Gary Stanley (‘Ancient’ on #dragonflybsd) has posted a patch against the most recent DragonFly sources that adds SCSI domain validation. It ought to work on older releases, too.
The name isn’t exciting, but SCSI domain validation ensures your SCSI bus runs as fast as possible. If you have the hardware for it, try it out.
Also, if you’re in a testing mood, Matthew Dillon has posted a new version of kern_objcache.c, using spinlocks instead of tokens, coming from a longer conversation detailing locking models in DragonFly.
It’s been reported that most every flavor of BSD (including DragonFly) has a FireWire bug allowing a local user to dump all system memory by passing a negative value to an ioctl. This is reported as part of the Month of Kernel Bugs, though that project’s web page doesn’t list it.
Joerg Sonnenberger pointed out that it isn’t a problem on i386 systems, as copyout checks that the argument doesn’t intrude into userland or beyond address space.
Porting/coding machine Sepherosa Ziehau has added stge(4) support, which works with a number of gigabit ethernet cards.
As Joerg Sonnenberger recently described, his ‘stable‘ group of binary pkgsrc packages comes from the regular pkgsrc quarterly releases, and the ‘current‘ batch comes from whatever is in pkgsrc at the time of the build.Â Stick with ‘stable’ for the most dependable results.
Joerg Sonnenberger’s presentation and paper (PDF) about pkgsrc, from EuroBSDCon 2006, is available (See abstract).Â Jeffrey Hsu, another DragonFly developer, also gave two presentations.
The virtual kernel work Matthew Dillon is doing will help support architectures other than x86 someday, but the work isn’t complete yet.
I post this in part because I see people ask “Does DragonFly support the AMD64?” relatively often. There’s also other platforms that are becoming more common (ARM) or less (PowerPC) that would be nice to support.
Of course, AMD64 is a relative term, since it certainly works on AMD64 – you’re reading this web page served from such a system now.
Today brought a number of commits for support of disk controllers and various networking chipsets.
The call for papers (check the list of people’s titles at the end of that document) has gone out for EuroBSDCon 2007.
From Sascha Wildner in #dragonflybsd: when you’re rebuilding parts of the world, use wmake instead of buildworld; the correct environment will be used, but the build will go much faster.
A combination of software upgrades and me changing my apache config on shiningsilence.com led to some unplanned downtime this weekend – sorry!
Matthew Dillon expounded a bit on the reasoning and method behind his kernel file reorganization.
Out of a conversation on users@, Oliver Fromme gave a list of the ports used by NFS. Someday, you may be on the other side of a firewall wondering what those ports are…
‘walt’ gave some details on configuring X to use DMPS, so that your monitor turns off when your console is inactive for extended periods.Â Good for conservation, but not as fun as the alternative.
Sascha Wildner gave an update on support for GCC 4.1: the short version is: almost working.Â Read his post for details.
Victor Balada Diaz has been working on improvements to jail(8); he has a patch that has IPv6 support and allows for multiple IP support.Â Please try, and give feedback.
Chris Kuethe has an article up at Undeadly that talks about using altq on OpenBSD (also possible on DragonFly) to prioritize traffic leaving his (zippy) home network through his (relatively poky) calbe modem.
lukemftpd (the ftp server nowadays known as tnftpd) was removed from the base DragonFly system by Perter Avalos.Â It wasn’t built by default, and it’s still available in pkgsrc if you need it.Â Note that the server version was removed, but the client version, lukemftp (also now known as tnftp) has been updated.
Thanks to Andreas Hauser, it’s possible to reach the EFNet #dragonflybsd channel through SILC at silc://silc.fortunaty.net.Â (Use this.)
Will Backman wrote a journal of his experiences at the recent NYCBSDCon. (Thanks, Undeadly)
Matthew Dillon is planning major cleanup in the kernel files, in part because it’s been historically inconsistent, and in part to support virtual kernels. The part that will affect most people is a new location for the kernel config file, and ‘i386’ is now the more relevant ‘pc32’.Â (or maybe not.Â I’ll post when it’s defined.)