdrill (may be I
cannot find it),  a BPF prog is used to kprobe to
sock_queue_err_skb() and print out the value of
serr->ee.ee_data.

Both the packetdrill and the bcc BPF script is attached at the end of
this commit message.

BPF Output Before Fix:
~~~~~~
      <...>-2056  [001] d.s.   433.927987: : ee_data:1459  #incorrect
packetdrill-2056  [001] d.s.   433.929563: : ee_data:1459  #incorrect
packetdrill-2056  [001] d.s.   433.930765: : ee_data:1459  #incorrect
packetdrill-2056  [001] d.s.   434.028177: : ee_data:1459
packetdrill-2056  [001] d.s.   434.029686: : ee_data:14599

BPF Output After Fix:
~~~~~~
      <...>-2049  [000] d.s.   113.517039: : ee_data:1459
      <...>-2049  [000] d.s.   113.517253: : ee_data:14599

BCC BPF Script:
~~~~~~
#!/usr/bin/env python

from __future__ import print_function
from bcc import BPF

bpf_text = """
#include <uapi/linux/ptrace.h>
#include <net/sock.h>
#include <bcc/proto.h>
#include <linux/errqueue.h>

#ifdef memset
#undef memset
#endif

int trace_err_skb(struct pt_regs *ctx)
{
	struct sk_buff *skb = (struct sk_buff *)ctx->si;
	struct sock *sk = (struct sock *)ctx->di;
	struct sock_exterr_skb *serr;
	u32 ee_data = 0;

	if (!sk || !skb)
		return 0;

	serr = SKB_EXT_ERR(skb);
	bpf_probe_read(&ee_data, sizeof(ee_data), &serr->ee.ee_data);
	bpf_trace_printk("ee_data:%u\\n", ee_data);

	return 0;
};
"""

b = BPF(text=bpf_text)
b.attach_kprobe(event="sock_queue_err_skb", fn_name="trace_err_skb")
print("Attached to kprobe")
b.trace_print()

Packetdrill Script:
~~~~~~
+0 `sysctl -q -w net.ipv4.tcp_min_tso_segs=10`
+0 `sysctl -q -w net.ipv4.tcp_no_metrics_save=1`
+0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0

0.100 < S 0:0(0) win 32792 <mss 1460,sackOK,nop,nop,nop,wscale 7>
0.100 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 7>
0.200 < . 1:1(0) ack 1 win 257
0.200 accept(3, ..., ...) = 4
+0 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0

+0 setsockopt(4, SOL_SOCKET, 37, [2688], 4) = 0
0.200 write(4, ..., 1460) = 1460
0.200 write(4, ..., 13140) = 13140

0.200 > P. 1:1461(1460) ack 1
0.200 > . 1461:8761(7300) ack 1
0.200 > P. 8761:14601(5840) ack 1

0.300 < . 1:1(0) ack 1 win 257 <sack 1461:2921,nop,nop>
0.300 < . 1:1(0) ack 1 win 257 <sack 1461:4381,nop,nop>
0.300 < . 1:1(0) ack 1 win 257 <sack 1461:5841,nop,nop>
0.300 > P. 1:1461(1460) ack 1
0.400 < . 1:1(0) ack 14601 win 257

0.400 close(4) = 0
0.400 > F. 14601:14601(0) ack 1
0.500 < F. 1:1(0) ack 14602 win 257
0.500 > . 14602:14602(0) ack 2
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Soheil Hassas Yeganeh <soheil.kdev@gmail.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: NSoheil Hassas Yeganeh <soheil@google.com>
Tested-by: NSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Last known hot point during SYNFLOOD attack is the clearing
of rx_opt.saw_tstamp in tcp_rcv_state_process()

It is not needed for a listener, so we move it where it matters.

Performance while a SYNFLOOD hits a single listener socket
went from 5 Mpps to 6 Mpps on my test server (24 cores, 8 NIC RX queues)
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When removing sk_refcnt manipulation on synflood, I missed that
using skb_set_owner_w() was racy, if sk->sk_wmem_alloc had already
transitioned to 0.

We should hold sk_refcnt instead, but this is a big deal under attack.
(Doing so increase performance from 3.2 Mpps to 3.8 Mpps only)

In this patch, I chose to not attach a socket to syncookies skb.

Performance is now 5 Mpps instead of 3.2 Mpps.

Following patch will remove last known false sharing in
tcp_rcv_state_process()

Fixes: 3b24d854 ("tcp/dccp: do not touch listener sk_refcnt under synflood")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Goal: packets dropped by a listener are accounted for.

This adds tcp_listendrop() helper, and clears sk_drops in sk_clone_lock()
so that children do not inherit their parent drop count.

Note that we no longer increment LINUX_MIB_LISTENDROPS counter when
sending a SYNCOOKIE, since the SYN packet generated a SYNACK.
We already have a separate LINUX_MIB_SYNCOOKIESSENT
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now ss can report sk_drops, we can instruct TCP to increment
this per socket counter when it drops an incoming frame, to refine
monitoring and debugging.

Following patch takes care of listeners drops.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, to avoid a cache line miss for accessing skb_shinfo,
tcp_ack_tstamp skips socket that do not have
SOF_TIMESTAMPING_TX_ACK bit set in sk_tsflags. This is
implemented based on an implicit assumption that the
SOF_TIMESTAMPING_TX_ACK is set via socket options for the
duration that ACK timestamps are needed.

To implement per-write timestamps, this check should be
removed and replaced with a per-packet alternative that
quickly skips packets missing ACK timestamps marks without
a cache-line miss.

To enable per-packet marking without a cache line miss, use
one bit in TCP_SKB_CB to mark a whether a SKB might need a
ack tx timestamp or not. Further checks in tcp_ack_tstamp are not
modified and work as before.
Signed-off-by: NSoheil Hassas Yeganeh <soheil@google.com>
Acked-by: NWillem de Bruijn <willemb@google.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For non-SACK connections, cwnd is lowered to inflight plus 3 packets
when the recovery ends. This is an optional feature in the NewReno
RFC 2582 to reduce the potential burst when cwnd is "re-opened"
after recovery and inflight is low.

This feature is questionably effective because of PRR: when
the recovery ends (i.e., snd_una == high_seq) NewReno holds the
CA_Recovery state for another round trip to prevent false fast
retransmits. But if the inflight is low, PRR will overwrite the
moderated cwnd in tcp_cwnd_reduction() later regardlessly. So if a
receiver responds bogus ACKs (i.e., acking future data) to speed up
transfer after recovery, it can only induce a burst up to a window
worth of data packets by acking up to SND.NXT. A restart from (short)
idle or receiving streched ACKs can both cause such bursts as well.

On the other hand, if the recovery ends because the sender
detects the losses were spurious (e.g., reordering). This feature
unconditionally lowers a reverted cwnd even though nothing
was lost.

By principle loss recovery module should not update cwnd. Further
pacing is much more effective to reduce burst. Hence this patch
removes the cwnd moderation feature.

v2 changes: revised commit message on bogus ACKs and burst, and
            missing signature
Signed-off-by: NMatt Mathis <mattmathis@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NSoheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are some cases where rtt_us derives from deltas of jiffies,
instead of using usec timestamps.

Since we want to track minimal rtt, better to assume a delta of 0 jiffie
might be in fact be very close to 1 jiffie.

It is kind of sad jiffies_to_usecs(1) calls a function instead of simply
using a constant.

Fixes: f6722583 ("tcp: track min RTT using windowed min-filter")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNikolay Borisov <kernel@kyup.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNikolay Borisov <kernel@kyup.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Refactor and consolidate cwnd and rate updates into a new function
tcp_cong_control().
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This change enables congestion control to update cwnd based on
not only packet cumulatively acked but also packets delivered
out-of-order. This makes congestion control robust against packet
reordering because it may raise cwnd as long as packets are being
delivered once reordering has been detected (i.e., it only cares
the amount of packets delivered, not the ordering among them).
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A small refactoring that gets number of packets cumulatively acked
from tcp_clean_rtx_queue() directly.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch changes the accounting of how many packets are
newly acked or sacked when the sender receives an ACK.

The current approach basically computes

   newly_acked_sacked = (prior_packets - prior_sacked) -
                        (tp->packets_out - tp->sacked_out)

   where prior_packets and prior_sacked out are snapshot
   at the beginning of the ACK processing.

The new approach tracks the delivery information via a new
TCP state variable "delivered" which monotically increases
as new packets are delivered in order or out-of-order.

The reason for this change is that the current approach is
brittle that produces negative or inaccurate estimate.

   1) For non-SACK connections, an ACK that advances the SND.UNA
   could reset the DUPACK counters (tp->sacked_out) in
   tcp_process_loss() or tcp_fastretrans_alert(). This inflates
   the inflight suddenly and causes under-estimate or even
   negative estimate. Here is a real example:

                   before   after (processing ACK)
   packets_out     75       73
   sacked_out      23        0
   ca state        Loss     Open

   The old approach computes (75-23) - (73 - 0) = -21 delivered
   while the new approach computes 1 delivered since it
   considers the 2nd-24th packets are delivered OOO.

   2) MSS change would re-count packets_out and sacked_out so
   the estimate is in-accurate and can even become negative.
   E.g., the inflight is doubled when MSS is halved.

   3) Spurious retransmission signaled by DSACK is not accounted

The new approach is simpler and more robust. For SACK connections,
tp->delivered increments as packets are being acked or sacked in
SACK and ACK processing.

For non-sack connections, it's done in tcp_remove_reno_sacks() and
tcp_add_reno_sack(). When an ACK advances the SND.UNA, tp->delivered
is incremented by the number of packets ACKed (less the current
number of DUPACKs received plus one packet hole).  Upon receiving
a DUPACK, tp->delivered is incremented assuming one out-of-order
packet is delivered.

Upon receiving a DSACK, tp->delivered is incremtened assuming one
retransmission is delivered in tcp_sacktag_write_queue().
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently the cwnd is reduced and increased in various different
places. The reduction happens in various places in the recovery
state processing (tcp_fastretrans_alert) while the increase
happens afterward.

A better sequence is to identify lost packets and update
the congestion control state (icsk_ca_state) first. Then base
on the new state, up/down the cwnd in one central place. It's
more clear to reason cwnd changes.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The retransmission and F-RTO transmission currently happen inside
recovery state processing (tcp_fastretrans_alert) but before
congestion control.  This refactoring moves the logic after both
s.t. we can determine how much to send (cwnd) before deciding what to
send.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When we acknowledge a FIN, it is not enough to ack the sequence number
and queue the skb into receive queue. We also have to call tcp_fin()
to properly update socket state and send proper poll() notifications.

It seems we also had the problem if we received a SYN packet with the
FIN flag set, but it does not seem an urgent issue, as no known
implementation can do that.

Fixes: 61d2bcae ("tcp: fastopen: accept data/FIN present in SYNACK message")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RFC 7413 (TCP Fast Open) 4.2.2 states that the SYNACK message
MAY include data and/or FIN

This patch adds support for the client side :

If we receive a SYNACK with payload or FIN, queue the skb instead
of ignoring it.

Since we already support the same for SYN, we refactor the existing
code and reuse it. Note we need to clone the skb, so this operation
might fail under memory pressure.

Sara Dickinson pointed out FreeBSD server Fast Open implementation
was planned to generate such SYNACK in the future.

The server side might be implemented on linux later.
Reported-by: NSara Dickinson <sara@sinodun.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RFC 4015 section 3.4 says the TCP sender MUST refrain from
reversing the congestion control state when the ACK signals
congestion through the ECN-Echo flag. Currently we may not
always do that when prior_ssthresh is reset upon receiving
ACKs with ECE marks. This patch fixes that.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This commit fixes a corner case in tcp_mark_head_lost() which was
causing the WARN_ON(len > skb->len) in tcp_fragment() to fire.

tcp_mark_head_lost() was assuming that if a packet has
tcp_skb_pcount(skb) of N, then it's safe to fragment off a prefix of
M*mss bytes, for any M < N. But with the tricky way TCP pcounts are
maintained, this is not always true.

For example, suppose the sender sends 4 1-byte packets and have the
last 3 packet sacked. It will merge the last 3 packets in the write
queue into an skb with pcount = 3 and len = 3 bytes. If another
recovery happens after a sack reneging event, tcp_mark_head_lost()
may attempt to split the skb assuming it has more than 2*MSS bytes.

This sounds very counterintuitive, but as the commit description for
the related commit c0638c24 ("tcp: don't fragment SACKed skbs in
tcp_mark_head_lost()") notes, this is because tcp_shifted_skb()
coalesces adjacent regions of SACKed skbs, and when doing this it
preserves the sum of their packet counts in order to reflect the
real-world dynamics on the wire. The c0638c24 commit tried to
avoid problems by not fragmenting SACKed skbs, since SACKed skbs are
where the non-proportionality between pcount and skb->len/mss is known
to be possible. However, that commit did not handle the case where
during a reneging event one of these weird SACKed skbs becomes an
un-SACKed skb, which tcp_mark_head_lost() can then try to fragment.

The fix is to simply mark the entire skb lost when this happens.
This makes the recovery slightly more aggressive in such corner
cases before we detect reordering. But once we detect reordering
this code path is by-passed because FACK is disabled.
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Patch 3759824d ("tcp: PRR uses CRB mode by default and SS mode
conditionally") introduced a bug that cwnd may become 0 when both
inflight and sndcnt are 0 (cwnd = inflight + sndcnt). This may lead
to a div-by-zero if the connection starts another cwnd reduction
phase by setting tp->prior_cwnd to the current cwnd (0) in
tcp_init_cwnd_reduction().

To prevent this we skip PRR operation when nothing is acked or
sacked. Then cwnd must be positive in all cases as long as ssthresh
is positive:

1) The proportional reduction mode
   inflight > ssthresh > 0

2) The reduction bound mode
  a) inflight == ssthresh > 0

  b) inflight < ssthresh
     sndcnt > 0 since newly_acked_sacked > 0 and inflight < ssthresh

Therefore in all cases inflight and sndcnt can not both be 0.
We check invalid tp->prior_cwnd to avoid potential div0 bugs.

In reality this bug is triggered only with a sequence of less common
events.  For example, the connection is terminating an ECN-triggered
cwnd reduction with an inflight 0, then it receives reordered/old
ACKs or DSACKs from prior transmission (which acks nothing). Or the
connection is in fast recovery stage that marks everything lost,
but fails to retransmit due to local issues, then receives data
packets from other end which acks nothing.

Fixes: 3759824d ("tcp: PRR uses CRB mode by default and SS mode conditionally")
Reported-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Allow accepted sockets to derive their sk_bound_dev_if setting from the
l3mdev domain in which the packets originated. A sysctl setting is added
to control the behavior which is similar to sk_mark and
sysctl_tcp_fwmark_accept.

This effectively allow a process to have a "VRF-global" listen socket,
with child sockets bound to the VRF device in which the packet originated.
A similar behavior can be achieved using sk_mark, but a solution using marks
is incomplete as it does not handle duplicate addresses in different L3
domains/VRFs. Allowing sockets to inherit the sk_bound_dev_if from l3mdev
domain provides a complete solution.
Signed-off-by: NDavid Ahern <dsa@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dmitry provided a syzkaller (http://github.com/google/syzkaller)
generated program that triggers the WARNING at
net/ipv4/tcp.c:1729 in tcp_recvmsg() :

WARN_ON(tp->copied_seq != tp->rcv_nxt &&
        !(flags & (MSG_PEEK | MSG_TRUNC)));

His program is specifically attempting a Cross SYN TCP exchange,
that we support (for the pleasure of hackers ?), but it looks we
lack proper tcp->copied_seq initialization.

Thanks again Dmitry for your report and testings.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Tested-by: NDmitry Vyukov <dvyukov@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

tcp_send_rcvq() is used for re-injecting data into tcp receive queue.

Problems :

- No check against size is performed, allowed user to fool kernel in
  attempting very large memory allocations, eventually triggering
  OOM when memory is fragmented.

- In case of fault during the copy we do not return correct errno.

Lets use alloc_skb_with_frags() to cook optimal skbs.

Fixes: 292e8d8c ("tcp: Move rcvq sending to tcp_input.c")
Fixes: c0e88ff0 ("tcp: Repair socket queues")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Acked-by: NPavel Emelyanov <xemul@parallels.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch implements the second half of RACK that uses the the most
recent transmit time among all delivered packets to detect losses.

tcp_rack_mark_lost() is called upon receiving a dubious ACK.
It then checks if an not-yet-sacked packet was sent at least
"reo_wnd" prior to the sent time of the most recently delivered.
If so the packet is deemed lost.

The "reo_wnd" reordering window starts with 1msec for fast loss
detection and changes to min-RTT/4 when reordering is observed.
We found 1msec accommodates well on tiny degree of reordering
(<3 pkts) on faster links. We use min-RTT instead of SRTT because
reordering is more of a path property but SRTT can be inflated by
self-inflicated congestion. The factor of 4 is borrowed from the
delayed early retransmit and seems to work reasonably well.

Since RACK is still experimental, it is now used as a supplemental
loss detection on top of existing algorithms. It is only effective
after the fast recovery starts or after the timeout occurs. The
fast recovery is still triggered by FACK and/or dupack threshold
instead of RACK.

We introduce a new sysctl net.ipv4.tcp_recovery for future
experiments of loss recoveries. For now RACK can be disabled by
setting it to 0.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch is the first half of the RACK loss recovery.

RACK loss recovery uses the notion of time instead
of packet sequence (FACK) or counts (dupthresh). It's inspired by the
previous FACK heuristic in tcp_mark_lost_retrans(): when a limited
transmit (new data packet) is sacked, then current retransmitted
sequence below the newly sacked sequence must been lost,
since at least one round trip time has elapsed.

But it has several limitations:
1) can't detect tail drops since it depends on limited transmit
2) is disabled upon reordering (assumes no reordering)
3) only enabled in fast recovery ut not timeout recovery

RACK (Recently ACK) addresses these limitations with the notion
of time instead: a packet P1 is lost if a later packet P2 is s/acked,
as at least one round trip has passed.

Since RACK cares about the time sequence instead of the data sequence
of packets, it can detect tail drops when later retransmission is
s/acked while FACK or dupthresh can't. For reordering RACK uses a
dynamically adjusted reordering window ("reo_wnd") to reduce false
positives on ever (small) degree of reordering.

This patch implements tcp_advanced_rack() which tracks the
most recent transmission time among the packets that have been
delivered (ACKed or SACKed) in tp->rack.mstamp. This timestamp
is the key to determine which packet has been lost.

Consider an example that the sender sends six packets:
T1: P1 (lost)
T2: P2
T3: P3
T4: P4
T100: sack of P2. rack.mstamp = T2
T101: retransmit P1
T102: sack of P2,P3,P4. rack.mstamp = T4
T205: ACK of P4 since the hole is repaired. rack.mstamp = T101

We need to be careful about spurious retransmission because it may
falsely advance tp->rack.mstamp by an RTT or an RTO, causing RACK
to falsely mark all packets lost, just like a spurious timeout.

We identify spurious retransmission by the ACK's TS echo value.
If TS option is not applicable but the retransmission is acknowledged
less than min-RTT ago, it is likely to be spurious. We refrain from
using the transmission time of these spurious retransmissions.

The second half is implemented in the next patch that marks packet
lost using RACK timestamp.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

a helper to prepare the main RACK patch
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove the existing lost retransmit detection because RACK subsumes
it completely. This also stops the overloading the ack_seq field of
the skb control block.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Kathleen Nichols' algorithm for tracking the minimum RTT of a
data stream over some measurement window. It uses constant space
and constant time per update. Yet it almost always delivers
the same minimum as an implementation that has to keep all
the data in the window. The measurement window is tunable via
sysctl.net.ipv4.tcp_min_rtt_wlen with a default value of 5 minutes.

The algorithm keeps track of the best, 2nd best & 3rd best min
values, maintaining an invariant that the measurement time of
the n'th best >= n-1'th best. It also makes sure that the three
values are widely separated in the time window since that bounds
the worse case error when that data is monotonically increasing
over the window.

Upon getting a new min, we can forget everything earlier because
it has no value - the new min is less than everything else in the
window by definition and it's the most recent. So we restart fresh
on every new min and overwrites the 2nd & 3rd choices. The same
property holds for the 2nd & 3rd best.

Therefore we have to maintain two invariants to maximize the
information in the samples, one on values (1st.v <= 2nd.v <=
3rd.v) and the other on times (now-win <=1st.t <= 2nd.t <= 3rd.t <=
now). These invariants determine the structure of the code

The RTT input to the windowed filter is the minimum RTT measured
from ACK or SACK, or as the last resort from TCP timestamps.

The accessor tcp_min_rtt() returns the minimum RTT seen in the
window. ~0U indicates it is not available. The minimum is 1usec
even if the true RTT is below that.
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently ca_seq_rtt_us does not use Kern's check. Fix that by
checking if any packet acked is a retransmit, for both RTT used
for RTT estimation and congestion control.

Fixes: 5b08e47c ("tcp: prefer packet timing to TS-ECR for RTT")
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

At the time of commit fff32699 ("tcp: reflect SYN queue_mapping into
SYNACK packets") we had little ways to cope with SYN floods.

We no longer need to reflect incoming skb queue mappings, and instead
can pick a TX queue based on cpu cooking the SYNACK, with normal XPS
affinities.

Note that all SYNACK retransmits were picking TX queue 0, this no longer
is a win given that SYNACK rtx are now distributed on all cpus.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

One 32bit hole is following skc_refcnt, use it.
skc_incoming_cpu can also be an union for request_sock rcv_wnd.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

inet_reqsk_alloc() is used to allocate a temporary request
in order to generate a SYNACK with a cookie. Then later,
syncookie validation also uses a temporary request.

These paths already took a reference on listener refcount,
we can avoid a couple of atomic operations.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are multiple races that need fixes :

1) skb_get() + queue skb + kfree_skb() is racy

An accept() can be done on another cpu, data consumed immediately.
tcp_recvmsg() uses __kfree_skb() as it is assumed all skb found in
socket receive queue are private.

Then the kfree_skb() in tcp_rcv_state_process() uses an already freed skb

2) tcp_reqsk_record_syn() needs to be done before tcp_try_fastopen()
for the same reasons.

3) We want to send the SYNACK before queueing child into accept queue,
otherwise we might reintroduce the ooo issue fixed in
commit 7c85af88 ("tcp: avoid reorders for TFO passive connections")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If a listen backlog is very big (to avoid syncookies), then
the listener sk->sk_wmem_alloc is the main source of false
sharing, as we need to touch it twice per SYNACK re-transmit
and TX completion.

(One SYN packet takes listener lock once, but up to 6 SYNACK
are generated)

By attaching the skb to the request socket, we remove this
source of contention.

Tested:

 listen(fd, 10485760); // single listener (no SO_REUSEPORT)
 16 RX/TX queue NIC
 Sustain a SYNFLOOD attack of ~320,000 SYN per second,
 Sending ~1,400,000 SYNACK per second.
 Perf profiles now show listener spinlock being next bottleneck.

    20.29%  [kernel]  [k] queued_spin_lock_slowpath
    10.06%  [kernel]  [k] __inet_lookup_established
     5.12%  [kernel]  [k] reqsk_timer_handler
     3.22%  [kernel]  [k] get_next_timer_interrupt
     3.00%  [kernel]  [k] tcp_make_synack
     2.77%  [kernel]  [k] ipt_do_table
     2.70%  [kernel]  [k] run_timer_softirq
     2.50%  [kernel]  [k] ip_finish_output
     2.04%  [kernel]  [k] cascade
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In this patch, we insert request sockets into TCP/DCCP
regular ehash table (where ESTABLISHED and TIMEWAIT sockets
are) instead of using the per listener hash table.

ACK packets find SYN_RECV pseudo sockets without having
to find and lock the listener.

In nominal conditions, this halves pressure on listener lock.

Note that this will allow for SO_REUSEPORT refinements,
so that we can select a listener using cpu/numa affinities instead
of the prior 'consistent hash', since only SYN packets will
apply this selection logic.

We will shrink listen_sock in the following patch to ease
code review.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Ying Cai <ycai@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

long term plan is to remove struct listen_sock when its hash
table is no longer there.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

tcp_syn_flood_action() will soon be called with unlocked socket.
In order to avoid SYN flood warning being emitted multiple times,
use xchg().
Extend max_qlen_log and synflood_warned fields in struct listen_sock
to u32
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>