linux.git/net/ipv4/tcp_input.c, branch v4.20

tcp: address problems caused by EDT misshaps

2018-11-25T01:41:37+00:00

When a qdisc setup including pacing FQ is dismantled and recreated,
some TCP packets are sent earlier than instructed by TCP stack.

TCP can be fooled when ACK comes back, because the following
operation can return a negative value.

    tcp_time_stamp(tp) - tp->rx_opt.rcv_tsecr;

Some paths in TCP stack were not dealing properly with this,
this patch addresses four of them.

Fixes: ab408b6dc744 ("tcp: switch tcp and sch_fq to new earliest departure time model")
Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

tcp: defer SACK compression after DupThresh

2018-11-21T23:49:52+00:00

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

tcp: Fix SOF_TIMESTAMPING_RX_HARDWARE to use the latest timestamp during TCP coalescing

2018-11-20T18:32:11+00:00

During tcp coalescing ensure that the skb hardware timestamp refers to the
highest sequence number data.
Previously only the software timestamp was updated during coalescing.

Signed-off-by: Stephen Mallon 
Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

tcp: add tcp_reset_xmit_timer() helper

2018-10-24T02:42:44+00:00

With EDT model, SRTT no longer is inflated by pacing delays.

This means that RTO and some other xmit timers might be setup
incorrectly. This is particularly visible with either :

- Very small enforced pacing rates (SO_MAX_PACING_RATE)
- Reduced rto (from the default 200 ms)

This can lead to TCP flows aborts in the worst case,
or spurious retransmits in other cases.

For example, this session gets far more throughput
than the requested 80kbit :

$ netperf -H 127.0.0.2 -l 100 -- -q 10000
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 127.0.0.2 () port 0 AF_INET
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

540000 262144 262144    104.00      2.66

With the fix :

$ netperf -H 127.0.0.2 -l 100 -- -q 10000
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 127.0.0.2 () port 0 AF_INET
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

540000 262144 262144    104.00      0.12

EDT allows for better control of rtx timers, since TCP has
a better idea of the earliest departure time of each skb
in the rtx queue. We only have to eventually add to the
timer the difference of the EDT time with current time.

Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

2018-10-04T04:00:17+00:00

Minor conflict in net/core/rtnetlink.c, David Ahern's bug fix in 'net'
overlapped the renaming of a netlink attribute in net-next.

Signed-off-by: David S. Miller

tcp: start receiver buffer autotuning sooner

2018-10-01T22:45:26+00:00

Previously receiver buffer auto-tuning starts after receiving
one advertised window amount of data. After the initial receiver
buffer was raised by patch a337531b942b ("tcp: up initial rmem to
128KB and SYN rwin to around 64KB"), the reciver buffer may take
too long to start raising. To address this issue, this patch lowers
the initial bytes expected to receive roughly the expected sender's
initial window.

Fixes: a337531b942b ("tcp: up initial rmem to 128KB and SYN rwin to around 64KB")
Signed-off-by: Yuchung Cheng 
Signed-off-by: Wei Wang 
Signed-off-by: Neal Cardwell 
Signed-off-by: Eric Dumazet 
Reviewed-by: Soheil Hassas Yeganeh 
Signed-off-by: David S. Miller

tcp/dccp: fix lockdep issue when SYN is backlogged

2018-10-01T22:42:13+00:00

In normal SYN processing, packets are handled without listener
lock and in RCU protected ingress path.

But syzkaller is known to be able to trick us and SYN
packets might be processed in process context, after being
queued into socket backlog.

In commit 06f877d613be ("tcp/dccp: fix other lockdep splats
accessing ireq_opt") I made a very stupid fix, that happened
to work mostly because of the regular path being RCU protected.

Really the thing protecting ireq->ireq_opt is RCU read lock,
and the pseudo request refcnt is not relevant.

This patch extends what I did in commit 449809a66c1d ("tcp/dccp:
block BH for SYN processing") by adding an extra rcu_read_{lock|unlock}
pair in the paths that might be taken when processing SYN from
socket backlog (thus possibly in process context)

Fixes: 06f877d613be ("tcp/dccp: fix other lockdep splats accessing ireq_opt")
Signed-off-by: Eric Dumazet 
Reported-by: syzbot 
Signed-off-by: David S. Miller

tcp: up initial rmem to 128KB and SYN rwin to around 64KB

2018-09-29T18:22:22+00:00

Previously TCP initial receive buffer is ~87KB by default and
the initial receive window is ~29KB (20 MSS). This patch changes
the two numbers to 128KB and ~64KB (rounding down to the multiples
of MSS) respectively. The patch also simplifies the calculations s.t.
the two numbers are directly controlled by sysctl tcp_rmem[1]:

  1) Initial receiver buffer budget (sk_rcvbuf): while this should
     be configured via sysctl tcp_rmem[1], previously tcp_fixup_rcvbuf()
     always override and set a larger size when a new connection
     establishes.

  2) Initial receive window in SYN: previously it is set to 20
     packets if MSS <= 1460. The number 20 was based on the initial
     congestion window of 10: the receiver needs twice amount to
     avoid being limited by the receive window upon out-of-order
     delivery in the first window burst. But since this only
     applies if the receiving MSS <= 1460, connection using large MTU
     (e.g. to utilize receiver zero-copy) may be limited by the
     receive window.

With this patch TCP memory configuration is more straight-forward and
more properly sized to modern high-speed networks by default. Several
popular stacks have been announcing 64KB rwin in SYNs as well.

Signed-off-by: Yuchung Cheng 
Signed-off-by: Wei Wang 
Signed-off-by: Neal Cardwell 
Signed-off-by: Eric Dumazet 
Reviewed-by: Soheil Hassas Yeganeh 
Signed-off-by: David S. Miller

tcp: introduce tcp_skb_timestamp_us() helper

2018-09-22T02:37:59+00:00

There are few places where TCP reads skb->skb_mstamp expecting
a value in usec unit.

skb->tstamp (aka skb->skb_mstamp) will soon store CLOCK_TAI nsec value.

Add tcp_skb_timestamp_us() to provide proper conversion when needed.

Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

2018-09-13T05:22:42+00:00