netdev - [PATCH] reduce the spinlock conflict during massive connect

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <9b38c346-0035-4c12-21e7-dbde9f961c8a@gmail.com>
Date:   Mon, 6 Nov 2017 10:28:21 +0800
From:   Liu Yu <liuyu924@...il.com>
To:     netdev@...r.kernel.org
Cc:     "\"David" "S." "Miller\"" "<davem@...emloft.net>;Alexey" Kuznetsov
         "<kuznet@....inr.ac.ru>;Hideaki" YOSHIFUJI 
        <yoshfuji@...ux-ipv6.org>
Subject: [PATCH] reduce the spinlock conflict during massive connect

From: Liu Yu <allanyuliu@...cent.com>

When a mount of processes connect to the same port at the same address
simultaneously, they are likely getting the same bhash and therefore
conflict with each other.

The more the cpu number, the worse in this case.

Use spin_trylock instead for this scene, which seems doesn't matter
for common case.

Signed-off-by: Liu Yu <allanyuliu@...cent.com>
---
 net/ipv4/inet_hashtables.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/net/ipv4/inet_hashtables.c b/net/ipv4/inet_hashtables.c
index e7d15fb..cc11ec7 100644
--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -581,13 +581,17 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 other_parity_scan:
 	port = low + offset;
 	for (i = 0; i < remaining; i += 2, port += 2) {
+		int ret;
+
 		if (unlikely(port >= high))
 			port -= remaining;
 		if (inet_is_local_reserved_port(net, port))
 			continue;
 		head = &hinfo->bhash[inet_bhashfn(net, port,
 						  hinfo->bhash_size)];
-		spin_lock_bh(&head->lock);
+		ret = spin_trylock(&head->lock);
+		if (unlikely(!ret))
+			continue;
 
 		/* Does not bother with rcv_saddr checks, because
 		 * the established check is already unique enough.
-- 
1.7.1