Matthew Dillon has a rather lengthy writeup of the needs of a filesystem in a clustered latent environment.Â (i.e. DragonFly’s goal)
Month: January 2007
Found on the web: WarpBSD, a “project to incorporate OS/2 support into FreeBSD”, though it sounds like the vkernel now makes DragonFly a better choice.
If you want to read a very dense chunk of text, the “December 2006 Pkgsrc Changes Summary” is out.
UnixReview.com has 2 new articles up: “Parsing Web Form Input in CGI Shell Scripts“, which deals with the crazy notion of shell scripts handling interactive web pages, and “New Year, More Security Challenges“, which covers some U.S. federal law changes for 2007 that require computer data as part of the discovery phase of a lawsuit.
I’ve changed the theme, and added some elements to make this match the other DragonFly sites a bit more.Â If I’ve removed a link you need, please let me know.
An ongoing conversation about virtual kernels led to a description of just how virtual kernel and real kernel memory usage interacts; they are surprisingly well synchronized.
Matthew Dillon has synchronized Preview with the latest bleeding-edge code,Â to match up with the large number of commits lately.
Matthew Dillon reports changes to the kernel configuration file are needed now.Â Also, if you are running bleeding-edge code, a full buildworld/buildkernel is required on the next upgrade.
Branching for 1.8 will happen very soon; as soon as ACPI is ready.Â The release date has not slipped.
Joerg Sonnenberger posted to pkgsrc-users@ how he’s coming along in transitioning Xorg from the monolithic version currently in pkgsrc to the new 7.x modular version. (Short version: not yet, but soon)
And for your daily vkernel news…Â Virtual kernels now are secure by default, meaning no loading kernel modules and no writing to kernel memory.Â This is disabled by a command line flag at virtual kernel start time.Â Also, Matthew Dillon realized that the virtual kernel automatically presents a safe way to cluster.
Matthew Dillon’s latest status report indicates he’s added asynchronous I/O to virtual kernels, for a significant speed boost. Also: the 1.8 branch is tomorrow, with release in two weeks as planned.
Sepherosa Ziehau’s support for networking with virtual kernels has been committed. His commit message includes the exact instructions on how to get networking set up.
Now on UnixReview.com: Examining the Novell Certified Linux Professional 10 Certification, a review of the Komodo IDE (which I can’t get to work on DragonFly, darnit), and a book review of “Cisco Network Admission Control Volume II: NAC Framework Deployment and Troubleshooting“.
Virtual kernel development is proceeding quickly enough that I already have to summarize two days of development. Virtual kernels are able to finish a complete buildworld, at relatively good speed for an unoptimized system a few days old. This will be improved soon, with more room for development. Also, networking is working well, even with multiple virtual network interfaces.Â Oh, and various stat functions now work too.
Petr Janda wants someone to create a pkgsrc package for mod_cfg_ldap.Â Any takers?
Sepherosa Ziehau has already managed to construct a virtual network connection to match virtual kernels, ridiculously quickly.
So, there’s a Gentoo/FreeBSD project, that attempts to graft the two systems together.Â The lead developer in that project misread the old 4-clause BSD license on some older files, and paniced repeatedly.Â (It even made it to the howling wasteland.)Â Of course, the problem is not actually a problem – it’s caused by worry about a clause that was removed years ago.Â Dru Lavigne has a nice writeup, and Wes Peters summed it up best in comments: “Now everybody get back to work. This is a 7-year-old nonissue…”
Matthew Dillon reports it is now possible to boot a virtual kernel and log in using that virtual system. The big remaining step (other than bugfixing) is a virtual network interface.
Matthew Dillon is renaming some I/O calls.Â It shouldn’t cause major problems, but as always, make sure to do a complete buildworld/buildkernel when next upgrading your bleeding-edge system.
Matthew Dillon is making trapframe changes – it will require a complete buildworld if you are following bleeding edge code.Â Read his post for more details.
Matthew Dillon has posted a list of what remains on his virtual kernel work, along with the news that it can partially boot – see his post for the progress.Â It should be ready shortly after the next release.Â If you want to help, one of the needed things is a virtual network interface, perhaps similar to Qemu’s tap.
TLS system calls are being renamed by Matthew Dillon.Â If you’re running HEAD (bleeding edge code), this will require both a kernel and world rebuild on your next update.
Simon ‘corecode’ Schubert has committed his giant sound update, supporting (among other things) High Definition Audio and bringing in changes from FreeBSD.Â Aside from a change in the generic kernel module name, it should work the same.
A few of the mirrors out there have DragonFly source available through rsync; Peter Avalos describes the correct command to retrieve it from theshell.com.Â (Note: read that post for details before trying it yourself.)
One of the eternal chicken-and-egg problems is kernel modification.Â Sometimes, a freshly installed system requires a different kernel, but you can’t download the source to build that new kernel until those changes are made.Â However, kernel source will be included with the 1.8 release, so this should theoretically not be a problem.
Matthew Dillon posted a list of the various ways testing could be done over the next 4 weeks for the 1.8 release.Â Help out, if you’ve got the inclination.
Simon ‘corecode’ Schubert, having recently finished his thesis, has started up work on 1:1 userland threading again, with a status report in his latest commit.