On Tue, Nov 17, 2015 at 09:26:57PM +0000, Short, Todd wrote:
> Embedded systems don�t have the luxury of mbuf-type of buffer scheme (as
> you describe for NSS). Many have ethernet-frame sized buffers in
> locked/pinned memory that read in a whole ethernet frame, and then strip
> off headers by advancing pointers into the frame. This minimizes copies,
> and the goal is to have a zero-copy network stack.
Right, both IP and TCP add padding after options to align the
payload on a 32-bit boundary. So the TLS record layer always starts
at a 32-bit aligned address, which ensures that the record payload
is always misaligned. Padding the record header with zeros to the
next multiple of 32-bits could plausibly improve the performance
of bare-metal zero-copy implementations.
--
Viktor.
_______________________________________________
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls
