On Tue, 2025-07-29 at 10:12 +0200, Hannes Reinecke wrote:

[snip]...

quoted

diff --git a/Documentation/networking/tls-handshake.rst

b/Documentation/networking/tls-handshake.rst
index 6f5ea1646a47..cd984a137779 100644

--- a/Documentation/networking/tls-handshake.rst
+++ b/Documentation/networking/tls-handshake.rst

@@ -169,7 +169,8 @@ The synopsis of this function is:

  .. code-block:: c
  
    typedef void	(*tls_done_func_t)(void *data, int status,
-                                   key_serial_t peerid);
+                                   key_serial_t peerid,
+                                   size_t tls_record_size_limit);
  
  The consumer provides a cookie in the @ta_data field of the
  tls_handshake_args structure that is returned in the @data
parameter of

Why is this exposed to the TLS handshake consumer?
The TLS record size is surely required for handling and processing
TLS
streams in net/tls, but the consumer of that (eg NVMe-TCP, NFS)
are blissfully unaware that there _are_ such things like TLS records.
And they really should keep it that way.

So I'd really _not_ expose that to any ULP and keep it internal to
the TLS layer.

Hey Hannes,

Sorry for the delay in response, and thanks for the feedback! Yeah I
agree it was a bad approach from me. It definitely makes more sense to
keep things in the TLS layer. I will try to address this in V2.

Regards,
Wilfred