Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 20 Aug 2009 19:10:02 +0200
From:      marta carbone <marta@freebsd.org>
To:        soc-status@freebsd.org
Subject:   ipfw, status update and final report
Message-ID:  <b62a99a60908201010o23175a7dkdf015b7efff2e3e4@mail.gmail.com>

next in thread | raw e-mail | index | archive | help
During the last week I worked on the ipfw userland configuration tool
and on the userland-kernel interface.

The default behavior of the ipfw configuration tool is to silently accept
different syntax to specify the rules. Now it has a new "strict"
option which implements stronger error checking.

The userland-kernel interface allowed to transfer the whole ruleset for
each userland request.=A0 Rules transfer is now split in two parts,
static and dynamic, allowing to request only a subset of rules. This
reduces the size of transfers and contention on kernel data structures.


Now that the project is at the end, I summarized the work done so far.

- the ipfw and dummynet source code was moved in a separate directory;
- the ipfw_chk() function, composed by a very huge switch statement,
=A0 is now implemented by a dispatching table, making the code more
=A0 readable. Before an after this change, I did a set of performance
=A0 measurements to profile the ipfw_chk() execution times;
- the microinstruction compiler is now built as a library, splitting
=A0 the ipfw compilation and decompilation functions from the socket I/O
=A0 functions;
- the userland ipfw configuration tool is built using such `libipfw'
=A0 library;
- the userland-kernel rule interface allows distinct transfers for
=A0 static or dynamic rules;
- the ipfw userland code has a new options to implement strict
=A0 checks on rule parsing.

The code related to this work was uploaded on the perforce server,
on the soc2009/marta_ipfw project.

marta



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?b62a99a60908201010o23175a7dkdf015b7efff2e3e4>