On Thu, 15 Jun 2023 10:36:51 +0200
Marcin Szycik [off-list ref] wrote:

On 14.06.2023 18:54, Stephen Hemminger wrote:

quoted

On Wed, 14 Jun 2023 11:17:58 +0200
Marcin Szycik [off-list ref] wrote:
  

quoted

+static void flower_print_pfcp_opts(const char *name, struct rtattr *attr,
+				   char *strbuf, int len)
+{
+	struct rtattr *tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_MAX + 1];
+	struct rtattr *i = RTA_DATA(attr);
+	int rem = RTA_PAYLOAD(attr);
+	__be64 seid;
+	__u8 type;
+
+	parse_rtattr(tb, TCA_FLOWER_KEY_ENC_OPT_PFCP_MAX, i, rem);
+	type = rta_getattr_u8(tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE]);
+	seid = rta_getattr_be64(tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_SEID]);
+
+	snprintf(strbuf, len, "%02x:%llx", type, seid);
+}
+  

NAK you need to support JSON output.
Also whet if kernel gives partial data.  

Hmm... when we were adding GTP opts parsing, you requested to remove JSON
support [1]. Are you sure you want us to add it here?

Could you elaborate about partial data? Are there any similar functions
which properly handle partial data?

[1] https://lore.kernel.org/netdev/20220127091541.6667d4d1@hermes.local (local)

The handling of JSON in flower needs lots of cleanup work, not related to this patch.
So changing my mind. This patch is OK, it just continues existing unique model of
handling.

Ideally, there would not be all the special case handling of JSON in flower.
And the code would be more robust about getting partial attributes back from
the kernel.  Other parts of tc are not as brittle.