cvs-src summary for August 16-23
mjohnston at skyweb.ca
Tue Aug 24 08:41:12 PDT 2004
FreeBSD cvs-src summary for 16/08/04 to 23/08/04
This is a regular weekly summary of FreeBSD's cutting-edge development.
It is intended to help the FreeBSD community keep up with the fast-paced
work going on in FreeBSD-CURRENT by distilling the deluge of data from
the CVS mailing list into a (hopefully) easy-to-read newsletter. This
newsletter is marked up in reStructuredText_, so any odd punctuation
that you see is likely intended for the reST parser.
.. _reStructuredText: http://docutils.sourceforge.net/rst.html
You can get old summaries, and an HTML version of this one, at
http://www.xl0.org/FreeBSD/. Please send any comments to Mark Johnston
(mark at xl0.org).
If you would like to get the summary without subscribing to current@,
please send mail to freebsd-cvs-summary-subscribe at lists.enderunix.org.
Thanks to Omer Faruk Sen and EnderUNIX for hosting this list.
For Lukasz Dudek and Szymon Roczniak's Polish translations of these
summaries, which may lag the English ones slightly, please see
RELENG_5 branched; HEAD becomes 6-CURRENT
Two small commits by Scott Long (scottl) this week marked two major,
connected events in FreeBSD's history. A new branch, RELENG_5, was
created, marking the transition of FreeBSD 5 to "stable" status. This
"stable" tag means that 5.x releases will become the new production
FreeBSD releases. This is the first new stable branch since RELENG_4 was
branched in March 2000, marking the start of the 4.x branch.
The second commit changed the major number of the CVS HEAD from 5 to
6. -CURRENT FreeBSD now uses the major number 6, the first new major
number since March 2000. This will be the new development branch, leading
towards the eventual release of FreeBSD 6.0.
Thank you and congratulations to all the FreeBSD contributors! Rather
than mention all the roles and teams that make the project work, I highly
recommend you read the `Contributors to FreeBSD`_ article.
.. _`Contributors to FreeBSD`:
Rewritten floppy driver
Poul-Henning Kamp (phk) committed a rewrite of the floppy driver, making
it work with GEOM and not require the Giant system lock in SMP systems.
The new driver also fixes some bugs in the old code and adds some new
features, like sysctl tunables for many drive-related parameters.
make universe now parallelizable
Ruslan Ermilov (ru) enhanced make universe, the Makefile target that
builds world for all the architectures, to be able to build the worlds in
parallel. He gives the following examples of how to use the new build
Builds worlds sequentially, each world sequentially.
make universe JFLAG=-j4
Builds worlds sequentially, each world in parallel.
make -j4 universe
make -j4 universe JFLAG=-j2
Builds four worlds in parallel, each world will be
built in parallel too. World parallelization is
set to four in the first synopsis, and to two in
make -j4 universe JFLAG=-B
Builds worlds in parallel, each world sequentially.
ipfw converted to use PFIL_HOOKS
Andre Oppermann (andre) converted ipfw to use the PFIL_HOOKS packet filter
API. This doesn't change the use or ABI of ipfw, but it does mean that
the PFIL_HOOKS kernel option is required to use the IPFIREWALL option.
Netgraph made 64-bit ready
Julian Elischer (julian) committed code to align the Netgraph message
fields so it will work on 64-bit machines. This change requires a
recompile of any code that uses Netgraph.
The meaning of MFC
In Julian Elischer (julian)'s commit described in `Netgraph made 64-bit
ready`_, he noted, "MFC after: 2 days".
Ruslan Ermilov (ru) replied, "We need to be careful with using "MFC" now.
What does it mean, RELENG_5, RELENG_4, or both? ;)"
Ken Smith (kensmith) followed up, "This problem will continue for a while
too unfortunately. :-(". He clarified, "shortly after the 5.X release
RELENG_5 becomes -stable. HEAD as always will be -current. But RELENG_4
will still be an active development branch and the target of at least
some normal developer-generated commits until we release 4.11. After
4.11 is released the current plan is to turn over RELENG_4 (as well as
RELENG_4_11) to so@ and it becomes an Errata Branch."
Eivind Eklund (eivind) answered, "I suggest replacing "MFC" with 'MT4'
(merge to 4.x) and 'MT5' (merge to 5) for the case where we are indicating
Ruslan pointed out, "Changing to this would lose the 'from' information,
which is quite useful, as we sometimes merge not only from -CURRENT." He
suggested the format "MFC [into <branch>] (after|in):".
Eivind clarified, "I meant for the markers we put into the commit messages
for the reminder script, and that's explictly from the branch the change
is commited to." Regarding Ruslan's suggestion, he said, "I see some
slight disadvantages: It is more to type, and it include information (that
this is -current) that is both unnecessary (it is known from the ids) and
may be wrong if we change release engineering practices."
M. Warner Losh (imp) also replied to Ken's post, noting, "Of course MFC
After just generates email to the person doing the commit as a reminder."
Handling routing message overflows
Robert Watson (rwatson) made a commit allowing the size of the routing
message queue to be set with the sysctl net.route.netisr_maxqlen. He
mentioned, "Default the maximum depth to 256 rather than IFQ_MAXLEN due to
the downsides of dropping routing messages."
Andre Oppermann (andre) replied, "A rtmessage should never ever be
dropped. That would wedge the synchronized state of any userland routing
Peter Grehan (grehan) noted, "That only works with infinite memory or if
you can flow-control the sources of the messages."
Robert also answered Andre's post, saying, "That as may be, but we've
always been able to drop them [ . . . ] as we're often unable to block
when such a message is generated." He asked, "Do you have any measurement
of how deep that socket buffer gets in practice on high volume routers?"
Brooks Davis (brooks) suggested, "It might be useful to add a RTM_FOOBAR
(or similar :-) message to send when we are forced to drop a message."
Important bug fixes
Security hole in lukemftpd fixed
David O'Brien (obrien) imported an updated version of lukemftpd, fixing
a remotely-exploitable `security hole`_. Since lukemftpd isn't compiled
or used in a default install, FreeBSD isn't generally vulnerable to this
hole. The only release including lukemftpd was 4.7; even in 4.7, though,
it was not enabled by default.
.. _`security hole`:
Other bug fixes
Robert Watson (rwatson) fixed a bug that was causing "out of data"
warnings on large UDP packets sent over the loopback interface.
Maxime Henrion (mux) fixed a bug in the lnc driver for AMD Lance/PCnet
Ethernet cards. This corrects a problem with network support in VMware.
More information about the freebsd-current