The first release candidate of FreeBSD 5.4 is now available for general testing.
FreeBSD 5.4-RC1 Released
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
It looks like there is still some significant features either “In progress” or “Not done” yet.
http://www.freebsd.org/releases/5.4R/schedule.html
Particularly, improvements to the ULE scheduler
seems like those ule improvements are quite the long time coming
The made vast improvement to the ULE Scheduler since 5.3. It is now very usable, but unfortunately still has some problems. They also have same changes in 6.0-current that haven’t been MFC’d yet.
for 5.4?
Thnxs
I’m running current on a laptop with no real problems, but single cpu isn’t really the test. Is anyone running -CURRENT on a dual cpu system, if so, how does it seem to compare to 5.3+?
As I recall from various mailing list posts 6.0-Release is not all that far off.
Actually, I was surprised how short that todo has gotten, compared to the lengthy list we saw with the 5.2 and 5.3 release candidates. Notice that all of the “in progress” features are targeted to be complete for the full 5.4 release. It seems the only serious feature that might not be ready for 5.4 are the ULE updates, but remember that even the 4BSD scheduler is greatly improved in the 5.x series.
To those of you carping about the unfinished features: there’s nothing so “gratifying” to a developer after pouring heart and soul into a project than to hear the first words be complaints about the one or two things not done. I have been there more times than I can tell, and it is really wearying. These guys have moved mountains, and you are complaining about the few remaining pebbles in your way.
I can’t comment on the 5.4 RC or stable branch yet, but I am running 5.3 on a p4 3.0 with Hyperthreading turned on. The system sees that as dual CPU. I haven’t yet run into any serious problems, and that includes trying things like make world with -j8.
Not that that’s really a great indicator, but the machine has been smoking fast (ULE turned on), with KDE constantly running and doing heavy compiling.
That sounds pretty good– The last time I tried Hyper Threading I got hard locks, that was on a P4 3.0 and a supermicro motherboard.
Before long I am setting up a dual cpu Xeon, also with hyper threading, I am interested to see how I fair on that system.
On my personal systems, throughout the recent lifetime of 5.x, FreeBSD has spread from a single 4.2-STABLE firewall to several 5.x servers, two laptops, three workstations, and a few other odd machines. Now I only have one dual boot to windows, and a deficiency of space. But You are right alot of these boards get filled up with talk about ULE and where’s this or that, at least for me its b/c FreeBSD has seated itself like an addiction in my mind, forcing me to find bigger and better hardware for it to feed upon. It’s coming faster now and every release seems years beyond its predecessor and that is awesome, in the original size and scope of the word.
Not happy I just installed 5.3 on a server at work and now they release this.
Oh well I can always update it.
I’m install FreeBSD 5.3 Release on my old machine (Dual PIII Xeon 700MHz, 1 GB RAM). After update to 5.4, I got 5.4 Stable (#uname -srm -> FreeBSD 5.4-STABLE i386). It’s work with make -j15!! and compile custom kernel with SMP & ULE scheduler. Then I install Apache2.0 with worker.c, it can handle concurent 10000 static request(benchmark using AB->Apache Benchmark) but still die when request a heavy CGI/Perl page…:-P even only 1000 request. Maybe we need perl thread-safe program….:-p.
Anyway, congratulations to FreeBSD team!!!
How did you enable the ULE scheduler in the kernel?
Add SCHED_ULE to your kernel config, comment out SCHED_4BSD and recompile the kernel.
FreeBSD hanged himself during initial boot on a AMD64 3000+ Asus K8N mobo with 2 80 GB SATAS.
Not happy I just installed 5.3 on a server at work and now they release this.
Oh well I can always update it.
That’s the whole point of FreeBSD; updating is not a showstopper ;-).
It’ll be interesting to see how much FreeBSD improves for 6.0 since by then the word that P.H.K. got donatations for should be all merged. *excited*
Anyone know if the Direct Rendering Manager has been fixed in -CURRENT yet? PHK knowingly made changes that broke it a couple months back and, last I heard, no one had fixed it yet.
You need to disable apic from bios or boot once without acpi and disable it from /boot/loader.conf with hint.acpi.apic.0.disabled=”1″ It’s a bug in nforce chip that they haven’t been able to resolve in 5.3 at least.
There’s a patch from jhb on the cvs-src@ mailing list that fixes it. Not sure if he committed it.
How many cds do I need to download, I can see 2 that I might need. Please help
Just disc 1 should be adequate
you only need the first iso. from there you can install over the net or from your CD. (i’m not sure what the “boot only” iso will give).
you can install the minimal needed for a complete freebsd OS from the first cd. make sure you also install the ports collection, then once your running freebsd at the comand line, you can install everything you need using the ports collection or you may prefer packages.
You must read the docs to learn how to choose “install media” and/or how to use ports or packages.
Off course the second CD will save you time and hassle with installing those gui’s. I prefer to build from sources so the second CD would be a waste of time.
i’m not sure what the “boot only” iso will give
The “boot only” ISO contains no install sets (tarballs). They’re just bootable, nothing more (like the floppies). You’ll have to get the rest via FTP or somewhere else.
Full cd = boot-only + install sets. The other cd’s are stuffed with binary packages (for use with pkg_add), they’re not required for a base install.
I can’t comment on the 5.4 RC or stable branch yet, but I am running 5.3 on a p4 3.0 with Hyperthreading turned on. The system sees that as dual CPU. I haven’t yet run into any serious problems, and that includes trying things like make world with -j8.
You’ll most likely want to disable HTT in the BIOS and run a non-SMP kernel. The FreeBSD scheduler is not HTT aware. It treats the logical CPU as if it were a full-fledged CPU, and will schedule tasks as if there were two FPUs, two ALUs, two caches, and so on. This will lead to performance decreases for most workloads.
It’s just not worth it to enable HTT.
I’m install FreeBSD 5.3 Release on my old machine (Dual PIII Xeon 700MHz, 1 GB RAM). After update to 5.4, I got 5.4 Stable (#uname -srm -> FreeBSD 5.4-STABLE i386). It’s work with make -j15!! and compile custom kernel with SMP & ULE scheduler. Then I install Apache2.0 with worker.c, it can handle concurent 10000 static request(benchmark using AB->Apache Benchmark) but still die when request a heavy CGI/Perl page…:-P even only 1000 request. Maybe we need perl thread-safe program….:-p.
You didn’t update to 5.4, you updated to 5-STABLE. Big different.
In order to get 5.4 you need to cvsup your source tree using tag=RELENG_5_4. This will get you 5.4-RC1 (labelled 5.4-PRERELEASE in uname output). Once the release is official, you will see 5.4-RELEASE in uname output. And once a security fix hits the tree, you’ll see 5.4-RELEASE-pX in uname output.
You updated using tag=RELENG_5 which gets you the latest development code in the 5.x series. Right now, this is 5.4-STABLE, or just 5-STABLE. This is not the 5.4 release code.
I agree with you. I chose tag=RELENG_5 because when 5.4-RELEASE coming, the stable branch will be 5.4-STABLE. I’m prefer using stable branch….:-)