Select lacp egress ports based on NUMA domain

This change creates an array of port maps indexed by numa domain
for lacp port selection. If we have lacp interfaces in more than
one domain, then we select the egress port by indexing into the
numa port maps and picking a port on the appropriate numa domain.

This is behavior is controlled by the new ifconfig use_numa flag
and net.link.lagg.use_numa sysctl/tunable (both modeled after the
existing use_flowid), which default to enabled.

Reviewed by:	bz, hselasky, markj (and scottl, earlier version)
Sponsored by:	Netflix
Differential Revision:	https://reviews.freebsd.org/D20060

2 files changed, 21 insertions, 1 deletions

diff --git a/sbin/ifconfig/ifconfig.8 b/sbin/ifconfig/ifconfig.8
index 7d94bca57b20..92b319869cbd 100644
--- a/sbin/ifconfig/ifconfig.8
+++ b/sbin/ifconfig/ifconfig.8
@@ -28,7 +28,7 @@
 .\"     From: @(#)ifconfig.8	8.3 (Berkeley) 1/5/94
 .\" $FreeBSD$
 .\"
-.Dd June 27, 2018
+.Dd May 3, 2019
 .Dt IFCONFIG 8
 .Os
 .Sh NAME
@@ -2497,6 +2497,22 @@ Use the RSS hash from the network card if available.
 Set a shift parameter for RSS local hash computation.
 Hash is calculated by using flowid bits in a packet header mbuf
 which are shifted by the number of this parameter.
+.It Cm use_numa
+Enable selection of egress ports based on the native
+.Xr NUMA 4
+domain for the packets being transmitted.
+This is currently only implemented for lacp mode.
+This works only on
+.Xr NUMA 4
+hardware, running a kernel compiled with the
+.Xr NUMA 4
+option, and when interfaces from multiple
+.Xr NUMA 4
+domains are ports of the aggregation interface.
+.It Cm -use_numa
+Disable selection of egress ports based on the native
+.Xr NUMA 4
+domain for the packets being transmitted.
 .It Cm lacp_fast_timeout
 Enable lacp fast-timeout on the interface.
 .It Cm -lacp_fast_timeout
diff --git a/sbin/ifconfig/iflagg.c b/sbin/ifconfig/iflagg.c
index 89b9ce097f35..3b9e900605b5 100644
--- a/sbin/ifconfig/iflagg.c
+++ b/sbin/ifconfig/iflagg.c
@@ -130,6 +130,8 @@ setlaggsetopt(const char *val, int d, int s, const struct afswtch *afp)
 	switch (ro.ro_opts) {
 	case LAGG_OPT_USE_FLOWID:
 	case -LAGG_OPT_USE_FLOWID:
+	case LAGG_OPT_USE_NUMA:
+	case -LAGG_OPT_USE_NUMA:
 	case LAGG_OPT_LACP_STRICT:
 	case -LAGG_OPT_LACP_STRICT:
 	case LAGG_OPT_LACP_TXTEST:
@@ -303,6 +305,8 @@ static struct cmd lagg_cmds[] = {
 	DEF_CMD_ARG("lagghash",		setlagghash),
 	DEF_CMD("use_flowid",	LAGG_OPT_USE_FLOWID,	setlaggsetopt),
 	DEF_CMD("-use_flowid",	-LAGG_OPT_USE_FLOWID,	setlaggsetopt),
+	DEF_CMD("use_numa",	LAGG_OPT_USE_NUMA,	setlaggsetopt),
+	DEF_CMD("-use_numa",	-LAGG_OPT_USE_NUMA,	setlaggsetopt),
 	DEF_CMD("lacp_strict",	LAGG_OPT_LACP_STRICT,	setlaggsetopt),
 	DEF_CMD("-lacp_strict",	-LAGG_OPT_LACP_STRICT,	setlaggsetopt),
 	DEF_CMD("lacp_txtest",	LAGG_OPT_LACP_TXTEST,	setlaggsetopt),