The superfast filesystem Reiser4 has been pushing for entry to the mainline kernel tree for sometime now, but has been hampered by incompatibilites with the core filesystem layer. After learning of plans to include Reiser4 as early as 2.6.19, we decided to get the low down on whether Reiser4 would, after years of development, finally make it into mainline.
Reiser4, the successor to the popular Reiserfs by Hans Reiser, has been in development for a number of years now and, for almost as long, Reiser has been pushing for inclusion of the filesystem in mainline -- that is, to make it into the official kernel release. While reiserfs has been part of the kern since 2.4.1, Reiser4 remains a feature offered only in third party kernel patchsets.
For those living under rocks, Reiser4 is the only Linux filesystem of its kind based off 'dancing B+trees', a faster form of B-Trees commonly used in modern filesystems today (the operation of which I couldn't explain even with a degree in 'genius'), and has developed into a fast, scalable, and extensible filesystem making such inroads that its continued development is partly sponsored by DARPA (the US Defense Advanced Research Projects Agency).
While still in development, the core inhibitors for Reiser4's inclusion have possibly been more political than technical. Reiser is a passionate and outspoken developer who has, at times, rubbed some of the most respected kernel developers the wrong way. Still, he has listened to criticism and slowly but surely Reiser4 has been prepared to merge into mainline, thanks to reviews of the code and feedback by kernel developers.
So it was with interest that rumours recently spread on news and mailing lists that, finally, an inclusion of Reiser4 into mainline was imminent in the upcoming 2.6.19 or 2.6.20 kernels. This would be a milestone for Reiser4, and about more than just giving the filesystem a formal acceptance into the kernel tree -- right now, the only way to use Reiser4 is to patch the kernel by hand, or use a pre-patched kernel that uses Andrew Morton's '-mm' sources, which has been the unofficial home of Reiser4 for some time now. Additionally, inclusion in the mainline will mean increased adoption through distributions bundling the stable kernel releases -- as it stands now many distributions, even the big names, are reluctant to support an unofficial patch to the kernel.
To see if the rumours were true we talked to Andrew Morton, one of the key developers reviewing and providing feedback on Reiser4, about the filesystem's inclusion into mainline. Morton confirmed there are plans to finally integrate the filesystem, but that it wasn't going to be happening just yet, stating there is some still some ground to cover and a 2.6.20 release is unlikely. According to Morton there's still some work to be done before it's ready, saying "It's missing important functionality such as direct-io and extended attribute support. We can probably provisionally merge it in that state... [but] it needs further review by other filesystem developers."
So, for all the Reiser4 fans, it's unfortunately not happening just yet. In the meantime, you can as usual find the latest release in Andrew Morton's patchsets.
Reiser4 as integrated into the 2.6.18-viper1 experimental kernel.