All of lore.kernel.org
 help / color / mirror / Atom feed
From: Eric Dumazet <edumazet@google.com>
To: "David S . Miller" <davem@davemloft.net>
Cc: netdev <netdev@vger.kernel.org>,
	Eric Dumazet <edumazet@google.com>,
	Eric Dumazet <eric.dumazet@gmail.com>
Subject: [PATCH net-next 3/4] tcp: tsq: add shortcut in tcp_tasklet_func()
Date: Fri,  2 Dec 2016 10:25:58 -0800	[thread overview]
Message-ID: <1480703159-2327-4-git-send-email-edumazet@google.com> (raw)
In-Reply-To: <1480703159-2327-1-git-send-email-edumazet@google.com>

Under high stress, I've seen tcp_tasklet_func() consuming
~700 usec, handling ~150 tcp sockets.

By setting TCP_TSQ_DEFERRED in tcp_wfree(), we give a chance
for other cpus/threads entering tcp_write_xmit() to grab it,
allowing tcp_tasklet_func() to skip sockets that already did
an xmit cycle.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/tcp_output.c | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 76be79437595..9143c52b3105 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -767,19 +767,19 @@ static void tcp_tasklet_func(unsigned long data)
 	list_for_each_safe(q, n, &list) {
 		tp = list_entry(q, struct tcp_sock, tsq_node);
 		list_del(&tp->tsq_node);
+		clear_bit(TSQ_QUEUED, &tp->tsq_flags);
 
 		sk = (struct sock *)tp;
-		bh_lock_sock(sk);
-
-		if (!sock_owned_by_user(sk)) {
-			tcp_tsq_handler(sk);
-		} else {
-			/* defer the work to tcp_release_cb() */
-			set_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags);
+		if (!sk->sk_lock.owned &&
+		    test_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags)) {
+			bh_lock_sock(sk);
+			if (!sock_owned_by_user(sk)) {
+				clear_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags);
+				tcp_tsq_handler(sk);
+			}
+			bh_unlock_sock(sk);
 		}
-		bh_unlock_sock(sk);
 
-		clear_bit(TSQ_QUEUED, &tp->tsq_flags);
 		sk_free(sk);
 	}
 }
@@ -884,7 +884,7 @@ void tcp_wfree(struct sk_buff *skb)
 		if (!(oval & TSQF_THROTTLED) || (oval & TSQF_QUEUED))
 			goto out;
 
-		nval = (oval & ~TSQF_THROTTLED) | TSQF_QUEUED;
+		nval = (oval & ~TSQF_THROTTLED) | TSQF_QUEUED | TCP_TSQ_DEFERRED;
 		nval = cmpxchg(&tp->tsq_flags, oval, nval);
 		if (nval != oval)
 			continue;
@@ -2229,6 +2229,8 @@ static bool tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle,
 		    unlikely(tso_fragment(sk, skb, limit, mss_now, gfp)))
 			break;
 
+		if (test_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags))
+			clear_bit(TCP_TSQ_DEFERRED, &tp->tsq_flags);
 		if (tcp_small_queue_check(sk, skb, 0))
 			break;
 
-- 
2.8.0.rc3.226.g39d4020

  parent reply	other threads:[~2016-12-02 18:27 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-12-02 18:25 [PATCH net-next 0/4] tcp: tsq: performance series Eric Dumazet
2016-12-02 18:25 ` [PATCH net-next 1/4] tcp: tsq: add tsq_flags / tsq_enum Eric Dumazet
2016-12-02 18:25 ` [PATCH net-next 2/4] tcp: tsq: remove one locked operation in tcp_wfree() Eric Dumazet
2016-12-02 18:25 ` Eric Dumazet [this message]
2016-12-02 22:12   ` [PATCH net-next 3/4] tcp: tsq: add shortcut in tcp_tasklet_func() Eric Dumazet
2016-12-02 18:25 ` [PATCH net-next 4/4] tcp: tsq: avoid one atomic in tcp_wfree() Eric Dumazet
2016-12-02 20:57 ` [PATCH net-next 0/4] tcp: tsq: performance series Eric Dumazet

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1480703159-2327-4-git-send-email-edumazet@google.com \
    --to=edumazet@google.com \
    --cc=davem@davemloft.net \
    --cc=eric.dumazet@gmail.com \
    --cc=netdev@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.