There is no reason for stopping all TX queues from dev_watchdog()

Not only this stops feeding the NIC, it also migrates all qdiscs
to be serviced on the cpu calling netif_tx_unlock(), causing
a potential latency artifact.

Signed-off-by: Eric Dumazet <edumazet@google.com>

These are not fast path, there is no point in inlining them.

Also provide netif_freeze_queues()/netif_unfreeze_queues()
so that we can use them from dev_watchdog() in the following patch.

Signed-off-by: Eric Dumazet <edumazet@google.com>

In following patches, dev_watchdog() will no longer stop all queues.
It will read queue->trans_start locklessly.

Signed-off-by: Eric Dumazet <edumazet@google.com>

tx_timeout_show() assumed dev_watchdog() would stop all
the queues, to fetch queue->trans_timeout under protection
of the queue->_xmit_lock.

As we want to no longer disrupt transmits, we use an
atomic_long_t instead.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: david decotigny <david.decotigny@google.com>

Eric Dumazet says:

====================
net: prot_inuse and sock_inuse cleanups

Small series cleaning and optimizing sock_prot_inuse_add()
and sock_inuse_add().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

This is distracting really, let's make this simpler,
because many callers had to take care of this
by themselves, even if on x86 this adds more
code than really needed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net->core.sock_inuse is a per cpu variable (int),
while net->core.prot_inuse is another per cpu variable
of 64 integers.

per cpu allocator tend to place them in very different places.

Grouping them together makes sense, since it makes
updates potentially faster, if hitting the same
cache line.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MPTCP hard codes it, let us instead provide this helper.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sock_prot_inuse_add() is very small, we can inline it.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
gro: get out of core files

Move GRO related content into net/core/gro.c
and include/net/gro.h.

This reduces GRO scope to where it is really needed,
and shrinks too big files (include/linux/netdevice.h
and net/core/dev.c)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

Move gro code and data from net/core/dev.c to net/core/gro.c
to ease maintenance.

gro_normal_list() and gro_normal_one() are inlined
because they are called from both files.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/core/gro.c will contain all core gro functions,
to shrink net/core/skbuff.c and net/core/dev.c

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This helper is used once, no need to keep it in fat net/core/skbuff.c

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/netdevice.h became too big, move gro stuff
into include/net/gro.h

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Eric Dumazet says:

====================
tcp: optimizations for linux-5.17

Mostly small improvements in this series.

The notable change is in "defer skb freeing after
socket lock is released" in recvmsg() (and RX zerocopy)

The idea is to try to let skb freeing to BH handler,
whenever possible, or at least perform the freeing
outside of the socket lock section, for much improved
performance. This idea can probably be extended
to other protocols.

 Tests on a 100Gbit NIC
 Max throughput for one TCP_STREAM flow, over 10 runs.

 MTU : 1500  (1428 bytes of TCP payload per MSS)
 Before: 55 Gbit
 After:  66 Gbit

 MTU : 4096+ (4096 bytes of TCP payload, plus TCP/IPv6 headers)
 Before: 82 Gbit
 After:  95 Gbit
====================

Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sk_rx_dst/sk_rx_dst_ifindex/sk_rx_dst_cookie are read in early demux,
and currently spans two cache lines.

Moving them close to sk_refcnt makes more sense, as only one cache
line is needed.

New layout for this hot cache line is :

struct sock {
	struct sock_common         __sk_common;          /*     0  0x88 */
	/* --- cacheline 2 boundary (128 bytes) was 8 bytes ago --- */
	struct dst_entry *         sk_rx_dst;            /*  0x88   0x8 */
	int                        sk_rx_dst_ifindex;    /*  0x90   0x4 */
	u32                        sk_rx_dst_cookie;     /*  0x94   0x4 */
	socket_lock_t              sk_lock;              /*  0x98  0x20 */
	atomic_t                   sk_drops;             /*  0xb8   0x4 */
	int                        sk_rcvlowat;          /*  0xbc   0x4 */
	/* --- cacheline 3 boundary (192 bytes) --- */

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Under pressure, tcp recvmsg() has logic to process the socket backlog,
but calls tcp_cleanup_rbuf() right before.

Avoiding sending ACK right before processing new segments makes
a lot of sense, as this decrease the number of ACK packets,
with no impact on effective ACK clocking.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Testing timeo before sk_err/sk_state/sk_shutdown makes more sense.

Modern applications use non-blocking IO, while a socket is terminated
only once during its life time.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp recvmsg() (or rx zerocopy) spends a fair amount of time
freeing skbs after their payload has been consumed.

A typical ~64KB GRO packet has to release ~45 page
references, eventually going to page allocator
for each of them.

Currently, this freeing is performed while socket lock
is held, meaning that there is a high chance that
BH handler has to queue incoming packets to tcp socket backlog.

This can cause additional latencies, because the user
thread has to process the backlog at release_sock() time,
and while doing so, additional frames can be added
by BH handler.

This patch adds logic to defer these frees after socket
lock is released, or directly from BH handler if possible.

Being able to free these skbs from BH handler helps a lot,
because this avoids the usual alloc/free assymetry,
when BH handler and user thread do not run on same cpu or
NUMA node.

One cpu can now be fully utilized for the kernel->user copy,
and another cpu is handling BH processing and skb/page
allocs/frees (assuming RFS is not forcing use of a single CPU)

Tested:
 100Gbit NIC
 Max throughput for one TCP_STREAM flow, over 10 runs

MTU : 1500
Before: 55 Gbit
After:  66 Gbit

MTU : 4096+(headers)
Before: 82 Gbit
After:  95 Gbit

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

TCP uses sk_eat_skb() when skbs can be removed from receive queue.
However, the call to skb_orphan() from __kfree_skb() incurs
an indirect call so sock_rfee(), which is more expensive than
a direct call, especially for CONFIG_RETPOLINE=y.

Add tcp_eat_recv_skb() function to make the call before
__kfree_skb().

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use some unlikely() hints in the fast path.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_poll() and tcp_ioctl() are reading tp->urg_data without socket lock
owned.

Also, it is faster to first check tp->urg_data in tcp_poll(),
then tp->urg_seq == tp->copied_seq, because tp->urg_seq is
located in a different/cold cache line.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_segs_in() can be called from BH, while socket spinlock
is held but socket owned by user, eventually reading these
fields from tcp_get_info()

Found by code inspection, no need to backport this patch
to older kernels.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use INDIRECT_CALL_INET() to avoid an indirect call
when/if CONFIG_RETPOLINE=y

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When reading large chunks of data, incoming packets might
be added to the backlog from BH.

tcp recvmsg() detects the backlog queue is not empty, and uses
a release_sock()/lock_sock() pair to process this backlog.

We now have __sk_flush_backlog() to perform this
a bit faster.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp_memory_allocated and tcp_sockets_allocated often share
a common cache line, source of false sharing.

Also take care of udp_memory_allocated and mptcp_sockets_allocated.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

(struct proto)->sk_forward_alloc is currently only used by MPTCP.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move sk_bind_phc next to sk_peer_lock to fill a hole.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gso_size can be moved after tclass, to use an existing hole.
(8 bytes saved on 64bit arches)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Instead of using a full netdev_features_t, we can use a single bit,
as sk_route_nocaps is only used to remove NETIF_F_GSO_MASK from
sk->sk_route_cap.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We were only using one bit, and we can replace it by sk_is_tcp()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move sk_is_tcp() to include/net/sock.h and use it where we can.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For TCP flows, inet6_sk(sk)->saddr has the same value
than sk->sk_v6_rcv_saddr.

Using sk->sk_v6_rcv_saddr increases data locality.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For some reason, I forgot to change __tcp_v6_send_check() at
the same time I removed (ip_summed == CHECKSUM_PARTIAL) check
in __tcp_v4_send_check()

Fixes: 98be9b1 ("tcp: remove dead code after CHECKSUM_PARTIAL adoption")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If packet is going to be coalesced, sk_sndbuf/sk_rcvbuf values
are not used. Defer their access to the point we need them.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>