From 76958568803fa23c334e3bb17f9823191b53bae1 Mon Sep 17 00:00:00 2001
From: Robert Marko <robimarko@gmail.com>
Date: Fri, 18 Mar 2022 22:02:01 +0100
Subject: [PATCH] switchdev: attempt to fix roaming
Try and solve the roaming issue by trying to replicate what NSS bridge
module is doing, but by utilizing switchdev FDB notifiers instead of
adding new notifiers to the bridge code.
It basically gets triggered once a new FDB entry has been discovered
and then it will look for VSI ID and simply erase the existing FDB
by MAC and VSI ID.
Signed-off-by: Robert Marko <robimarko@gmail.com>
---
nss_dp_switchdev.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 59 insertions(+)
diff --git a/nss_dp_switchdev.c b/nss_dp_switchdev.c
index e72cfbc..4ff4128 100644
--- a/nss_dp_switchdev.c
+++ b/nss_dp_switchdev.c
@@ -24,6 +24,8 @@
#include "nss_dp_dev.h"
#include "fal/fal_stp.h"
#include "fal/fal_ctrlpkt.h"
+#include "fal/fal_fdb.h"
+#include "ref/ref_vsi.h"
#define NSS_DP_SWITCH_ID 0
#define NSS_DP_SW_ETHTYPE_PID 0 /* PPE ethtype profile ID for slow protocols */
@@ -348,10 +350,62 @@ static int nss_dp_switchdev_event(struct notifier_block *unused,
return NOTIFY_DONE;
}
+static int nss_dp_switchdev_fdb_event(struct net_device *netdev,
+ struct switchdev_notifier_fdb_info *fdb_info)
+{
+ struct nss_dp_dev *dp_priv = (struct nss_dp_dev *)netdev_priv(netdev);
+ fal_fdb_entry_t entry;
+ a_uint32_t vsi_id;
+ sw_error_t rv;
+
+ rv = ppe_port_vsi_get(NSS_DP_SWITCH_ID, dp_priv->macid, &vsi_id);
+ if (rv) {
+ netdev_err(netdev, "cannot get VSI ID: %d\n", rv);
+ return notifier_from_errno(rv);
+ }
+
+ memset(&entry, 0, sizeof(entry));
+ memcpy(&entry.addr, fdb_info->addr, ETH_ALEN);
+ entry.fid = vsi_id;
+
+ rv = fal_fdb_entry_del_bymac(NSS_DP_SWITCH_ID, &entry);
+ if (rv) {
+ netdev_err(netdev, "FDB entry delete failed with MAC %pM and fid %d\n",
+ &entry.addr, entry.fid);
+ return notifier_from_errno(rv);
+ }
+
+ return notifier_from_errno(rv);
+};
+
+static int nss_dp_fdb_switchdev_event(struct notifier_block *nb,
+ unsigned long event, void *ptr)
+{
+ struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
+
+ /*
+ * Handle switchdev event only for physical devices
+ */
+ if (!nss_dp_is_phy_dev(dev)) {
+ return NOTIFY_DONE;
+ }
+
+ switch (event) {
+ case SWITCHDEV_FDB_ADD_TO_DEVICE:
+ return nss_dp_switchdev_fdb_event(dev, ptr);
+ }
+
+ return NOTIFY_DONE;
+}
+
static struct notifier_block nss_dp_switchdev_notifier = {
.notifier_call = nss_dp_switchdev_event,
};
+static struct notifier_block nss_dp_switchdev_fdb_notifier = {
+ .notifier_call = nss_dp_fdb_switchdev_event,
+};
+
static bool switch_init_done;
/*
@@ -366,6 +420,11 @@ void nss_dp_switchdev_setup(struct net_device *dev)
return;
}
+ err = register_switchdev_notifier(&nss_dp_switchdev_fdb_notifier);
+ if (err) {
+ netdev_dbg(dev, "%px:Failed to register switchdev notifier\n", dev);
+ }
+
err = register_switchdev_blocking_notifier(&nss_dp_switchdev_notifier);
if (err) {
netdev_dbg(dev, "%px:Failed to register switchdev notifier\n", dev);
--
2.35.1
Add this to the NSS-DP as the last patch and enable switchdev in the kernel config.
Ok, so I gotta squeeze in an AP, trust me my desk is so full of boards.
I am hoping that there is an even easier way as I need to be able to quickly test