When a kernel-to-user Netlink buffer overflows, the kernel reports
ENOBUFS without passing along an actual message. When it does this,
we should immediately try again, because we know that there is a
message waiting, instead of reporting the error to the caller.
This improves the OVS response rate to "hping3 --flood" traffic by
a few percentage points in my testing.
Signed-off-by: Ben Pfaff <b...@nicira.com>
---
lib/dpif-linux.c | 10 ++++++++++
1 files changed, 10 insertions(+), 0 deletions(-)
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index 256c9d6..9d84edd 100644
--- a/lib/dpif-linux.c
+++ b/lib/dpif-linux.c
@@ -150,6 +150,7 @@ struct dpif_linux {
};
static struct vlog_rate_limit error_rl = VLOG_RATE_LIMIT_INIT(9999, 5);
+static struct vlog_rate_limit enobufs_rl = VLOG_RATE_LIMIT_INIT(60, 5);
/* Generic Netlink family numbers for OVS. */
static int ovs_datapath_family;
@@ -1138,6 +1139,15 @@ dpif_linux_recv(struct dpif *dpif_, struct dpif_upcall
*upcall,
error = nl_sock_recv(upcall_sock, buf, false);
if (error) {
+ if (error == ENOBUFS) {
+ /* ENOBUFS typically means that we've received so many
+ * packets that the buffer overflowed. Try again
+ * immediately because there's almost certainly a packet
+ * waiting for us. */
+ VLOG_ERR_RL(&enobufs_rl, "%s: lost packet with hash %d",
+ dpif_name(dpif_), dpif->ready_mask);
+ continue;
+ }
if (error == EAGAIN) {
break;
}
--
1.7.2.5
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
