Month: May 2006
Simon ‘corecode’ Schubert has added a change that makes cvs ask for confirmation before using a filename when adding a commit message.
As Matthew Dillon posted, SMP builds may be broken for the next few days, so rebuild with caution.
Matthew Dillon warned that he is committing a lot of work on multiprocessor support over the next few days; if you are one of the people who run bleeding-edge versions of DragonFly (1.5 from CVS, or ‘HEAD’), there will probably be some instability. It’s not called bleeding-edge for nothing…
BSD memory management systems are legendary for handling stress well; however, there’s a limit on how much paging can happen and still have a responsive system.Â Matthew Dillon has put in a possible improvment for low-memory solutions.
Sepherosa Ziehau has his version of the rtw(4) driver available for testing.Â This is taken fromt he NetBSD driver, and is used in newer wireless cards.
Matthew Dillon has found that an extended form of spinlock is useful for the MP goals he has to do before continuing his VFS work.
Matthew Dillon has removed support for ibcs2 and svr4 emulation.Â It hasn’t been touched in 10 years…Â Are there even binaries that still require that anymore?
Matthew Dillon has revamped the system include files in DragonFly, so now including the correct files is much simpler.
There were a number of interesting commits today: Sepherosa Ziehau’s new 802.11 framework, taken in part from FreeBSD 6, is now committed, and he’s also updated the man pages to match. (minor yet very important!) His ath(4) driver will be following soon. Also, Matthew Dillon has moved the LWKT from a token system to spinlocks – see the commit message for details. Finally, there are some side benefits for DragonFly from the Coverity scan of FreeBSD.
Matthew Dillon found some problems in his ongoing vnode work.Â Apparently, the way to solve them is to make other portions of the code multiprocessor safe.
This week, on UnixReview.com: the book reviews of “Extrusion Detection: Security Monitoring for Internal Intrusions” and “Linux Patch Management“, and more on security certification.
If you’re concerned about (or involved in) device documentation, there’s a new wiki site called Vendor Watch, which lists the state of efforts to get different hardware vendors to document their hardware in a way that makes it usable for open source efforts.
wiki.dragonflybsd.org is down, along with gobsd.com.Â The wiki was on a separate server from the rest of dragonflybsd.org, so the rest of the domain is fine, but there’s currently no details on when the wiki will be running again, as the hosting company has apparently taken the server offline.
Because of recent changes to the Java licensing scheme, it’s now possible to include Java as part of a packaging system.Â It’s available now for pkgsrc, for some versions of NetBSD.Â Other pkgsrc platforms (like DragonFly) will probably follow suit.
Matthew Dillon’s starting/continuing work on that aforementioned clustering by breaking out the journaling protocols into a module he’s calling “SYSLINK“.
Matthew Dillon’s decided to use the journaling work that was done previously on DragonFly to handle communication between the kernel and a VFS, and also between machines in a cluster. He typed up a very detailed explanation that shows where a lot of the groundwork has been done.Â (Plus, a followup.)
Marcin Jessa pointed out that since it’s possible to compile the DesktopBSD tools on FreeBSD, it may also be possible on DragonFly.
One of the proposed projects for Google’s Summer of Code 2006 is a rewrite of pkg_install, which encompasses the various utility programs used for pkgsrc. The proposal is by Joerg Sonnenberger, who has commit access to both DragonFly and pkgsrc, and has made an astounding quantity of packages work on DragonFly.
Sascha Wildner’s removing a whole lot of kernel options.Â Speak up if you are using them…Â though if you are, they probably don’t work.
Anyone want to write a new devfs? (That’s device file system, if you haven’t seen the term before.) A discussion about tracking disks and their appropriate mount points ended with Matthew Dillon noting that at this point, the DragonFly system is cleaned up enough that this would be an approachable task for someone with experience.