Le Wed, 13 Aug 2025 14:44:27 +0200,
Oleksij Rempel [off-list ref] a écrit :

quoted

+int pse_ntf_cb(const struct nlmsghdr *nlhdr, void *data)
+{
+	const struct nlattr *tb[ETHTOOL_A_PSE_MAX + 1] = {};  

s/ETHTOOL_A_PSE_MAX/ETHTOOL_A_PSE_NTF_MAX ?

Thanks, well spotted.

quoted

+	struct nl_context *nlctx = data;
+	DECLARE_ATTR_TB_INFO(tb);
+	u64 val;
+	int ret, i;
+
+	ret = mnl_attr_parse(nlhdr, GENL_HDRLEN, attr_cb, &tb_info);
+	if (ret < 0)
+		return MNL_CB_OK;
+
+	if (!tb[ETHTOOL_A_PSE_NTF_EVENTS])
+		return MNL_CB_OK;
+
+	nlctx->devname = get_dev_name(tb[ETHTOOL_A_PSE_HEADER]);  

s/ETHTOOL_A_PSE_HEADER/ETHTOOL_A_PSE_NTF_HEADER ?

Thanks, well spotted.

quoted

+	if (!dev_ok(nlctx))
+		return MNL_CB_OK;
+
+	open_json_object(NULL);
+	print_string(PRINT_ANY, "ifname", "PSE event for %s:\n",
+		     nlctx->devname);
+	open_json_array("events", "Events:");
+	val = attr_get_uint(tb[ETHTOOL_A_PSE_NTF_EVENTS]);  

we have here uint but val is u64, is it as expected?

Yes, same behavior in Linux using nla_put_uint().

quoted

+	for (i = 0; 1 << i <= ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR; i++)
+		if (val & 1 << i)
+			print_string(PRINT_ANY, NULL, " %s",
+				     pse_events_name(val & 1 << i));  

Hm, may be it is better to not limit to ETHTOOL_PSE_EVENT_SW_PW_CONTROL_ERROR
and report unknow numeric value. It will keep even old ethtool at least
partially usable.

Ok, I will loop until UINT_MAX then.

Regards,
-- 
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com