linux-toradex.git/net/core/sock.c, branch v6.18-rc4

net: move sk->sk_err_soft and sk->sk_sndbuf

2025-09-23T00:55:24+00:00

sk->sk_sndbuf is read-mostly in tx path, so move it from
sock_write_tx group to more appropriate sock_read_tx.

sk->sk_err_soft was not identified previously, but
is used from tcp_ack().

Move it to sock_write_tx group for better cache locality.

Also change tcp_ack() to clear sk->sk_err_soft only if needed.

Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250919204856.2977245-3-edumazet@google.com
Signed-off-by: Jakub Kicinski

net: move sk_uid and sk_protocol to sock_read_tx

2025-09-23T00:55:24+00:00

sk_uid and sk_protocol are read from inet6_csk_route_socket()
for each TCP transmit.

Also read from udpv6_sendmsg(), udp_sendmsg() and others.

Move them to sock_read_tx for better cache locality.

Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250919204856.2977245-2-edumazet@google.com
Signed-off-by: Jakub Kicinski

udp: add udp_drops_inc() helper

2025-09-18T08:17:10+00:00

Generic sk_drops_inc() reads sk->sk_drop_counters.
We know the precise location for UDP sockets.

Move sk_drop_counters out of sock_read_rxtx
so that sock_write_rxtx starts at a cache line boundary.

Signed-off-by: Eric Dumazet 
Reviewed-by: Willem de Bruijn 
Reviewed-by: David Ahern 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250916160951.541279-9-edumazet@google.com
Reviewed-by: Jakub Kicinski 
Signed-off-by: Paolo Abeni

net: call cond_resched() less often in __release_sock()

2025-09-05T02:16:51+00:00

While stress testing TCP I had unexpected retransmits and sack packets
when a single cpu receives data from multiple high-throughput flows.

super_netperf 4 -H srv -T,10 -l 3000 &

Tcpdump extract:

 00:00:00.000007 IP6 clnt > srv: Flags [.], seq 26062848:26124288, ack 1, win 66, options [nop,nop,TS val 651460834 ecr 3100749131], length 61440
 00:00:00.000006 IP6 clnt > srv: Flags [.], seq 26124288:26185728, ack 1, win 66, options [nop,nop,TS val 651460834 ecr 3100749131], length 61440
 00:00:00.000005 IP6 clnt > srv: Flags [P.], seq 26185728:26243072, ack 1, win 66, options [nop,nop,TS val 651460834 ecr 3100749131], length 57344
 00:00:00.000006 IP6 clnt > srv: Flags [.], seq 26243072:26304512, ack 1, win 66, options [nop,nop,TS val 651460844 ecr 3100749141], length 61440
 00:00:00.000005 IP6 clnt > srv: Flags [.], seq 26304512:26365952, ack 1, win 66, options [nop,nop,TS val 651460844 ecr 3100749141], length 61440
 00:00:00.000007 IP6 clnt > srv: Flags [P.], seq 26365952:26423296, ack 1, win 66, options [nop,nop,TS val 651460844 ecr 3100749141], length 57344
 00:00:00.000006 IP6 clnt > srv: Flags [.], seq 26423296:26484736, ack 1, win 66, options [nop,nop,TS val 651460853 ecr 3100749150], length 61440
 00:00:00.000005 IP6 clnt > srv: Flags [.], seq 26484736:26546176, ack 1, win 66, options [nop,nop,TS val 651460853 ecr 3100749150], length 61440
 00:00:00.000005 IP6 clnt > srv: Flags [P.], seq 26546176:26603520, ack 1, win 66, options [nop,nop,TS val 651460853 ecr 3100749150], length 57344
 00:00:00.003932 IP6 clnt > srv: Flags [P.], seq 26603520:26619904, ack 1, win 66, options [nop,nop,TS val 651464844 ecr 3100753141], length 16384
 00:00:00.006602 IP6 clnt > srv: Flags [.], seq 24862720:24866816, ack 1, win 66, options [nop,nop,TS val 651471419 ecr 3100759716], length 4096
 00:00:00.013000 IP6 clnt > srv: Flags [.], seq 24862720:24866816, ack 1, win 66, options [nop,nop,TS val 651484421 ecr 3100772718], length 4096
 00:00:00.000416 IP6 srv > clnt: Flags [.], ack 26619904, win 1393, options [nop,nop,TS val 3100773185 ecr 651484421,nop,nop,sack 1 {24862720:24866816}], length 0

After analysis, it appears this is because of the cond_resched()
call from  __release_sock().

When current thread is yielding, while still holding the TCP socket lock,
it might regain the cpu after a very long time.

Other peer TLP/RTO is firing (multiple times) and packets are retransmit,
while the initial copy is waiting in the socket backlog or receive queue.

In this patch, I call cond_resched() only once every 16 packets.

Modern TCP stack now spends less time per packet in the backlog,
especially because ACK are no longer sent (commit 133c4c0d3717
"tcp: defer regular ACK while processing socket backlog")

Before:

clnt:/# nstat -n;sleep 10;nstat|egrep "TcpOutSegs|TcpRetransSegs|TCPFastRetrans|TCPTimeouts|Probes|TCPSpuriousRTOs|DSACK"
TcpOutSegs                      19046186           0.0
TcpRetransSegs                  1471               0.0
TcpExtTCPTimeouts               1397               0.0
TcpExtTCPLossProbes             1356               0.0
TcpExtTCPDSACKRecv              1352               0.0
TcpExtTCPSpuriousRTOs           114                0.0
TcpExtTCPDSACKRecvSegs          1352               0.0

After:

clnt:/# nstat -n;sleep 10;nstat|egrep "TcpOutSegs|TcpRetransSegs|TCPFastRetrans|TCPTimeouts|Probes|TCPSpuriousRTOs|DSACK"
TcpOutSegs                      19218936           0.0

Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250903174811.1930820-1-edumazet@google.com
Signed-off-by: Jakub Kicinski

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

2025-09-04T20:33:00+00:00

Cross-merge networking fixes after downstream PR (net-6.17-rc5).

No conflicts.

Adjacent changes:

include/net/sock.h
  c51613fa276f ("net: add sk->sk_drop_counters")
  5d6b58c932ec ("net: lockless sock_i_ino()")

Signed-off-by: Jakub Kicinski

net: lockless sock_i_ino()

2025-09-03T23:08:24+00:00

Followup of commit c51da3f7a161 ("net: remove sock_i_uid()")

A recent syzbot report was the trigger for this change.

Over the years, we had many problems caused by the
read_lock[_bh](&sk->sk_callback_lock) in sock_i_uid().

We could fix smc_diag_dump_proto() or make a more radical move:

Instead of waiting for new syzbot reports, cache the socket
inode number in sk->sk_ino, so that we no longer
need to acquire sk->sk_callback_lock in sock_i_ino().

This makes socket dumps faster (one less cache line miss,
and two atomic ops avoided).

Prior art:

commit 25a9c8a4431c ("netlink: Add __sock_i_ino() for __netlink_diag_dump().")
commit 4f9bf2a2f5aa ("tcp: Don't acquire inet_listen_hashbucket::lock with disabled BH.")
commit efc3dbc37412 ("rds: Make rds_sock_lock BH rather than IRQ safe.")

Fixes: d2d6422f8bd1 ("x86: Allow to enable PREEMPT_RT.")
Reported-by: syzbot+50603c05bbdf4dfdaffa@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/68b73804.050a0220.3db4df.01d8.GAE@google.com/T/#u
Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Reviewed-by: Sebastian Andrzej Siewior 
Link: https://patch.msgid.link/20250902183603.740428-1-edumazet@google.com
Signed-off-by: Jakub Kicinski

net: use dst_dev_rcu() in sk_setup_caps()

2025-08-30T02:36:32+00:00

Use RCU to protect accesses to dst->dev from sk_setup_caps()
and sk_dst_gso_max_size().

Also use dst_dev_rcu() in ip6_dst_mtu_maybe_forward(),
and ip_dst_mtu_maybe_forward().

ip4_dst_hoplimit() can use dst_dev_net_rcu().

Fixes: 4a6ce2b6f2ec ("net: introduce a new function dst_dev_put()")
Signed-off-by: Eric Dumazet 
Reviewed-by: David Ahern 
Link: https://patch.msgid.link/20250828195823.3958522-6-edumazet@google.com
Signed-off-by: Jakub Kicinski

net: add sk->sk_drop_counters

2025-08-28T11:14:50+00:00

Some sockets suffer from heavy false sharing on sk->sk_drops,
and fields in the same cache line.

Add sk->sk_drop_counters to:

- move the drop counter(s) to dedicated cache lines.
- Add basic NUMA awareness to these drop counter(s).

Following patches will use this infrastructure for UDP and RAW sockets.

sk_clone_lock() is not yet ready, it would need to properly
set newsk->sk_drop_counters if we plan to use this for TCP sockets.

v2: used Paolo suggestion from https://lore.kernel.org/netdev/8f09830a-d83d-43c9-b36b-88ba0a23e9b2@redhat.com/

Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250826125031.1578842-4-edumazet@google.com
Signed-off-by: Paolo Abeni

net: add sk_drops_read(), sk_drops_inc() and sk_drops_reset() helpers

2025-08-28T11:14:50+00:00

We want to split sk->sk_drops in the future to reduce
potential contention on this field.

Signed-off-by: Eric Dumazet 
Reviewed-by: Kuniyuki Iwashima 
Link: https://patch.msgid.link/20250826125031.1578842-2-edumazet@google.com
Signed-off-by: Paolo Abeni

net: set net.core.rmem_max and net.core.wmem_max to 4 MB

2025-08-21T02:35:00+00:00

SO_RCVBUF and SO_SNDBUF have limited range today, unless
distros or system admins change rmem_max and wmem_max.

Even iproute2 uses 1 MB SO_RCVBUF which is capped by
the kernel.

Decouple [rw]mem_max and [rw]mem_default and increase
[rw]mem_max to 4 MB.

Before:

$ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
net.core.rmem_default = 212992
net.core.rmem_max = 212992
net.core.wmem_default = 212992
net.core.wmem_max = 212992

After:

$ sysctl net.core.rmem_default net.core.rmem_max net.core.wmem_default net.core.wmem_max
net.core.rmem_default = 212992
net.core.rmem_max = 4194304
net.core.wmem_default = 212992
net.core.wmem_max = 4194304

Signed-off-by: Eric Dumazet 
Reviewed-by: Neal Cardwell 
Link: https://patch.msgid.link/20250819174030.1986278-1-edumazet@google.com
Signed-off-by: Jakub Kicinski