Slashdot Mirror
Info About Kernel 2.3
Peter Hernberg writes "Linuxtoday has a interesting little tidbit from Linus about 2.3. "
Talks about the last fixes in the 2.2.8 patch as well as what is
happening first in the 2.3.0 series.
And just a note to help curb future flamage directed towards
me and my mediocre website from appearing on the kernel
mailing lists: 2.3.x is the devel series kernels. Don't touch
them unless you are a kernel developer, a mental patient (or both)
The appearance of 2.3.0 "merely" means that the Linux kernel development process continues. That's great. It's also good to see the fork happen much sooner than the 2.1.0 fork did.
Beyond that, the fuss and flames must be because of new Linux users. Otherwise, they'd already know the scoop about 2.odd.n kernels, and this silly discussion wouldn't be occurring.
To get real:
Is there a preliminary menu of features for 2.3? I've heard of LVM and ext3, to name two. There have been some rumblings of extending SMP to higher numbers of CPUs. Someone needs to say "GGI" and start another flame war...
I rather hope that 2.4 doesn't take as long as 2.2 did. Not out of any sense of hurry, but because the long gestation of 2.2 tended to put feature creep pressure on the 2.0 series.
Is that if you get a cycle as long as the last one was, you're practically FORCED to run one if you need features you can't find in the current stable ones. The 2.1.xx series gave me MUCH faster (Like, 3 times faster) disk access and better SMP performance too. If they worked (And they ususally did) they worked 100% -- nice and stable and much much faster. If they were broken, they would generally not boot at all and I'd just go back to the last one I was using.
Hopefully they'll narrow the feature set between stable kernel releases so we can get a stable kernel with features we need on a timely basis.
>>Don't touch them unless you are a kernel developer, a mental patient (or both)
This is of course crap. Everybody willing to test
development kernels is of course welcome to do so,
and a wide test by the user base is a very important element of the linux development process! If people don't test early the releases will be buggy, so to get good releases there need to be hordes of alpha/beta testers.
As the old GNU saying is "the contribution of users to GNU is testing". This applies 100% to Linux kernel development too.
Of course it is not recommended to run development kernels on production machines, but e.g. on your workstation it can be done without many problems. Even development kernels are usually more stable than most Microsoft releases @) If something doesn't work just fix it, of if you can't, report it with a detailed bug report (of course it is usually a good idea to check the l-k archives first if the bug has been already reported - that is often the case with "obvious" compile time problems)
If some misguided people on l-k say otherwise, just ignore them. Or did you see any prominent kernel developer in this silly thread?
-Andi Kleen
After Q3Test _hardlocked_ my SMP machine twice (with an ne2000 no less) I was getting a bit concerned as well. This bugfix is very encouraging :)
That's the only difference between 2.2.8 and 2.3.0. The 2.3.1prex patches contain further changes.
Uhhh, yeah, that's what Linus says in the page that's linked to...
I remember reading a quote where Linus _liked_ that people still run old kernels. He liked the fact that people didn't just _have_ to upgrade their kernels. Often, an older kernel suited someone just fine and they saw no reason to upgrade.
I think that's a pretty big complement to a programmer. That's basically like saying that what you had two years ago is good enough for me, what you have now must be fantastic!
Of course, all of that being said, you don't _have_ to have a reason to upgrade. You can do it just because you want to. That's reason enough.
If there are NO patches OR warnings on /. for a given kernel inside of one week of release, it is probably a stable kernel, with no catastrophic bugs.
If the last patch > 100K, even when bzipped, there are enough changes to make unexpected side-effects, or even simple typos, likely.
If the last two patches are If these suggest that a development kernel is going to be stable, I'll probably upgrade to it. I like playing with the new features. Mind you, if these suggest the kernel is liable to destroy the compputer, raid the fridge and pull funny faces at my boss, I'd probably upgrade anyway.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
....it begins to have fully working features such as full USB support, for example.
And even then I'll deliberately keep back a few revisions - which will allow me to see if there are any serious gotchas in any particular release (like filesystem corruption).
And even then, I'll be running it in a nice separate partition, and keep my main partition with a 2.2.xx kernel!
Because we can! :p~
The problem is that the market is flooded with all sorts of NE2000 clones at two for a penny, each of them having its own perculiarities and shipped with its own windows driver.
My honest advice would be to stay away from this minefield. Anecdotal evidence (of which there is far too much on /.) suggests that some NE2K clones make a dangerous playground - transfer rates would vary from 20 to 400kb/s, depending on which machine I was talking to. Another NE2000 clone, costing less than a tenner, would produce all sorts of wierd timeout errors and would not talk to anything, computer included. The former was (I think) due to the onboard ethernet electronics - windows produces the same results - and the latter was a pile of poo anyway.
My advice: Buy a different ethernet card and avoid no-name NE2000 clones like the plague. At risk of starting an ethernet flame war^H^H^H^H^H^H^H^H^H discussion thread I'd recommend any 3com card and the Kingston KNE100TX (based on the dec Tulip 21x4x) - they've both been more than satisfactory (980kb/s average for a 300MB file yestarday), and I'm sure there are many other cards out there well up to the job.
Too bad I haven't yet worked out how to construct a TCP/IP loopback with two cards in one box :-(
2.0.38 is due out around the end of the year, actually.
I ate something that disagreed with me. Maybe I should have cooked him first.
oh... oh yeah... those people
;-)
i used to be like that, i guess, for
a little while, until i broke my system
and had to, like, reformat...
heh... that sux...
linuxnewbie.com
Wow. You aparently read a different post than the one I typed. Let me try again...
I'm not telling anyone to NOT upgrade their kernel. I'm just question what the value of upgrading for the sake of upgrading is. If your machine is crashing, by all means, update the kernel. If you enjoy upgrading kernels or like learning about the code or whatever, upgrade the kernel.
All I was saying is there's a lot of people upgrading just to upgrade, just to have the latest version. They're machines dont crash, they dont read the changelogs, yet they MUST have the latest kernel... Why?
-Rich
I never said avoid bugfixes and updates. Clearly if your machine is being attacked or is in danger of being attacked upgrade your machine. All I'm saying is there's people out there who upgrade just for the sake of upgrading. That, I dont get.
-Rich
Edit your Makefile and make it any version you :-)
want...
-Rich
I dont see why people should be so thrilled about a new kernel unless it actually does something for them. I used to download the kernel every new patch for 0.99 and 1.0, but those were the days when each patch really addressed something important. If it aint broke, dont fix it, for the most part it aint broke, so I aint fixing it :-)
I cant even remember the last time one of my machines crashed due to a kernel bug '95 maybe. Possibly '94.
-Rich
Will Alan Cox still maintain the 2.0 kernel series? Seems like developing 2.3 while maintaining 2.0 and 2.2 would be a lot of work kernel?
Does anyone still maintain/use the old 1.2 kernel?
cpeterso
Then don't read it. Obviously.
~ Artificial intelligence is no match for natural stupidity ~
Mine has a couple times in the last few months. I always assumed it was one of vmware's modules since I was actively using that each time. I stopped using vmware and the crashes stopped. But Linus says:
Yep, describes my hardware and what happened exactly. I hope this means I can start using vmware again without fear...
Say hello to zMac.
the final 2.0.37 prepatch is almost done; When that's out, that'll be it for the 2.0 series. 1.2 et al is no longer being 'officially' maintained, but feel free to roll your own
-- Post No Gravy