On 9/13/25 12:27 AM, Jakub Kicinski wrote:

On Thu, 11 Sep 2025 20:05:01 +0000 Asbjørn Sloth Tønnesen wrote:

quoted

+int ynl_attr_validate_payload(struct ynl_parse_arg *yarg,
+			      const struct nlattr *attr, unsigned int type)
+{
+	return __ynl_attr_validate(yarg, attr, type);
+}

Why not expose __ynl_attr_validate() to the callers?
I don't think the _payload() suffix is crystal clear, we're still
validating attr, _payload() makes it sound like we're validating
what's inside attr?

I didn't wanna call __ynl_attr_validate() directly, as the only __ynl_*
function in ynl-priv.h is __ynl_attr_put_overflow(), and that is only
used in other static functions within that file. I agree, that _payload()
might not be the best given that we currently don't look deeper than
validating that the length a bit, so maybe _length() would have been
better.

In v4, I have changed it to just expose __ynl_attr_validate() in
ynl-priv.h, and changed ynl_attr_validate() to an inline function.