Re: [RFC PATCH net-next v2 01/10] net: phy: Introduce ethernet link topology representation

Hello Andrew,

On Tue, 21 Nov 2023 01:24:47 +0100
Andrew Lunn [off-list ref] wrote:

quoted

+int link_topo_add_phy(struct link_topology *lt, struct phy_device *phy,
+		      enum phy_upstream upt, void *upstream)
+{
+	ret = xa_alloc_cyclic(&lt->phys, &phy->phyindex, pdn, xa_limit_32b,
+			      &lt->next_phy_index, GFP_KERNEL);
+	if (ret)
+		goto err;
+
+	return 0;  

It looks like that could be just return xa_alloc_cyclic(...);

Indeed, nice catch

quoted

diff --git a/include/linux/link_topology.h b/include/linux/link_topology.h  

I think this filename is too generic. Maybe phy_link_topology.h, or
move it into include/net.

quoted

+struct phy_device *link_topo_get_phy(struct link_topology *lt, int phyindex);
+int link_topo_add_phy(struct link_topology *lt, struct phy_device *phy,
+		      enum phy_upstream upt, void *upstream);
+
+void link_topo_del_phy(struct link_topology *lt, struct phy_device *phy);  

What is the locking for these functions? Are you assuming RTNL? Maybe
add ASSERT_RTNL(); into them to make this clear.

Indeed I assume it, I'll add an assert and document that clearly then.

quoted

diff --git a/include/linux/link_topology_core.h b/include/linux/link_topology_core.h  

Again, i think this filename is too generic.

quoted

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index a16c9cc063fe..7021a0d3d982 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h

@@ -40,7 +40,6 @@
 #include <net/dcbnl.h>
 #endif
 #include <net/netprio_cgroup.h>
-
 #include <linux/netdev_features.h>  

Whitespace change.

quoted

 #include <linux/neighbour.h>
 #include <uapi/linux/netdevice.h>

@@ -52,6 +51,7 @@
 #include <net/net_trackers.h>
 #include <net/net_debug.h>
 #include <net/dropreason-core.h>
+#include <linux/link_topology_core.h>
 
 struct netpoll_info;
 struct device;

@@ -2405,6 +2405,7 @@ struct net_device {
 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
 	struct netprio_map __rcu *priomap;
 #endif
+	struct link_topology	link_topo;
 	struct phy_device	*phydev;
 	struct sfp_bus		*sfp_bus;
 	struct lock_class_key	*qdisc_tx_busylock;

diff --git a/include/linux/phy.h b/include/linux/phy.h
index 3cc52826f18e..d698180b1df0 100644
--- a/include/linux/phy.h
+++ b/include/linux/phy.h

@@ -543,6 +543,8 @@ struct macsec_ops;
  * @drv: Pointer to the driver for this PHY instance
  * @devlink: Create a link between phy dev and mac dev, if the external phy
  *           used by current mac interface is managed by another mac interface.
+ * @phyindex: Unique id across the phy's parent tree of phys to address the PHY
+ *	      from userspace, similar to ifindex. It's never recycled.
  * @phy_id: UID for this device found during discovery
  * @c45_ids: 802.3-c45 Device Identifiers if is_c45.
  * @is_c45:  Set to true if this PHY uses clause 45 addressing.

@@ -640,6 +642,7 @@ struct phy_device {
 
 	struct device_link *devlink;
 
+	int phyindex;  

Is this int, or unsigned int? Is a negative value possible and legal?

You're right this should be unsigned. I've also missed properly
documenting correct values and their meaning (like phyindex 0 would be
no index assigned).

quoted

+enum phy_upstream {
+	PHY_UPSTREAM_MAC,
+	PHY_UPSTREAM_SFP,
+	PHY_UPSTREAM_PHY,
+};  

Please document what these actually mean.

I'll do this as well.

Thank you for reviewing,

Maxime

       Andrew