Coda is an advanced networked filesystem. It has been developed at CMU since 1987 by the systems group of M. Satyanarayanan in the SCS department.
C C++ Other eC Shell Makefile Other
Failed to load latest commit information.
coda-src Do not use a ViceFetch MultiRPC call to fetch file data Jan 23, 2017
configs Remove Kerberos 4/5 authentication Sep 29, 2016
doc Remove (most) of the __BSD44__ defines Aug 1, 2007
lib-src Bump versions for Coda-6.11.0 Jan 23, 2017
m4 Install a modules-load.d configuration file Oct 28, 2016
tools Don't try to canonicalize paths for now Jan 23, 2017
utils-src Normalize exit codes to stdlib defined EXIT_SUCCESS/EXIT_FAILURE Jan 6, 2017
.gitignore Various cleanups for the rpc2 stub generator. Aug 17, 2016
.travis.yml Trying to get a travis build to work. Jun 1, 2016
AUTHORS Convert Coda build to automake Aug 1, 2007
COPYING Allow automated license categorization. Oct 13, 2016
ChangeLog Bump version to 6.1.2, update changelog. Oct 19, 2006
INSTALL Remove mention of obsolete install targets from documentation. May 29, 2008
LICENSE Allow automated license categorization. Oct 13, 2016
LICENSE.lgpl Allow automated license categorization. Oct 13, 2016 Install a modules-load.d configuration file Oct 28, 2016
NEWS Move per volume store identifier/uniquefier to a global structure. Feb 7, 2008
README.BSD Remove RCS Id tags Sep 7, 2006
README.Solaris Point to solaris-coda for build instructions of solaris-coda. Feb 4, 2000 Remove bit about keeping commits separate Sep 23, 2016
README.nt Remove win95 (djgpp compiled, dos binary client) code. Sep 19, 2006 Fix cast from const char * in nt_msg_write. Dec 12, 2007 We couldn't find venus.conf/realms files anymore Aug 2, 2007 Bump version for Coda-6.11.1 Jan 23, 2017

Coda Distributed File System Build Status

Coda is an advanced networked filesystem. It has been developed at CMU since 1987 by the systems group of M. Satyanarayanan in the SCS department.

About this repository

This repository combines, to a best effort, the history from the official CVS repositories of Coda, as well as the supporting LWP, RPC2 and RVM libraries. The CVS repositories at this point only useful as a historical reference and further development will only be committed to the Git repository. The supporting libraries can be found in their respective directories under lib-src/.

The reasons the supporting libraries and their history have been merged with the Coda source tree are twofold. They were originally part of the Coda source tree but were seperated out as over time other non-Coda users of the libraries emerged. By now Coda has outlived all other external users and merging them back to consolidate the development history seemed to make more sense.

For the past several Coda releases there have been new releases of the LWP, RPC2 and RVM libraries. Coordinating building and packaging of a single source tree is more straightforward than building four different ones, especially when there is a specific build/install ordering dependency and a new build could break a currently installed version. The current combined tree will ensure that everything builds and installs with the same library versions.



Coda requires a working C/C++ development environment with gcc, gcc-c++, autotools, libtool, automake, pkg-config, flex, and bison. We require development headers for readline, ncurses5, and optionally the lua5.1 library.

On Redhat/Fedora/CentOS systems

$ yum install gcc gcc-c++ autoconf automake libtool pkgconfig flex bison readline-devel ncurses5-devel lua-devel

On Debian/Ubuntu and derived systems

$ apt-get install build-essential automake libtool pkg-config flex bison libreadline-dev libncurses5-dev liblua5.1-0-dev


$ ./
$ ./configure --prefix=/usr --with-lua
$ make
$ sudo make install


Configuring the Coda client

You need a configuration file in /etc/coda/venus.conf. The quickest way to get started is to run sudo venus-setup which will copy an example file with a suggested configuration, it will also create the directories that Coda uses for the local file cache and logs.

The first time the Coda client is started we need to force it to clean and initialize its 'recoverable virtual memory' (persistent state), this is done by running sudo venus -init.

Running the Coda client

The Coda client needs access to the file system services provided by the 'coda' kernel module, so to successfully start the client we have to make sure this module is loaded first. Once the kernel module is available, the client should be able to start and will automatically start running in the background as soon as it successfully mounts the file system on /coda.

$ sudo modprobe coda
$ sudo venus

Initially nothing will be visible under the /coda mountpoint, however any lookup for a Coda realm name, either through the file system, or through obtaining Coda credentials, will lead to a Coda server discovery request that, when successful, will make a permanent realm root appear.

$ ls /coda/
$ clog
password: guest
$ ls /coda

Because Coda persists its state in recoverable memory, the root will be visible even when Coda is restarted. To cleanly shut down the system, first unmount the file system and only then stop the daemon process, this way if any process has open references to files in Coda the unmount will fail and we avoid lost data.

$ sudo umount /coda && sudo killall venus

Configuring the Coda server



Although the source of LWP, RPC2, and RVM has been merged back into the main Coda repository, we are still trying to keep them mostly independent. This means that they retain their own build infrastructure and library versioning.

When updating supporting library sources make sure to properly follow the libtool library versioning guidelines. The version info is set using the CODA_LIBRARY_VERSION macro in and consists of current:revision:age values which update according to the following basic rules.

  • If the library source code has changed at all since the last update, then increment revision (‘c:r:a’ becomes ‘c:r+1:a’)
  • If any interfaces have been added, removed, or changed since the last update, increment current, and set revision to 0.
  • If any interfaces have been added since the last public release, then increment age.
  • If any interfaces have been removed or changed since the last public release, then set age to 0.

Aside from this, we try to minimize overall version number changes so we skip updating if the corresponding version number change has already been applied since the last tagged stable release. This is different from libtool guidelines where they suggest updating only immediately before a public release.

The assumption here is that version updates are easily forgotten right before a release so it is better to update early. Developers also don't have to care as much because they can run their binaries directly from the build tree in which case libtool will make sure the right library is used, or will use make install which updates everything at the same time.