summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJason Wang <jasowang@redhat.com>2012-05-02 11:42:32 +0800
committerMichael S. Tsirkin <mst@redhat.com>2012-05-02 18:22:24 +0300
commitdbf34207c62bdec16b49721d119647c470a3443c (patch)
tree156779e1562f655cee6f383474d44e8960a818de
parentc460f0573941cb28dc7f35595679c3508f0ce41f (diff)
downloadop-kernel-dev-dbf34207c62bdec16b49721d119647c470a3443c.zip
op-kernel-dev-dbf34207c62bdec16b49721d119647c470a3443c.tar.gz
vhost_net: re-poll only on EAGAIN or ENOBUFS
Currently, we restart tx polling unconditionally when sendmsg() fails. This would cause unnecessary wakeups of vhost wokers and waste cpu utlization when evil userspace(guest driver) is able to hit EFAULT or EINVAL. The polling is only needed when the socket send buffer were exceeded or not enough memory. So fix this by restarting polling only when sendmsg() returns EAGAIN/ENOBUFS. Signed-off-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-rw-r--r--drivers/vhost/net.c3
1 files changed, 2 insertions, 1 deletions
diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
index 35abe90..f54b1d5 100644
--- a/drivers/vhost/net.c
+++ b/drivers/vhost/net.c
@@ -257,7 +257,8 @@ static void handle_tx(struct vhost_net *net)
UIO_MAXIOV;
}
vhost_discard_vq_desc(vq, 1);
- tx_poll_start(net, sock);
+ if (err == -EAGAIN || err == -ENOBUFS)
+ tx_poll_start(net, sock);
break;
}
if (err != len)
OpenPOWER on IntegriCloud