[PATCH] af_unix: optimize unix_dgram_poll()

[Eric Dumazet <eric.dumazet@gmail.com>]

Le vendredi 29 octobre 2010 à 13:46 -0700, Davide Libenzi a écrit :

> Also, why not using the existing wait->key instead of adding a poll2()?

Indeed, if wait is not null, we have in wait->key the interest of
poller. If a particular poll() function is expensive, it can test these
bits.

Thanks !

Note: I chose the 'goto skip_write' to make this patch really obvious.

[PATCH] af_unix: optimize unix_dgram_poll()

unix_dgram_poll() is pretty expensive to check POLLOUT status, because
it has to lock the socket to get its peer, take a reference on the peer
to check its receive queue status, and queue another poll_wait on
peer_wait. This all can be avoided if the process calling
unix_dgram_poll() is not interested in POLLOUT status. It makes
unix_dgram_recvmsg() faster by not queueing irrelevant pollers in
peer_wait.

On a test program provided by Alan Crequy :

Before:

real    0m0.211s
user    0m0.000s
sys     0m0.208s

After:

real	0m0.044s
user	0m0.000s
sys	0m0.040s

Suggested-by: Davide Libenzi <davidel@xmailserver.org>
Reported-by: Alban Crequy <alban.crequy@collabora.co.uk>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 net/unix/af_unix.c |    4 ++++
 1 file changed, 4 insertions(+)

diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 3c95304..dcb84fe 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -2090,6 +2090,9 @@ static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
 			return mask;
 	}
 
+	if (wait && !(wait->key & (POLLWRBAND | POLLWRNORM | POLLOUT)))
+		goto skip_write;
+
 	/* writable? */
 	writable = unix_writable(sk);
 	if (writable) {
@@ -2111,6 +2114,7 @@ static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
 	else
 		set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
 
+skip_write:
 	return mask;
 }