cvs-src summary for November 22-29

Mon Nov 29 20:32:58 PST 2004

FreeBSD cvs-src summary for 22/11/04 to 29/11/04
++++++++++++++++++++++++++++++++++++++++++++++++
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).

For Lukasz Dudek and Szymon Roczniak's Polish translations of these
summaries, which may lag the English ones slightly, please see
http://mocart.pinco.pl/FreeBSD/.

.. contents::

============
New features
============
New ACPI PCI linking code
-------------------------
John Baldwin (jhb) committed a major rewrite of the ACPI PCI linking
code, which assigns IRQs to and does other handling of PCI devices.  This
new code supports multiple IRQs and other resources per link device and
improves on the IRQ assignment algorithm.  If you are using tunables
to override default IRQ settings, note that the new code routes entire
PCI link devices, rather than individual devices using that link.  John
writes, "For example, to adjust the IRQ that \_SB_.LNKA uses, one would
set 'hw.pci.link.LNKA.irq=10' from the loader."

http://www.freebsd.org/cgi/mid.cgi?200411232226.iANMQioc077290

Low-level TCP information available from sockets
------------------------------------------------
Robert Watson (rwatson) added support for some parts of the TCP_INFO
socket option, as found in Linux 2.6.  This allows processes to query
sockets for some low-level TCP information, like the send, bandwidth, and
congestion windows.

http://www.freebsd.org/cgi/mid.cgi?200411261858.iAQIwlXV032915

periodic security output now includes packets blocked by PF
-----------------------------------------------------------
Max Laier (mlaier) committed code to the daily security run to display
packets blocked by PF, in the way that packets blocked by ipfw are.  A new
variable, named daily_status_security_pfdenied_enable, is now supported in
/etc/periodic.conf; it defaults to YES.  This code was submitted by Clive
Lin (clive).

http://www.freebsd.org/cgi/mid.cgi?200411241841.iAOIfsY5054400

New -m (month) flag added to ncal
---------------------------------
Garrett Wollman (wollman) added support to ncal, which displays calendars
on the console, for a new month flag.  If -m <month_number> is passed to
ncal, it will display the given month of the current year.

http://www.freebsd.org/cgi/mid.cgi?200411232257.iANMvHaD078657

New failure-detection algorithm for Netgraph
--------------------------------------------
Gleb Smirnoff (glebius) committed a new failure detection algorithm
for Netgraph, a modular system to hook together networking functions,
called NG_ONE2MANY_FAIL_NOTIFY.  The new algorithm detects downstream
components that are offline and removes them from or adds them to its list
of downstreams appropriately.

http://www.freebsd.org/cgi/mid.cgi?200411230945.iAN9jmSW027271

Per-jail fstab support
----------------------
Maxime Henrion (mux) changed rc.conf to support an fstab file for each jail.  
He
gave this example::

    jail_foo_mount_enable="YES"
    jail_foo_fstab="/etc/fstab.foo"

He also noted that the second line has no effect in this case, since
/etc/fstab.<jailname> is the default when no fstab is specified.

The code Maxime committed was submitted by Jeremie Le Hen.

http://www.freebsd.org/cgi/mid.cgi?200411232009.iANK9w8G066191

=================
Discussion topics
=================
Beastie boot menu removed
-------------------------
[ This is something of an inflammatory issue, so as you read the summary,
please keep in mind that cvs-src posts represent only some people's
reaction to commits, and that usually the people that post are the
objectors to any commit.  Specifically, the fact that the replies to this
commit were generally in opposition should /not/ be construed as a general
consensus that it was a bad thing.  I have no clue whether more people
consider the menu a bad idea than a good one, but that would be a matter
for more-formal discussion on another list. ]

Scott Long (scottl) removed the Beastie boot menu, saying, "It offends
some and annoys everyone else, and I'm frankly tired of the controversy.
When people ask me why FreeBSD isn't user-friendly, I'll tell them that I
tried.  RIP."

This post prompted many replies; in general opposition to eliminating
the menu were David Xu (davidxu), Jiawei Ye, John Birrell, Johan Karlson
(johan), Gleb Smirnoff (glebius), Alexey Dokuchaev (danfe), Stijn Hoop,
Frank Mayhar, John Baldwin (jhb), and Danny J. Zerkel.  Garance A. Drosihn
(gad), Mike Silbersack (silby), Dag-Erling Smorgrav (des), Robert Watson
(rwatson), Bjoern A. Zeeb (bz), and Roman Kurakin (rik) generally
suggested that the boot menu would be as useful without the ASCII-art
Beastie, for reasons ranging from possible offense to slow display, and
that a plain text menu might be a suitable way to revive it.

Scott noted, in response to this last suggestion, "I had patches in the
works to make the image optional. [ . . . ] However, there were a quite a
few complaints about the timer, text, options, and keys that had nothing
to do with the image."

In response to Scott's comments earlier about ease-of-use, M. Warner Losh
(imp) suggested, "One could easily have a 'safeboot' command, which would
disable all the things that the safe boot menu item used to do."

Scott replied, "safeboot isn't the same as 'disable acpi'.  It was
actually quite useful to disable ACPI but still leave ATA-DMA and SMP
enabled."

Warner responded, "Agreed.  A few wrappers would go a long way is all I'm
saying.."

Scott asked, "Where would these 'wrappers' be documented?"

In another conversation, Poul-Henning Kamp (phk) suspected that it was his
removal of the Beastie menu from NanoBSD that prompted Scott's commit; he
said, "The reason why I disable the beastie menu in some of my nanobsd
builds are that a couple of those systems sit with a 2400bps GSM phone as
console connection. [ . . . ] whatever happens in nanobsd should not be
taken as indicative of anything related to user-friendlyness [ . . . ]."

Scott replied, "It actually had nothing to do with nanobsd.  It has
everything to do with many months of conversations with many people."

In another reply to Scott's initial commit, Eivind Eklund (eivind)
suggested, "I've for a long time been thinking of the benefits of creating
a 'user friendliness officer' similar to the security officer / tech
advisory board we have now.  As it is, we end up with bikesheds over UI
issues all the time, grumbling, and a user interface that's the result of
who shouts loudest."

http://www.freebsd.org/cgi/mid.cgi?200411290132.iAT1WAse056942

=================
Committer changes
=================
Jean-Sebastien Pedron (dumbbell) has joined as a new src committer.
Jean-Sebastien has implemented read-only ReiserFS support for FreeBSD,
which he will continue to work on (among other things) as a committer.
Maxime Henrion (mux) will be his mentor.

http://www.freebsd.org/cgi/mid.cgi?200411292130.iATLUrgT033674

===================
Important bug fixes
===================
Crash or load failure with sk driver on Gigabit-Ethernet-only cards fixed
-------------------------------------------------------------------------
Bjoern A. Zeeb (bz) fixed a bug in the networking code that was preventing
the sk driver, which supports the SysKonnect chip found on many Gigabit
Ethernet boards, from loading properly.  The bug would show up when the
driver was used with boards that supported only Gigabit Ethernet, instead
of Gigabit, Fast, and regular Ethernet.  This fix closes PRs 63313_, 71733_,
and 73725_.

.. _63313: http://www.freebsd.org/cgi/query-pr.cgi?pr=63313
.. _71733: http://www.freebsd.org/cgi/query-pr.cgi?pr=71733
.. _73725: http://www.freebsd.org/cgi/query-pr.cgi?pr=73725

http://www.freebsd.org/cgi/mid.cgi?200411261942.iAQJgeXo039249

Long-standing TCP bug with ignored RSTs fixed
---------------------------------------------
Mike Silbersack (silby) corrected a bug in the TCP code that would cause
the stack to ignore RST (reset) packets if the receive window was 0 bytes
in size.  This could have caused connections to hang open when they should
have been disconnected.  The bug description and fix, as well as a test
case, were submitted by Michiel Boland.

http://www.freebsd.org/cgi/mid.cgi?200411251904.iAPJ4KXf045915

Crash in PF when using bidirectional NAT on dynamic interfaces fixed
--------------------------------------------------------------------
Daniel Hartmeier (dhartmei) fixed a bug in PF, the OpenBSD packet filter,
that can cause a crash.  The crash was caused by bidirectional NAT rules
in the form "binat from ... to ... -> (interface)" were used with a
dynamic interface.  The bug was discovered by kos at bastard dot net, and
analyzed by Pyun YongHyeon (yongari).

http://www.freebsd.org/cgi/mid.cgi?200411240043.iAO0hYv4082483