Re: [RFC PATCH net-next v2 06/10] net: ethtool: Introduce a command to list PHYs on an interface

On Tue, 21 Nov 2023 02:40:27 +0100
Andrew Lunn [off-list ref] wrote:

quoted

+ethnl_phy_reply_size(const struct ethnl_req_info *req_base,
+		     struct netlink_ext_ack *extack)
+{
+	struct phy_device_node *pdn;
+	struct phy_device *phydev;
+	struct link_topology *lt;
+	unsigned long index;
+	size_t size;
+
+	lt = &req_base->dev->link_topo;
+
+	size = nla_total_size(0);
+
+	xa_for_each(&lt->phys, index, pdn) {
+		phydev = pdn->phy;
+
+		/* ETHTOOL_A_PHY_INDEX */
+		size += nla_total_size(sizeof(u32));
+
+		/* ETHTOOL_A_DRVNAME */
+		size += nla_total_size(strlen(phydev->drv->name));
+
+		/* ETHTOOL_A_NAME */
+		size += nla_total_size(strlen(dev_name(&phydev->mdio.dev)));
+
+		/* ETHTOOL_A_PHY_UPSTREAM_TYPE */
+		size += nla_total_size(sizeof(u8));
+
+		/* ETHTOOL_A_PHY_ID */
+		size += nla_total_size(sizeof(u32));
+
+		if (phy_on_sfp(phydev)) {
+			/* ETHTOOL_A_PHY_UPSTREAM_SFP_NAME */
+			if (sfp_get_name(pdn->parent_sfp_bus))
+				size += nla_total_size(strlen(sfp_get_name(pdn->parent_sfp_bus)));  

Have you tried a modular build?

Now that you mention it, no. I did try with CONFIG_SFP disabled, but
not as a module, I'll add it to my test suite.

sfp_get_name() could be in a module, and then you will get linker
errors. It is all a bit messy calling into phylib from ethtool :-(

This might actually be the only function you need? If so, its small
enough you can move it into a header as a static inline function.

It's the only one indeed, so as add it as a header function then.

Thanks,

Maxime

       Andrew