Soeren Schmidt announced this weekend that ATAng has been merged into the FreeBSD -current kernel tree. Described by the author as “rather radical changes to the ATA driver”, ATAng offers a number of impovements over the old ATA driver. This includes removal of the ‘GIANT’ lock, an improved framework that supports newer ATA controllers, merging of ATA and ATAPI code, and the removal of numerous bugs.
Can’t wait to try it out.
I’m pretty n00b to BSD so… What was the ‘GIANT’ lock?
The Big Giant Lock is the coarsest lock in the FreeBSD kernel, and locks the entire kernel for exclusive access. Unfortunately it is still used by many drivers:
http://www.freebsd.org/projects/busdma/
Will it support my VIA onboard controller? It’s that VIA one that Linux has special “workarounds” for.
in the old good days freebsd was very selective about changing important parts of the kernel, they used to have a good development and testing cycle, even they prefer fix the current schema and not introduce a completely new one, now that’s gone, that’s why i stick with 4.8, even 2.2.8 IS the EVER stable, hopefully someday the good hackers (hubbard, matt) will return and fix the VERY unstable freebsd in these days.
I’ve been using FreeBSD 5-branch since the release of 5.0 and never had any stability issues, even with the first release of the NVIDIA drivers. I am using 5.1-RELEASE-p2 now and it works like a champ.
Which is why 4.8 is “stable” and not the 5.x branch or the -current branch. 5.x introduces some incredible new things which required radical changes. Personally I’m excited to know that new development is still occuring in FreeBSD.
I can’t agree with you. 5-CURRENT has never given me any unstabilities since 5.0-RELEASE. ATAng has been tested on the current@ mailing list for about a month before it was merged in CVS. And also, this is what -CURRENT is; things are tried out here first. I wouldn’t exactly call this a ‘path to instability’. First, because it isn’t true, and second, because the goal of -CURRENT is not to be the most stable branch. I’d call this the ‘path to a kick-ass OS’.
“I’ve been using FreeBSD 5-branch since the release of 5.0 and never had any stability issues”
That’s a lie! probably you never put a FreeBSD 5 with intensive Network & Disk I/O on a UP/SMP box, you’ll began with starving processes not getting enough CPU, simple test: storm your box with sufficient I/O (disk/net) and when it finish, you’ll have a “slower machine”, and probably it will deadlock to death.
Hey Otto … if he hasn’t ever used the system in a way which agrivated it’s instability it is NOT A LIE for him to say he’s never had a stability problem … and you know what, neither have I …
I know that I’m not pushing mine hard, because right now I run 4.8 on my “real” server, and 5.0 and 5.1 on my two desktop machines … but none of my machines is dual proc, nor do I get much traffic … all I’m really pushing is apache, ftp, and cvs, and those are kinda the vanilla server pieces …
still, don’t call somebody a liar who isn’t, instead, mention the areas that are unstable without the untrue name calling (by all means flame away, WHEN IT’S TRUE)!
another thing, they ARE backporting the good stuff to 4.x when appropriate, and easy … so it reall is the path to great OS …
That’s a lie!
And so is everything else which doesn’t agree with your opinion.
probably you never put a FreeBSD 5 with intensive Network & Disk I/O on a UP/SMP box
Isn’t “UP/SMP” just slightly redundant?
So basically what you’re asking is, do we have any I/O bound applications running on FreeBSD 5?
I do. My backup server is running FreeBSD 5. Every night I start an rsync process for every system on our network (so ~30 instances) and all operate simultaneously, writing to a two drive vinum array with UFS2+soft updates.
you’ll began with starving processes not getting enough CPU
Starving processes not getting enough CPU? We’re not beta testing Linux’s O(1) scheduler here…
simple test: storm your box with sufficient I/O (disk/net) and when it finish, you’ll have a “slower machine”, and probably it will deadlock to death.
Okay, well if you can suggest a better way to create an “I/O storm” than 30 instances of rsync, I’m willing to give it a try.
… giving a whole new ‘next generation’ meaning to ‘bsd is dead’ …
“That’s a lie!
And so is everything else which doesn’t agree with your opinion.”
Is not MY opinion, look at the freebsd site, they DON’T endorse the use of FreeBSD-5.x in production environments, look at the mailing lists, are plagued of stability problems, anyway can you show us your `uname && uptime`? i don’t think you’ll do…
Face it! FreeBSD 5.1-RELEASE is STILL in development as 5.0-RELEASE was, that’s not my fault nor yours but stability problems exists.
Isn’t “UP/SMP” just slightly redundant?
No it’s not. However, you can’t have a UP/SMP box. UP = 1 processor and SMP = more then 1 processor.
No it’s not. However, you can’t have a UP/SMP box. UP = 1 processor and SMP = more then 1 processor.
Using the canonical interpretation of “or” for /, we arrive at a UP/SMP box being a computer with one or more processors. However, any computer will have one or more processors, thereby making it redundant.
Q.E.D.
“What was the ‘GIANT’ lock?”
I found this page to be informative:
http://people.freebsd.org/~fsmp/SMP/SMPmeeting.html
Hope it helps…
Is not MY opinion, look at the freebsd site, they DON’T endorse the use of FreeBSD-5.x in production environments
Perhaps because they don’t want their mailing lists inundated with requests for help from those attempting to use 5.x in a production environment.
look at the mailing lists, are plagued of stability problems, anyway can you show us your `uname && uptime`? i don’t think you’ll do…
Not that this is really pertainent to anything, but:
FreeBSD avalanche.atmos.colostate.edu 5.1-RELEASE FreeBSD 5.1-RELEASE #0: Mon Jul 21 12:49:58 MDT 2003 [email protected]:/usr/src/sys/i386/compile/AVALANCHE i386
11:45AM up 25 days, 1:33, 3 users, load averages: 0.01, 0.01, 0.00
Face it! FreeBSD 5.1-RELEASE is STILL in development as 5.0-RELEASE was, that’s not my fault nor yours but stability problems exists.
So what would your preferable development model for an operating system be? Maintenance releases only?
Let’s look at another operating system that’s plagued with stability problems… Solaris.
Sun has bundled a number of components with Solaris which are not production ready, and many are installed per default. The first of these are the power management packages, which are responsible for a number of deadlocks on my Blade 2000s here. The next of these would be the Fair Share Scheduler.
Go ahead, run “dispadmin -d FSS” and tell me how long your Solaris system lasts before crashing.
Regardless, we have a number of people here running FreeBSD 5.x in production environments without issue, using it for CPU and disk I/O bound tasks you claim would bring the system to a screeching halt. Don’t you think your initial claim was a bit overembellished… to the point of being a lie?
“I’ve been using FreeBSD 5-branch since the release of 5.0 and never had any stability issues, even with the first release of the NVIDIA drivers.”
I’m going to go out on a limb and call you extremely lucky, instead of extremely untruthful
I am running FreeBSD 5.1 right now, with the latest nVidia drivers, on well supported hardware, and I am having no issues. When I used *either* FreeBSD 5.0 *or* the beta nVidia drivers (which if I remember correctly didn’t even support 5.0-CURRENT), very bad days resulted.
There was one point when I attempted to change virtual terminals only to be greeted by something that looked like a multi-colored lava lamp, and the times when I was running FreeBSD 4.8 with those flaky old nVidia drivers causing such horrendous lockups that a hard reboot was required.
FreeBSD 5.0 was in no way stable, as a general rule.
Hi. I installed the base system, copied to Cd from the downloaded iso. from there I installed kde3.1.3 per instructions at kde.org. KDE works fine, though i have not tested all apps.
also installed gnome 2.4 which was a chore ’cause i kept running into errors during the make install and would have to go back to the makefile and remove a dependency, seven of them, but i got gnome up and running though without gstplay, which i can find nowhere on the net. also gdm will not boot into the greater so i have to use kdm to get to gnome.
also installed mozilla 1.5a1,1 which works fine except will crash at flash intensive sights like ifilm.com or atomfilms.
mplayer works like a charm, much better than it does in linux.realplayer is grand. netscape4.8, browser only is fast and does do atomfilms.com. gaim is working. kmail works.
distributed folding is working.
freebsd is fine!!!! i feel i’m on the cutting edge with it.
watch out using that webcollage in xscreensaver while in kde . you might come home to find your wallpaper has been replaced with webcollage. honestly, the kicker was working and the task bar and i could launch apps all the while with webcollage going full speed in the background. kpm did not reveal xscreensaver as running. kind of scarry, so i rebooted. BUT it would be nice if this effect could be achieved in a legit manner.
Perhaps because they don’t want their mailing lists inundated with requests for help from those attempting to use 5.x in a production environment.
what a nonsense.
11:45AM up 25 days, 1:33, 3 users, load averages: 0.01, 0.01, 0.00
jaja, are you kidding! show me one with load more than 10 sustained (even linux-2.6.0-test4 can do that) and i will belive you (i’ll assume you’ll not edit it).
So what would your preferable development model for an operating system be? Maintenance releases only?
Even i don’t care what my preferable development model is, what i care is that FreeBSD-5 is not stable and was not supposed to reach the -RELEASE point, it should stick with the -CURRENT tag, anyway that one WAS a political move, not a technical one, you could pretend that everything is OK with FreeBSD and that is not true.
Are you the Scott Long’s babe louding to every one not happy with his decisions?
Let’s look at another operating system that’s plagued with stability problems… Solaris.
Is pertinent this?… just FYI i ran NetBSD & Linux on sparcs and they are truly stable hardware, solaris is not stable with opensource software, that is a fact, just ran OpenLDAP on solaris and count the days before you get db corruption.
Don’t you think your initial claim was a bit overembellished… to the point of being a lie?
I don’t think so.
What I say is in no way a lie. It may not apply to you and your particular experinces with the 5.x-branch, but it certainly applies to me. I am am using FreeBSD 5.x on my home desktop box and it has never given me any trouble, contrary to my experiences with RedHat, Debian and Gentoo.
I am using my box for running SETI, playing games (Quake + mission packs, Quake2, Quake3 + mission pack, Enemy Territory, Unreal Tournament and ZSNES), web browsing, irc, email, music and video playback (including watching DVDs), ripping auidio CDs and DVDs, etc, etc…
My system is an AMD T-Bird 1400 Mhz, Gigabyte-7DXR motherboard, 512 mb DDR ram, Creative GF4 Ti4200…
so how do I go about installing this instead of going thru the entire buildworld process?
Okay, I believe you. Some people are just lucky
I believe you either have to upgrade your existing installation to -CURRENT (using build/installworld) or wait for 5.2-RELEASE, which should occur in a few months.
Way to go Soeren! I wish you could have the time to respond to emails though… Although, I have to admit, he’s really a magic guy – I once had a broken SI0680 UltraATA controller that tried to destroy disks in Windows XP, but just silently refused to give any access errors in FreeBSD. Maybe even burncd will work too…. Ooh. Burncd is cool.
Oh wait, since it has been committed to the kernel tree, shouldn’t compiling and installing the kernel suffice?
No. The userland and the kernel must be kept in sync, otherwise “bad things” are to be expected.
Personally I think this is good news, news like this show that it isn’t only Linux that makes progress.
Btw, if you really want a stable OS you can always run DOS or NetWare. If 4.8 is the release that FreeBSD calls *stable*, you should not complain that 5.1 has some stability issues. Isn’t it the idea to support 4.x until 5.x has become really stable too?
You can’t really add nice things without creating at least minor problems, I think·
Thanks Bascule & Kingston for the info on the GIANT lock.
As for the instability of 5.1… It might be stable on some configurations, but maybe not on yours! No need to flame each other! It’s a work in progress so you shouldn’t expect it to be stable as a rock, anyway.
You can’t really add nice things without creating at least minor problems, I think·
That’s exactly what I think. Those new features might be buggy, but they will eventually be stable. Being conservative is good, but it shouldn’t stop evolution.
I can only vouch for the excellent stability of 5.x on my computers.
FreeBSD jaguar.gen.lu.se 5.0-RELEASE-p7 FreeBSD 5.0-RELEASE-p7 #0: Fri Apr 4 14:16:25 CEST 2003 [email protected]:/usr/obj/usr/src/sys/Jaguar_SMP_5.0.kernel i386
1:13AM up 130 days, 8:30, 2 users, load averages: 0.00, 0.00, 0.00
Now this production server is running the (according to posts) rather unstable 5.0 RELEASE. I have yet to see a crash on it. It runs phylogenetic analyses which are very CPU & disk IO intensive when they are run (usually runtime for the analyses are 20-72 hours). Used a little to little during summertime (hey it’s summer!).
FreeBSD gateway.chreo.net 5.1-RELEASE FreeBSD 5.1-RELEASE #1: Wed Jul 30 05:22:38 CEST 2003 [email protected]:/usr/obj/usr/src/sys/FreeBSD_5.1_Gateway i386
1:31AM up 13 days, 22:56, 1 user, load averages: 0.00, 0.00, 0.00
This is a router/firewall/simple_server installation running on an old K6 233 that I finally got around to replace my hardware router from a known manufacturer. A new install since last month that gave me a nice boot in netspeed from around 0.6MB/s to well above 3.5MB/s in download speed from my work computer located on a different net 13 hops away (according to a trace). Last reboot was due to a change in MTA.
Now I have also installed 5.1 on my laptop and runs analyses on that one. This install actually crapped out on me once… let me tell you; I did a custom kernel compile and installed it, booted and had the kernel throw up on me. Oook, remove -O2 from optimizations, recompile, install and have been working nicely ever since. Now a little reseach by me first would have revealed that I should not have used -march=pentium4 as that is known to produce BROKEN code on that gcc version. Simply put: my mistake, not FreeBSD 5.1 as the default install worked 100%. Now this one have a bad uptime (hey it’s a laptop) so no good figures in that department.
Now that said I can only agree with Bascules comments. With my setups 5.x have been reliable as 4.x (except on the laptop where 4.7 and 4.8 crapped out during installation). Now I would NOT recomment putting 5.1 on critical servers but I sure recommend it on non-critical production servers. As with anything like this: YMMV! My setups are NOT your setups.