Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 25 May 2004 22:00:05 -0500
From:      Mark Johnston <mjohnston@skyweb.ca>
To:        current@freebsd.org
Subject:   cvs-src summary for May 17-24
Message-ID:  <200405252200.05399.mjohnston@skyweb.ca>

next in thread | raw e-mail | index | archive | help
Due to a Canadian statutory holiday yesterday, our entire editorial department 
was out of the office.  He was mostly moving.  :)

Also, I haven't yet produced the HTML version of this summary, since the 
computer I use for that is inconveniently in-a-box.  I hope to upload it 
within the next couple of days.

FreeBSD cvs-src summary for 17/05/04 to 24/05/04
++++++++++++++++++++++++++++++++++++++++++++++++
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
============
geom management utility for GEOM added
--------------------------------------
Pawel Jakub Dawidek (pjd) committed geom, a tool to work with GEOM classes
(libraries that provide functionality through the GEOM disk layer.)  It
works generically by loading a library specific to the desired class at
runtime.

http://www.freebsd.org/cgi/mid.cgi?200405201009.i4KA9uQ1058841

New GEOM libraries added
------------------------
Pawel Jakub Dawidek (pjd) added a STRIPE library, which implements RAID 0
striping, to GEOM.  He also added a CONCAT library, which, in conjunction
with the geom utility mentioned above, replaces the gconcat management
tool.

http://www.freebsd.org/cgi/mid.cgi?200405201020.i4KAKnEO061087

ACPI-CA 20040514 imported
-------------------------
Nate Lawson imported Intel's ACPI-CA ACPI library and utilities, version
20040514.  Upgrades to ACPI-CA generally improve performance and
compatibility on laptops and new motherboards that use ACPI.

http://www.freebsd.org/cgi/mid.cgi?200405250234.i4P2YttN017430

libthr threading library made async-signal-safe
-----------------------------------------------
Mike Makonnen (mtm) modified libthr, one of the system threading
libraries, to make it safe for use with asynchronous signal delivery.
This means that some libthr functions can be called from a signal handler.

http://www.freebsd.org/cgi/mid.cgi?200405201206.i4KC6Hwr091950

smbmsg SMBus message utility added
----------------------------------
Joerg Wunsch (joerg) committed a new tool, called smbmsg, that can send
and receive messages on the SMBus (System Management Bus).

http://www.freebsd.org/cgi/mid.cgi?200405171056.i4HAu6aX016271

eeprom tool for Open Firmware added
-----------------------------------
Marius Strobl (marius) committed eeprom, a new utility used to display or
modify the system configuration stored in EEPROM or NVRAM on Open Firmware
machines.  It has been tested on Sun, but should work on any Open Firmware
platform.  eeprom uses some code from NetBSD.

http://www.freebsd.org/cgi/mid.cgi?200405221656.i4MGu50k062998

Kernel modules working on amd64
-------------------------------
Peter Wemm (peter) tweaked the kernel linker to avoid a potential panic.
This was the last in a series of commits in an ongoing project to make
loadable kernel modules work on the amd64 platform.  Peter also enabled
kernel modules in the amd64 build.

http://www.freebsd.org/cgi/mid.cgi?200405172124.i4HLOeTY074301
http://www.freebsd.org/cgi/mid.cgi?200405172213.i4HMDElo085895

=================
Discussion topics
=================
Naming of the cpu_sched_exit function
-------------------------------------
Thomas Moestl (tmm) fixed a bug in the cpu_sched_exit function.  Julian
Elischer (julian) followed up, suggesting that Thomas rename the function
to cpu_exit2, since it's unrelated to the scheduler.  Thomas clarified
that the function must be called with the scheduler lock held, and does
tasks related to thread switching, making the name appropriate.  Julian
responded, arguing that switching has nothing to do with the scheduler.
John Baldwin (jhb) answered, stating that switching between threads is all
the scheduler does.  Julian explained that the scheduler interface decides
which task to run, but the actual switching code is independent of the
scheduler.

http://www.freebsd.org/cgi/mid.cgi?200405201841.i4KIf730084094

Saving and restoring PCI variables
----------------------------------

M. Warner Losh (imp) tweaked the PCI code to save and restore a number
of typically read-only PCI fields across suspends.  Scott Long (scottl)
responded, asking whether the work would be better handled in a device
driver, since it would benefit only buggy hardware.  Warner posted a piece
of the PCI specification discussing saving and restoring values, and Scott
pointed out that it only specifically mentions the command register and
base address register.  Bruce M. Simpson (bms) replied, asking whether the
behavior could be kept and made optional, since he had once had a laptop
that would lose its PCI configuration when suspending.

Warner answered Scott's post, noting that the spec specifically says
"but not limited to", implying more needs to be done.  He also addressed
Bruce's post, saying that adding an option would make things worse, and
setting out a requirement to see a real device that needs the code changed
before it would be changed.  He also pointed out that Linux and Windows
do the save and restore.  Nate Lawson (njl), in a subthread, agreed with
the new code, saying it is important to save and restore all the fields.
Warner clarified that he is talking only about selected (supposedly
read-only) headers, not all of them.  He also pointed out a contradiction
in the specs: the PCI standard states that some fields are read-only, but
the IDE standard says that one, progif, is writable.  Nate responded that
he thinks only progif would then have to be restored.  Warner replied
that they were both just guessing based on hunches, which is why he saved
and restored everything.  Gavin Atkinson suggested that the restore could
be done only when necessary.  Warner asked how one could tell when it is
necessary, adding that he had still not heard of any chip at all that has
problems when it is restored with the new code.  He will watch for actual
problems to appear, then fix them if they do.

http://www.freebsd.org/cgi/mid.cgi?200405210636.i4L6aadV059034

=================
Committer changes
=================
Oliver Eikemeier (eik) has been given src committer privileges in addition
to his previous ports committership.  Joerg Wunsch (joerg) and Ruslan
Ermilov (ru) will be his mentors.

http://www.freebsd.org/cgi/mid.cgi?200405201525.i4KFPCLr036097

===================
Important bug fixes
===================
CVS security problem fixed
--------------------------
Dag-Erling Smorgrav (des) fixed a problem in the CVS server code that
would allow malicious clients to run arbitrary code on a server running
in CVS pserver mode.  Servers running in RSH or SSH mode only are
not affected.  `FreeBSD-SA-04:10.cvs`_ was issued to describe this
vulnerability.  Dag-Erling also merged the fix to -STABLE, 4.7 to 4.10,
and 5.0 to 5.2.

.. _`FreeBSD-SA-04:10.cvs`: 
ftp://ftp.freebsd.org/pub/FreeBSD/CERT/advisories/FreeBSD-SA-04:10.cvs.asc

http://www.freebsd.org/cgi/mid.cgi?200405190611.i4J6BjwD055409

"Thundering herd" problem with Apache re-fixed
----------------------------------------------
Paul Saab (ps) re-fixed a bug in the socket code that was originally fixed
in 1997, but reintroduced when SYN caching was added.  The bug causes a
`thundering herd`_ problem, where every process listening on a socket is
awakened when a single event comes in, then all but one immediately go
back to sleep.  This shows up most frequently when running Apache.  Paul
also merged this fix to -STABLE and 4.10.

.. _`thundering herd`: 
http://www.catb.org/~esr/jargon/html/T/thundering-herd-problem.html

http://www.freebsd.org/cgi/mid.cgi?200405190022.i4J0MBN1074059

===============
Other bug fixes
===============
Ian Dowse (iedowse) committed a patch by Daan Vreeken fixing a crash when
two of the same USB Ethernet devices were connected.

http://www.freebsd.org/cgi/mid.cgi?200405231235.i4NCZPXn053175

Robert Watson (rwatson) added proper locking to the NFS server code,
making a number of cleanups to it as well.

http://www.freebsd.org/cgi/mid.cgi?200405240406.i4O46EtI086608



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200405252200.05399.mjohnston>