cvs-src summary for July 12-19

From: Mark Johnston <mjohnston_at_skyweb.ca>
Date: Mon, 19 Jul 2004 20:47:48 -0500
FreeBSD cvs-src summary for 12/07/04 to 19/07/04
++++++++++++++++++++++++++++++++++++++++++++++++
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
============
libthread_db thread debugging library added
-------------------------------------------
David Xu (davidxu) added libthread_db, a library for debugging threaded
processes with GDB.  Currently, it supports only libpthread, but Marcel
Moolenaar (marcel) will be working to add support for libthr and libc_r,
the other threading libraries.

http://www.freebsd.org/cgi/mid.cgi?200407150336.i6F3aZIw036415

bsdtar becomes the default system tar
-------------------------------------
Tim Kientzle (kientzle) changed the system tar from the old GNU tar to
his new bsdtar program.  You can use the old GNU tar as gtar, or define
WITH_GTAR when building and installing the world to make gtar the default.

bsdtar is built on Tim's libarchive, which supports many different
streaming archive formats, auto-detecting what format you're trying to
extract.  It also supports file flags, ACLs, and arbitrary device numbers.

For more information on libarchive and bsdtar, you can visit `the
libarchive page`_.

.. _`the libarchive page`: http://people.freebsd.org/~kientzle/libarchive/

http://www.freebsd.org/cgi/mid.cgi?200407170603.i6H63lkv029158

Adaptive mutexes enabled
------------------------
Scott Long (scottl) enabled the code supporting adaptive mutexes, which
improve performance when locking in many circumstances.  It has been
enabled by default on AMD64, but it is now enabled everywhere.  This
also removes the kernel option ADAPTIVE_MUTEXES and creates one for
NO_ADAPTIVE_MUTEXES in its place.

http://www.freebsd.org/cgi/mid.cgi?200407181559.i6IFx42N094234

ACPI support for floppy drives added
------------------------------------
Nate Lawson (njl) committed ACPI support for floppy disk drives.

http://www.freebsd.org/cgi/mid.cgi?200407151638.i6FGc7Sb061577

RADIUS accounting support in PPP enhanced
-----------------------------------------
Brian Somers (brian) committed a patch to PPP to improve the RADIUS
accounting support.  There is now a "set rad_alive N" command that enables
periodic reports of RADIUS accounting data to the server.  The code for
this was mostly submitted by alx_at_sm.ukrtel.net.

http://www.freebsd.org/cgi/mid.cgi?200407170107.i6H17rub019304

===============
Notable changes
===============
fla driver for M-Systems DiskOnChip removed
-------------------------------------------
Poul-Henning Kamp (phk) removed the fla driver for M-Systems DiskOnChip
devices.

http://www.freebsd.org/cgi/mid.cgi?200407081638.i68GcbMo019165

Biba and MLS MAC _single labels renamed to _effective
-----------------------------------------------------
Robert Watson (rwatson) renamed the label _single to _effective in the
code implementing the Biba and MLS mandatory access control policies.  The
new _effective label is similar to the use of "effective" in traditional
UNIX access control.

http://www.freebsd.org/cgi/mid.cgi?200407160203.i6G23oaR078694

Sound device drivers renamed
----------------------------
Seigo Tanimura (tanimura) renamed the sound drivers to make them more
consistent, with "sound" being the generic sound driver, and "snd_*" being
device-specific drivers.  He has also renamed the "snd_pcm" kernel module
to "sound" for consistency.

http://www.freebsd.org/cgi/mid.cgi?200407160358.i6G3wPIl082809

make world now requires DESTDIR
-------------------------------
Marcel Moolenaar (marcel) committed a patch to the system Makefile that
requires a DESTDIR variable, for destination directory, to be specified
when using the "make world" command.  This is to protect people who
unknowingly try to use "make world" to upgrade their system, instead of
following the instructions in /usr/src/UPDATING, since make world does not
upgrade the kernel and can leave the system in an unusable or unbootable
state.

http://www.freebsd.org/cgi/mid.cgi?200407160251.i6G2pSts080488

=================
Discussion topics
=================
kldunload and Newbus
--------------------
Poul-Henning Kamp (phk) made a commit to "Give kldunload a -f(orce)
argument."  This adds new semantics for modules to be asked whether they
can unload (called "quiesce") and then unloaded forcefully.

Nate Lawson (njl) replied, "Have you kept up on the newbus discussions?
The tentative plan was to add quiesce functionality to it as part of
device_detach(). [ . . . ] I think duplicating this at multiple layers is
not a good idea and the module level is not the right layer to implement
it."

Scott Long (scottl) said, " It typically won't be duplicated.  This
particular work is good for things like GEOM modules that don't fit into
newbus."

Poul-Henning replied, "I wish I could grasp from where this 'newbus fits
all' attiude comes, I certainly see nothing anywhere that could explain
it."

Scott answered, "NetBSD has most things under newbus."

Poul-Henning replied, "It is not realistic to expect us to ever tie
netgraph modules, tty linedisciplines, network protocols, filesystems, vm
objects, vnodes and geom classes together with newbus. [ . . . ] And since
all of these things are involved in loadable modules, I think we should
stop wasting time dreaming about the day the world will be one bit tangle
of newbus."

Scott said, "My original statement was just a hypothetical and nothing
more. The newbus cabal/bikeshed/whatever is not about to come crashing
down on your head.", to which Poul-Henning answered "I know it isn't, and
that's was my point."

M. Warner Losh (imp) also replied to Poul-Henning's previous post, saying,
"I think you should spent less time attacking newbus."  Doug Rabson (dfr)
added, "And besides, newbus isn't going to take over the world, kobj is
;-)"

In another branch, Warner responded to Poul-Henning's question about
"newbus fits all", saying, "Too many years of having weird, odd-ball
attachments maybe? [ . . . ] Frankly, I dislike your obstructionist
attitudes towards newbus."

Poul-Henning answered, "Obstructionist ? That's a pretty harsh word for
somebody who has wasted a couple of months trying to write a newbus-ified
GEOM."

Warner clarified, "It just seems that whenever the topic of newbus comes
up, you don't want to use it."

Poul-Henning responded, "having tried it out [ . . . ], I appreciate the
size of the gap between what newbus can today and what it needs to learn
before the fully newbusified kernel can be a realistic dream.  That is not
obstructionist, that is being realistic."

Nate followed up to Poul-Henning's post about the newbus-ified GEOM,
saying, "I'll happily help review patches/designs for this."

Poul-Henning responded, "No need to.  It doesn't work because newbus
doesn't support multiple interitance."

Some highly-technical discussion of newbus between Poul-Henning and Warner
followed, in which they concluded that the multiple inheritance added a
few months ago would meet Poul-Henning's needs.  Robert Watson (rwatson)
added a suggestion at the end that further discussion of the Newbus
architecture would be well-suited to the arch_at_ list.

A second, shorter thread also spawned from Nate's original post, with
Poul-Henning explaining that the quiesce functionality is needed at the
module layer because "one kld can contain multiple modules, and since the
modules get to veto an unload with MOD_UNLOAD, I don't really see how you
can come up with something that doesn't have a MOD_QUIESCE."

Nate answered ,"there doesn't have to be MOD_QUIESCE.  The normal
MOD_UNLOAD request will trigger (multiple calls to) device_detach via a
TBD kernel interface."

Poul-Henning replied, "This is not about 'eject ed0', this is about 'evict
the ed driver'.  There is only one sane way to implement that:  Pass the
flag to the module and have the module decided if it is possible."

Colin Percival (cperciva) observed, "this will be useful from the point of
view of security as well.  When there is a security advisory concerning
an optional module (compat/linux being the most recent example) it would
be nice to have a more effective workaround than 'well, you can try
`kldunload linux`, but it probably won't work... if it doesn't work,
you'll have to reboot the system'."

Brian Fundakowski Feldman (green) pointed out, "'Well, you can try
'kldunload -f linux', but it probably won't work... if it doesn't work, I
guess a crash isn't really much worse than a reboot.'"

http://www.freebsd.org/cgi/mid.cgi?200407131936.i6DJaxrp074391

=================
Committer changes
=================
Julian Elischer (julian) introduced Gleb Smirnov (glebius) as a new src
committer.

http://www.freebsd.org/cgi/mid.cgi?200407140650.i6E6om5m093425

Murray Stokely (murray) retired Valentino Vaschetto (logo)'s commit bit
for safekeeping.

http://www.freebsd.org/cgi/mid.cgi?200407150845.i6F8j5gH047574

===================
Important bug fixes
===================
Several rpc.lockd bugs fixed
----------------------------
Michael Reifenberger (mr) committed a patch from Bjorn Gronvall that fixes
some lockups and crashes within rpc.lockd, the NFS locking server.  This
closes `PR 61718`_.

.. _`PR 61718`: http://www.freebsd.org/cgi/query-pr.cgi?pr=61718

http://www.freebsd.org/cgi/mid.cgi?200407161930.16GJUxuw009400

===============
Other bug fixes
===============
Christian S. J. Peron (csjp) applied a patch to the file descriptor
allocation code to prevent a jailed root user from using up the last few
file descriptors, which are reserved for root.

http://www.freebsd.org/cgi/mid.cgi?200407141904.i6EJ4VKD016422

Max Laier (mlaier) committed a fix to the pf packet filter, which
improves routing loop detection.  The fix was written by Daniel Hartmeier
(dhartmei) and imported from OpenBSD.

http://www.freebsd.org/cgi/mid.cgi?200407171715.i6HHFGXJ052457

Soeren Schmidt (sos) committed a fix for newer VIA IDE hard drive
controllers, which would hang when their timing setting was reprogrammed
to a higher speed than UDMA33.  This close `PR 69180`_, although he didn't
use the included patch.

.. _`PR 69180`: http://www.freebsd.org/cgi/query-pr.cgi?pr=69180

http://www.freebsd.org/cgi/mid.cgi?200407171448.i6HEmkMp044631

Brian Somers (brian) fixed the "set ifaddr" command, which had previously
been doing nothing, in ppp.  This closes `PR 62050`_.

.. _`PR 62050`: http://www.freebsd.org/cgi/query-pr.cgi?pr=62050

http://www.freebsd.org/cgi/mid.cgi?200407150942.i6F9gH1Y049408
Received on Mon Jul 19 2004 - 23:48:16 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:02 UTC