cvs-src summaries misposted

From: Mark Johnston <mjohnston_at_skyweb.ca>
Date: Mon, 29 Mar 2004 10:00:00 -0600
I had unsubscribed from -current a couple of weeks ago, forgetting all
about the fact that I wouldn't be able to post the cvs-src summaries.  I
just realized this now, but it looks like I broke last week's posting,
too - my apologies to anyone that thought I had disappeared.  Anyway,
I've resubscribed and resolved to use nomail in the future.  Sorry for
the gap.

I'm attaching the summaries for March 21st and 28th in plain text; as
usual, they're also available at http://www.xl0.org/FreeBSD/ (where they
were actually posted on time.)

Mark

FreeBSD cvs-src summary for XX/XX/XX to YY/YY/YY
++++++++++++++++++++++++++++++++++++++++++++++++
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
============
PC Card ATA/SATA controller support
-----------------------------------
Søren Schmidt added support to the ATA code for detaching PCI ATA
controllers.  This allows PC card (Cardbus) ATA and Serial ATA controllers
to be used, with similar speeds to PCI controllers.

http://docs.freebsd.org/cgi/mid.cgi?200403151203.i2FC3mrZ046442

Intel ACPI-CA 20040311 imported
-------------------------------
Nate Lawson (njl) imported Intel's ACPI-CA ACPI support code, version
20040311.  ACPI-CA upgrades like this one improve hardware support,
especially on laptops.

http://docs.freebsd.org/cgi/mid.cgi?200403181742.i2IHgK8r043974

Aironet Arlan 655 wireless NIC support
--------------------------------------
Max Khon (fjoe) committed the arl driver for the Aironet Arlan 655
wireless network adapter.  He also added the arlconfig program to control
Airolan cards.

http://docs.freebsd.org/cgi/mid.cgi?200403152224.i2FMOSQh002806

Mixer settings saved and restored across reboots
------------------------------------------------
John Baldwin (jhb) added an rc script to save the audio mixer settings
at shutdown and restore them when starting up.  Also, the command
``/etc/rc.d/mixer reload`` can be used to restore the last saved settings.

http://docs.freebsd.org/cgi/mid.cgi?200403152309.i2FN9I5n012369

===============
Notable changes
===============
RAIDFrame removed
-----------------
Scott Long (scottl) removed the RAIDFrame software RAID driver, since it
hasn't been working since the GEOM disk API was introduced, and nobody
has volunteered to fix it.  For software raid, you should use the `Vinum`_
package, which is included with the base system.

.. _`Vinum`: http://www.vinumvm.org/

http://docs.freebsd.org/cgi/mid.cgi?200403161223.i2GCNh6a010477

=================
Discussion topics
=================
Opinions on the Intel C Compiler
--------------------------------
As a follow-up to last week's commit of `ICC support for kernel building`,
David O'Brien (obrien) noted that Intel C Compiler (ICC) version 8 will
build code that either doesn't work or disables all optimizations on AMD
processors.  Alexander Leidinger (netchild) responded that Intel has found
some problems with ICC on AMD processors; David stated that Intel is
deliberately breaking compatibility with AMD chips.  Alexander suggested
that different releases could be built for different processors, allowing
the use of the ICC optimizations without the incompatibility problems.
Scott Long (scottl) brought up the point that even if release kernels were
built with ICC, users often rebuild the kernels, so they would lose the
speed benefits of ICC.  Garance A. Drosihn (gad) pointed out that even if
releases are still built with GCC, the ability to build the system with
other compilers provides a benchmark to compare with GCC.

.. `ICC support for kernel building`: http://excel.xl0.org/FreeBSD/14-03-04.html#intel-c-compiler-support-added-for-kernel-building

http://docs.freebsd.org/cgi/mid.cgi?20040315033213.GA40858

Finding a home for dvcfg.h and physio_proc.h
--------------------------------------------
Last week, Peter Wemm (peter) moved the files machine/dvcfg.h and
machine/physio_proc.h from various machine-dependent locations to a common
machine-independent area in compat/netbsd.  Yoshihiro TAKAHASHI (nyan)
pointed out that the files came from from NetBSD/pc98, but aren't actually
for NetBSD compatibility, and that they aren't PC98 specific.  David
O'Brien (obrien) suggested that sys/dev/<device_name>/ would be a good
place for the files, but he didn't know what that device name should be.
Noriaki Mitsunaga (non) agreed that the files are not PC98 specific, and
suggested sys/ or dev/dev/ for the files.  Peter specified that he didn't
want to put the files under sys/, and suggested dev/include.  M. Warner
Losh (imp) agreed.  John Baldwin (jhb) wondered whether the files even
need to exist; he suggested that the outdated APIs could be replaced by
other code with similar functionality.

http://docs.freebsd.org/cgi/mid.cgi?200403131946.i2DJkRit022925

Specifying the serial console and debug port without hints
----------------------------------------------------------
Marcel Moolenaar (marcel) committed code to replace the use of hints
to select a serial console and debug port with the loader variables
hw.uart.console and hw.uart.dbgport, since hints have a side effect of
reserving a unit number for ISA or ACPI devices, but not PCI, and they
don't have enough flexibility to select non-standard ports.  John-Mark
Gurney (jmg) asked for more explanation, saying that hints are supposed
to tie down a device so it doesn't change between reboots, and that they
should work on all bus types.  Marcel explained that hints are used
to tell the kernel about devices it can't always detect on its own;
the PCI bus doesn't have this problem, so it doesn't need hints.  He
pointed out that the console settings are used before device listings are
taken on busses, so a unit number would be meaningless.  M. Warner Losh
(imp) agreed, adding that this scheme works only for devices with fixed
addresses.  Marcel pointed out that devices without fixed addresses can't
be used as low-level consoles.

http://docs.freebsd.org/cgi/mid.cgi?200403200214.i2K2E3ps052217

=================
Committer changes
=================
Nate Lawson (njl) introduced Mark Santcroos (marks), who he will be mentoring.

===================
Important bug fixes
===================
Denial-of-service vulnerabilities in OpenSSL patched
----------------------------------------------------
Jacques A. Vidrine (nectar) committed patches to the branches for 5.2,
5.1, 4.8, 4.9, and -STABLE to fix the denial-of-service vulnerabilities
reported in the `OpenSSL security advisory`_ of March 17th.  He also
imported OpenSSL 0.9.7d, which includes the fix, to -CURRENT.

.. _`OpenSSL security advisory`: http://www.openssl.org/news/secadv_20040317.txt

Import of fix: http://docs.freebsd.org/cgi/mid.cgi?200403171211.i2HCB8AN067605
OpenSSL 0.9.7d: http://docs.freebsd.org/cgi/mid.cgi?200403171550.i2HFo0Rm023201

Information leak in process separation fixed
--------------------------------------------
Pawel Jakub Dawidek (pjd) committed a fix to the kernel code for
separating processes.  This fix closes a hole that allowed processes to
circumvent restrictions on which processes can see which, like the sysctl
security.bsd.see_other_uids and the jail subsystem.

http://docs.freebsd.org/cgi/mid.cgi?200403171319.i2HDJihF089490

Installation on disks and filesystems larger than 1 TB fixed
------------------------------------------------------------
John Baldwin (jhb) committed code submitted by Julian Elischer (julian)
that changes sysinstall to use 64-bit integers for disk addresses.  This
allows sysinstall to work properly on disks and filesystems larger than 1
TB.

http://docs.freebsd.org/cgi/mid.cgi?200403161707.i2GH76qZ087362

panic: lockmgr: locking against myself (kern_lock.c:370) fixed
--------------------------------------------------------------
Alexander Kabaev (kan) fixed a bug in the FFS code that could cause a
panic under heavy disk activity while a filesystem snapshot exists; for
instance, during background fsck.  This commit closes `PR 64206`, reported
by Nik Clayton (nik).

.. _`PR 64206`: http://www.freebsd.org/cgi/query-pr.cgi?pr=64206

http://docs.freebsd.org/cgi/mid.cgi?200403162206.i2GM6WWF061249

===============
Other bug fixes
===============
John Baldwin (jhb) fixed bsdlabel to create working, bootable disklabels
for the Alpha.

http://docs.freebsd.org/cgi/mid.cgi?200403152310.i2FNAYO7012470

Bill Paul (wpaul) mostly fixed his NDISulator code for the Intel Centrino
2200BG wireless chip.  The only remaining problem is that the driver
can use up too much stack space, causing a panic; adding ``options
KSTACK_PAGES=8`` to the kernel configuration file is a temporary fix.

http://docs.freebsd.org/cgi/mid.cgi?200403202339.i2KNdhQI069350

FreeBSD cvs-src summary for XX/XX/XX to YY/YY/YY
++++++++++++++++++++++++++++++++++++++++++++++++
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
============
truss enhanced to decode more structures
----------------------------------------
David Malone (dwmalone) committed code to truss, which traces the system
calls a program is executing, that allows it to decode many structures
passed to and from system calls, producing more useful output.  The code
was initially contributed by Dan Nelson in `PR 52190`_, and has been
modified somewhat by David.

.. _`PR 52190`: http://www.freebsd.org/cgi/query-pr.cgi?pr=52190

http://docs.freebsd.org/cgi/mid.cgi?200403231237.i2NCb3vw085027

Build target added to build the toolchain only
----------------------------------------------
Ruslan Ermilov (ru) added a target to the system Makefile called
"toolchain", which does the same thing as buildworld, but stops before
actually building the world; i.e., it rebuilds only the tools and
libraries.  With these changes, individual parts of world can be built
alone, or the kernel can be built without the world.  To find out how to
do this, see a `follow-up from Ruslan`.  The toolchain target should be
used only if you're sure you know what you're doing; the only supported
upgrade path is still as `documented in the Handbook`.

.. _`follow-up from Ruslan`: http://docs.freebsd.org/cgi/mid.cgi?20040324151245.GD42844
.. _`documented in the Handbook`: http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/makeworld.html

http://docs.freebsd.org/cgi/mid.cgi?200403240826.i2O8QMoF087714

Multicast group support added to netstat
----------------------------------------
Bruce M. Simpson (bms) added support to netstat to print the multicast
groups that the system belongs to.  This function is called automatically
when netstat is used, so no new command-line flags are necessary.

http://docs.freebsd.org/cgi/mid.cgi?200403250843.i2P8hx6g047670

pkill and pgrep commands added
------------------------------
Garance A. Drosehn added the pkill and pgrep commands from NetBSD,
applying a number of patches written by Mario Sergio Fujikawa Ferreira
(lioux) for the sysutils/pkill port. pgrep is used to search the process
table on the system, like ps | grep, and pkill is used to kill processes
matching a given pattern, like killall.

http://docs.freebsd.org/cgi/mid.cgi?200403251908.i2PJ8m91006049

===============
Notable changes
===============
debug.mpsafenet temporarily out of order
----------------------------------------
Robert Watson (rwatson) added the following note to UPDATING:

The debug.mpsafenet tunable controls whether the kernel Giant lock is held
across the lower levels of the network stack, and by default is turned
off.  In the few days following 20040322, the behavior of debug.mpsafenet
will change such that this tunable controls Giant over all levels of the
network stack.  If you are currently setting debug.mpsafenet to 1, you
should set it back to 0 (the default) again during the change-over.  An
additional note will be added to UPDATING when sufficient locking is
merged to permit this to take place.

http://docs.freebsd.org/cgi/mid.cgi?200403221637.i2MGbXxb090535

UTF-8 versions added for all supported locales
----------------------------------------------
Tim J. Robbins (tjr) committed Unicode UTF-8 versions of all of the
supported system locales.  The UTF-8 locales were previously available in
Hye-Shik Chang (perky)'s misc/utf8locale port.  Most of the work for this
commit was done by Hye-Shik in that port.  For a precise list of the exact
locales affected, please see the commit message below.

http://docs.freebsd.org/cgi/mid.cgi?200403270814.i2R8EGiO061708

Diskless boot system split up
-----------------------------
Brooks Davis (brooks) split the diskless rc script, which sets up
filesystems for a diskless system, into hostname, resolve, tmp, and
var pieces.  He has also improved the scripts to create /tmp and /var
automatically if they aren't writable, rather than triggering when / is
NFS mounted.

http://docs.freebsd.org/cgi/mid.cgi?200403232322.i2NNMZLe056410

Major Firewire enhancements; also MFC'ed
----------------------------------------
Hidetoshi Shimokawa (simokawa) committed many enhancements to the Firewire
code, improving compatibility with 4.x and DragonFly BSD, improving device
probing, and fixing other bugs in both the Firewire and the SBP-2 (serial
Bus Protocol) support required for Firewire-based storage devices.  This
work was also merged to 4.x.

http://docs.freebsd.org/cgi/mid.cgi?200403262317.i2QNHBgY030234

i386 pipe performance significantly improved
--------------------------------------------
Alan Cox (alc) removed some unnecessary code from the pipe system,
adjusting the optimized case.  This eliminates the sysctls
kern.ipc.maxpipekvawired and kern.ipc.amountpipekvawired, and speeds up
pipes on i386 significantly.  For instance, on a 2.4GHz Xeon, Alan saw the
pipe bandwidth increase from 1050 MB/s to 1850 MB/s.

http://docs.freebsd.org/cgi/mid.cgi?200403271950.i2RJoN9P033780

Most PicoBSD floppy image scripts removed
-----------------------------------------
Luigi Rizoz (luigi) removed the net, router, isp, and dial floppy types
for PicoBSD, a compact build of FreeBSD.  In -CURRENT, only a single base
floppy type will be supported, with users building on that to suit their
needs.

http://docs.freebsd.org/cgi/mid.cgi?200403240751.i2O7peAO079157
http://docs.freebsd.org/cgi/mid.cgi?200403240753.i2O7rCg0079966
http://docs.freebsd.org/cgi/mid.cgi?200403240754.i2O7s0mH080045
http://docs.freebsd.org/cgi/mid.cgi?200403240755.i2O7t9Nq080220

=================
Discussion topics
=================
doscmd moved from base system to ports tree
-------------------------------------------
Dag-Erling Smorgrav removed doscmd, a basic DOS emulator, from the base
system, since it has been moved to ports.  This sparked a few threads
of confusion.  David O'Brien (obrien) asked whether there was really
consensus for the change; Dag-Erling said that there had been.  Garance
A. Drosihn (gad) disagreed, saying that he had thought it was going to
stay until 6.0.  Julian Elischer agreed with Garance.  Doug Barton (dougb)
gave his impression of the conversation, ending with Tim J.  Robbins
(tjr)'s suggestion that waiting until 6.0 could work, but if that wasn't
the plan, doscmd should be removed before 5.3.  Julian, John Baldwin
(jhb), and Tim followed up, saying that each had different impressions of
the original discussion, but that they all agreed that doscmd should stay
gone.  Garance and Dag-Erling also posted their impressions of the initial
discussion, differing again on the details, but again not arguing that
doscmd should return.

Alexey Dokuchaev (danfe) expressed his disappointment at the removal,
since he had planned to improve doscmd.  Dag-Erling clarified that doscmd
is still available in the projects repository; Alexey answered that he had
missed that.

http://docs.freebsd.org/cgi/mid.cgi?200403232227.i2NMRQn5042762

=================
Committer changes
=================
Wilko Bulte (wilko), acting as the core secretary, moved Hiten Pandya
(hmp) from being a doc-only committer to a src committer.  Dag-Erling
Smorgrav (des) will be his mentor.

http://docs.freebsd.org/cgi/mid.cgi?200403231935.i2NJZ4BU098762

===================
Important bug fixes
===================
Local denial-of-service bug in VM system fixed
----------------------------------------------
Tim J. Robbins (tjr) fixed a bug in the VM code that allowed a user to
force the kernel to allocate an unlimited amount of memory, causing
a panic or system hang.  Tim has also seen the bug occur without any
malicious user input.  I'm not sure whether other FreeBSD versions are
affected by the bug.

http://docs.freebsd.org/cgi/mid.cgi?200403230837.i2N8bZe4025918

Crash when using setsockopt on IPv6 sockets eliminated
------------------------------------------------------
Hajimu Umemoto committed a fix to the IPv6 code that prevents a crash when
setsockopt() is given invalid parameters.  Before the fix, regular users
could crash the system using setsockopt() on v6 sockets.  This problem was
reported by Katsuhisa ABE in `PR 61513`_.

.. _`PR 61513`: http://www.freebsd.org/cgi/query-pr.cgi?pr=61513

http://docs.freebsd.org/cgi/mid.cgi?200403261952.i2QJqIrp078521

Information leak in process separation fix MFC'ed
-------------------------------------------------
Pawel Jakub Dawidek (pjd) MFC'ed the fix to the information leak problem
`mentioned last week`_.  Again, this fix prevents processes from
circumventing restrictions on seeing one another.

.. _`mentioned last week`: http://excel.xl0.org/FreeBSD/21-03-04.html#information-leak-in-process-separation-fixed

http://docs.freebsd.org/cgi/mid.cgi?200403262207.i2QM7SPX014969

===============
Other bug fixes
===============

Bill Paul (wpaul) committed a fix to the VM subsystem that allows threads
to be created with extra stack pages.  This bug was preventing his fixing
the `problem with the Intel Centrino 2200BG`_ wireless networking chip
mentioned last week.  With the VM problem fixed, the Centrino problem has
been eliminated.

.. _`problem with the Intel Centrino 2200BG`: http://excel.xl0.org/FreeBSD/21-03-04.html#other-bug-fixes

http://docs.freebsd.org/cgi/mid.cgi?200403220028.i2M0ScPc047763

Thomas Moestl (tmm) fixed the isp driver, supporting Qlogic-based SCSI
and fibre channel cards, to work properly on Sparc-64-based systems.  The
problem was reported, and the fix tested, by Marius Strobl.

http://docs.freebsd.org/cgi/mid.cgi?200403232341.i2NNfdlU060190

Bruce M. Simpson (bms) fixed the madvise() system call to work under Linux
binary emulation; it was previously unimplemented.  The fix was submitted
by Matthias Schuendehuette, who reports that it makes the Linux binaries
of StarOffice 7.0 work.

http://docs.freebsd.org/cgi/mid.cgi?200403282143.i2SLhRK6021263

Max Laier (mlaier) imported fixes to the PF packet filter from OpenBSD's
stable branch.  The first fix makes binat work for incoming connections
when a netblock is used as the source in the rule, and the second corrects
a kernel panic when an anchor ruleset is empty.  The first bug was
reported to OpenBSD by Karl O. Pinc.

http://docs.freebsd.org/cgi/mid.cgi?200403281931.i2SJVnQc086555
Received on Mon Mar 29 2004 - 05:59:57 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:49 UTC