FreeBSD Status Report Jan-Mar 2005
scottl at samsco.org
Thu Apr 21 10:52:26 PDT 2005
-------------- next part --------------
January-April 2005 Status Report
The first quarter of 2005 has been extremely active in both
FreeBSD-CURRENT and -STABLE. With FreeBSD 5.4 in the final RC stage
and an anticipated branch of FreeBSD-6 this summer we have seen a lot
of performance improvements in 5 and a couple of exciting new features
The report turnout was extremely good and it seems that the webform
provided by Julian Elischer has made it more enjoyable to write
reports. Many thanks to Julian for providing this. We also like to get
your attention to the open tasks section provided in some reports.
On special note, please take a look at the report about the upcoming
BSDCan in Ottawa. There will be lots of interesting FreeBSD related
talks and activities. If you enjoy reading these reports, you will
love the conference. See you there!
Thanks to all the reporters, we hope you enjoy reading.
* Common Address Redundancy Protocol - CARP
* FreeBSD Java Project
* FreeBSD Release Engineering
* GELI - GEOM class for providers encryption
* GSHSEC - GEOM class for handling shared secret
* Secure Updating
* FreeBSD Dutch Documentation Project
* Coverity Code Analysis
* Filesystem journalling for UFS
* Infrastructure Cleanup
* Interrupt Latency
* Low-overhead performance monitoring for FreeBSD
* Many subdirs for UFS
* Status Report for FreeBSD ATA driver project
* Storage driver SMPng locking
* IPv6 Support for IPFW
* Move ARP out of routing table
* netgraph(4) status report
* Removable interface improvements.
* Support for telephone hardware (aka Zaptel)
* Wireless Networking Support
* Pipe namespace added to portalfs
* ARM Support for TS-7200
* PowerPC Port
* XenFreeBSD - FreeBSD on Xen
* Ports Collection
* Update of the Linux userland infrastructure
Vendor / 3rd Party Software
* OpenBSD packet filter - pf
* twa driver
* FreeBSD Security Officer and Security Team
* IMUNES - a FreeBSD based kernel-level network topology emulator
ARM Support for TS-7200
Contact: John-Mark Gurney <jmg at FreeBSD.org>
I have been working on getting FreeBSD/arm running on the TS-7200. So
far the board boots, and has somewhat working ethernet (some
unexplained packet loss). I can netboot from a FreeBSD/i386 machine,
and I can also mount msdosfs's on CF.
1. Figuring out why some small packets transmit with error
2. EP93xx identification information to properly attach various
Contact: Thomas Quinot <thomas at freebsd.org>
ATAPI/CAM integration with the new ATA (mkIII) framework is now
completed. ATAPI/CAM is now available as a loadable module
(atapicam.ko). It is also independant from the native ATAPI drivers
again, as was the case before mkIII.
Thanks to Scott Long and Søren Schmidt for their participation in the
Contact: Dan Langille <dan at langille.org>
BSDCan made a strong debut in 2004 . The favorable reception gave us a
strong incentive for 2005 . We have been rewarded with a very
interesting program and a higher rate of registrations.
Percentage-wise, we have more Europeans than last year as they have
decided that the trip across the Atlantic is worth taking. We know
they won't be disappointed. See you at BSDCan 2005!
1. volunteers needed for the conference
Common Address Redundancy Protocol - CARP
Contact: Max Laier <mlaier at FreeBSD.org>
Contact: Gleb Smirnoff <glebius at FreeBSD.org>
CARP is an alternative to VRRP. In contrast to VRRP it has full
support for IPv6 and uses crypto to protect the advertisements. It was
developed by OpenBSD due to concerns that the HSRP patent might cover
VRRP and CISCO might defend its patent. CARP has, since then, improved
a lot over VRRP.
CARP has been committed to HEAD and MFCed to RELENG_5. It will be
available in upcoming 5.4-RELEASE.
Big thanks to all users who provided testing and reported bugs to Max
and Gleb. Daniel Seuffert has donated hardware to Max for this
project. Gleb's work was sponsored by Rambler .
1. Improve vlan(4) support. Test ng_eiface(4).
2. Improve locking, consider removing interface layer.
Coverity Code Analysis
Contact: Sam Leffler <sam at freebsd.org>
There has been an ongoing effort to review the kernel source code
using Coverity's source code analysis tools (http://www.coverity.com).
These tools check for a variety of problems such as null pointer
dereference, use-after-free of allocated variables, invalid array
references, etc. This work is a joint project between FreeBSD and
Two passes have been completed over the 6-current kernel source code
base and all significant problems have been corrected. These runs were
done in February and March of this year. A few reports of minor
problems await response from outside groups and will be resolved in
time for the first 6.x release. Another analysis run over the kernel
will happen soon. We are looking for a way to use these tools on a
regular basis as they have been helpful in improving the code base.
Thanks to Coverity for their help and especially Ted Unangst. Several
developers have been especially helpful in resolving reports:
Poul-Henning Kamp, David Schultz, Pawel Jakub Dawidek, George V.
Neville-Neil, and Matthew Dodd.
Contact: Nate Lawson <njl>
The cpufreq project was committed to 6-CURRENT in early February and
has undergone bugfixes and updates. It will soon be MFCd to 5-STABLE.
The cpufreq driver provides a unified kernel and user interface to CPU
frequency control drivers. It combines multiple drivers offering
different settings into a single interface of all possible levels.
Users can access this interface directly via sysctl(8), by indicating
to power_profile that it should switch settings when the AC line state
changes, or by using powerd(8).
For example, an absolute driver offering frequencies of 1000 Mhz and
750 Mhz combined with a relative driver offering settings of 100% and
50% would result in cpufreq providing levels of 1000, 750, 500, and
Colin Percival helped with powerd(8), which provides automatic control
of CPU frequencies. The adaptive mode is especially interesting since
it attempts to respond to changes in system load while reducing power
Current hardware drivers include acpi_perf (ACPI CPU performance
states), est (Intel Enhanced SpeedStep for Pentium-M), ichss (Intel's
original SpeedStep for ICH), and powernow (AMD Powernow! K7 and K8
support). Other drivers for relative hardware include acpi_throttle
(ACPI CPU throttling) and p4tcc (Pentium 4 Thermal Control Circuitry)
Thanks to Bruno Ducrot for the powernow driver, Colin Percival for the
est driver, and the many testers who have sent in feedback.
1. We'd appreciate someone with a Transmeta CPU converting the
existing longrun driver to the cpufreq framework. It would also be
good if someone wrote a VIA Longhaul driver. See the Linux
arch/i386/kernel/cpu/cpufreq directory for examples.
2. Various other architectures, including ARM, have CPU power control
that could be implemented as a cpufreq driver.
3. The powerd(8) algorithm is rather simple and we'd appreciate more
help in testing it and alternative algorithms with various
workloads. The -v flag causes powerd to report frequency
transitions and print a summary of total energy used upon
termination. This should help testers profile their algorithms.
Contact: George Neville-Neil <gnn at neville-neil.com>
On the protocol conformance tool I have finally made some progress
getting a scriptable packet library using libnet, and SWIG. This will
hopefully become a port that can then be used to do conformance
testing on protocol stack changes. Qing Li has separately taken up the
ARP rewrite and that will be taken out of the Dingo project pages.
1. Many :-)
Contact: Eric Anholt <anholt at FreeBSD.org>
A DRM update was finally committed to -current on 2005-04-15, after
jhb@ did the necessary fix to vm_mmap. New development drivers were
added for mach64 and r300 (see URL for info). The nearly-finished code
for savage and i915 were also added, but left disconnected from the
build. However, the most visible change is likely the support for
texture tiling, color tiling, and HyperZ on Radeons, which (with
updated userland) likely provide a 50-75% framerate increase in many
1. Find someone with newbus knowledge to figure out why the i915
won't attach to drmsub0.
2. Finish porting the savage driver.
3. Integrate busdma code from Tonnerre (NetBSD).
Filesystem journalling for UFS
Contact: Scott Long <scottl at freebsd.org>
It's time to bite the bullet and admit that fsck is no longer scalable
for modern storage capacities. While a healthy debate can still be had
on the merits and data integrity guarantees of journalling vs.
SoftUpdates, the fact that SoftUpdates still requires a fsck to ensure
consistency of the filesystem metadata after an unclean shutdown means
uptime is lost. While background fsck is available, it saps system
performance and stretched the fsck time out to hours.
Journalling provides a way to record transactions that might not have
fully been written to disk before the system crashed, and then quickly
recover the system back to a consistent state by replaying these
transactions. It doesn't guarantee that no data will be lost, but it
does guarantee that the filesystem will be back to a consistent state
after the replay is performed. This contrasts to SoftUpdates that
re-arranges metadata updates so that inconsistencies are minimized and
easy to recover from, though recovery still requires the traditional
full filesystem scan.
Journalling is a key feature of many modern filesystems like NTFS,
XFS, JFS, ReiserFS, and Ext3, so the ground is well covered and the
risks for UFS/FFS are low. I'm aware that groups from CMU and RPI have
attempted similar work in the past, but unfortunately the work is
either very outdates, or I haven't had any luck in contacting the
groups. Is this absence, I've decided to work on this project myself
in hopes of having a functional prototype in time for FreeBSD 6.0.
The approach is simple and journals full metadata blocks instead of
just deltas or high-level operations. This greatly simplifies the
replay code at the cost of requiring more disk space for the journal
and more work within the filesystem to identify discreet update
points. An important design consideration is whether to make the
journal data and code compatible with the UFS2 filesystem, or to start
a new UFS3 derivative. Since the latter presents a very high barrier
to adoption for most people, I'm going to try to make it a compatible
option for UFS2. This means that the journal blocks will likely appear
as an unlinked file to legacy filesystem and fsck code, and will be
treated as such. This will allow seamless fallback to using fsck,
though once the unlinked journal data blocks are reclaimed by fsck,
the user will have to take action to re-create the journal file again.
One key piece of journalling is ensuring that each journal transaction
is fully written to disk before the associated metadata blocks are
written to the filesystem. I plan to adopt the buffer 'pinning'
mechanism from Alexander Kabaev's XFS work to assist with this. This
will allow the journalling subsystem fine-grained control over which
blocks get flushed to disk by the buffer daemon without having to
further complicate the UFS/FFS code. One consideration is how
Softupdates falls into this and whether it is multually exclusive of
journalling or if it can help provide transaction ordering
functionality to the journal. Research here is on-going.
Some preliminary work can be found in Perforce in the
//depot/user/scottl/ufsj/... tree or at the URL provided. Hopefully
this will quickly accelerate.
FreeBSD Dutch Documentation Project
Contact: Remko Lodder <remko at FreeBSD.org>
The FreeBSD Dutch Documentation Project is a ongoing project in
translating the English documentation to the Dutch language. Currently
we have translated almost the entire handbook, and more to come. If
you want to help out by review the Dutch documents, or you want to
help translating the remainders of the handbook or other documents,
feel free to contact me at remko at FreeBSD.org
1. Translate the English handbook, then review the Dutch handbook
2. Translate the English FAQ, then review the Dutch FAQ
3. Translate the English Articles, then review the Dutch Articles
FreeBSD Java Project
Contact: Greg Lewis <glewis at FreeBSD.org>
Contact: Alexey Zelkin <phantom at FreeBSD.org>
The FreeBSD Java Project released its initial support for JDK 1.5.0
with patch set 1 "Sabretooth" in January. The initial release featured
support for both FreeBSD 5.3/i386 and 5.3/amd64. Since then
preliminary support for FreeBSD 4.11/i386 has been added and several
bug fixes have been made. Updates in the coming months will add
support for the browser plug in and Java Web Start, which were not in
the initial release.
1. Volunteers to look into some serious problems with JDK 1.5.0 on
FreeBSD Release Engineering
Contact: RE Team <re at freebsd.org>
FreeBSD 4.11, the final formal release of the 4.x series, was released
on 25 Jan 2005. Many thanks to the all of the developers and users
over the past 5 years who made it successful. While no more releases
are planned, the security team will continue to support it through
security update patches until 2007. Developers are also free to commit
bug fixes and low-risk features to the RELENG_4 branch for the
FreeBSD 5.4 is going through its final release candidate stages and is
expected to be released in late April. Its focus is mostly bug fixes
and minor feature and performance improvements, so it is an excellent
target for those looking to upgrade from previous versions or to give
FreeBSD a try for the first time. FreeBSD 5.5 will be release in about
4-6 months after 5.4.
FreeBSD 6.0 is rapidly approaching also. In contrast to FreeBSD 5.0,
the goal is to take a more incremental approach to major changes, and
not wait for years to get as many features in as possible. FreeBSD 6.0
will largely be an evolutionary change from the 5.x series, with the
largest changes centered around multi-threading and streamlining the
filesystem and device layers. Feature freeze and code freeze for 6.0
are coming up in May and June, and we hope to have 6.0 stable and
ready for release in July or August.
The release engineering team has also started doing monthy informal
snapshots of the 6-CURRENT and 5-STABLE trees. These are intended to
increase the exposure of new features and get more users involved in
testing and providing feedback. Snapshots can be found at
FreeBSD Security Officer and Security Team
Contact: Security Officer <security-officer at FreeBSD.org>
Contact: Security Team <security-team at FreeBSD.org>
In January 2005, Warner Losh (Security Officer Emeritus) stepped down
from the FreeBSD Security Team in order to better devote his time to
other projects. In March, Colin Percival was named as a second Deputy
Security Officer, joining Dag-Erling Smørgrav in that position. The
current Security Team membership is published on the web site.
So far in 2005, four security advisories have been issued concerning
problems in the base system of FreeBSD, three of which were specific
to FreeBSD. The Vulnerabilities and Exposures Markup Language (VuXML)
document has continued to be updated by the Security Team and the
Ports Committers documenting new vulnerabilities in the FreeBSD Ports
Collection. As of April 17, 127 entries have been added in 2005
bringing the FreeBSD VuXML file up to a total of 422 entries.
In the past months both the VuXML web site and the FreshPorts VuXML
integration have been improved. The VuXML web site has had a face lift
and, among other things, each package now has a separate web page
which lists all documented vulnerabilities for the particular package.
CVE information is now also included directly on the VuXML web site.
Finally, the first few months of 2005 also saw FreeBSD 4.8 -- the
first release to be offered "extended support" -- reach its designated
End of Life. The currently supported releases are FreeBSD 4.10, 4.11,
Contact: Dan Langille <dan at langille.org>
This is the first status report for FreshPorts. FreshPorts started in
early 2000 and now contains over 170,000 commits. FreshPorts is
primarily concerned with port commits, but actually processes and
records all commits to the FreeBSD source tree. Its sister site,
FreshSource uses the same database as FreshPorts but has a wider
reporting scope. In recent months, FreshPorts has been enhanced to
process and include VuXML information. In addition, RESTRICTED and
NO_CDROM have been added to list of things that FreshPorts keeps track
of. For unmantained ports, we recently added this message:
There is no maintainer for this port.
Any concerns regarding this port should be directed to the FreeBSD
Ports mailing list via ports at FreeBSD.org
FreshPorts, with direct and indirect support from the FreeBSD
community, continues to evolve and to provide a great tool for users
and developers alike.
1. Provide a copy/paste method for updating watch lists
2. improvement of query times for "People watching this port, also
3. pagination of commits within a port
4. pagination of watch lists
5. create an RSS feed for individual watch lists
GELI - GEOM class for providers encryption
Contact: Pawel Jakub Dawidek <pjd at FreeBSD.org>
GELI is a GEOM class used for GEOM providers encryption. I decided to
work on this, as I needed some feature, which cannot be found in
similar projects. Here is the list of features, I found interesting:
* makes use of crypto(9)
* if there is a crypto hardware available, GELI will run
cryptography on it automatically; if not, it starts dedicated
kernel thread and do crypto software work in there
* supports many cryptographic algorithms (AES, Blowfish, 3DES)
* is able to take key components from many sources at once (user
entered passphrase, random bits from a file, etc.)
* allows to encrypt root partition
* user will be asked for the passphrase before root file system is
* uses "PKCS #5: Password-Based Cryptography Specification Version
2.0" for user passphrase protection (optional)
* allows to use two independent keys (e.g. "user key" and "company
* is fast
* GELI does simple sector-to-sector encryption
* allows to backup/restore Master Keys, so when user have to quickly
destroy keys, it is able to get the data back by restoring keys
from the backup
* provider can be configured at attach time to automatically detach
on last close (so user don't have to remember to detach after
unmounting file system)
* allows to attach provider with a random, one-time keys
* useful for swap partitions and temporary file systems
1. Code audit/review is more than welcome!
GSHSEC - GEOM class for handling shared secret
Contact: Pawel Jakub Dawidek <pjd at FreeBSD.org>
GSHSEC is a GEOM class used for handling shared secret data between
multiple GEOM providers. For every write request, SHSEC class splits
the data using XOR operation with random data, so N-1 providers gets
just random data and one provider gets the data XORed with the random
data from the other providers. All of the configured providers must be
present in order to reveal the secret. The class is already committed
to HEAD and RELENG_5 branches.
if_bridge from NetBSD
Contact: Andrew Thompson <andy at fud.org.nz>
This project aims to import the bridging code and interface from
NetBSD and OpenBSD. The bridge is a cloned interface which can be
modified by ifconfig and brconfig. It supports assigning an IP address
directly to the bridge (e.g. bridge0) instead of one of the member
interfaces, and can be used with tcpdump to inspect the bridged
packets. The code also supports spanning tree (802.1D) for loop
detection and link redundancy. Any pfil(9) packet filter can be used
to filter the bridged packets.
1. Testing performance and functionality against the existing bridge
code. Testers welcome!
IMUNES - a FreeBSD based kernel-level network topology emulator
Contact: Miljenko Mikuc <miljenko at tel.fer.hr>
Contact: Marko Zec <zec at tel.fer.hr>
IMUNES is a scalable kernel-level network topology emulator based on
FreeBSD. In IMUNES each virtual node operates on its private instance
of network stack state variables, such as routing tables, interface
addresses, sockets, ipfw rules etc. Most if not all existing FreeBSD
application binaries, including routing protocol daemons such as
quagga or XORP, can run unmodified within the context of virtual nodes
with no noticeable performance penalty. Complex network topologies can
be constructed by connecting the virtual nodes through netgraph-based
link-layer paths. A GUI tool allows for simple and intuitive network
topology specification, deployment and management. The current version
of IMUNES is based on FreeBSD 4.11-RELEASE and supports IPv4.
Contact: Warner Losh <imp at freebsd.org>
Contact: Takahashi Yoshihiro <nyan at freebsd.org>
Unglamorous cleanup of the code base continues. The focus of recent
efforts have been to reduce the number of machine #ifdefs that are in
the machine independent code. In addition, we're also trying to
increase code sharing between pc98 and i386 ports and reduce the
number of #ifdef PC98 instances in the tree.
In addition, a number of cleanup tasks are underway for different
parts of the kernel that are more complicated than necessary.
Recently, the pccard code's allocation routines were simplified to
reassign ownership of resources more directly than before. The search
is on for other areas that can benefit from cleanup.
1. On pc98, there's no such thing as an ISA bus. It is desirable to
move to having cbus appear in the probe messages. This would also
allow for additional segregation of pc98 specific code in the
drivers and eliminate many ifdefs. Ideally, isa and cbus would
share a common newbus ancestor class so their similarities can be
exploited (they both have PNPBIOS enumeration methods, for
2. cbus devices can have complicated resources. There's support for
vectors of resources. Yet there's no support for populating a
vector of resources from the plug and play information. Doing so
would help the complex world of pc98 a lot, and the odd edge cases
in i386 (floppy, ata) a little.
3. The hints mechanism provides a way to associate hardware with
drivers and resource that would otherwise be completely unknown to
the system. A refinement in the hints mechanism to allow matching
of driver instances to resources is desirable. This would allow
one to hardwire sio0 to 0x2f8, even when the serial device in the
plug and play resource list (or acpi resource list) is listed
second. A further refinement could also be wiring sio0 to "port B"
as defined by acpi or some other enumeration method. Chances are
good that these seemingly related concepts may need separate
implementations due to the decision points for unit assignment.
4. Pccard, cardbus and usb probe their devices after interrupts are
enabled. It would be desirable to hook into new kernel APIs to
allow the mounting of root to be put off until those systems know
that they are done with their initial probe of the devices present
Contact: Warner Losh <imp at freebsd.org>
I've setup a test system to measure interrupt latency on FreeBSD 5.3
and current. So far I've measured the baseline latency for a 300MHz
embedded cyrix based single board computer. I've tried a number of
different strategies to optimize the interrupt path. Most of these
strategies resulted in some improvement of the time it takes to get
from the start of the interrupt servicing to the driver's ISR. These
improvements turned out to be about 1-2% of the processing times on
this single board computer, but a wash on faster machines. However,
the time between when the interrupt should happen, and when FreeBSD
starts to service the interrupt is the dominant factor in these
measurements. Despite the fact that these are fast interrupt handlers
(so the scheduler is out of the loop), I routinely see average
latencies of 18us, with large variations (on the order of 5us standard
1. I need to measure the latencies with 4.x and current to
characterize the differences more precisely. I'm especially
interested in the effects on interrupt latency that the
elimination of mixed mode will cause.
2. I need to characterize different parts of our ISR routines to see
if some of the variation I've seen so far can be reduced by
improved coding techniques.
3. I need to re-run my tests with 5.4 and summarize my results in a
IPv6 Support for IPFW
Contact: Brooks Davis <brooks at FreeBSD.org>
In April 18th, I committed support for IPv6 to IPFW. This support was
written by two student of Luigi's, Mariano Tortoriello and Raffaele De
Lorenzo. I updated it to use PFIL_HOOKS and fixed a few minor issues.
As of this commit, IP6FW should be considered deprecated in favor of
IPFW. It should be possible to MFC this change to 5.x, but that is not
2. IP6FW to IPFW migration guide.
3. Patches relative to 5-STABLE.
Contact: David Xu <davidxu at freebsd.org>
libthread is a pure 1:1 threading library, it had stayed in my
perforce branch for a long time, recent it was imported into source
tree and replaced libthr. The purpose of the work is to improve 1:1
threading on FreeBSD, the library is designed in mind that simplest is
best, currently it can run almost all of the applications libpthread
can run, but gives you better SMP performance. The library size is
smaller than libpthread.
Currently it supports i386, AMD64, sparc64 and ia64 and may support
alpha, powerpc and arm. I didn't do many tests on sparc64 and ia64, I
only tested it on FreeBSD cluster machines. For i386, I always used
LDT, but know that Peter committed GDT code, and now there is no 8191
threads limitation anymore.
libthread_db was updated to support debugging the new libthr. It is an
assistant library used by gdb to debug threaded process, that
understands internal detail of thread libraries. I have improved it a
bit to support event reports for libthr, currently it can report
thread creation and death events. That means a thread that was created
and died will be reported to the user regardless if you are tracking
it or not.
1. I am working on thread creation performance, currently it needs
considerable number of libc functions and syscalls to create a
thread, I would like to introduce a syscall to create a thread in
atomicaly. That means one syscall will setup thread entry, tls,
and signal mask and PTHREAD_SCOPE_PROCESS/SYSTEM; in future maybe
even CPU affinity masks, when userland entry code is executed, the
thread is already fully setup.
2. Process shareable synchronization objects. In Current FreeBSD does
not support this specification. The idea about the shareable mutex
and others is like other systems did, one can use mmap() to create
a shared memory page, and put a pthread synchronization object in
the page, multiple processes use the shared object to control
resource access. I am not working on it, if someone is interested,
please let me know.
Low-overhead performance monitoring for FreeBSD
Contact: Joseph Koshy <jkoshy at FreeBSD.org>
Many modern CPUs have on-chip performance monitoring counters (PMCs)
that can be used to count low-level hardware events like instruction
retirals, branch mispredictions, cache and TLB misses and the like.
PMC architectures and capabilities vary between CPU vendors and
between CPU generations from the same vendor, making the creation of
portable applications difficult. This project attempts to provide a
uniform API for applications to use, and the necessary infrastructure
to "virtualize" and manage the available PMC hardware resources. The
creation of performance analysis tools that use this infrastructure is
also part of the project's goals.
Work since the last status report:
* Support for Intel
Pentium-Pro/Pentium-II/Pentium-III/Pentium-M/Celeron style PMCs
has been added.
* The Pentium-4/HTT machine dependent layer has been overhauled.
* A Python language interface to the C library interface pmc(3) has
* Many bugs have been fixed and documentation has been updated.
1. The code needs to be tested on Intel Pentium-M, Celeron, Pentium
II and Pentium Pro CPUs.
Many subdirs for UFS
Contact: David Malone <dwmalone at freebsd.org>
I'm currently looking at the limit on the number of subdirectories a
directory can have in UFS. There is currently a limit of 32K
subdirectories because of the 16 bit link count field in both struct
stat and the on-disk inode format. The thread above shows that dirhash
provides acceptable performance for directories with 100k
subdirectories using a prototype patch. Two options for allowing many
subdirectories seem to exist: changing the link counting scheme for
directories and expanding the link count field. The prototype patch
implements the first scheme and there are plans to investigate the
second scheme (which may require an ABI change).
Move ARP out of routing table
Contact: Qing Li <qingli at freebsd.org>
I have finished the basic functionality for both IPv4 and IPv6. The
userland utilities ("arp" and "ndp") have been updated. I have tested
the changes with "make buildworld". I have been testing the new code
in a production environment and things appear to be stable. Gleb
Smirnoff (glebius at FreeBSD.org) has provided review comments and I have
incorported these feedback into the patch. I have discussed the IPv6
changes with two of the core KAME developers during the last IETF
meeting in March 2005. They indicated that these changes may result in
divergence from the KAME project but that is not necessarily a bad
1. I am waiting for review feedback from my mentor Andre. I need
locking experts to help me fix my giant-lock shortcut. I am hoping
to send out the code for wider review soon.
netgraph(4) status report
Contact: Gleb Smirnoff <glebius at FreeBSD.org>
This report covers period since August 2004 until April 2005.
New nodes. Two new nodes have been added to base FreeBSD distribution.
ng_netflow(4) node, which implements NetFlow version 5 accounting of
IPv4 packets. ng_ipfw(4) node, which diverts packets from ipfw(4) to
netgraph(4) and back. A well known ng_ipacct node has been added to
SMP. Nodes, which need to allocate unique names have been protected
with mutex in RELENG_5, and subr_unit allocator in HEAD. Nodes, which
need to run periodical jobs were reworked to use mpsafe ng_callout()
API. ng_tty(4) node has been overhauled to be compatible with
debug.mpsafenet=1. NetGraph ISR and callout are now declared MPSAFE in
NetGraph flow control. Two nodes ng_ether(4) and ng_cisco(4) have been
improved to emit flow control messages to upstream node, when state of
link changes. New link failure detection method have been introduced
in ng_one2many(4) node - listening to these flow control messages from
1. more SMP testing of many nodes
2. review locking of graph restructuring
3. ng_nat node - an in-kernel natd(8)
4. make ng_bridge(4) multithreaded
OpenBSD packet filter - pf
Contact: Max Laier <mlaier at freebsd.org>
OpenBSD is about to release version 3.7 . There are patches available
to catch up with the development done in OpenBSD 3.6 and 3.7. These
patches are in an early stage, but ready for testing, please help.
Otherwise there was not much activity on pf, as it already is quite
stable. Other work, such as CARP and if_bridge are having impact on pf
in FreeBSD however, please see the respective reports.
1. Alpha/Betatesting of the 3.7 import
2. Testing with if_bridge
Pipe namespace added to portalfs
Contact: Diomidis Spinellis <dds at FreeBSD.org>
A new sub-namespace, called pipe, has been added to portalfs. The pipe
namespace executes the named command, starting back at the root
directory. The command's arguments can be provided after the command's
name, by separating them with spaces or tabs. Files opened for reading
in the pipe namespace will receive their input from the command's
standard output; files opened for writing will send the data of write
operations to the command's standard input. The pipe namespace allows
us to perform scatter gather operations without using temporary files,
create non-linear pipelines, and implement file views using symbolic
Contact: Mark Linimon <linimon at FreeBSD.org>
As this report was being written, the 5.4 release was ongoing.
A new charter for the Ports Management (portmgr) team was approved by
core and has been posted at the URL above. In addition, two other new
pages describe the policies of the team, and the range of QA
activities both during and between releases.
Due to being absent from email discussions for some time, Oliver
Eikemeier (eik) was moved to non-voting status on portmgr.
We have added several new and very active committers recently; this is
helping us to keep the PR count low even with the large numbers of new
ports that have been added.
Several more iterations of infrastructure changes have been tested on
the cluster and committed; see /usr/ports/CHANGES for details.
Updates have occurred to x.org, GNOME, KDE, and perl.
There have been some updates to the Porter's Handbook, but more
sections are still in need of updates to include recent changes in
The ports collection now contains almost 12,750 ports.
1. Further progress has been made in cracking down on ports that
install files outside the approved directories and/or do not
deinstall cleanly (see "Extra files not listed in PLIST" on
pointyhat ) and this will remain a focus area. We appreciate
everyone who has sent in PRs or committed fixes.
2. Demand for new features and revisions for bsd.port.mk is still
very high and the portmgr team is trying to work through them all.
3. We still have a large number of PRs that have been assigned to
committers for some time (in fact, they constitute the majority).
One goal of portmgr in the coming months is to try to reduce this
number, and we would like to ask our committers to help us out as
much as possible.
Contact: Peter Grehan <grehan at FreeBSD.org>
Progress continues. X.Org 6.8.1 server has been up and running on a
number of different Macs, and the work is being merged into 6.8.2.
There have been successful installs on Mac Minis
Removable interface improvements.
Contact: Brooks Davis <brooks at FreeBSD.org>
This project is an attempt to clean up handling of network interfaces
in order to allow interfaces to be removed reliably. Current problems
include panics if Dummynet is delaying packets to an interface when it
I am currently working to remove struct ifnet's from device driver
structures to allow them to be managed properly upon device removal. I
believe I have removed all known instances of casting a struct ifnet
pointer to something else (except that that are just magic values and
not real struct ifnets.) I will begin committing these changes to the
tree shortly and will then add a new function if_alloc() that will
allocate struct ifnets. if_detach() will be modified to destroy them.
Contact: Colin Percival <cperciva at FreeBSD.org>
Shortly before the ports freeze for FreeBSD 5.4, I released a new
version of Portsnap. In addition to being secure and more efficient
than CVSup, this latest version distributes INDEX, INDEX-5, and
INDEX-6 files, thereby eliminating the need to run "make fetchindex"
and ensuring that the ports INDEX will match the existing ports tree.
In addition, portsnap builds have now moved onto hardware managed by
the FreeBSD project, thereby sharply increasing portsnap's chances of
survival if I get hit by a bus.
In early February hardware problems caused both FreeBSD Update and
Portsnap to stop functioning for a few days, but those were resolved
thanks to a server donated by layeredtech.com.
I intend bring Portsnap into the FreeBSD base system before the end of
the month, followed by FreeBSD Update a few months later.
Status Report for FreeBSD ATA driver project
Contact: Søren Schmidt <sos at FreeBSD.org>
ATA mkIII has been committed to -current after a couple of month
testing as patches post on -current and 5-stable. I will continue to
provide patches for 5-stable for those that need up-to-date ATA
Here a short rehash of what mkIII brings:
ATA is now fully modular so each part can be loaded/unloaded at will
to provided the wanted functionality.
Much improved SATA support that support hotplug events on controllers
that support it (Promise, SiS, nVidia so far) ie the system will
automagically detect when SATA devices come and go and add/delete
device entries etc.
Much improved ATA RAID support. The ata-raid driver has been largely
rewritten to take advantage of the features the improved
infrastructure provides, including composite ATA operations etc. The
rebuild functionality has been changed to rebuild on userland reads,
so a simple dd of the entire array will get it rebuild (what
atacontrol now does). This means that the resources used for this can
be better tailored to the actualy usage pattern if needed. ATA RAID
now supports 10+ different RAID metadata formats, so most BIOS defined
ATA RAID arrays can be picked up and used. The number of metadata
formats that can be created from within FreeBSD is still limitted
though and is not a high priority feature right now.
The lowlevel infrastructure of the ATA driver has been refined even
further to support "strange" chipsets much more easily and in most
case transparent to the higher levels. This to easy ports to new
platforms where ATA controllers doesn't necessarily have the x86
Lots of bug fixes and corrections all over the driver proper. The
rework of the infrastructure has revealed bugs and deficiencies that
has been fixed in the process of modulerising ATA and making the
infrastructure more generic, and hopefully easier to understand.
The work continues to keep ATA on top of new chipsets and other
advancements in the ATA camp. SATA ATAPI support is in the works and
so are support for NCA/TCQ (tags). Donations of unsupported hardware
is the way to get it supported as I'm way out of my budget for new
hardware for the next decade or so according to my wife :)
1. Lots of testing wanted, especially SATA and RAID support
Storage driver SMPng locking
Contact: Scott Long <scottl at freebsd.org>
Several storage drivers have been taken out from under the Giant mutex
in the past few months. Thanks to sponsorship from FreeBSD Systems,
Inc and ImproWare, AG, Switzerland , the LSI MegaRAID (AMR) and
IBM/Adaptec ServeRAID (IPS) drivers have been locked. SMPng locking is
a key step in improving the performance of system drivers in FreeBSD
5.x and beyond, and both of these drivers are showing the benefits of
this. FreeBSD 5.4 will contains these improvements when it is
Similar work is ongoing with the 3WARE Escalade (TWE) driver, and
preliminary patches have been made available to testers. I hope to
have this driver complete in time for the next FreeBSD release.
Unfortunately, most benefits can only be gained from pure block
storage drivers such as the ones mentioned here due to the SCSI
subsystem in FreeBSD (CAM) not be locked itself at this time. It is
possible, however, to lock a CAM sub-driver and bring the driver's
interrupt handler out from under Giant for a partial gain. The Sun
FAS366 SCSI driver (ESP) operates like this. Volunteers to lock other
drivers or to tackle locking CAM are gladly accepted, so please
contact me if you are interested.
Support for telephone hardware (aka Zaptel)
Contact: Maxim Sobolev <sobomax at FreeBSD.org>
Contact: Oleksandr Tymoshenko <gonzo at pbxpress.com>
Contact: Max Khon <fjoe at FreeBSD.org>
During the last 2 months lot of progress has been made. Existing
support for TDM400 (FXO/FXS) has been significantly improved. Drivers
for PRI and BRI cards have been added and now should be considered
1. More testing of PRI/BRI drivers.
2. Add support for channelized DS3 card(s).
Contact: Vinod Kashyap <vkashyap at amcc.com>
A newly re-architected twa(4) driver was committed to 6 -CURRENT on
04/12/2005. Highlights of this release are:
1. The driver has been re-architected to use a "Common Layer" (all
tw_cl* files), which is a consolidation of all OS-independent
parts of the driver. The FreeBSD OS specific portions of the
driver go into an "OS Layer" (all tw_osl* files). This
re-architecture is to achieve better maintainability, consistency
of behavior across OS's, and better portability to new OS's
(drivers for new OS's can be written by just adding an OS Layer
that's specific to the OS, by complying to a "Common Layer
Programming Interface (CLPI)" API. If there's interest in porting
the 3ware driver to any other OS, you may contact ctchu at
amcc.com to get a copy of the CLPI specifications.
2. The driver takes advantage of multiple processors. It does not
need to be Giant protected anymore.
3. The driver has a new firmware image bundled, the new features of
which include Online Capacity Expansion and multi-lun support,
among others. More details about 3ware's 9.2 release can be found
Update of the Linux userland infrastructure
Contact: Alexander Leidinger <netchild at FreeBSD.org>
Contact: Emulation Mailinglist <freebsd-emulation at FreeBSD.org>
The update to RedHat 8 as discussed in the last status report went
smoothly (just some minor glitches which got resolved fast).
As a next step a cleanup/streamlining and the possibility of
overriding the default linux base is in progress. This depends on
changes which need at least one testrun on the ports build cluster, so
the final date for those changes depends upon the availability of the
1. Refactoring the common RPM code into bsd.rpm.mk.
2. Determining which up-to-date Linux distribution to use as the next
default linux base. Important criteria:
+ RPM based (to be able to use the existing infrastructure)
+ good track record regarding availability of security fixes
+ packages available from several mirror sites
+ available for several hardware architectures (e.g. i386,
amd64, sparc64; Note: not all architectures have a working
linuxolator for their native bit with, but as long as there
are no userland bits available, no motivation regarding
writting the kernel bits will arise)
3. Moving the linuxolator userland to an up-to-date version (see
Wireless Networking Support
Contact: Sam Leffler <sam at freebsd.org>
Several new drivers by by Damien Bergamini were brought into the tree:
iwi, ipw, ral, and ural.
WPA-PSK support for the ndis driver was contributed by Arvind
A new tx rate control algorithm for the ath driver was contributed by
John Bicket. It will become the default algorithm shortly.
Work on multi-bss support is going on outside the cvs tree. A
presentation on this work will be given at BSDCan 2005 and the slides
for the talk will be made available after.
1. Drivers other than ath and ndis need updates to support the new
2. hostapd needs work to support the IAPP and 802.11i
preauthentication protocols (these are simple conversions of
existing Linux code).
3. The openbsd dhclient program has been ported but needs a developer
that will maintain it once it is brought into cvs.
XenFreeBSD - FreeBSD on Xen
Contact: Kip Macy <kmacy at fsmware.com>
FreeBSD 5.3 runs on the stable and the development branches of xen and
is now checked into both trees. Over the next couple of weeks I will
be adding improvements for better batching of page table updates and
1. FreeBSD support for running as Domain 0, i.e. running as the
hosting operating system.
2. FreeBSD support for VM checkpoint and migration.
More information about the freebsd-stable