diff --git a/en/projects/ideas/ideas.xml b/en/projects/ideas/ideas.xml index e9d5843eb9..629dfa2142 100644 --- a/en/projects/ideas/ideas.xml +++ b/en/projects/ideas/ideas.xml @@ -1,2090 +1,2090 @@ - $FreeBSD: www/en/projects/ideas/ideas.xml,v 1.125 2009/03/20 17:35:23 jkoshy Exp $ + $FreeBSD: www/en/projects/ideas/ideas.xml,v 1.126 2009/03/21 12:08:41 rwatson Exp $ Embedded Reduced FreeBSD kernel size for embedded

Technical Contact: Warner Losh

The FreeBSD kernel has been optimized over the years for a server or workstation environment. Memory is plentiful in these environments, so little attention was given to the size of the kernel. There's a number items in the kernel that can be made optional without reducing affecting the functionality needed in an embedded environment. These include things like not compiling in strings into the kernel, less agressively inlining code, making some non-optional features optional and investigating compile time flags. This task requires identifying potentially optional kernel content and building the infrastructure to make that content optional.

requirements:

  • Strong C language programming skills.
  • Understanding of system calls and general kernel architecture.
 NAND Flash driver support

Technical Contact: Warner Losh

-

Add support for nand flash support.

+

Add support for NAND flash devices.

Requirements:

  • Good C programming skills
  • Ability to understand FreeBSD subsystems like geom, device, etc.
 Make creating a bus easier

Technical Contact: Warner Losh

There's about a dozen busses in the tree now that manage resources and activate children. They are far too hard to create. We need to abstract out the basics for these buses and provide a way to allow these buses to be a subclass of this new base class.

Requirements:

  • Good C programming skills
  • Good knowledge of the NewBus configuration system in FreeBSD
 Variable hints

Technical Contact: Warner Losh

Often times in the embedded world, you know what kind of built-in devices are on a SoC (System on a Chip) only because you know the specific model of that SoC. It is desirable to have a mechanism that code on these machines can use to load one of several sets of hints, which can then be used to populate the bus.

Requirements:

  • Good C programming skills
 ARM cleanup

Technical Contact: Warner Losh

Adding a new board to the arm code is a lot harder than it needs to be. A lot of benefit could be had by creating tables for memory ranges, etc, and having more generic initialization code. Much of this can also be Machine Independent (MI).

Requirements:

  • Good C programming skills
  • Good refactoring skills
 PPC/ARM/MIPS bring up

Technical Contact: Warner Losh

There's a number of SoCs that are in consumer grade routers, etc that have chips that are supported by FreeBSD, or nearly supported by FreeBSD. Pick one and bring FreeBSD up on it. Integrate it into the tree.

Requirements:

  • Good C programming skills
  • Good kernel debugging skills
 Overhaul the config system

Technical Contact: Warner Losh

Right now the kernel is built twice: once for the static modules in the kernel, and once for the dynamically loaded. There's also inconsistent dependency tracking. We should fix this. Bikeshed included, along with three colors of paint.

Requirements:

  • Good C programming skills
  • Good kernel debugging skills
  • Ability to withstand Bikesheds
 File System FAT (msdosfs) infrastructure work

Technical Contact: Robert Watson

The FreeBSD FAT implementation, msdosfs, offers scope for a number of projects:

  • General cleanup.
  • Introduce appropriate locking to make the file system operate without the Giant lock (MPSAFE).

It is unclear to what extent the last of these items, arguably the most useful, will require modifying surrounding infrastructure such as BIO, GEOM, and VM.

Requirements:

  • Strong C programming skills.
  • Familiarity with concurrent programming techniques.
  • Familiarity with FAT file system layout.
  • Familiarity with virtual file system and virtual memory.
 Improve the performance of dump/restore

A performance evaluation of the split cache (as is) and an unified cache (like e.g. NetBSD) would be interesting. More details in this mail to the hackers mailing list. Additional improvements are welcome too.

Requirements:

  • Knowledge of C programming.
  • Basic understanding of backup/restore procedures.
 Extend UFS2 with on-disk indexing

Technical Contact: David Malone

The section 8.3 Naming of the book Design and Implementation of the FreeBSD Operating System describes the current approach of name lookups in UFS2 and a possible extension/improvement by utilizing on-disk indexing in a backward compatible way. While dirhash has eliminated many of the performance problem associated with UFS directory lookups, it still has to build an index on each access. On-disk indexing would improve this situation. Another possible improvement would be to make dirhash's memory usage dynamic, rather than having a fixed memory limit.

Requirements:

  • Knowledge of C programming.
  • Basic understanding of filesystems.
  • The book Design and Implementation of the FreeBSD Operating System.
 Analyze NetBSD's ext2fs regarding valuable improvements

FreeBSD has an implementation of the ext2fs filesystems but it contains some files under the GPL which make it undesirable, among other things, to use it in the GENERIC kernel. Ext2fs is a rather simple but practical filesystem and NetBSD has had for a while an implementation based on UFS1 sources. The NetBSD implementation needs to be analyzed regarding features and performance. If it is on par or better with our GPLed implementation, it should be ported to FreeBSD.

Requirements:

  • Knowledge of C programming.
  • Basic understanding of filesystems.
 Implement co-location for UFS2

While FreeBSD's FFS implementation is pretty much state-of-the-art, in addition to softupdates, Greg Granger proposed other strategies that would be useful, especially when working with small files. Quoting Greg Ganger: "The key insight for why current file systems perform poorly is that locality is insufficient - exploiting disk bandwidth for small data objects requires that they be placed adjacently". Explicit grouping, in particular, seems to provide important performance improvements without less implementation complexity than embedded inodes. As this changes the on-disk structure, care needs to be taken that the implementation is backwards compatible.

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • Knowledge of the UFS2 filesystem.
 MDFS lockups

Fix MDFS lockups when using async operation modes. Revision 1.115 of md.c has a discussion of the problem.

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • Knowledge of the VFS and VMA subsystems.
 MPSAFE filesystem work

Technical contact: Kris Kennaway

Take a filesystem and MPSAFE it. e.g. ext2fs, ntfs, coda, etc.

 Kernel DDB/gdb scripting

Technical contact: Kris Kennaway

The goal would be to develop scripts that automatically run a standard suite of useful debugging commands in DDB upon panic and save in a textdump. Might be too short on its own, so could be combined with a project to write gdb macro equivalents of the DDB command set, extending the macros John Baldwin has. New DDB commands and macros could also be implemented, e.g. for inspecting other common data structures.

 FIFO optimizations

Technical contact: Kris Kennaway

Evaluate the possibility of merging the FIFO implementation with the pipe implementation for improved performance. Care would need to be taken to avoid regressions, so part of this project should be attention to previous and existing FIFO bug reports, and writing of conformance testing to verify correct behaviour. Possible extensions might include a re-evaluation of some of the performance tradeoffs made in the pipe code in light of modern CPUs.

 Timecounter Performance Improvements

Technical contact: Kris Kennaway

The gettimeofday syscall is a performance bottleneck in certain applications. An approach taken by other operating systems is to export the time counter to userland via a shared page, and to update it periodically (a prototype implementation is available). For some time consumers this is sufficient resolution. Other consumers need higher resolution. On the x86 architecture the TSC timecounter can be read from userland. However depending on the hardware there may be issues with synchronization between CPUs, as well as interaction with CPU frequency changes. With care it can be used as a delta against the timestamp updated by the kernel to provide improved resolution and avoid the need for the syscall.

 Automated kernel crash reporting system

Technical contact: Xin LI, Howard SU

In some recent operating systems, it is common that crashes are automatically reported to its vendor, which is very helpful for finding hidden problems that can not be easily triggered by usual test cases. Newer GNOME applications also has similar functionalities.

This project would consist two parts. One is some improvements over the current savecore rc.d script to teach it how to collect necessary information (of course, automatic reporting has to be explicitly enabled by individual system administrators, and should have at least three options: not to send out anything at all as a default, send out after administrator confirmation, and automatically send all necessary information). The FreeBSD kernel in 8-current has the textdump feature which may be interesting to use for this part

Another part after the first one is finished is the server side one, which will keep a database of backtraces where similar (call stack minus addresses) reports are kept together and be considered as a "vote", to make it possible for developers and release engineers to focus on the most commonly triggered issues.

Requirements:

  • Strong knowledge of C.
  • Understanding of kernel debugging.
  • Knowledge about web application as well as database.
  • Understanding of user privacy protection.
 Avoiding syscall overhead

Technical contact: Kris Kennaway

setproctitle() calls are a serious performance bottleneck in a default pgsql configuration (they are called at least once per query, which might be thousands of times per second - I measured a performance impact of about 33% on sysbench).

One idea for avoiding the syscall (and global sysctl lock) overhead for this kind of thing would be a memory page shared between kernel and userland which libc could read/write to access things like the process title. There are potentially many other data values that could be optimized by a similar method. This is presumably a well established technique in other OSes.

This project requires mentoring/review/planning with someone with significant VM experience to make sure this approach works properly. Done incorrectly, this could result in fairly massive security holes, performance issues (perhaps not visible in simple benchmarks), etc.

Requirements:

  • Strong knowledge of C.
  • Understanding of kernel VM.
 CPU online/offline project

Technical contact: John Baldwin

The project would need to extend the current CPU states of absent and present to include absent, offline, and online. A set of event handlers (probably using EVENTHANDLER(9)) will need to be created to allow subsystems to be notified of CPU state changes. At a minimum, the functionality of the current 'hlt_cpus_mask' should be reimplemented in both the 4BSD and ULE schedulers using this mechanism. Note that the ability to disable all but one thread within a core must still be supported easily. (This is currently done on x86 via the machdep.hyperthreading_allowed sysctl.) Each subsystem that uses CPU_ABSENT() will need to be evaluated to see if it needs updating to handle online vs offline CPUs. Further improvements may include teaching UMA to free per-CPU buckets when a CPU is taken offline.

Requirements:

  • Strong knowledge of C.
  • Experience with multi-threaded programming.
  • Experience with kernel programming.
 Document all sysctls

Technical contacts: Mathieu Arnold, Brad Davis

The sysctl(8) utility retrieves kernel states and allows processes with appropriate privilege to change kernel states. On request it is able to display description lines which document the kernel state. Unfortunately not every sysctl is documented. This task is possible to share with other volunteers. mat has done some development in Perforce, in the mat_sysctl_cleanup branch.

  • Find every undocumented sysctl in the kernel.
  • Try to determine what this sysctl is for and document it.

Requirements:

  • Ability to read and understand foreign C code.
 Document the sound subsystem

Technical contacts: Alexander Leidinger, Ariff Abdullah

  • Add sound subsystem related section 9 manual pages, so far no sound subsystem related manual pages exists.
  • Add an example driver in share/examples which allows to write a new driver. For this purpose the example driver should contain enough documentation as comments and/or pointers to documentation in man-section 9. This work can be based upon this template.
  • Rewrite the sound subsystem chapter in the FreeBSD Architecture Handbook. The rewrite should contain an overview of the available parts in the sound subsystem and how they interact (data flow, dependencies, ...) and fit together. Additionally it should contain links to already available documentation (official standards, section 9 manual pages, ...).
  • A wiki page documenting everything related to the sound subsystem in FreeBSD has been created.

Requirements:

  • Ability to read and understand foreign C code.
  • Documentation writing skills.
 DTrace

Technical contact: John Birrell

URL: Perforce repository, DTrace for FreeBSD

DTrace is a dynamic tracing facility designed by Sun Microsystems and released in Solaris 10. They have since released the major part of Solaris under the banner of OpenSolaris and the Common Development and Distribution License (CDDL) 1.0. John Birrell has created an initial port and should be contacted for information on what tasks remain to be done; two possible areas of work are:

  • We need a clean CTF implementation for FreeBSD to avoid the license (CDDL) that Sun has on their code. John will write a specification about the file format and the Summer of Code project is to implement that and write tests for the implementation without looking at the Sun code.

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • A good understanding of the FreeBSD kernel.
 DWARF2 call frame information

A debug kernel is not able to show stack traces with cross exceptions anymore. This is because we do not emit any dwarf2 call frame information for any assembler code, since gdb switched to the dwarf2 format. A volunteer should annotate every assembler file [*.[sS]] with dwarf2 call frame information.

Requirements:

  • Knowledge of assembly code.
  • Knowledge of ".cfi_*" pseudo-ops to insert dwarf2 frame descriptors.
 Kernel support for linux DVB device drivers

Technical contact: Luigi Rizzo

In early 2007 we started a project was started to support the building of linux device drivers on FreeBSD. This was done through an in-kernel emulation layer, which implements part of the linux kernel API on top of the FreeBSD kernel API. The initial implementation was good enough to support a few USB webcam drivers, and is documented here. The code is actually available as a port, devel/linux-kmod-compat, and a popular driver that uses this infrastructure is multimedia/linux-gspca-kmod .

We would like to use a similar approach to add support for DVB devices, which are widely supported in Linux but not in FreeBSD. In particular we expect the project to provide, within the FreeBSD kernel, enough of linux compatibility to build the core components of the drivers/media/ linux kernel, and then a few device drivers including one for a PCI DVB card (e.g. saa7134-based).

Before the start of the project, a Summer of Code applicant is expected to i) become familiar with the approach used by linux-kmod-compat; ii) set up a proper test environment, with a couple of DVB devices supported by linux, and a working linux installation so that one can compare results; iii) become familiar with the architecture of the linux code in drivers/media.
Probably the attention should be focused in PCI devices, because at this stage the USB stack is in a transition phase and would pose some additional difficulties.

Expected results are a working porting infrastructure, a working linux-dvb-kmod device driver, and a working application to demonstrate that the driver is working as expected. We suggest to look at "kaffeine" for which a FreeBSD port already exists.

Extend ktrace/kdump output

Technical contact: Alexander Leidinger

The ktrace(1) facility allows to monitor what running processes do. It allows to determine if a process is stuck or if it still does useful work. The goal of this item is to look at the kernel interfaces, add missing "pieces" (e.g. syscall's) to the ktrace output and to extend the output with "decoded" (translating hex/dec values into human readable information, e.g. O_RDONLY in the case of open(2)) information. Some work has been completed and committed, but a few parts still remains. More information is available here.

Also, a related project would be to modify ktrace to write to pipes. Currently the ktrace infrastructure requires the dump output go to a file. It would be useful to be able to instead have it write to pipe, or in fact any type of file descriptor.

Requirements:

  • Knowledge of C.
  • Good knowledge of POSIX interfaces or how to use man(1).
  • No fear to look into the kernel sources.
 Fast syscall support for FreeBSD/i386

Technical contact: Attilio Rao, Jeff Roberson

The instruction pair sysenter and sysexit can contribute to certain performance improvements when a syscall is made on IA32. There is however no implementation of this available for FreeBSD, so a volunteer would have to add sysenter/sysexit support to the kernel. This needs to be properly evaluated and benchmarked though, so a complete implementation should therefore also contain informative benchmarks which shows a clear improvement in performance. It is also important to stress the fact that this project is of research quality and measures should be taken to ensure that no regressions are introduced. Another interesting extension to this project would be to investigate and evaluate the possibility to use mmx/xmm registers to gather syscalls arguments. David Xu has some work in progress in his sysenter branch in the perforce repository.

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • Ability to write and understand x86 assembly.
 Generic input device layer

Technical contact: Philip Paeps

WIP: http://wiki.freebsd.org/GenericInputDeviceLayer

The kernel is lacking a generic input device layer analogous to the Linux 'input core' layer. Having such a layer would make it easy to write e.g. touchscreen support (Philip Paeps has some work-in-progress regarding pointer devices and touchscreen support, but not enough time to also cover keyboard support or other generic features). This project was worked on as part of Google Summer of Code 2007, and you can find more information on the FreeBSD.org wiki

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
 Interactive Splash Screen

Improve upon / replace the existing static VESA splash screen support in FreeBSD, with a script-driven back-end, which allows animation in the loading graphics. This would greatly improve the bootup experience for desktop users, while providing graphical feedback to the startup of the kernel / system services. Additionally this could be used to replace the beastie.4th menu, with a VESA driven graphical loader screen.

 PCI-Hotplug support

Technical contact: Bruce M. Simpson

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • A good understanding of low-level access of the hardware.
  • A good understanding of FreeBSD device drivers.
 Collective limits on set of processes (a.k.a. jobs)

Technical contact: Brooks Davis

In SGI's Irix operating system, there is a concept of a job which is a collection of processes. These processes share a set of resource limits similar to those accessible through the set/getrlimit and getrusage system calls. Schedulers such as Sun Grid Engine currently implement tracking the processes that make up a job and enforcing collective limits on them in an adhoc manner. Having first class kernel support would be useful.

It seems most likely that implementing something like the Irix interface would be the most useful approach since that would enable us to leverage existing code. Implementers will need to make sure that the job construct has no negative performance implications when not enabled (ideally, processes that are not part of jobs should be unaffected) and quantify the impact on perfomance when enabled.

References: Irix Man Pages mentioning jid_t

Requirements:

  • Good knowledge of C.
  • Kernel awareness.
  • Ability to implement functionality from interface specifications.
 Geom-based Disk Schedulers

Technical contact: Luigi Rizzo

In a 2005 GSoC project, "Pluggable Disk Schedulers", Emiliano Mennucci explored the feasibility of pluggable disk schedulers for FreeBSD. The project was successful, but we could not explore certain approaches (e.g. "anticipation", where requests are delayed hoping that some future ones can served without a seek) due to architectural limitations the kernel had at the time.

Since then, the GEOM infrastructure has become available on FreeBSD for interacting with disk I/O requests. GEOM has enabled us to work on disk schedulers in a much more flexible way, allowing a much faster development of disk scheduling algorithms. With Fabio Checconi, we have developed a prototype implementation of some anticipatory schedulers, see GEOM_SCHED.

GEOM_SCHED works within the geom layer, i.e. above the device driver where queueing of requests may actually occur. The way GEOM_SCHED does scheduling is by limiting the number of outstanding requests to the device, and the performance implications of this approach need to be measured. An alternative approach is to push the scheduler (using the same algorithms developed in GEOM_SCHED, and most likely the same code) within the device drivers. This less general (as it needs to be replicated in all drivers) but it may be an interesting thing to do e.g. for some popular device drivers such as ATA.

The proposed SoC work can address one or more of the following aspects:

  • implement suitable classifiers for disk requests;
  • implement techniques for the auto-tuning of the scheduler parameters;
  • measure the performance implications of doing scheduling above the device driver, and possibly design and implement a suitable mechanism to push the GEOM_SCHED module within the device driver itself.
Ultimately, we expect to end up with a production quality subsystem for use in FreeBSD.

References: The Pluggable Disk Schedulers SoC project, Patches

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • Knowledge of GEOM (or interest in getting familiar with it).
 Remove procfs dependencies

Technical contact: Olivier Houchard

Someone needs to finish the support for PT_SYSCALL in the ptrace() subsystem and remove the need for procfs in gcore. Removing the procfs(5) dependency from ps -e is also desirable.

Requirements:

  • C knowledge.
  • Understanding of kernel debugging interfaces.
 Suspend to disk

Implement a suspend/resume from disk mechanism. Possibly use the dump functions to dump pages to disk, then use ACPI to put the system in S4 or power-off. Resume would require changes to the loader to load the memory image directly and then begin executing again.

Requirements:

  • Good knowledge of C.
  • Understanding of the hardware/software interface.
  • A laptop that works with ACPI.
  • Kernel awareness.
 Sync FreeBSD i386 boot code with DragonFly

Technical contact: John Baldwin

DragonFly invested a lot of time to clean up and document it. Additionally they fixed some bugs. Interesting files in the DragonFly CVS are sys/boot/i386/bootasm.h, sys/boot/i386/bootasmdef.c, sys/boot/boot0/*, sys/boot/boot2/*, sys/boot/i386/btx/*, sys/boot/i386/cdboot/*, sys/boot/i386/libi386/amd64_tramp.S, sys/boot/i386/libi386/biosdisk.c and sys/boot/i386/loader/main.c. An interested volunteer has to compare and evaluate both implementations and port interesting/good parts.

Requirements:

  • Ability to read and understand foreign C code.
  • Ability to write C code.
  • Knowledge of i386 assembly.
  • Knowledge of BIOS interfaces.
  • Knowledge of low-level boot behavior.
 Make optional kernel subsystems register themselves via sysctl

Technical contact: Kris Kennaway

Currently there is no way for e.g., a port makefile to tell whether things like FreeBSD 5.x compatibility are present on the system (just installing the compat5x port is not enough, you need a kernel built with COMPAT_FREEBSD5). All such optional kernel features need to register themselves with the FEATURE macro so that the userland can easily query whether a given feature is present. So far not all kernel features are using this infrastructure.

There needs also to be a way to spoof those values, e.g., when the ports build cluster is building for older FreeBSD versions in a jail. Suport for this is not available in the FEATURE macro.

Requirements:

  • Good knowledge of C.
  • Kernel awareness.
 EFI support for FreeBSD/i386 and FreeBSD/amd64

Technical contact: Rui Paulo

Finish EFI support on the i386/amd64 ports. Final work should be able to boot a FreeBSD kernel into single user mode, at least.

Requirements:

  • Good knowledge of C.
  • Deep understanding of the boot process and EFI.
  • A system running EFI 1.0 (Intel Macs, for example)
 Networking csup improvements

Technical contact: Ulf Lilleengen

URL's: csup homepage, CVSweb

csup is a port of the cvsup high-speed CVS repository replication application from the original Modula-3 to the C lanaguage. It is now distributed with FreeBSD, but is missing some important features that would make useful projects to work on:

  • Add support for authentication.
  • Working rsync support.
  • Optimize rcsfile handling.
  • Create a library out of the ports that might be of use in a C language csupd.
  • Add support for shell commands sent by the server.
  • Add missing support for various CVSup options: -D, -a (requires authentication support), -e and -E (requires shell commands support) and the destDir parameter.
  • Work on a new csupd.

Requirements:

  • Strong knowledge of C.
  • Good knowledge of POSIX standards.
  • Ability to work with multi-threaded applications.
 TCP/IP regression test suite

Technical contact: Robert Watson, George V. Neville-Neil

Design and implement a wire level regression test suite to exercise various states in the TCP/IP protocol suite. Ideally with both IPv4 and IPv6 support.

Requirements:

  • Strong TCP/IP knowledge.
 Passive libpcap based TCP session anomaly detector

Technical contact: Andre Opperman.

Listens on an interface and tracks all TCP sessions it sees. In the normal case only general information is carried forward (seq#/ack#, negotiated SYN/ACK features, etc). Whenever an anomaly happens - that is a duplicate ACK, SACK response, out-of-order segment, retransmission or others; it captures those packets into a tcpdump file for later deep inspection with Wireshark or other tools. This tool is to be deployed on live hosts and passive monitors to collect reliable condensed data about real-world behavior of TCP on the global Internet. Currently no such quantitative data exist and contribution of such a tool that can be easily run is a significant step in helping further development of TCP algorithms.

Difficulty: Medium, good familiarity with the TCP RFCs is necessary and detection of many edge cases has to be implemented correctly.

 WPA2 preauthentication in hostapd

Technical contact: Sam Leffler

WPA2 is the authentication protocol defined as part of the IEEE 802.11i specification. This protocol is now commonly used to authenticate wireless stations to access points. Part of this protocol is the ability to pre-authenticate a station with one or more access points so that roaming can happen quickly. FreeBSD lacks support for this aspect of the protocol in the hostapd program used to construct a WPA-enabled access point. This task would port the Linux code that exists to support pre-authentication in hostapd. This mostly involves rewriting some user-mode multicast code and testing the result.

Requirements:

  • Good knowledge of C.
  • Wireless networking fundamentals.
  • WPA-capable wireless network setup.
 802.11 Fuzzing and Testing

Technical contact: Sam Leffler

Build a "packet fuzzer" tool that can be used to build test suites to improve reliability of the 802.11 code against garbage data. There are various tools out but we're not aware of any good ones that work with 802.11 and are generally available. The basic idea is to write a packet injector/playback tool that's driven by a scripting language. Then you need to build up a database of test cases. It's also possibly important to do time-based playback.

Requirements:

  • Good knowledge of C.
  • Wireless networking fundamentals.
  • WPA-capable wireless network setup.
 SCPS, Space Communication Protocol Standards

SCPS is a protocol suite designed to allow communication over challenging environments. Originally developed jointly by NASA and DoD's USSPACECOM, these protocols are used for commercial, educational, and military environments. A student project in this area would involve implementing various network protocols according to specification (SCPS File Protocol, similar to FTP; SCPS-Transport Protocol, based on TCP; and others.)

 Implement TCP UTO

Technical contact: Rui Paulo

Implement TCP UTO (User Timeout Option) as defined by RFC 5482.

Requirements:

  • Good knowledge of C and TCP.
  • Able to understand the FreeBSD TCP/IP stack.
  • A testbed with at least two machines.
 Ports Add .db support to pkg_tools

Technical contact: Kris Kennaway

pkg_create(1) and friends use flat databases (aka ordinary files and directories in /var/db/pkg) to maintain their data. This makes it cumbersome and/or impossible to do efficient lookups of data on installed packages and makes certain operations very slow. portupgrade has the right idea of hashing this into a berkeley db file, but it uses tools that are not in the base system (ruby).

A self-contained project would be to add similar (preferably compatible) code into pkg_tools directly, possibly also extending the data that is stored and allowing for more flexible querying with tools like pkg_info (e.g. replicating the pkg_which utility of portupgrade). Adding mutual exclusion to protect concurrent pkg_add/delete operations from corrupting database state is also important.

Requirements:

  • Knowledge of C.
  • Basic understanding of the use of berkeley db.
 Cleanup of USE and WITH variables

Technical contact: Erwin Lansing

Make these more consistent. WITH_* should be user-settable variables while USE_* only is for internal use in the ports.

Requirement:

  • Strong knowledge of shell and make code.
  • A basic understanding of the inner workings of the ports tree.
 Package tools improvements

The pkg_* tools, which deal with the installation of pre-build binary package of ports, could do with a code cleanup or maybe even a rewrite from scratch. Some features of the ports tree are not supported by the pkg_* tools, e.g. versioned dependencies.

Requirements:

  • Strong knowledge of C code.
  • A basic understanding of the inner workings of the ports tree.
 Parallelization in the Ports Collection

Technical contact: Pav Lucistnik

Add locking of write access to PKG_DBDIR (/var/db/pkg), to allow several port builds run in parallel without clobbering the package data. Should be done both in makefiles and in C tools like pkg_install and pkg_delete. A simple flock(2) approach over the whole database comes to mind.

The next step is the parallelization of dependency building. Have the port build it's dependencies in parallel, automatically depending on number of CPUs in the machine, or manually specified by user (make -j3 install clean). Some kind of split screen should be devised, so user can easily watch the process and interact with it (make config screens, for example). Attention must be paid to prevent deadlocks.

Allow for situation when two ports want to build and install common dependency. One of the ports have to wait on the other to install it before proceeding.

Requirements:

  • Strong knowledge of make and shell code.
  • Strong knowledge of C code.
  • Good understanding of the inner design of the Ports Collection.
 Ports license auditing infrastructure

Technical contact: Kris Kennaway, Brooks Davis

Develop and deploy infrastructure for annotating license conditions that apply to third party software in the ports collection. For example, identifying ports provided under the GPL version 3 license, or under licenses that do not permit redistribution or which impose non-standard requirements. Part of this project will involve exploring methods for automatically classifying licenses using HP's fossology tool (http://www.fossology.org/) or other mechanisms.

Requirements:

  • Familiarity with bsd.port.mk and related ports collection infrastructure.
 Complete (a.k.a. Fat) packages

Technical contact: Brooks Davis

When bootstrapping systems it would be useful to be able to create a single package file that contains one or more packages and all the required dependent packages. This is conceptually similar to, but different from PC-BSD's PBI package format. PBI's contain a private copy of all dependencies, fat packages would contain each individual package and once installed it would be as though each package was individually installed in the usual manner.

This project would consist of additions to the pkg_tools to support creation and installation of a new package file format and to ports to build these packages.

Requirements:

  • Strong knowledge of C code.
  • A basic understanding of the inner workings of the ports tree.
 Security Audit kernel event sources

Technical contact: Robert Watson

A number of kernel security subsystems, such as IPFW and pf, generate security log data. This task involves identifying potential sources of security event information in the kernel and modifying kernel subsystems to log that information using the kernel security event auditing system. User and programmer documentation of audit may be found on the TrustedBSD Documentation Page. There are also extensive manual pages relating to audit in FreeBSD. This project will require careful security analysis and kernel programming, and will likely need some re-working of the kernel audit framework (which is currently entirely focused on gathering user and kernel system call audit data).

Requirements:

  • Strong C programming skills.
  • Familiarity with concurrent programming techniques.
  • General understanding of TCP/IP firewalls.
  • Willingness to read the CC/CAPP specification.
 Distributed audit / log shipping daemon

Technical contact: Robert Watson
WIP: http://wiki.freebsd.org/DistributedAuditDaemon

Create a tool to securely and reliably ship log files to remote hosts. The main focus is to manage per-machine audit records and submit them to a central site for processing and long-term archiving/management. Ideally with support for SSL (or the like) so they do not travel on the wire in the clear.

Requirements:

  • Strong (portable) C programming skills.
  • Knowledge of the audit subsystem.
  • OpenSSL knowledge a plus.
 Libfetch authentication support

Technical contact: Poul-Henning Kamp

Currently libfetch only supports basic HTTP authentication, which is generally frowned upon because it transmits the username and password on the wire (base64 encoded). Add RFC2617 digest authentication.

 Mandatory Access Control

Technical contact: Robert Watson

FreeBSD 5.0 was the first FreeBSD release to ship with support for Mandatory Access Control (MAC), an access control technology allowing system administrators to implement multi-level security, integrity protection, and other "mandatory" policies. Policies may be compiled into the kernel, or loaded as loadable kernel modules. Later revisions of FreeBSD and the MAC Framework enhanced MAC support, and additional policy modules were made available, such as a port of the SELinux FLASK/TE framework available as a third party policy module. However, many of the sample MAC modules included with FreeBSD are considered experimental examples of what the technology can be used for, rather than production policies. For example, the Biba integrity policy can be deployed in production, but requires significant tuning to do so effectively.

This task involves a general review of the MAC Framework and Policy modules, with the goal of identifying improvement areas. It also involves specific cleanups, optimizations, and completeness work on specific policy modules -- most importantly, the Biba and MLS sample labeled policy modules. Work there includes improving memory overhead and efficiency; for example, moving from allocating complete labels for every labeled object to referencing common label storage where labels are identical, which occurs a great deal of the time in most systems. Other cleanups include moving towards a canonical/extensible on-disk label storage format, adding regression tests, investigating interactions with user applications, and writing documentation.

Requirements:

  • Strong C programming skills.
  • Familiarity with OS security policies, including discretionary and mandatory access control.
  • Familiarity with concurrent programming techniques.
  • Willingness to read the CC/CAPP specification.
 Path-based file system MAC policy

Technical contact: Robert Watson, TrustedBSD discuss mailing list

The TrustedBSD MAC Framework makes it easy to extend the FreeBSD kernel security model using pluggable modules, which has provided support for "traditional" mandatory access control, as well as allowed a number of companies to build local security policy extensions for workstation, server, and appliance products; one such example is Apple's "Seatbelt" policy for Mac OS X. However, the base system MAC policies are difficult to use, primarily because they either don't extend security meta-data (ugidfw) or they require extensive labeling (Biba, MLS). This project idea involves crafting a new OS security policy extension along the lines of ugidfw, the file system firewall, but using path names instead of existing file properties to manage protection. This requires careful thinking about what a file system path "is" in the UNIX environment, as well as regarding what sorts of policies would be useful.

Requirements:

  • Strong C programming skills.
  • Familiarity with OS security policies, including discretionary and mandatory access control.
 Security regression tests

Technical contact: Robert Watson

FreeBSD is undergoing constant and active improvement to all of its critical subsystems, from file systems to the network stack. With any change, there is a risk of introducing bugs or regressions. The goal of this task is to produce a security regression test suite, which encapsulates requirements regarding system security properties and tests that they (still) hold. Areas to test include file system access control, privilege, authentication, cryptography, process containment, and more. There are some current tests along these lines in the FreeBSD regression test tree, but they are both incomplete and and inadequate. New tests must be created; existing tests must be completed and updated.

Requirements:

  • Strong C programming skills.
  • High tolerance for writing test code.
  • High tolerance for reading API specifications.
  • Rigorous and devious mindset.
 Audit and Jail

Technical contact: Robert Watson, Christian Peron

The TrustedBSD Audit implementation allows fine-grained monitoring of processes in a FreeBSD install. This task extends the Audit implementation to have specific support for Jails:

  • teach the audit kernel code how to handle the security properties of jails,
  • implements independent audit configurations and a trail for each jail (possibly in addition to a global jail),
  • tags audit records with BSM zone tags to indicate which jail they came from,
  • write a test suite to confirm that it all works properly.

Requirements:

  • Strong C programming skills.
 A New Audit Parsing API

Technical contact: Robert Watson, TrustedBSD audit mailing list

The current OpenBSM audit parsing API has a number of limitations, not least that it can't handle little endian BSM records that may come from Solaris x86 systems, in terms of ABI robustness in the presence of new record types, ability to process trails generated non-locally in terms of supporting uid/gid->name translation, and in terms of incrementally processing a byte stream from, for example, socket sources without using the C FILE API.

This task would consider existing audit parsing APIs in the industry, including POSIX.1e, relevant Open Group specs, and in-use APIs on other systems such as Solaris, Linux, Windows NT, and others, in order to first identify an existing candidate API or design a new candidate API, then implement the API and adapt existing audit applications to use it. The task would also document the API using man pages, create an audit parsing tutorial document, create a test suites, and require interaction with the OpenBSM and FreeBSD communities to identify audit parsing requirements.

If successful, the results of this work would be integrated into OpenBSM, the open source BSD-licensed audit framework shipped with FreeBSD and Mac OS X.

Requirements:

  • Strong C programming skills.
  • Past coursework or reading in the area of computer security.
 NT/Linux to BSM Conversion Tool

Technical contact: Robert Watson, TrustedBSD audit mailing list

The BSM (Basic Security Framework) audit trail format is the de facto industry standard for portable operating system audit trails, being supported on Solaris, FreeBSD, and Mac OS X. However, many other audit trail formats exist that are less portable, including audit trail formats local to Windows NT and Linux.

This task would create BSD-licensed conversion tools to import audit trails from other systems and convert them to BSM format so that they can be inspected and managed using the OpenBSM tool set. This would require the creation of BSD-licensed parsers for audit trail formats of interest, designing and documenting a semantic mapping to the BSM trail format, and writing conversion utilities using the new parsers, semantic mapping, and BSM generation routines in OpenBSM. A key part of this work would be to rigorously understand and document the mapping and its limitations (for example, perhaps some Windows NT concepts can't be represented in BSM in a natural way, such as larger SID than UID fields, which is important information for anyone using BSM trails from NT systems). A test suite is also required.

If successful, the results of this work would be integrated into OpenBSM, the open source BSD-licensed audit framework shipped with FreeBSD and Mac OS X.

Requirements:

  • Strong C programming skills.
  • Past coursework or reading in the area of computer security.
 Application-Specific Audit Trails

Technical contact: Robert Watson, TrustedBSD audit mailing list

Currently, FreeBSD's audit trail support is a system-level facility, intended to track system-level security events. However, as the affordability of computers has gone up, they are increasingly single-application systems, and applications are increasingly playing a role in the Trusted Code Base (TCB) of larger integrated computing systems. As such, it is desirable to support reliable and secure auditing of application events that do not correspond to the traditional TCB components, such as events from databases, web servers, or virtual machines. However, these events need to be handled separately, as they have quite different security and selection requirements.

This task would extend the FreeBSD kernel's audit framework to allow multiple concurrent audit trails to be maintained, reflecting different "slices" of the system. The base slice would be the existing system audit trail, submitted to by the kernel itself, as well as using the audit(2) system call, and stored in /var/audit. New slices would be represented by named special devices in /dev, with ownership and permissions reflecting their desired application security properties; each slice would have its own audit worker and queue, with its own pipe reliability properties and target trail files. This would require extending OpenBSM's auditd and libbsm to support multiple trails as well. Finally, two or more sample applications would be modified to submit application-level auditing, such as the authentication routines in Apache and PostgreSQL. This design would have several interesting properties, not least that audit trails submitted by applications would be protected from application tampering after submission, as the kernel would mediate access using the TCB. Analysis of potential designs, security issues, documentation, and a test suite are required.

If successful, the results of this work would be integrated into the FreeBSD kernel and OpenBSM.

Requirements

  • Strong C programming skills.
  • Past coursework or reading in the area of computer security.
  • Past coursework in the area of operating system kernels.
 Userland / Installation Tools BSD-licensed ELF Tools

Technical contact: Joseph Koshy, Kai Wang

Create BSD-licensed versions of ELF processing tools (e.g., ld, dbx, as and others) using the ELF(3) and GELF(3) API set. Identify overlapping functions in those tools and create a library out of the common functions. Identify parts which can be generated by tools (e.g., machine code parser generators) to support our Tier-1 and Tier-2 architectures.

References:

Requirements:

  • Knowledge of C.
 BSD-licensed Text-Processing Tools

Technical contact: Diomidis Spinellis, Gábor Kövesdán

Create/port BSD-licensed versions of one or more of the text processing tools that are currently missing from the FreeBSD distribution: sort, diff, groff/troff and the grep family. Licensed versions of some or all of these tools are already included in OpenBSD, so this task involves more porting and feature completion than development from scratch. Emphasis should be placed on performance, standards-compliance, and support for handling wide character sets.

Regarding groff/troff, there exist the OpenSolaris versions at SourceForge which at least do not come with a viral license like the current GNU versions we use. Additionally this implementation has support for common vector fonts and unicode. If those utilities are option-compatible or not has to be analyzed. A port of this is already available as textproc/heirloom-doctools. Alternativly, OpenBSD's mdocml can replace groff's mdoc functionality and may be sufficent for our purposes.

Requirements:

  • Knowledge of C.
 Build options improvements

Technical contact: Alexander Leidinger, Gardner Bell

The new "delete-old" and "delete-old-libs" target in /usr/src for 6.1 and -CURRENT should be extended to support the WITHOUT_* knobs, e.g. WITHOUT_RESCUE or WITHOUT_CRYPT, and delete files which are covered by those knobs. Some switches have already been covered. You can view a list of all switches and what effect they have here.

Requirements:

  • Time to build and install the world several times.
  • A way to determine which files were not touched by an installworld.
 KDE front-ends to the freebsd-update(8)utility

Technical contact: Colin Percival

The freebsd-update(8) utility is used to fetch, install, and rollback binary updates to the FreeBSD base system. A nice project would be to develop a graphical front-end for freebsd-update(8), using the QT toolkit. A GTK frontend was developed as part of GSoC 2007 and exists at berlios; the QT frontend could maybe share common functions/classes and design ideas.

Requirements:

  • Experience writing KDE applications
 IPv6 User Land Cleanup

Technical contact: George V. Neville-Neil

Many userland network utilities do not work correctly with IPv6.

  • rpc.statd(8) is not IPv6 clean.
  • rpc.rquotad(8) is not IPv6 clean.
  • who(1) truncates IPv6 addresses in its output.

This project could also include a broader survey of other network services in /usr/bin and /usr/sbin to make sure they're all IPv6 clean.

 lint(1) improvements from OpenBSD

OpenBSD has some improvements to lint(1) which may be beneficial to have.

Requirements:

  • Good knowledge of C.
 Libprocstat and libnetstat

Technical contact: Robert Watson

Create, similar to libmemstat, wrapper libraries to support monitoring and management applications to avoid direct use of kvm. Three parts to the project: for each of the above, add kernel support to export data in a less ABI-sensitive way using sysctl, write a library to present the information in an extensible way to applications, and update applications to use the library instead of reaching directly into kernel memory / consuming sysctls. The goal is to allow the kernel implementation to change without breaking applications and requiring them to be recompiled, and to allow monitoring functions to be extended without breaking applications. This should also facilitate writing new classes of monitoring and profiling tools.

Requirements:

  • Good knowledge of C.
 NDMP data server

URL: The NDMP Initiative

The NDMP initiative was launched to create an open standard protocol for network-based backup for network-attached storage. Major commercial storage systems come with a compliant service. This allows major commercial backup systems to backup such NAS devices. Including a NDMP disk server into FreeBSD would allow to play nice out of the box (modulo some configuring) regarding backups in a corporate environment.

  • Evaluate the existing revisions of the NDMP standard.
  • Choose an appropriate revision (after checking of supported versions in commercial backup systems).
  • Implement at least a NDMP data server.
  • Bonus: implement a NDMP tape server (to allow attached tapes to be used).

Requirements:

  • Access to a commercial backup system with NDMP support (mostly for interoperability testing; since a NDMPcopy application seems to be available, this is not a hard requirement).
  • Good knowledge of a programming language which is included in the base system.
  • Knowledge about UFS snapshots.
 Port prebind from OpenBSD

The OpenBSD prebind is a secure implementation of prelinking that is compatible with address space randomization. Prelinking allows to speed up application startup when a lot of libraries are involved. This should show a noticeable effect with e.g. GNOME/KDE.

Requirements:

  • Good C knowledge (reading and writing).
 Proxy auto-config file support for libfetch

A proxy auto-config (PAC) file contains a JavaScript function "FindProxyForURL(url, host)" that determines which HTTP or SOCKS proxy, if any, to use to access a given URL. In most application the file may be specified manually or discovered using the Web Proxy Autodiscovery Protocol. Support for PAC files in libfetch would make fetch more versitle.

Supporting PAC files nominally requires a fairly complete JavaScript implementation. There appear to be no BSD Licensed JavaScript implementations so one will likely need to be written. A minimalist implementation of the language with commonly used constructs such as if/else, string comparison, and functions would be sufficient in many cases.

References:

Requirements:

  • Strong knowledge of secure C programming.
 PXE Installer

It would be great to have a bundled PXE installer. This would allow one to boot an install server from a FreeSBIE live CD-ROM on one box, set the BIOS on subsequent boxes to PXE boot, and then have the rest happen by magic. This would be very helpful for installing cluster nodes, etc.

Markus Boelter is working on a bundled PXE installer as part of his BSDInstaller project within the Google Summer of Code 2006. The PXE Installer is working but some non-PXE related issues have to be solved before it can enter the tree.

Requirements:

  • Good PXE knowledge.
 Regression testing system

Technical contact: Alexander Leidinger, Nik Clayton

Nik has written a regression test infrastructure using Perl. More of the regression tests should be made to work with libtap.

  • Many of the existing tests should be moved from using assert() to using ok() and friends from libtap.
  • More regression tests should be written.

Porting LTP might also be a good idea.

Requirements:

  • Good knowledge of scripting languages (Perl preferred).
  • Good knowledge of software testing.
 Sysinstall
  • Ask for network configuration before install - so you do not have to configure the net twice.
  • Make a guess of the timezone based upon country and keyboard.

Requirements:

  • Good C knowledge (reading and writing).
  • No fear regarding "naturally grown" code.
 Tar output mode for installworld

Technical contact: Colin Percival, Tim Kientzle

Instead of installing using install, mkdir, mtree, etc, directly construct a tarball. This would allow creating install distributions without root access, as setuid etc would never hit the local disk. This would require some retrofitting of our installation mechanisms.

Bsdtar now (8-current, 20080101) has a feature that allows it to create a tar archive from a description provided in the form of an mtree file. This description can specify owner, permissions, and contents for each entry and does not require the files on disk to already have correct ownership. This should make it possible to build a FreeBSD distribution as a non-root user. Talk to Tim Kientzle for details of the new bsdtar features and look at NetBSD, which has a similar facility, for ideas about how to proceed.

Requirements:

  • No fear regarding our installation system.
 Unicode support in vi

Technical info:: on J.R. Oldroyd's Unicode Support on FreeBSD page

Many base system utilities grew multibyte support in 2004. It would be nice to continue this trend by teaching vi(1) to display and edit documents in UTF-8 encoding. The above referenced page contains info of what is needed to improve the Unicode support in vi(1).

Requirements:

  • Knowledge of C.
 Improve cron(8) and atrun(8)

Currently, cron(8) and atrun(8) are outdated in their implementation. Here are some directions for improvement:

  • Update cron(8) to ISC cron with security fixes from OpenBSD.
  • Integrate the atrun(8) functionality into cron(8), as it was done in NetBSD.

Requirements:

  • Strong knowledge of the C language and Unix API.
 Improve Wine support in FreeBSD

Technical contact: Kris Moore

This last summer Tijl Coosemans did excellent work getting wine to a very usable point on FreeBSD. However, there are some issues which still need to be addressed at http://wiki.freebsd.org/Wine. This would be a big improvement for our desktop users.

 BSNMP enhancements

Technical contact: Shteryana Shopova, Bjoern A. Zeeb

BSNMP is a portable SNMP framework consisting of a daemon, modules and tools. It includes libraries that ease the development of loadable modules for configuring and monitoring various subsystems. You can find more information about BSNMP on the wiki. Some project ideas, but not limited to that list, can be found here..

Requirements:

  • Knowledge of C.
  • Some familiarity with SNMP/MIBs desirable.
 Userspace pthread mutex lock contention profiling and lock order verification tool

Technical contact: Jeff Roberson

This task would create an extension to the threading library that would allow application developers to measure and locate lock ordering and lock contention problems within their application. Such a tool is invaluable in debugging application deadlocks and creating high-performance multithreaded software. Existing lock ordering and profiling tools exist in the FreeBSD kernel, and could be used as the model for the userspace implementation. We would recommend beginning with profiling due to its immediate usefulness in optimizing performance, and to allow improvements in kernel scheduling to better manage user application lock contention.

Requirements:

  • Strong knowledge of C.
  • Threading experience.
  • Strong debugging skills.
 FreeBSD port of NetworkManager

Technical contact: Joe Marcus Clarke

This task is to port NetworkManager to FreeBSD. Porting NetworkManager will also require some core userland changes to FreeBSD, especially to ifconfig.

Requirements:

  • C programming skills
  • Knowledge of the FreeBSD net80211 wireless stack
 Fix FreeBSD support in GNOME's system-tools-backends

Technical contact: Joe Marcus Clarke

This task is to fix FreeBSD support in sysutils/system-tools-backends.

Requirements:

  • Perl programming skills
  • Knowledge of the FreeBSD system configuration
 Extend hal support on FreeBSD

Technical contact: Joe Marcus Clarke

This task is to add hal support for some additional subsystems. In particular FreeBSD is lacking support for the ieee1394 (i.e. Firewire), bluetooth, and printer. Adding support for these subsystems will require changes to the FreeBSD kernel. Those interested should use the latest HAL Specification as a guide.

Requirements:

  • C programming skills
  • Knowledge of the FreeBSD system internals
 Finish support for compilation of i386 binaries on amd64

Technical contact: Konstantin Belousov

The FreeBSD binutils and gcc on amd64 are capable to build 32bit binaries. What we miss is the proper includes for i386 accessed when #include <machine/some.h> is used, for the -m32 invocation of the cc. Project would provide an installation of both amd64 and i386 includes in non-conflicting way, and changes to the gcc to select proper location for machine/ based on the compilation target architecture.

Requirements:

  • C and shell programming skills
  • Knowledge of the FreeBSD build and installation infrastructure
  • Some knowledge of the gcc internals