FreeBSD 4.9 Stability Update
Dan writes "FreeBSD Release Engineering Team's Scott Long would like to get a new poll on the stability and readiness of 4.9. The belief is that the last of the PAE-induced instability was resolved on August 31. Is anyone still experiencing unusual crashes, corruption, etc, on a system that is running with up-to-date sources? Now is the time to speak up and get the problems resolved. Scott is also requesting help with testing. In response to this, we are adding our own poll. Please vote and add comments as appropriate to help Scott."
Drivers and applications can run into some issues with the way memory is windowed and accessed via PAE. It could be that the drivers that have issues access memory in a way that isn't too friendly to PAE or things are hardcoded so that they could end up with memory violations.
/me shrugs
It's kind of like making drivers work properly in non-SMP and SMP mode, mostly how interrupts are handled. It can get even trickier when you throw in NUMA or ccNUMA found in the AMD64 architecture.
Okay, education point here (if you know, then it's cool. If not, here's something you should know).
The -STABLE branch is NOT STABLE! In the FreeBSD development cycle, the most stable systems run -RELEASE. Major development is done on -CURRENT. Working ideas are then moved into -STABLE (to stablize) -- a -STABLE system is a development system and is not guaranteed to boot. When the -STABLE branch proves itself to be very stable and contains enough new functionality, a -RELEASE snapshot is made. -STABLE should be renamed to -BETA or something similar (it actually was at one point, but this was retracted when a lot of people complained that they didn't want a -BETA branch, they wanted it to be stable).
For more information, please read about -CURRENT and -STABLE and what they really are at http://freebsd.org/handbook/current-stable.html.
Unless you're doing development and if these are production servers, I suggest that you run -RELEASE on them.
www.sitetronics.com/wordpress
From what I remember from the discussions on the mailing list the PAE import seems to have various instabilities even if not running PAE. This needs to be ironed out before a release. And seemingly this IS already fixed.