Slashdot Mirror
Linux Kernel 2.6.0-test10 Released
antrix angler writes "Linus Torvalds released the 2.6.0-test10 Linux kernel today, tentatively calling it the "stoned beaver" release. Linus plans to hand the kernel over to Andrew Morton in a few weeks, and then it will be up to Andrew to decide when we see the final 2.6.0 stable kernel. Download it from a mirror."
i had some serious problems with my usb mouse in test9 (while it worked in test4 and before), and i don't see any mouse related fixes in the ChangeLog, so for now I guess usb mouse users should stick with older releases.
i really hope this gets fixed before 2.6.0, especially since it worked before..
Here
Let's share our bandwidht!!
--
"Res publica non dominetur"
Is 2.6 really noticably faster than 2.4 for regular desktop use (X responsiveness, etc...)?
Yes, it definately is.
Note that Linus says preemption should be turned off for now, there are odd problems.
-- Note: If you don't agree with me, don't bother replying. I won't read it.
Well I've been using the 2.6 series for about the last 3 months on my desktop. Not had one crash and its been under heavy load. Definitly shaping up well.
Now if I could just get iptables working right
Rus
Cheap UK and US VPS
Preemption is one of those things thats great under ideal circumstances, but in something less it hurts more than it helps. Preemption, for anyone that doesnt know, allows one task to jump to the top of the runqueue whenever it needs to run anything if it has a higher priority than everything else. This works great until you run something like SoulSeek or Bittorrent(both wxwindows python apps..) which end up with a much higher priority than they deserve. This then preempts all of your other tasks making your system highly unresponsive.
Pain lasts, kid. Its how you know you're alive. Sometimes I think this growing up thing is just pain management-TheMaxx
Nope, you do not /have/ to enable devfs in 2.6. I'm using it on several machines just fine that do not have kernel support for devfs or the userspace devfsd component.
/vastly/ improved.
I've had both good and bad experiences with devfs+devfsd, but any problems I have had have been quickly debugged and fixed (or I've bashed my forehead for being pebkac error). Generally I err on the side of caution (except for this 2.6.0-test deal, since more testers can't hurt), so I've decided to remain without devfs+devfsd on several machines.
Yes, the build system in 2.6.0-test is
I know I'm anonymous, but can anyone tell me whether I have to/ really really should turn on devfs if I upgrade to 2.6?
I've tried using devfs under 2.4 several times but have never succeeded in getting it to work with my crufty (been around since slink) debian box.
And since I use ALSA and XFS the 2.6 kernel would simplify compiling desktop kernels no end.
I don't know about 2.6, but I use 2.4.x with devfs on two stable boxes ("servers" running on obsolete desktop hardware) and two unstable boxes (a desktop and a Powerbook), and they all work fine. Tab-completing commands is so much nicer when you only have device nodes for hardware you actually have; it's also handy to be able to see (say) whether your CD drive was detected properly, or how many partitions a hard disk has, by looking at the device nodes.
If you're using at least Debian 3.0 stable (woody), install devfsd, install a devfs kernel, reboot, and everything should Just Work.
What specific problems do you have with it?
Not at all. In fact, devfs has been declared obsolete in 2.6; on those machines where I'm testing 2.6, I'm using a plain old /dev until I have time to figure out how sysfs and udev are supposed to work.
Now, if you want to use devfs, all you have to do is: 1) install devfsd, 2) compile a new kernel with devfs enabled and set to mount on boot, 3) install said kernel, 4) reboot. Doing 1 before 2 is the tricky stuff for me, I always forget that. You may also have to adjust some permissions and/or create some nonstandard devices you made yourself, but I've been using devfs on all my machines with 2.4 and it's working perfectly.
BTW, I'm not faking Linus, here is the signature :)
--
"Res publica non dominetur"
These explained a little:
... thousands?
udev presentation (PDF), Google HTML version.
Detailed paper on udev (PDF), Google HTML version.
devfs works fine for me, but since some people (see second link) want thousands of disks I guess I'm not the target market. I mean
Slashdot looked deep within my soul and assigned
me a number based on the order in which I joined
One devfs gotcha in 2.6, which caught me out for a while, is that the devpts (pseudo-tty) functionality has been separated out. In 2.4 kernels, mounting /dev using devfs automatically mounted /dev/pts too; however, this behaviour has changed in 2.6, and you have to mount /dev/pts explicitly:
On some systems (such as Gentoo), the rc scripts are smart enough to do this automatically.
Tubal-Cain smokes the white owl.
No. This was fixed in 2.4.19 (well, /really/ fixed in .20 since .19 really just disabled it completely) and in 2.5. See this page.
I'm running 2.6.0-test10 at this very second, and have been running every test version since test2 except for test8 (screwed up my sound system for some reason). I gave udev and hotplug a try but found hotplug to be straying behind. They had a few notes on their web site about patches for pciutils and usbutils to add something that probed all the devices (especially for coldplugging), but also said that you won't need them for 2.6.x due to the new sysfs filesystem (mkdir /sys; mount none /sys -t sysfs). But then doesn't say anything else about hotplug utility support in 2.6.x. So without a working hotplug, udev is kind of useless (since it interfaces with the hotplug).
:).
ATM I'm using murasaki as my hotplug facility, as I've personally had the best luck with it (that's really the only reason), especially on the 2.6 test kernels. I'm also still using devfsd (which will be obsoleted by udev).
This is the extent of my knowledge about new things like udev, etc. in the new kernel. So anyone should feel free to correct any innacuracies, omissions or blatant stupidities
- Sil
A new interrupt controller dreamed up by Intel to enable multiprocessing. It seems to cause a disproportionate amount of problems under Linux (a dedicated server I run would crash something like every 4 hours before I enabled noapic)
Unless you've got more than one CPU, it's more trouble than it's worth. The old 'legacy' 8259 interrupt controller (or the interface your system chipset supplies that emulates it anyway) works just fine for single-CPU applications.
But the easiest way to change it is to simply type:
...and let debconf allow you to choose the nice value with the normal debconf configuration dialog.
dpkg-reconfigure xserver-common
JFS seems to be abanodoned. Recenctly I've found serious bug in JFS and submitted it to their BTS. After 10 days there's no comments, no followups...
Actually my friend, as you'll see from pages like this, it actually often spends much less time in the kernel. One major slowdown always present is the tenfold increase in timer interrupts. As you can see from the web page I referred to though, this doesn't always result in a net slowdown.
Is slashdot getting slow...?
Story is up 12 hours after release on bitkeeper!?
WTF?
Warning: This sig contains a small bug. ==> *
Note that this patch doesn't apply directly to any of the 2.6 kernels; I just make the change by hand, since it is only one line. I have no idea why Linus isn't including this patch; it has been available for months, and it isn't exactly huge. It does fix the Radeon frame buffer issues.
Also note that /. is munging the code; it insists on inserting a "&nbs p;" that shouldn't be in there. I can't seem to get rid of it. Gotta love buggy software.
I agree, if you want a pretty business name, go wrap the Stoned Beaver up in a distro or just call it by it's version number. You're free to do so.
That's intentional, and is part of the anti-page-widening-post code. It prevents really long lines causing the page to overflow.
1) open web broswer and goto www.google.com
2) type "building 2.6.0 with redhat"
3) browse results.
Or, goto kerneltrap.org, for eg.
http://kerneltrap.org/node/view/799 - a nice article, but slightly outdated now.
The biggest gotya is the requirement for a new modutils prior to running the newer kernel. Read the article for more info.
- I stole your sig.
I'd rather have badgers on shrooms instead.
Escher was the first MC and Giger invented the HR department.
I hashed this one out with Patrick and Rusty Russell, and Rusty was the one who caught it.
oh fuck yes
2.4 was completely unacceptable. 2.6 is what allowed me to ditch Windows.
I understand there are patches to 2.4 that can give you most of the same stuff, but I really didn't feel like dealing with that.
When someone might yell at me, it has to be OpenBSD.
you want only one of the two threading ports to respond to hardware events. Not a good idea, it's best to spread that out... especially since a thread could starve the other on scarce shared processor resources on the PIV
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
That's wrong. "sysfs - _The_ filesystem for exporting kernel objects." (from Documentation/filesystems/sysfs.txt). sysfs exports kernel data structures to userspace and in the long run should leave /proc for processes like it was meant. it is no replacement for devfs.
but there already exists a userspace daemon called udev which simulates devfsd behaviour by taking the information found in sysfs.