Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Linux Software

Linux 2.2.15 Released 146

Posted by CowboyNeal from the waiting-for-2.4 dept.
Fluid Donkey was the first of many to let us know that kernel 2.2.15 has come down the wire. It's a fairly large patch (just over a meg), but it contains many updates, including support for Intelligent Input/Output (I20) devices. You can find it in the usual places. (ftp.<two-letter-country-code>.kernel.org if you forgot)
This discussion has been archived. No new comments can be posted.

Linux 2.2.15 Released More

Linux 2.2.15 Released

Comments Filter:

  • Re:Easy kernel upgrading? (Score:1)

    by Anonymous Coward
    kernel upgrading is a bitch. remember this:

    keep rescue disks and back everything up.

    I had many problems upgrading on Redhat. I had to make many manual changes. the lib/modules directory wasn't created initially. a whole host of stuff.

    As far as information. The README in the kernel source tells you how to do the kernel recompile. You just need to Cover Your Ass by throwing the existing kernel and lilo configuration on disk somewhere before you go experimenting.

    I'm too tired, so I'm probably not making sense.

  • Re:Who works on this? (Score:1)

    by Anonymous Coward
    My fav:

    Name: Douglas Stanley Jr.
    Email: IT-Manager@MontgomeryAirport.org
    Location:
    Occupation:

    This is one of main weaknesses that plagues open-source projects. This kind of software is a blessing, but it can be frustrating to wait for upgrades. Linus Torvalds may claim that his job isn't hindering further development of the 2.4 Kernel, but the fact is that it IS problem for most working people. One advantage that proprietary software has is what I call "market discipline"; if a company sets a schedule, and sets budget and manpower levels adequate to the task, they can usually finish the task in the desired amount of time

    Kinda like Slashdot's "IT Consultant" troll, but apparently for real.

  • Root RAID (Score:1)

    by Anonymous Coward
    I'm running root RAID on one of my boxes, which is still running a copy of 2.2.11. I was going to update it to 2.2.15 tonight, but remembered the root RAID patches that I had to apply before. I checked out ftp.kernel.org [kernel.org]'s raid patches, but the last one available was 8/24/99 for 2.2.11. Is this patch ok to apply against 2.2.15, or is the latest root RAID now incoporated into 2.2.15, or what? Any help would be appreciated...
  • Someone already does this, it's just not up the instant the kernel's released. Try kernelnotes.org [kernelnotes.org] and wait for the link to show up.

    :wq!

  • I don't normally do "me too" posts, but I gotta add a bit to this one...

    My router ran 2.0.37 before it died. (massive failure of 9 year old hd) The new one got a 2.2 kernel because it was a backup/test box, and wanted it up in mimimum time. But, it uses no more than is available in 2.0. 2.2 was on it, and it worked.
    In short, use what works. Don't upgrade for upgrade's sake.

    Thou on a more interesting note, the same box is running debian potato. Talk about a moving target there. :) (Elvii forgets what he just said about not upgrading for upgrade's sake. :)

    bash: ispell: command not found

  • Re:Who works on this? (Score:4)

    by Eric Green ( 627 ) on Wednesday May 03, 2000 @06:45PM (#1093456) Homepage
    Some of us need a stable kernel and Linux distribution because we need to get work done, we don't need to twiddle with the Linux kernel all day long. My router, for example, is still running the *2.0.37* kernel... there just hasn't been any reason to upgrade it, it just plain works, why mess with stuff that works?

    Most of the changes in 2.2.15 were actually back-ported from the 2.3 series kernels. For example, I helped Kai with some changes to the 2.3 series tape driver, and those changes have been backported to the 2.2.x series kernels (but not in time for 2.2.15, sigh... maybe 2.2.16).

    -E

  • Red Hat release a new version every 6 months...usually April and October. Every year it's been like this. So...don't grab the latest release in September either. :)

  • Re:Now I'm a tad bit miffed... (Score:3)

    by The Man ( 684 ) on Wednesday May 03, 2000 @06:17PM (#1093458) Homepage
    Well, there were 20 prereleases. And Alan's diary [linux.org.uk] did note that he had sent a 2.2.15 to Linus. And there was already a 16pre1 out there. A bit more checking and you wouldn't have had this problem. Oh well, it's fun anyway, right? :)

  • I have always wondered what it is and what it does?

    Well, since you don't want to take the 5 seconds on Google to search for "i2o", I did it for you.

    Check out www.i2osig.org [i2osig.org], or go straight to their Q&A section [i2osig.org].

    More importantly would it give me any speed improvements on an IDE system?

    I'd guess it might, if you had an I2O IDE controller. Check out this list of products [i2osig.org].

  • Re:Forgive my impatience, but (Score:3)

    by Stormie ( 708 ) on Wednesday May 03, 2000 @06:22PM (#1093460) Homepage

    2. Why are additions being thrown in? Are they at least defaulted to "N"?

    I would assume they default to "N" (most things do) but I can't swear to that.

    As for the why - they are being thrown in because people wish them to be thrown in, and because it pleases the benevolent lord of the 2.2 kernel, Alan Cox, to grant these people their wish.

    Seriously though - the additions being "thrown in" are device drivers, for devices that didn't exist when 2.2 was released. People own these devices and need to use them - you can't just tell them to use a dev kernel, or wait for 2.4. You need to implement these drivers as quickly as possible, test them hard (2.2.15 had about 20 -pre versions before this final release, you know), and then get them out there!

    I suppose you could adopt a "bug fixes only" attitude towards changes in the stable major version, but you would really really need to have much more frequent major releases for that to make sense.

  • I2O is pretty naff from a software standpoint. The i960 that is used for the IO processor (hey they used to use Z80s for this!) runs vxWorks. Couldn't they have used an open source "alternative" like RTEMS?
  • When they release 2.4.x! (or 2.5.x, for that matter...)

    But, in the meantime... over a meg? Anything really cool I should know about?

    If there's no Changelog, I guess I could just grep through the patch. Documentation is for wussies anyhow, right, guys? :)
    ---
    pb Reply or e-mail; don't vaguely moderate [ncsu.edu].
  • What size limit?

    Did you try typing "make bzimage" or "make bzlilo"?
    ---
    pb Reply or e-mail; don't vaguely moderate [ncsu.edu].

  • The Real Meaning of I2O... (Score:3)

    by jd ( 1658 ) <imipakNO@SPAMyahoo.com> on Thursday May 04, 2000 @02:19AM (#1093464) Homepage Journal
    Some time ago, scientists realised that the world's H2O (water) supply was becoming increasingly contaminated, depriving computers of necessary fluid to keep functioning.

    They therefore created "water++", or I2O, to fix the problem. I2O is "virtual water", and provides all the liquid refreshment your PC requires.

    Unfortunately, some motherboards don't have the necessary aqueducts to support I2O. In these cases, it is necessary to install suitable drivers who can run taxi services to and from the I2O manufacuring plants.

  • And Leno's people wrote this question, and intercut shots from Who Wants to be a Millionaire with Ed Hall sitting in front of a fake game set. Hal Duston
  • I tried Linux 2.1.15 and it no longer recognizes the Intel EEPRO 100B clone card (it uses the same chipset as the 100B). I looked at the patch and there are extensive rewrites for the eepro100.c driver. 2.1.14 and earlier worked just fine.

    (Rant) It's too bad Don Becker and team are no allowed to maintain the Ethernet drivers for Linux anymore--now we have a bunch of apparent amateurs breaking stuff. Oh well. :-(

  • hook, line, and sinker.

    come on, could he have been more obvious?
  • (This comments are from my severely coffee deprived brain, so it might be wrong on a few points).

    I believe DPT got stung with some of the code that was used because they got part of it from the I20 orginization and it has be be kept seperate from the kernel due to licencing issues. But, there has been quite a bit of work being done to get the I20 drivers in the kernel to run the DPT cards, and thus you wont have to use the one provided by DPT.

    I haven't look into it for a while, I'll guess I'll have to revisit it now.
  • Friend, the same copy of Solaris that runs my Sparc 20 can power an E10k. does an out of box linux distro power the same range of equipment?

    Actually, yes. I installed Linux from a single CD on my Sparc 4, Sparc 20 and an Ultra Enterprise 4000. The kernel has had support for the starfire for some time, too, although they wouldn't let me near ours with a Linux CD :-( Oh, and the same kernel image can be used on sun4c machines, too, which is more than can be said for Solaris. That said Solaris does currently scale better than Linux.

  • Same thing happened to me a few weeks ago. I was 500Mb into d/l redhat 6.1. What happens that sunday when I want to continue the d/l. Redhat 6.2. Happy d/l.
  • You are missing something.

    Odd numbered kernels are development kernels, while even numbered ones are "stable".

    There are lots of good reasons to use 2.2 series kernels, like not having weird-arsed crashes owing to obscure bugs under heavy load, being more confident that there are fewer unpublicised security holes, yadda yadda.You may not be living dangerously using a 2.3.x kernel, and if you have obscure hardware you might NEED to, and someone has to run them or they won't be properly tested, but sometimes you need the comfort of a stable series kernel.
  • I'm upgrading after trying out a DoS I found from (packetstorm I believe) on a friend's 2.2.14 machine, and royally pissing him off in the process. :)
  • I am just as American as the next umm... American, but I always download the kernel source from the Vatican mirror. I'm no Catholic, and maybe I'm nuts, but perhaps there's something... different about a kernel from the Papal State.

    --Ben
  • These don't seem to work with 2.2.15.
    Back to 2.2.14 I go!

    (BTW I get no video at all - blank screen - and can't swtich virtual consoles. Ctrl-Alt-Del fixes it though.)

    :-(
  • 2.0.37? I've got one machine that's still running 1.2.13! It just keeps running, and it's not directly on the net so I don't have to worry about remote crashes... It's hit 365 days uptime twice in the 4 years it's been around, and would have gone longer if it weren't for UPS maint and server room rearranging...
    Too bad I can't find a copy of SLS... I want to try running it in vmware to try to remember what it was like when I started long ago with 0.99pl5.

  • Funny.. (Score:1)

    by robbo ( 4388 )
    that you should mention stability. I just wrestled a 2.2 kernel bug to the ground that was causing my dual processor machines to crash- ok, actually, RedHat beat me to the punch, and it was their custom 2.2.14 kernel that was the problem. So they provide an updated version, also labelled 2.2.14. What gives? Why is redhat providing patched kernels of the same minor version? Why not just provide the *official* 2.2.14 kernel and send bug-fixing patches to AC? Talk about miffed- everyone's getting excited about 2.4 and we're still facing 2.2 bugs.

  • Re:linux is crushing solaris in the enterprise (Score:3)

    by drix ( 4602 ) on Wednesday May 03, 2000 @07:16PM (#1093477) Homepage
    Clickitty [dictionary.com] click. [dictionary.com]

    --
  • Your too busy to compile a new kernel yet you find time to read slashdot and post comments?

    I'd rather use windows then linux 2.0.x ;)
  • That's really cruel.

    I *just* finished a wipe and reinstall on my home
    desktop machine not 20 minutes ago, and now I gotta download and recompile the kernel again.

    arggghhh...
  • I have always wondered what it is and what it does? More importantly would it give me any speed improvements on an IDE system?
  • Being able to read file-systems is built into the Linux kernel; being able to create them or do things like export them via NFS can be done completely in userspace--Linux's kernelspace NFS-export is purely for speed.
    I've found that 2.2.x can't export things that it has mounted over NFS, but you be able to export whatever you want with a userspace daemon, until knfsd issues are resolved.
  • As usual, everyone loves to complain that they just installed the previous kernel when a new one comes out. Here's the solution: download patch, apply patch, make config, compile, vi lilo.conf, shutdown -r now. It takes all of an hour. It's not hard. Stop your bitching.
  • No changelog in the 2.2 directory.
    No changelog in linux/ of the tarball

    I don't understand why you expect to find a ChangeLog in these places, since none has ever existed there before.

    You can grab the changes from a Linux Today [linuxtoday.com] search. For example, here are the 2.2.15-pre17 changes [linuxtoday.com], which is pretty close to 2.2.15 proper.

    And eventually the changes will also be posted by Alan Cox on the Linux.Org.UK [linux.org.uk] home page along with the other 2.2.x release notes.

    Jeff

  • For people who can't afford the random breakage that crops up with 2.3.x on their production machines, it's pretty important for bugs to continue to be fixed in the 2.2 series. Also backporting new drivers is pretty nice.

    Damn straight. I tried one of the 2.3.x kernels a few weeks back (sorry forgot which one) because it had the agpgart.o module avalible directly in the kernel. It worked fine.... EXCEPT for some bizzare reason it would not umount my drives when I issued a shutdown.

    Ill stick with 2.2.X and a loadable agpgart.o for now.
  • Dont you mean 2.4.X or 2.6.X ?

    (I know I am just being a picky bastard! :)
    Or are you using a 2.3.X kernel already?
  • I think it's great that it's taken half as many patches to get 2.2 rock-stable as it took for 2.0 to get nice.

    Yes, 2.0 did have more time to grow, due to the long 2.1 cycle, but the frequency of 2.2 releases has been less than the 2.0 releases.

    More developers, fewer bugs, world domination. Brought to you by open development, a subsidiary of free software.
  • >hmmm... you don't need that much fixes.... Just rename win95 to win98 and sell it.... ;-)

    You also have to put the word 'internet' and the letters 'i' and 'e' all over the program
  • yes, this does happen to some degree. But by whinging you will still get taken down, and I hardly think that YOUR critique on "linux zeolots" is hardly constructive, in which case your post deserves to get moderated down.

    I'd rather see moderation groups, so that given a group of preferences, I can see the list moderated by people like me. That way you don't get this massive wave of people moderating posts down which they "simply don't like".

    my 2c.
  • > 1. This won't make it into potato/frozen.

    Most likely not. That's a vital component of a system, and the Debian folks are not going to put a shiny new kernel release into a frozen release that they're in the final testing phases of. Besides, (and you may get differing views on this), as a standard rule of thumb, the first thing you do after installing a system should always be to compile a fresh kernel. I expect the vast majority of Debian users do this as a rule.

  • Story submissions (Score:5)

    by zCyl ( 14362 ) on Wednesday May 03, 2000 @07:45PM (#1093490)
    > (By the way, what does it take to get Slashdot to notice a story you submit? I
    > submitted close to 9 stories as an "Anonymous Coward", 3 as Fervent (I was the first
    > to make mention of the Limp Biskit-Napster support thing) but no mention of my
    > name on the front page. Is Commander Taco some kind of malevolent dictator?) :P

    Haven't you heard? Before being processed, submitted stories are uploaded to a nearby windows workstation and stored in the Windows Registry. As everyone knows, the recovery rate for information stored in the Windows Registry is 1 out of 5. Next time, to account for this try submitting the same story 68 different times under different usernames. I think there's something about this procedure in the FAQ...
  • i20 drivers are in FreeBSD SNAP 128 and greater. Although they are there, and seem to be pretty stable, they still lack many features; no CDROM support and some bugs.
  • I might just add that i2o is pretty damn cool. There some flaws like i2o controller software currently only runs on Intel's i960 chip (That is off the top of my head) and manufactures are adding their "extensions" to the i2o protocol. However, i2o is pretty damn fast and suppose to be driver OS independent (That's right, the same driver will run on all OS's.)

    Right now, the only major flaw is that the throughput is faster than what PCI 32bit can handle. Ya, I know, most i2o controlers are not going to reach that limit, but big SCSI RAID setups will (9+ drives and such). I guess that what Big Iron is for.
  • Actually, in Europe if I recall correctly, it's a decimal comma, not a point.

    Yup, that's right - except in the UK, where it's a decimal point...

    Cheers,

    Tim
  • so we should stop maintaining a product once something new and better is on it's way? gee i wonder what other products [microsoft.com] that happened to...

    seriously, one of the reasons that linux is great is that there is a good amount of maintinance and increasing stability in the 2.0 series... (that's why this is flamebait!!!)

    NT 3.5 is still used in a few places because of security probs in 4.0. I know a few ppl who wish it was maintained as well as linux 2.0. now don't make me make another microsoft comparison again!

  • Sounds like the same problem I had, it's because the format of /proc/mounts has changed slightly in 2.3.x, and Redhat's halt scripts use this file to work out what needs to be unmounted. It's trivial to fix :)
  • a better way
    first make sure you /etc/lilo.conf look some thing lie this

    boot=/dev/hda
    map=/boot/map
    install=/boot/boot.b
    prompt
    timeout=50
    linear
    default=linux

    image=/boot/vmlinuz
    label=linux
    read-only
    root=/dev/hda2

    image=/boot/vmlinuz.old
    label=linux.old
    read-only
    root=/dev/hda2

    yours will be a bit differant but the main thing is the linux.old entry

    then cd /usr/src/linux
    cp ..//.conf .
    make oldconfig
    make dep && make clean && make bzImage && make modules && make modules_install && make install && /sbin/lilo && reboot

    then you can go get a coffee

  • What I believe you want is this [linuxdoc.org], the Kernel HOW-TO.

    And yes, Commander Taco is a malevolent dictator.. refused to post a submission of mine on the DMCA rally [slashdot.org] that happened 2 days ago.

    Of course, since sites like Slashdot apparently don't find this worthy of the front page, many people had NO idea it happened, or if it even happened at all, including myself. Remember last time when there was a DMCA protest, Slashdot gave like 2 days notice? I sent this in -weeks- in advance, and Slashdot posts nothing. What's up with that?

    Anyway, off my soapbox, and happy kernel'ing.

  • > I have always wondered what it is and what it does?

    It's a utility program that converts *.i files to the *.o format.

    --
  • You're forgetting to actually re-run lilo!

    this would be a good thing to do before rebooting.
  • Uh... Alan Cox (maybe you've heard of him) is the one who coordinates the 2.2 maintenance releases.

    For people who can't afford the random breakage that crops up with 2.3.x on their production machines, it's pretty important for bugs to continue to be fixed in the 2.2 series. Also backporting new drivers is pretty nice.

    I know you're going to say "2.3.whatever is totally stable for me" but you probably can afford the risk of downtime. Also high loads seem to uncover bugs that most desktop users don't see.
  • > Unfortunatly I use redhat.

    Oh man, that is unfortunate!!!

    SLACKWARE RULEZ!!!

    ( Sorry I couldn't resist! )
    Johnny O
  • Hell, i've tried 2 or 3 times to get it just RAID working on 2.2.14 and i cant. i've got to keep backstepping to 2.2.13ac3 to get it to go. The annoying thing is i follow the HOWTO exactly and it wont work. All i have to change is the kernel version and it stops :(
    Any sites that have up to date RAID info would be great if you could post links. I've looked everywhere (well except for where they are).

  • Re:Who works on this? (Score:3)

    by EvlG ( 24576 ) on Wednesday May 03, 2000 @08:04PM (#1093503)
    Even more clueless users can be found on Borland's Kylix newsgroups. It's quite amusing to read the baffled Windows coders asking questions like:

    "What do you mean Kylix won't support DDE!? How else can we make apps communicate?!" and "I've never seen a Linux DLL, what about you?"

    I got a good laugh on an otherwise depressing day after reading that stuff :)

  • Wow! Defaults the way I like'em (Score:4)

    by Crutcher ( 24607 ) on Thursday May 04, 2000 @01:48AM (#1093504) Homepage
    Every damn piece of wackyness is turned off by default in this kernel, just the way I like it. About time things stoped getting defaulted on just cause they were some developers baby.

    On a lighter note, I went and bought the "Linux Core Kernel Comentary." I think this makes me a slashwhore, for running out to by it right after reading the review.

    But it means that I will hopefully be adding to the kernel by the fall. Will you?

    ---
    "Elegant, Commented, On Time; Pick any Two"
  • The correct link to Alan's diary is http://www.linux.org.uk/diary/ [linux.org.uk].

    Cheers.

  • # diff -r /usr/src/linux- /usr/src/linux-

    --
    "The crux of the biscuit is the Apostrophe(*)" - FZ
  • Oops! I forgot previewing...
    # diff -r /usr/src/linux-new.version.number /usr/src/linux-old.version.number

    --
    "The crux of the biscuit is the Apostrophe(*)" - FZ
  • -snip-
    Package: kernel-image-2.2.15
    Version: 2.2.15pre20-1
    -snip-

    The pre20 is already there. I guess we will include the real 2.2.15 there two.

  • The regular maintainers work on this. It is commonplace for chages made in the development branch (currently 2.3.x) to be "rolled" back into the stable branch (currently 2.2.x). Generally, only those changes that have been in 2.3.x for a while, and are considered "safe" will ever see 2.2.x. Bug fixes just get merged in, since they don't shake things up too badly. New functionality (read: drivers) will probably be marked experimental.

  • She just looks at me funny (Score:3)

    by Aravaipa ( 30801 ) on Wednesday May 03, 2000 @07:28PM (#1093510)
    When I say "Darling, guess what! Kernel 2.2.15 was just released hours ago and here we are still running 2.2.14!"

    How many people have upgraded software on the sole basis that the new version includes Hungarian support (Besides the Hungarians out there)?

  • Alright! We'll slashdot them with 404s! This could be a milestone for slashdot. A momentous day indeed.

  • Re:BTW its not on us.kernel.org (Score:3)

    by garver ( 30881 ) on Thursday May 04, 2000 @01:35AM (#1093512)

    There are actually a large number of us.kernel.org's. (Do an nslookup on it, I count 25+ currently.) They are all independent mirrors that have the same directory structure, or at least /pub/linux.

    You were probably unlucky and got one of the slow mirrors. Others may have gotten an updated one.

  • Check out http://people.redhat.com/mingo/raid-pa tches/ [redhat.com] and get raid-2.2.15-A0 [redhat.com]
  • Actually, the decimal system specifies a single decimal point!

    It's the morons that started the assanine idea that multiple decimal points belong in any kind of nomenclature thaqt are responsible for the confusion. "Best practices" is a zillion industries for a century or more have recognized the danger of using the decimal point as a saparator.

    Dashes are the traditional separator character choice where there needs to be a division for convenience: take phone numbers, for example, or the vast majority of part numbering schemes out there.

    [FLAME ON]BTW, this "cool, it's like the net" crud of using dots as separators in phone numbers is garbage. Although most people don't know it, there is a correct and standard way to write phone numbers: the IDDD standard, which specifies a leading + followed by country code ("1" for the US, Canada, and the Carribean) and phone number, separated into groups with dashes by whatever convention prevails in that country.[FLAME OFF]

  • "single decimal point" where? Actually, in Europe if I recall correctly, it's a decimal comma, not a point. That something has been done for plenty of time does not guarantee it's eternal existance unchanged.

    Second, a version number is not a decimal number, so your "single decimal point" standard has no bearing on it; it's quite clearly documented that it's MAJOR-VERSION (dot) MINOR-VERSION (dot) REVISION, which is much more legible than using dashes. Why? Because a dot, at least for CS people, has quite a diferent connotation than a dash.

    And third, why drive yourself up in a tizzy just for something as unimportant as the kernel version number separator character? Just because we/they don't think as you do? Is your consciosness so limited that it can't do context switches when it finds manifestations of new ideas? Relax, have a cup of coffee (or whatever) and get on with your life.

    -elf

  • Yeah. Grab a Slackware either. They don't release that often. They're just "slacking". =:-)

  • Well... some of the VPN stuff was updated in .13 and .14, so even patches to earlier versions get a little bit hairy. I think I went from .12 to .14 - I picked up the new version when I moved. System was up before that, and when I moved, I needed the new VPN functions, so I grabbed and recompiled. Only been down once since then (CPU upgrade). If there's nothing that really grabs you (aside from DDoS fixes - which you can patch and redo on the fly with modules), you generally don't need to upgrade every version... though back in college I was a little more obsessive - had to have the newest dev kernel on my backup box within a couple hours of release, or I felt behind... a little sick, but fun. Now I just want it to behave, and I keep my eyes open for whichever releases have the stuff I want...
  • What's wrong with:
    make bzlilo; make modules; make modules_install

    or getting more picky:
    make bzlilo; make modules_install (since it'll make the modules first anyway)

    bzlilo takes care of all of your stuff... and I still think that /boot stuff is nonsense - I 'fix' that first thing after installing a distro...
  • My year old celeron does it in ~3 minutes, give or take. I still remember when I got my pentium 100, and took my compiles down from ~1h15m(486/66) to ~12m. *That* was bliss.

    >I'd guess quite a lot of /. posters aren't kernel developers either ;)

    Gee, I hope not 8^) My best effort was fixing a network driver... non-Becker, of course ;-) Never got it quite stable enough to merge in, but it worked for me... as long as you waited a little bit between a network stop and start... the restart would trigger some nasty interrupts, and spew madness to the screen for a while... not anything I would feel comfortable submitting. Then I sold the card to a windows user friend. Works great 8^)
  • Besides, (and you may get differing views on this), as a standard rule of thumb, the first thing you do after installing a system should always be to compile a fresh kernel. I expect the vast majority of Debian users do this as a rule.

    Back in the Good Olde Days, sure. Now, I'd be willing to bet that a census would show the majority of Debian users out there are running Corel Linux with its (almost) single-click installation, and most Corel users would be terrified at the idea of compiling a kernel. Well, at lease those who know what a kernel IS. :)

    Note that word "most" in the last paragraph. If you're using Corel Linux on a quad 1GMHz system overclocked to 2G with liquid helium, working 24-7 to develop Gnu/Linux 3.0 and Gnome 3.0, please don't flame me for the generalization. :)

  • Bah. People today don't know they're born.

    My first Linux box took *8 days* to recompile the kernel (2Mb 386sx33). I ran a lot of Stuff for a fair while on a box that took about 28 hours to do a kernel build. You *really* thought about giving the right answers to 'make config' at that point.

    I think the current main box takes about 2 hours. I don't do it often enough to care much, but maybe I would if I was a kernel developer. I'd guess quite a lot of /. posters aren't kernel developers either ;)
  • You can use other processesors like StrongArm with IxWorks. We are using a MIPS proc with our own I2O compliant code (no IxWorks or VxWorks).
    It may not be open but it was pretty neat to see our board just magically work on Novel, Solaris, SCO, and even recent versions of Linux.
    If only MS would release their I2O support...they are on the committee! Now I have learned more than I even wanted about NT/Win2k device drivers.
  • It runs on other chips. We currently use IxWorks on a StrongArm. We are moving to our own shell running on a MIPS chip.

    As for speed...I am working on a FibreChannel Raid Adapter. We hit speeds of 190 MB/sec and 23,000 I/Os /sec. Imagine multiple boards on one PCI Bus! 32-bit, 33 Mhz PCI can only do 132 MB /sec. Even 64-bit, 66 Mhz PCI can only do 528 MB / sec. We saturate that with three cards. Infiniband is needed now!
  • Dude, if it's that much trouble, don't do it. The Linux police will not be coming to arrest you.
  • There's always room for modules.
  • Nautilus is the code name for the Alpha Processor Inc. (API) UP1000 motherboard, not the Gnome stuff.

    The Alpha directory is for the Alpha processor, not alpha as in beta.

    I work for API.
  • I wish they'd use a scheme like 2.2.009 instead of 2.2.9.

    When looking through FTP, and trying to find the latest release, it's REALLY annoying to have to look through the whole list to find it. The newest, 1.2.17 is actually quite a bit higher on the list than 1.2.6.

    If the project was named 1.2.017 and 1.2.006, we wouldn't have this problem. We would, however run into a problem of only allowing 1000 releases per minor version number, but why REALLY wants to see more than 1000 releases before a new version comes out?

  • Now I'm a tad bit miffed... (Score:4)

    by vectro ( 54263 ) <vectro@pipeline.com> on Wednesday May 03, 2000 @06:12PM (#1093529)
    OK, so I finally decide to upgrade my 2.2.10 kernel to 2.2.14 to fix a bug with my ethernet card stalling. And what happens, not 3 hours after I reboot into 2.2.14? They release 2.2.15!

    Sometimes I think I have worse timing than the Debian maintainers. (ducks)
  • Here are the past changelogs:

    http://www.linux.org.uk/VERSION/ relnotes.2211.html [linux.org.uk]
    http://www.linux.org.uk/VERSION/ relnotes.2212.html [linux.org.uk]
    http://www.linux.org.uk/VERSION/ relnotes.2213.html [linux.org.uk]
    http://www.linux.org.uk/VERSION/ relnotes.2214.html [linux.org.uk]

    Do you see the pattern here ? The changelog for 2.2.15 will most likely be available at:

    http://www.linux.org.uk/VERSION/ relnotes.2215.html [linux.org.uk]

    Be patient, have fun :-)

  • An hour?! What the heck are you compiling in there, a new car?

    A fresh compile and install on a P2@300mhz only takes about 20 minutes at most...

    Adam

    Click Here for Quality WebHosting [crimsonnet.net]

  • Re:What is I2O? (Score:4)

    by dsaxena ( 57330 ) on Wednesday May 03, 2000 @07:44PM (#1093532)
    [obdisclaimer: I worked on the drivers, so I'm somewhat biased] I2O is the "Intelligent I/O Software Architecuture". It's a really cool mechanism that allows for offloading of I/O handling tasks from the main CPU down to a separate proccessor (IOP). The idea of intelligent I/O isn't really new, but I2O is really cool b/c it is platform and OS independent. If a device suports I2O, and an OS supports I2O, it'll work. No special driver needed for the device. Intelligent I/O isn't that exciting of a thing at the moment, but it'll become really important in a few more years as more and more people start pounding those big porn servers ;) ~ Deepak
    --
    Deepak Saxena
  • It says that there are i20 drivers in there, is there any chance that the DPT SmartRAID V drivers might start shipping as a part of the kernel, or will we have to continue to hope that the supplied patches from DPT will work?
    treke
  • Unfortunatly I use redhat. I've had good luck with the drivers for 6.0, but had some problems getting the patches to work with 2.2.12 when I tried upgrading to redhat 6.1.
    treke
  • That wouldn't work especially well for a distribution that keeps its kernel image in /boot/vmlinuz instead of /vmlinuz.
    --
    No more e-mail address game - see my user info. Time for revenge.
  • did you try running Linux on a a PPC?

    no size limit! (at least, not for practical purposes :-)
  • Have access to a solaris box? Try "more /bin/clear" Shocked the hell out of me the first time I saw it.


    -- Bucket
  • What's the problem?

    Just get the patch (few minutes at most), apply the patch, do a

    'make bzlilo; make modules; make modules_install; reboot'

    and you have your new kernel. It's not as if it would cost you an hour or more.
  • OK, so, as I thought, eth16i should still be there, it's just that menuconfig for some reason, after I applied the patch.... lost it!?

    ahhhh... anyway.... back to vi .config

  • This is a bit offtopic, as fr THIS version of kernel. I have patched kernel 2.2.13 to 2.2.14 and recompiled it. During the config, (used menuconfig) I noticed there was no etherteam16i module anymore?! When was that discontinued?
    Also, where on earth can I find a list of changes between kernel versions? I mean, some changelog?

    Thanks in advance!

  • can i have her instead, i really need one.
  • Certainly S/390 and equivalents with Parallel Sysplex will allow this too (and don't you just love a computer you can shoot vast chunks off and it keeps on running...)

    More generally, I'd be suprised if most serious active-active cluster solutions didn't support this in some way: planned outages are still outages after all, and they can get seriously in the way of a 24/7 service.

    And loosely coupled clusters (Web farms etc) can of course do this sort of thing without problems.
    --
    Cheers

  • No changelog in the 2.2 directory.
    No changelog in linux/ of the tarball
    In 2.2.14, I find a total of 5.
    In drivers/char, drivers/scsi, drivers/sound, fs, and fs/hfs.
    And looking at the char and scsi ones, they're all pretty old stuff.
  • Luckykaa wrote:
    "Most people upgrade based on the assumption that any 0.0.01 improvement must be a huge difference.

    I still use a 2.2.9 kernel because I can't quite stop feeling that kernel 2.2.15 should come halfway between 2.2.1 and 2.2.2"

    I respond No!!!!!!!!!!!!!! Don't give into the Microsoft numbering scheme. They screwed EVERYTHING up!

    They numbered updates to Dos 6.2 as 6.21 and 6.22. That is INCORRECT. The correct numbering scheme is 6.2.1 or 6.2.2. Why? Because 21 is 2 + 19, not 2+1. 2 + less than one would be 2.x, like 2.1, 2.2, etc.

    That is how a decimal system works. Microsoft choose the other because it "looks right"

    However, by that logic, the upgade path allows Major and Minor X.Y and upgrades off that are tricky... X.YZZZZ However, that means that you can ONLY have 10 minor upgrades per major upgrade, and only 10 patches off that.

    That is CLEARLY a bad design. Especially because the correct idea:

    Major.Minor.Minor-Minor.Minor-Minor-Minor etc., etc.

    Allows infinite numbers of fixes. Within a single major release (Linux 2) you can have as many minor releases (2.0, 2.1, 2.2, 2.3, 2.4, 2.5, and so on, BEYOND 2.9.) Also, it means taht you can have updates on the minor ones.

    Your warped numbered scheme means that you CANNOT go beyond 2.2.9. I mean, 2.2.15 is the 15th revision of the 2.2 Major/Minor combo. Each portion between the decimal points is an integer, not the decimal part.

    Alex
  • I've found this site [gdr.free.fr] to be helpful. It's geared towards the PowerPC Linux [apple.com] distros, but should be applicable for almost anyone.

    John
  • generally your news stories have already been submitted about 1x times before you submit yours. Most of the time, they'll post news articles that are well written, and (of course) timely. Don't get miffed if your news piece doesn't make it. Most don't.

    Anyway - as far as upgrading your kernel. The best way to learn is to read the kernel HOWTO. It's a pretty good document, if not slightly antequated. It should be included in the dox on your box. If not...you can find it linked to at linux.org [htttp] - Just make sure to keep a bootable backup copy of your old kernel ready to go just in case something doesn't work right. - or you just plain fuck it up :)

    Happy kernel hacking!


    FluX
    After 16 years, MTV has finally completed its deevolution into the shiny things network
  • Most people upgrade based on the assumption that any 0.0.01 improvement must be a huge difference.

    I still use a 2.2.9 kernel because I can't quite stop feeling that kernel 2.2.15 should come halfway between 2.2.1 and 2.2.2
  • In other words, you can't wait till 2002. Me either.

    ZDnet ran an article [zdnet.com] on how the 2.4 Kernel release was pushed back. Of course if you look at the talkbacks all the windows zealots jumped all over it, because for once it wasn't Microsoft pushing back the date. Some even called it vaporware. (If you think people who post on Slashdot are clueless take a look at the comments here.)
  • ...but I found it on ca.kernel.org in both full source and patch form.
  • Can anyone point me to a few sites that demonstrate easy kernel upgrading? I'm relatively nacient to the Linux world (10 months), and while I have learned a surprisingly large amount of information rather quickly, recompiling something as basic as a kernel gives me the willies.

    (By the way, what does it take to get Slashdot to notice a story you submit? I submitted close to 9 stories as an "Anonymous Coward", 3 as Fervent (I was the first to make mention of the Limp Biskit-Napster support thing) but no mention of my name on the front page. Is Commander Taco some kind of malevolent dictator?) :P

Slashdot Top Deals

Adding features does not necessarily increase functionality -- it just makes the manuals thicker.

Close