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 2.4.0 Test2 Almost Ready for Prime Time 132

Posted by CmdrTaco from the gentleman-start-your-engines dept.
out of control sent us a quote from Linus from the kernel dev "There's a "test2" kernel out there now, integrating most of the -ac patches, and some code that wasn't in -ac. Normally, when you integrate almost 5MB of patches, bad things happen. This time, a miracle occurred. As I uploaded the resultant kernel, a specter of the holy penguin appeared before me, and said "It is Good. It is Bugfree". As if wanting to re-assure me that yes, it really =was= the holy penguin, it finally added "Do you have any Herring?" before fading out in a puff of holy penguin-smoke. Only a faint whiff of rancid fish remains as I type in these words.. In short, not only are most of Alan's patches integrated, I have it on higher authority that the result is perfect. So if it doesn't compile for you, you must be doing something wrong. Use a mirror.
This discussion has been archived. No new comments can be posted.

Linux 2.4.0 Test2 Almost Ready for Prime Time More

Linux 2.4.0 Test2 Almost Ready for Prime Time

Comments Filter:
  • Actually, "herring" is pronounced like "silly" with long "l" in Finnish (and written "silli").
  • Yeah, the people at the top of those projects could emit more zany quotes to get easy publicity.
  • so?
    slashdot is for ppl running free unicses (gnu/linux, gnu/hurd & *BSD)
    if you dont like it goto www.microsoft.com....
  • I'm using DevFS withOUT devfsd. It's a Good Thing(tm). Not sure why everyone shies away from it. TRY IT, damnit! a) YOU MUST COMPILE IN DEVFS AND SELECT THE AUTOMATICALLY MOUNT AT BOOT OPTIION! b) change inittab to have vc/1 ... vc/x instead of tty1...x. c) put vc/1...vc/x into /etc/securetty d) create an rc script to link /dev/tty0->/dev/vc/0 and /dev/tty7->/dev/vc/7 (goddamn X) e) edit your fstab (for me, I'm now on /dev/ide/host0/bus0/target0/lun0/part1, not /dev/hda1). f) make your kernels (that means change the root= param) g) any problems, bitch at Richard Gooch :P d


    -
  • pppd needs to be upgraded, too (or i needed to).
    the 2.4 series requires a development pppd, i think. of course, if you're not on dialup, you don't need to care...
    • So if it doesn't compile for you, you must be doing something wrong.

    The very first response to Linus's post was someone who got compilation errors, and many other people responded with the same problem. Apparently gcc 2.7.2.3, which is listed in the docs as the recommended compiler, won't compile it. :)

    The problem is in kernel/sched.c, which has

    • __cache_line_aligned spinlock_t runqueue_lock = SPIN_UNLOCKED;
    gcc 2.95.2 accepts that, but gcc 2.7.2.3 wants it to be
    • spinlock_t __cache_line_aligned runqueue_lock = SPIN_UNLOCKED;

    Gumbo

  • Or maybe he just had a bad haddock.
  • Dislaimer: As a maintener of one of the peripheral filesystems in linux, I can hardly be objective. However, the post I am replying also most certainly wasn't objective, so what the heck.

    Now there's a harebrained idea to add "generic" journaling functionality to the VFS. I assume this is so that when ext3 is finally ready, the VFS will support it well, and all other filesystems will have to then look like ext3.

    This is just plain FUD, sorry. Journaling support HAS to be added to VFS. MAIN reason why ext3 (or ReiserFS, but I'll get back later to that) is not going to be in 2.4.0 is because of all the VFS kludging they have to do. That is the same reason why you _CAN'T_ use the journaling with software RAID (which is combination which anything bigger than home user needs - both hardware rendundancy, and you don't have to bring a system down for days fsck-ing when it goes south)

    Microsoft make it nearly impossible to write new filesystems for Windows NT, because they want everyone to use NTFS. Viro's doing the same thing. So why is it tolerated in an open-source OS?

    Nobody is forcing them to cooperate with Linus and/or Viro. They are perfectly free to fork and release "Reiser's Linux" (which is what some distributions use anyway. SuSE comes to mind). I don't remember when I last used Linus vanilla kernel tree, except as a reference.

    Linux should have a generic, capable, stackable VFS that isn't tied to a specific filesystem, and doesn't provide special support for preferred filesystems.

    And that is exactly what Viro is doing. If it wasn't, don't you think Linus and would have noticed ?

    Adding to the problem is that the VFS is very poorly documented. Changes are made without any foreshadowing. The best documentation available is the source code for the Ext2 filesystem. And that is sad.

    There I have to agree. VFS is poorly documented. Part of it is because it is currently undergoing major recoding. When it is done, there will be better documentation, I'm sure. BTW, why don't YOU for example volunteer to write to some VFS documentation ? There is plenty of info on linux-fsdevel and I'm sure that others will be willing to help a cause by answering any questions that you can't answer by reading fsdevel mailing list. So it is just matter of putting some time and energy, no special kernel-hacking-knowledge is required.

    This post is not meant as a flame. The VFS is a serious issue. Linux could have had a journaling filesystem by now.

    Of couse it is not meant as a flame, it's meant as a FUD :-) VFS is a serious issue, and Linus does have a journaling filesystem by now (few of them actually that are working). They are just not in pristine Linus tree. But if you install whatever distribution, chances are that you WILL NOT be using Linus's tree anyway.

    As for Viro-Reiser flames: both of them are strong individuals and bully. Reiserfs team is MUCH better than Reiser himeself -- which especially shows when communicating with others. And although Viro usually does not have such open flame outbursts and name calling as Reiser practise, he will make implicit comments to provoke. But never by themselves, always backed up with facts about poor quality of code in reiserfs (outdated checks suited for 2.2.x kernels which changed completely, etc)

    Also, if Reiserfs gets rid of the kludges and fixes it's interface, it will go in Linus 2.4.x
    tree (but not in 2.4.0 -- this is Linus's comment)

    Although I've been bitten personally by VFS changes (filesystem which I maintain is also inode-less as is reiserfs; and VFS is still inode-centric, notwithstanding dentry-stuff), I do see that VFS changes are for the better.
  • use /dev/shm (or /shm, or something else), instead of /var/shm. /var/shm could confuse utilities that don't expect a virtual filesystem in /var (discussed heavily on linux-kernel).
  • I don't quite find it to be the case that Viro is the only one patching the VFS. The GFS group, Reiser, and several other groups have patches that go into the VFS to add support for simple things. One such very simple patch that ReiserFS requires, is to be able to pass additional data with an inode on read_inode. Unfortunately Viro doesn't seem to be willing to incorporate ANY solution to this problem. And any filesystem which does journaling, or has inode numbering semantics that differ from FFS derivatives (such as the one I am developing) need such context on reading an inode. Admitedly, Reisers patch isn't the most aesthetically pleasing in the world (read_inode2), but something similar could be added to read_inode, and have the default behavior of iget to pass NULL, and would solve everyones problem with minimal effort. Yes, every filesystem in existence would need to be patched, but that's allready been the case with alot of Viro's changes. I think he's broken every FS in the kernel at least 10 times, so using that as an exuse is bullshit. But with Viro being so hesitant to add a simple change to the VFS, that has been needed for months to make ReiserFS work, it really makes me hesitent to make my filesystem dependent on the VFS. So i have to write extra code to route around the damage. It's also interesting in terms of my original experiences when i started on my filesystem. I asked a question about the use of the generic_ip pointer, the only answer i got was that you can't, you have to patch the kernel and add your code to the union. So I ignored them, and have a working filesystem. But it's an attitude i see prevalent in the VFS area. I guess FS developers aren't excatly a dime a dozen, so it's an easy area for an elitist, silly, almost PROPRIETARY (in the sense of being someones property) approach to the development of a "VIRTUAL" interface. (and in case you are wondering, no, my FS hasn't been released yet. It should be soon, I'm just currently trying to get it into a stable state. If anyone is interested in it, please mail me (remove the obvious NOSPAM), and i'll give you an update when i've gotten it together.)

  • Re:Has anyone ever noticed? (Score:3)

    by Dan Jagnow ( 181761 ) on Saturday June 24, 2000 @09:15AM (#979161)

    It's almost as if there's a bunch of geeks in Vegas betting on when each development release of what kernel will be released.

    There are, and I just lost my shirt over this one. Now where will I hang my pocket protector?

  • Ya, forgot about that target=_blank deal. Here's the correct link: Slashdot Article about ReiserFS inclusion into Linux 2.4 [slashdot.org]
  • I suppose many of them are like me. At work I browse with a windows box. Of course in the course of the day I'm administering several linux boxen both as servers and desktops. I don't view slashdot with those boxes, but I'm still more interested in the linux content even though I browse with M$ (actually I've now made the 40hr/week plunge into using strictly linux, booting windows is an anomaly). If I were interested in M$ I'd stick with www.windrivers.com.

    I just can't figure out why some people are so dumb about the content on this site. It has always been that way, because that was the emphasis of the site, and just because some pc users have adopted the personality of nerds, doesn't mean this news for nerds site has to completely change its focus. If it sucks in your opinion go elsewhere.

  • Re:And in other news... (Score:3)

    by seppy ( 2431 ) on Saturday June 24, 2000 @09:32AM (#979164)
    Of course Anonymous Cowards are the most humourous, most informed, most experienced users out there. It's the slashdot paradox. Uhh, yeah...

    You should see the cookies slashdot has placed on my machine to track me!

    www.slashdot.org FALSE / FALSE 1236522993 user %2532%2534%2533%2531%253a%253a%256c%2575%2573%2565 %2572%2535
    www.slashdot.org TRUE / FALSE 1238388690 sexual_orientation 7699925.18704385
    .slashdot.org TRUE / FALSE 1238388690 sexual_orientation 7699925.18704385
    www.slashdot.org TRUE / FALSE 1238388845 soc_sec_num 1709714.88580108
    .slashdot.org TRUE / FALSE 1238388845 soc_sec_num 1709714.88580108
    www.slashdot.org TRUE / FALSE 1238408765 last_time_you_brushed_teeth 7630727.50531137
    .slashdot.org TRUE / FALSE 1238408765 last_time_you_brushed_teeth 7630727.50531137
    www.slashdot.org TRUE / FALSE 1238412245 high_school_gpa 1181069.01179999
    .slashdot.org TRUE / FALSE 1238412245 high_school_gpa 1181069.01179999
    www.slashdot.org TRUE / FALSE 1238412236 iq 4749934.93407965
    .slashdot.org TRUE / FALSE 1238412236 iq 4749934.93407965
    www.slashdot.org TRUE / FALSE 1238410318 religion 7864276.77143365
    .slashdot.org TRUE / FALSE 1238410318 religion 7864276.77143365
    www.slashdot.org TRUE / FALSE 1238412307 mothers_maiden_name 3729926.00314319
    .slashdot.org TRUE / FALSE 1238412307 mothers_maiden_name 3729926.00314319

  • Re:So what is the biggest feature of 2.4? (Score:4)

    by emir ( 111909 ) on Saturday June 24, 2000 @05:31AM (#979165)
    you might want to check joe pranevich's article on 2.4. its called "Wonderful world of Linux 2.4" and you can find it http://linuxtod ay.com/news_story.php3?ltsn=2000-05-13-003-04-NW-L F-KN [linuxtoday.com].

    i believe that udf support is added to 2.4. i dont believe that any journaling system (xfs/jfs/ext3/reiserfs) will make it into 2.4 but you can alway patch you kernel with reiserFS patch and get (imo) stable and fast journaling system. :)
  • I'm using ResierFS on the /home partition of my laptop, and it handles power failure, droppage, and other insults like a tank.

  • Re:Ready for testing yes, not prime time (Score:4)

    by axboe ( 76190 ) on Saturday June 24, 2000 @05:35AM (#979167) Homepage
    The I/O problems are fixed in 2.4.0-test2, which is what the announcement is about. VM is still shaky (which has an impact on I/O, naturally), but that is not related to block I/O.

    2.4.0-test2 contains a largely reworked I/O scheduling layer and several elevators to pick and choose from.
  • I hack regularly on the fs layer and vfs, so I feel I'm qualified to comment. There is some truth in the VFS deficiencies you point out, but I doubt that this is intentional. As I understand it, the prime directive for VFS development at this point is stability and all other issues are secondary. Imagine how much hay Microsoft could make of a cutting-edge, yet unstable VFS.

    the VFS has specific ext2 functionality built into it. It's a "virtual filesystem" as long as you can make your filesystem look and work like ext2.

    It's easy to understand why. VFS started life as a simple refactoring of, um, EXT. Actually pretty naive in conception, but it has worked well. I was able to work with it, and I didn't ask anyone's help. There are a number of documents out there that deal with VFS, notably TLK (The Linux Kerne) though admittedly in a somewhat superficial way. The rest of VFS I learned about with grep - if I were to do it again I'd use LXR.

    The problems you speak about (except for stacking) don't have a lot to do with VFS per se - they are really more problems with the buffer cache system. The buffer cache does its job pretty well as long as you're not doing something like journalling. As soon as you start journalling you run into questions about exactly what gets flushed to disk when and you expose areas in the buffer cache that were simply never designed with this in mind. What to do about them is an open question, and an interesting question. Look for this to be a main development center in 2.5

    The reason 2.4 has no journaling filesystem is that there are roadblocks in place to keep it that way. Ext3 will be the first journaling filesystem in Linux. Not because it will be the first journaling filesystem, or the best, but because it will be the one properly supported by the VFS ("Viro File System").

    I think a more accurate reason is that a journalling filesystem requires a massive development effort. EXT3 is being developed essentially by one man (perhaps this in itself supports your argument) and RieserFS by a team of about 3. I think they've done pretty well to move along as well as they have. 2.5 will have at least two journalling filesystems and they will no doubt be backported into 2.4 at an early stage.

    Adding to the problem is that the VFS is very poorly documented. Changes are made without any foreshadowing. The best documentation available is the source code for the Ext2 filesystem. And that is sad.

    Somebody who knows the VFS really well should go in and produce some definitive documentation. The trouble is, precisely the parts that need documenting are the ones that are in flux right now. The audience for this documentation is also tiny, though important. You also have to worry that if you document it thoroughly, you may be casting a bad design in stone. It's a problem. The payoff for better documentation will be more developers getting up to speed on the VFS quickly and as a result, more rapid filesystem evolution. Perhaps what we need is something more like a design manifesto for each new wave of VFS development.
    --

  • Hopefully the vision was true... (Score:4)

    by ConceptJunkie ( 24823 ) on Saturday June 24, 2000 @04:48AM (#979169) Homepage Journal
    ...unless the holy penguin was asking for RED herring.

  • Yay (Score:4)

    by Anonymous Coward on Saturday June 24, 2000 @04:48AM (#979170)
    Linux 2.4.0-pre alpha beta-open test 2.5 is almost ready for prime time, to be followed by Linux 2.4.0-pre alpha beta-open, test 2.6, up to pre alpha beta-open test X, which I'm sure will all have extensive coverage on Slashdot.

  • Has anyone ever noticed? (Score:3)

    by Anonymous Coward on Saturday June 24, 2000 @04:51AM (#979171)
    Has anyone ever noticed that Slashdot follows kernel releases like ESPN follows box scores? It's almost as if there's a bunch of geeks in Vegas betting on when each development release of what kernel will be released.
  • There was a bit of a head-to-head between the two memory patches; which one is in the test?

    Yours Truly,

    Dan Kaminsky
    DoxPara Research
    http://www.doxpara.com

  • Amen (Score:2)

    by 51M02 ( 165179 )
    I knew Linus has a pretty big sense of humor, I now have some proof. Any reference known of the Holy Penguin before this README ?

    Disclaimer: "These opinions are my own, though for a small fee they be yours too"
  • Could be that LinuxWorld is a time he'd like to be done with it and is in fact pacing himself against that deadline, consciously or not. But if you're suggesting a conspiracy or something ... well, since 2.4 was originally s'posed to be out a while ago and has been delayed for various reasons which have been covered on slashdot and elsewhere, it's a pretty tantalizing conspiracy;) When Linus walks around the Torvalds compound, with his true Dr. Evil head comfrtably in view (the little normal size head looks pretty uncomfortable, you must admit), he probably mutters "Where did I go wrong, self? This evil operating system I have unleashed was supposed to hurt people, not make them happy? What could I have done wrong. Oh, never mind ... let me tell them another release date and then renege depending on whether the code actually works well rnough ... No, let me make funny comments as I release updates so they do not suspefct my evil nature ... No, let me use self-deprecating humor and be nice to people to keep them off guard .. No, Drat. Works again! can't I ever get it wrong for once?!" "

    I'm excited by the fact that pretty soon a default install of Linux ought to include a journaling file system, decent USB support (which really is a function more of vendors like HP, once the code is in the kernel) and a powerful ip filtering system. Sounds like it it will be worth the wait! (And since I'm a relative free rider, I'd have no room to complain if for some reason 2.4 gives me an allergic reaction or something.)

    timothy

  • So what is the biggest feature of 2.4? (Score:2)

    by Anonymous Coward
    a journaling filesystem? (ext3 wasn't stable, last I checked, and SGI has been dragging their butt on xfs). Will the international kernel (the useful one with real crypto) debut at the same time as 2.4? NTFS write support yet? UDF write support for CDRWs? So what's new?
  • All the Windows kiddies have been ejaculating all over each other because the "Gold" release of Windows 98 Third Edition (ME) is being pirated as of June 23/00. It's not expected to hit the shelf until September.
  • Mmmm, we'd have to be careful - it could be that Linus didn't actually hear what the penguin said. His herring could be going...

  • What else do I have to upgrade? (Score:3)

    by fReNeTiK ( 31070 ) on Saturday June 24, 2000 @05:38AM (#979178)
    Hi,

    I've been sticking with 2.2(.16) until now, but if Holy Linux says it's perfect, I'd be willing to give 2.4 a try.

    However, I have to ask: What other parts of the system have to be upgraded in order to make a smooth transition? Do I only have to compile the new kernel put it in /boot and shutdown -r now or are there any libraries or tools I need to upgrade as well? I've heard that the ipchains code has been rewritten (again?)... What about USB?
  • my limited experience would like to indicate that your problems integrating the WD and Maxtor drives may have more to do with your motherboard bios than your OS.
  • I've heard Linus claim that before. Either

    1) He's overconfident.
    2) He's arrogant.
    3) He's being funny.

    I'd wager all three of them. :-)

    And 1.1.81 is officially BugFree(tm), so if you receive any bug-reports on it, you know they are just evil lies.
    -- Linus Torvalds

    I kinda like that quote. :)
  • Lookup the upgrade notes on this [kernelnotes.org]webpage.
  • Ok, I got test2 running and benchmarked it.

    I get 5.5 MB/s read and 2.0 MB/s write on my four disk software RAID 5. I usually (with 2.2) get 18 and 12 respectively. It may of course well be the software RAID playing in here.

    On an older (SCSI) disk I get 5.2 MB/s read and 3.9 MB/s write. This is pretty close to what I can expect from that disk. I guess this would indicate that it's only the semi-experimental software RAID-5 code that could need a little improvement.

    On the positive side, the system doesn't freeze for 5 second periods while I benchmark it :)

    It think it's safe to say that 2.4-test series are on the right track. There can't be many huge problems left, and the ones currently in the kernel doesn't keep people from testing - which is a very good thing.
  • Nothing is bug free. Every program has at least one bug in it, and can be reduced by at least one line. Corrollary: Every program can be reduced to one line with a bug in it.
    ---
  • ...i speak as one of many when i say that the whole cdrom burning infrastructure is pretty thrashed, but recovering.

    whether using scsi or atapi, on any distribution, and performing all the steps in the cdrom-writing howto, and using the lastest cdrecord, something is definately wrong.

    prior to about 6-8 months ago, i had never made a coaster (unless i did something stupid).

    lately, after a lot of kernel rebuilding and s/w updates, i have a debian install that is burning fairly solid.

    still occasional kernel messages and lockups, or the scsi just disappears, requiring a reboot.

    three machines, three distros, two different writers (1 scsi, 1 atapi), many updates and tests, and i'm down to about 1/10 coasters. a year ago, it was NO coasters.

    cd writing is important to a lot of people. something went wrong in the s/w infrastructure for burning cdroms several months ago, and it has yet to totally recover.
  • Hi!
    I tryed many 2.3.* and 2.4.* kernels and they all crash my computer after ~5-10 hours of work, while 2.2.* is perfectly fine. I can't even report a bug, because hell knows what to report :(
    My computer is Abit BP6 dual Cel-366, IBM at primary IDE and 2 cd-roms at the ATA66. One is new HP burner, second is some crappy I/O magic DVD. On PCI I have:
    00:00.0 Host bridge: Intel Corporation 440BX/ZX - 82443BX/ZX Host bridge (rev 03)
    00:01.0 PCI bridge: Intel Corporation 440BX/ZX - 82443BX/ZX AGP bridge (rev 03)
    00:07.0 ISA bridge: Intel Corporation 82371AB PIIX4 ISA (rev 02)
    00:07.1 IDE interface: Intel Corporation 82371AB PIIX4 IDE (rev 01)
    00:07.2 USB Controller: Intel Corporation 82371AB PIIX4 USB (rev 01)
    00:07.3 Bridge: Intel Corporation 82371AB PIIX4 ACPI (rev 02)
    00:09.0 Multimedia audio controller: Creative Labs SB Live! (rev 06)
    00:09.1 Input device controller: Creative Labs SB Live! Daughterboard (rev 06)
    00:0d.0 Multimedia video controller: Brooktree Corporation Bt878 (rev 02)
    00:0d.1 Multimedia controller: Brooktree Corporation Bt878 (rev 02)
    00:11.0 Ethernet controller: Realtek Semiconductor Co., Ltd. 8029
    00:13.0 Unknown mass storage controller: Triones Technologies, Inc.: Unknown device 0004 (rev 01)
    00:13.1 Unknown mass storage controller: Triones Technologies, Inc.: Unknown device 0004 (rev 01)
    01:00.0 VGA compatible controller: Matrox Graphics, Inc.: Unknown device 0525 (rev 03)

    Any idea?
    If someone help me to report the bug, that will be useful.
    It is quite possible that I have some hardware bugs, true. I was running memtest several times. It crashed once, but memory is fine.
  • I futzed around with test1 and the ac patches on my UT server at the office, only to find that the web ServerAdmin interface became hopelessly broken. I never found an explanation of exactly why that is, only posts indicating that yes, there is a problem. Otherwise test1-ac18 (the variant I toyed with) was a dream to work with. I even got the *gasp* DVD code from linuxvideo to work, as much as it was going to at that particular date.

    To get to the point... is there any indication that test2 resolves whatever phantom oddness it is that plagues the UT webadmin function? Or (more likely) does nobody see this as much of an issue?

    --- Karel P Kerezman

  • It's almost as if there's a bunch of geeks in Vegas betting on when each development release of what kernel will be released.

    There are, and I just lost my shirt over this one. Now where will I hang my pocket protector?

    How about a nice nipple ring with a aligator clip hanging from it?

  • the promise PDC20262 driver crashes all the way from test1 to -ac 22 ...
    I did send in multiple reports .. but noone seems to be interested.
    And no .. harware is ok, checked it with 3 different adapters, 3 maiboards, and 6 hardrives. All hang at boot :((
    so ... no 2.4 for me :(

    Samba Information HQ
  • Did anyone read this article [slashdot.org] about IBM's JFS? Isn't this a journaled file system for Linux or is it misnamed? It is under the GPL. Why not let the big corporate power pay for this? ;-) As you said, it takes a "massive development effort"...which IBM has the $$ for.
  • You are heading into the FUD department with your message...

    If you want non-GPL proprietary extensions to reiserfs, you can purchase them, too.
  • Hehe, thanks.

    I was actually going to check the release notes and other info I could dig up later on, but since this is slashdot and I'm lazy, I thought I'd just ask (RTFM, I know).
  • You gotta be kiddin' me :)

    test2-pre11 (or so) gave me 3 MB/s on my four disk SCSI RAID. That's about the speed of my laptop with 2.2. I'll go benchmark the real test2 tonight.

    Anyways, that means we're down to one hard problem right ?

    Axboe, you know more about this than me, if you feel like it, could you write a four-liner about what's going on in the VM currently ?

    (beware, hostile .dk takeover on SlashDot!)

  • How do you enable high UID support? I upgraded my system to 2.4-test1 a long time ago. It's very good but I still don't have high UID support. Does anyone know how to do this?

    For example, you still can't have a UID of 75000 without Linux thinking it's actually UID 9464.
  • linux:/tmp# chown 75000 file
    linux:/tmp# ls -al file
    -rw------- 1 9464 root 0 Jun 24 11:52 file
  • A question: what sucks with XFS? I'm just about to try it.
  • If you have a relatively new dist, I doubt if you will be required to update anything.

    You may need an extra line in /etc/fstab (for redhat) like :
    none /var/shm shm exec,dev,suid,rw 0 0

    In any event, try compiling it anyway. The scripts should warn you if there are library incompatibilities.

    Cheers.
  • These open-source monkeys aren't working to please themselves (if they were they would ditch linux and move to working on *BSD [flamebait]). People like Linus and Alan are working to serve us. Linus must answer to every complaint that we can think of (such as mine), and must still continue to develop great software!

    I believe that there is a major conspiracy involving VA Linux, Penguin Computing, Transmeta, LinuxOne, and Pepsi Co. The CEO's of these corporations are pushing for as much x.0 software to be released at LinuxWorld as possible. You must realize that this is only going to get worse, as these corporate thugs demand more high profile software be derailed from their respective timelines, and released for the hungry masses at LinuxWorld, attracting more and more people to this event. Eventually, when enough people have come to LinuxWorld, they will lock the doors, move the masses into cargo crates, and ship them to Redmond where the Linux lovers will be subjected to living in a Bill Gates brainwashing camp.

    Alan Cox and I have figured out their scheme, and are starting a group called the Linux Resistance Front. We will encourage the abolishment of software version numbers, and instead create a versioning system not unlike the UNIX timestamp system. According to our versioning system, Linux is currently at 9417522961. Not as sexy or attention gathering as 2.4.0-test2 is it? We are the only hope of fighting this unholy alliance.

    The time has come to open your eyes timothy.
  • As many years as the number of times ms windows crashes in a week.
  • SurfsUp said:

    The buffer cache does its job pretty well as long as you're not doing something like journalling. As soon as you start journalling you run into questions about exactly what gets flushed to disk when and you expose areas in the buffer cache that were simply never designed with this in mind. What to do about them is an open question, and an interesting question. Look for this to be a main development center in 2.5

    Well, that would be good news. Cleaning up the VFS in 2.5 would be a welcome development, seeing as more and more parts of the kernel are depending on the vfs (shared mem, devfs, proc, disk filesystems... etc).

    Take, for example, read_inode(). This function takes, as part of its single argument, a union of info for current filesystems. Filesystems not featured in this union must use the generic ip pointer. Now, the right way to do it is to make all filesystems use the generic pointer, rather than include fs-specific data in the vfs.

    I think if the vfs were more generic, then things like journaling would be less of an issue. The vfs should provide interfaces for filesystems without implementing any filesystem itself. As it is, it's a virtualized Ext2, as you said.

    I look forward to 2.5!
  • That list looks awfully out-of-date. For instance, I'm pretty sure that your modutils must be >= 2.3.11 these days. You'd be more likely to find an up-to-date requirement list in the kernel source tree itself, in the file Documentation/Changes (at least, that's where it is in 2.2.x).

  • You're a driver hacker, you say.

    Jeezus, you're talking to Jegg Garzik and you have the temerity to say that? Go have a look at the driver code in 2.4test - a pretty fair chunk of it has his fingerprints on it.

  • Not quite perfect. (Score:5)

    by 1010011010 ( 53039 ) on Saturday June 24, 2000 @06:03AM (#979202) Homepage
    Viro still holds a firm grip on the VFS, and refuses to make it stackable. Worse than that, the VFS has specific ext2 functionality built into it. It's a "virtual filesystem" as long as you can make your filesystem look and work like ext2.

    Now there's a harebrained idea to add "generic" journaling functionality to the VFS. I assume this is so that when ext3 is finally ready, the VFS will support it well, and all other filesystems will have to then look like ext3.

    Take a look at the enormous hacks the HFS and ReiserFS have had to make to work around Alexander Viro and his Virtual Ext2 Filesystem.

    Microsoft make it nearly impossible to write new filesystems for Windows NT, because they want everyone to use NTFS. Viro's doing the same thing. So why is it tolerated in an open-source OS?

    The reason 2.4 has no journaling filesystem is that there are roadblocks in place to keep it that way. Ext3 will be the first journaling filesystem in Linux. Not because it will be the first journaling filesystem, or the best, but because it will be the one properly supported by the VFS ("Viro File System").

    The Reiser-Viro flame wars aside, the filesystem cartel is doing serious damage to Linux. Linux should have a generic, capable, stackable VFS that isn't tied to a specific filesystem, and doesn't provide special support for preferred filesystems.

    Adding to the problem is that the VFS is very poorly documented. Changes are made without any foreshadowing. The best documentation available is the source code for the Ext2 filesystem. And that is sad.

    Maybe Linus will intercede to provide a better VFS. Maybe the Stark Fist of Removal will pay Viro a visit.

    This post is not meant as a flame. The VFS is a serious issue. Linux could have had a journaling filesystem by now.


  • I'd really like to see ReiserFS merged with the 2.4.0 kernel. I've been using ReiserFS 3.5.19 made available with the Linux-Mandrake 7.1 release, and I'm very happy with it. ReiserFS has survived several planned power outages and just keeps on ticking. By the way, the Linux-Mandrake installer even sets up the partition on which /boot exists with the notails option in /etc/fstab.

    Read the testimonials on the ReiserFS homepage [devlinux.com].

    A journaling filesystem is a very high profile Killer Feature. Having journaling in 2.4.0 would make Linux an even more obvious choice where data integrity is of paramount importance.

    Lets start a grassroots movement to have ReiserFS merged with 2.4.0!

  • That can't be good for your laptop.

  • 2.4 (Score:1)

    by prodeje ( 58779 )
    Why do I get the feeling that Linus is setting up the timeframe so this kernel can be announced at the LinuxWorld Expo?
  • I read [slashdot.org] that ReiserFS has made it into 2.4 in the nick of time! I've used it only peripherally, and it ran just fine.
  • I know this is offtopic, but I had to make a comment on this person's sig. I keep seeing that sig soliciting developers to go to www.beunited.com to help out BeOS. The only problem is, when I go to that site I am "turned away at the door" with a message telling me (in several different languages) that the site is under construction. I remember seeing the same thing when I went to that website a month or two ago.

    Either they don't really need developers that much or else the need a webmaster who can type with more than two fingers! Get with the program, guys!
  • Yeah, Viro is known for this sort of shit. You should read Him on USENET.

    Anyway, my biggest beef with Him currently is that He recently broke POSIX compliance in the filesystem by CHANGING symlink behavior. It's a very annoying change too.
  • How's the speed of ReiserFS? I use ReiserFS on my Mandrake system, but I just got a new hard drive so I can't tell if the speed up is due to that or ReiserFS.

  • Re:Not quite perfect. (Score:3)

    by be-fan ( 61476 ) on Saturday June 24, 2000 @06:26AM (#979210)
    Since Linux is so open and all, couldn't Reiser just patch VFS itself? Isn't Linux GPL'd to prevent just this sort of thing?
  • I haven't tried out 2.3 or 2.4, but I'm running devfs on 2.2 so I think I can answer B.

    Basically, you don't have to do anything to get it started. Once devfs support is compiled in, the kernel will populate /dev for you. However, you have to get used to the name changes. devfs adds namespaces, so you have either have to change stuff like your /etc/fstab (/dev/hda1 becomes /dev/discs/disc0/part1 or /dev/ide/host0/bus0/target0/lun0/part1, whatever you prefer), or you install devfsd (which is a user-space daemon) and set it up to keep 'legacy' devices in /dev populated for you (via symlinks).

    I suppose you could keep all your old devices in /dev if you wanted to (which isn't actually a bit idea to ease the strain until you get things figured out), but it kind of defeats the purpose of using devfs.

  • Oversimplification... (Score:3)

    by Christopher B. Brown ( 1267 ) <cbbrowne@gmail.com> on Saturday June 24, 2000 @07:09AM (#979212) Homepage
    If there were others doing VFS hacking, this would likely have some of the following effects:
    • They might trample on one another.

      I change this, you change that, we break each others' code.

      This is different from device drivers, which are pretty independent of one another; the pervasive use of VFS in ext2 means that changes have to filter through someone in order for there to be hope of coherency.

    • Linus may accept Al Viro's changes, even when they involve changes of VFS design, but be reluctant to accept others' changes to VFS design.

      Note that if Linus accepts changes from other people, as well as Al Viro, nothing stops Al from submitting patches that essentially reverse out others' changes in favor of his own. That would be not nice, to be sure.

      The side-effect of "patch preferences" is that if Linus accepts changes preferentially, those that aren't preferred won't necessarily take this gracefully, and may decide that there's no point to trying to work on VFS if their efforts are doomed to be ignored.

      The strong comments Hans Reiser has made indicate that he falls into the "won't take this gracefully" camp.

    Reiser has suggested that there's a "Red Hat" conspiracy; I don't believe that, but it is sure that there have been some disagreements between ReiserFS developers and Al Viro...
  • Pretty big words for an anonymous poster


    "I shoulda never sent a penguin out to do a daemon's work."
  • Mr. Reiser has been spouting an awful lot of paranoia lately on the list as well, claiming that his filesystem is being held up by some sort of Red Hat conspiracy... it's really sort of embarrasing to watch, IMHO.

    ---

  • Whoops - finger slipped. That Jeff Garzik, of course...

  • Re:Shared memory + performance. (Score:1)

    by Anonymous Coward
    > A) By now being a virtual filesystem, has shared memory taken a performance hit?

    No; the underlying mechanics of shared memory have not changed. The shmfs is only used to allow stuff like partial mappings of shm segments via mmap.

    > and I still think that the overhead introduced by making shared memory filesystem based causes a
    > performance hit.

    why? and where do you think the hit would be? in creating and destroying shm segments, sure, maybe it will be slightly faster or slightly slower, it's possible. but as to actually accessing the memory? memory is memory, it makes no difference...
  • Thanks for the tip! You rule!

  • Do what I did - put the boxes with big power switches on the front on top of your desk, and make sure any boxes under your desk are in cases with sliding covers. Helps a lot.
  • Must be in base 4749935.
  • "When a bug is found, Alan sends me a patch, and I sprinkle holy penguin pee on it, and it magically becomes official."

    There's one slight problem with this explanation; birds don't "pee". If you sprinkle "holy penguin pee" on the patch, you will also sprinkle it with, ah, let's call it "processed herring."
  • Shoot. That should be www.beunited.org. I'm a dumbass.
  • that would involve members of the various *bsd cores acquiring a sense of humor...
  • Why would you have a planned power outage, and not just shut the thing down???

  • Re:Not quite perfect. (Score:5)

    by jgarzik ( 11218 ) on Saturday June 24, 2000 @06:30AM (#979224) Homepage
    Not meant as a flame, eh? :-)

    Seriously, Al Viro is the only one standing up and doing VFS work, and informing Linus of his changes. I don't see anyone else actively hacking on the VFS, and then trying to push their changes through to Linus.

    I'm a driver hacker not a VFS hacker so I'm not gonna comment on whether the current changes are good or bad. But I will say that Al Viro is the most active at pushing VFS patches straight to Linus. Further, he does post things on linux-fsdevel describing his ideas and designs. It's little wonder his changes go in.

    If that situation needs to be changed, someone has got to sit down, code a better solution, and advocate it with Linus. Not just whine on Slashdot.

    Jeff

    P.S. I don't want to give the impression that Al Viro is the only one working on VFS. But merely wish to point out that he is the most active at pushing patches to Linus currently.

  • I have a couple of questions for those using the test kernel.
    A) By now being a virtual filesystem, has shared memory taken a performance hit? I was in an arguement with another guy in the Inferno post, and I still think that the overhead introduced by making shared memory filesystem based causes a performance hit.
    B) What about DevFS? How are people going to upgrade to that? Do you just nuke /dev and run a new copy of MAKEDEV?
    C) (Okay three questions) How's the performance? Some guy was saying that I/O is still terribly slow. I encountered I/O slowdowns (from 11.29MB/sec to 4.5MB/sec on my old harddrive as measured by hdparm) in dev kernels > 2.3.99-pre7, but have not encountered them in pre6 and below. Did something just break there?

  • Kernel (Score:1)

    by Wah ( 30840 )
    What's a kernel? Is he the captain? Dungeon Master, wait!!

    --
  • ("beware, hostile .dk takeover on SlashDot!)"

    Great....Danes.

    Thank you! I'll be here all week! Try the veal! Tip your waitress!

  • B) What about DevFS? How are people going to upgrade to that? Do you just nuke /dev and run a new copy of MAKEDEV?

    You don't need to nuke anything. Compile in devfs support, mounting /dev automatically, and compile & install devfsd. There is enough documentation with devfsd and in /usr/src/linux/Documentation/fs/devfs to make this easy. Then reboot. First /dev will be mounted, and then devfsd will create symlinks for all the old device names, so that your configuration won't break (depending on what's in /etc/devfsd.conf). You may need to pass 'root=/dev/whatever' to the kernel, but I can't quite remember. It's all in the documentation.

  • Come on, where is the smallest amount of journalistic integrity? By Slashdot actually including the "Bug-Free" section just creates more propaganda in the Linux camp. NOTHING is bug free. It's amazing how many lies have been manifested and actually believed with this kind of crud on Slashdot.
  • I found the fix by reading kernel traffic of all things. And honestly, I was taking their word for it.
  • your link is broken :(, but i'll try to find it later.

    anyway after reading kernel traffic for more than 3-4 months now i have hard time believing that linus will include reiserFS into 2.4 :( (quote from latest kerneltraffic "There is no need to delay reiserfs integration into 2.4 to accomplish a journaling API in 2.5." - Hans Reiser)
    it seems to me that lots of kernel developers have some kind of animosity towards reiserFS

    you might want to check this article [linuxcare.com] from latest kerneltraffic [linuxcare.com] its about standardizing journalling filesystem.
  • Whoops, that should be ...Documentation/filesystems/...

    Also, I forgot to mention that if something goes wrong, you can just pass devfs=nomount to the kernel at boot time, and get your old /dev back.

  • And in other news... (Score:3)

    by shlong ( 121504 ) on Saturday June 24, 2000 @04:55AM (#979234) Homepage
    And in other news, more fixes and features were added to Free|Net|OpenBSD last night, bringing them one step closer to their next release.


    "I shoulda never sent a penguin out to do a daemon's work."

  • Ready for testing yes, not prime time (Score:5)

    by Oestergaard ( 3005 ) on Saturday June 24, 2000 @04:57AM (#979235) Homepage
    The test series still have problems, notably with the VM and with I/O throughput.

    For most people the test series perform _very_ poor compared to the 2.2 series when it comes to disk thoughput. 2.4-test is as slow as 1/5 of 2.2 for some.

    But, 2.4-test is ready for testing. Definitely. Get the kernel, build it, run it, stress it. The developers need all the input they can get. If you have the time, then follow LKML from the archives (from kernelnotes.org or elsewhere), and respond with a benchmark/feedback every time a developer posts a patch.

    The 2.4 series has a large number of optimizations over the 2.2 series, so most of the kernel should run a lot better than 2.2. But if your disk throughput is low and your kernel swaps unnecessarily, those other optimizations get you nowhere. AFAIK the only performance-related problems in 2.4-test is I/O and VM related. Once these are fixed, 2.4 is going to be the leanest kernel of them all.

  • Herring? (Score:5)

    by domlayfield ( 85534 ) on Saturday June 24, 2000 @04:59AM (#979236)
    No, no, no. It's the finnish pronunciation which is causing all the problem. Just as "Linus" is really pronounced "Leanus", "herring" is, in fact, "hearing". So the penguin, who obviously doesn't speak very good english, was merely asking if Linus had heard correctly.

    Either that, or he was accusing him of whoring...

    XOX DOM
  • That's an easy one: USB.

    -Karl

  • Alexander Viro has simplified the VFS greatly during the 2.3 kernel cycle. Linux 2.4 has 37 filesystems integrated into the kernel, this was 28 in Linux 2.2. BUT the total line count of 2.4's fs code is down to 128 KLOCs, while its 166 KLOCs in 2.2! This simplification of the fs architecture is largely due to Alexander Viro's (and Linus') work. 35% more filesystems but 30% less total line count, this is a plain miracle.

    ext2fs is 4874 lines in 2.4, 5548 lines in 2.2, a 13% reduction. So in fact, contrary to your assertion, ext2fs was one of the filesystems which saw a *much smaller than average* benefit of 2.4's VFS enhancements.

    Hans (or whichever reiserfs developer you are), your whining is pityful. Trying to lobby your filesystem (which bears your name, now talk about being modest) by bashing another kernel hacker who has posted *so many* patches for the generic kernel while not adding even one copyright notice is just plain disgusting. Get a life Hans, when was the last time we saw *any* patch from you showing up on linux-fsdev or linux-kernel?

    Thank you Alexander Viro for your contributions, your hard work is very much welcome. Ignore the vultures :-)

  • Re:Not quite perfect. (Score:3)

    by 1010011010 ( 53039 ) on Saturday June 24, 2000 @06:37AM (#979246) Homepage
    Patching the VFS is a solution, but not a good one. It means your filesystem is much harder to maintain, and that it can never be included in the kernel distro.

    Most people use what's there when they run "make menuconfig", if they even compile their own kernels. The vast majority of Linux users would not compile their own kernel, but just use whatever RedHat/SuSE, etc. provide. Meaning a kernel component distributed as a patch will not be adopted, unless a distro adopts is as a standard non-standard part (like the TurboLinux clustering, and Reiser in SuSE).

    If a filesystem could be loaded as a module -- meaning it wouldn't need to patch the VFS -- then people could still use it easily, even with stock kernels. Just download the TGZ, RPM, DEB or whatever and install it. Because a patch is required, most people will not use it.

    Thus the lock-in.

  • I decided to try a 2.4 test kernel last night, but I had no idea which was current. I downloaded and tested 2.4 test2 without even realizing it was brand new.

    On the down side to this, it wouldn't compile. However, I have't compiled akernel in a long time, and as Linus said, I'm probably doing something wrong.
  • Al Viro is, interestingly, not the only one working on VFS code, not the only one documenting it, not the only one putting out patches and advocating changes, but it's only his stuff that makes it in.

    Al Viro is in charge of the VFS. To go straight to Linus is to circumvent Viro. If Linus accepts a patch that Viro would not, then Viro would be in the position of maintaining it. Which he wouldn't like. He'd probably resent it.

    People have already sat down and coded better solutions and attempted to advocate them. So far, it has not worked.

    You're a driver hacker, you say. For the sake of argument, let's asuume the networking interfaces in Linux heavily favored 3Com ethernet cards. So all other cards had to look like 3Com cards to work with the interface. And that the person in charge of the nic driver interface refused to change it to make it more generalized and useful to competing NICs. Even though you have written better code, advocated changes to existing code to make it more flexible, etc.

    You'd probably start by telling people that the current system is broken, right? Trying to create pressure from outside the kernel development community, because applying pressure from inside is not working. Hardly "whining." Thanks for the flip insult, though.


  • Yes, but what about the Master betas? Will Slashdot get their hands on the Master betas?

    (Sorry, had to...)

    Your Working Boy,

  • To me, the most exciting thing about the new kernels will be "netfilter". This thing is a lot stauncher than ipchains, but remains backwards compatible with both ipfwadm and ipchains if you choose to compile in the compatibility with them! I think this'll take packet-level firewalling under linux to a new level. (No pun intended).

  • What other parts of the system have to be upgraded in order to make a smooth transition?

    What I had to do, but certainly not everything:

    1 . Add shared mem fs to /etc/fstab

    none /var/shm shm defaults 0 0

    2 . Get new modutils (from kernel.org or mirror)

    3 . Get new kernel, compile, install, lilo, reboot!

    I'm not using devfs, however, and that may take extra steps.
  • YOu need to install Wine [winehq.com] in order to be able to use windows biniaries, it isn't akernel thing. As a sidenote, not all probably will work, even with wine.
  • I don't know, but they are for sure the ones that ruin performance in a very noticably way :)

    I mentioned that those were the only two areas (they're even somewhat related), because I think just about everything else is in place. Knowing that there are only one/two problem areas left, might help give an idea of where we stand today.

  • SOrt of... (Score:4)

    by jsydik ( 192763 ) on Saturday June 24, 2000 @05:12AM (#979263)
    The Linus interview in Linux Magazine (Issue 1.1) makes a reference: "When a bug is found, Alan sends me a patch, and I sprinkle holy penguin pee on it, and it magically becomes official."

  • Re:Grassroots movement to get ReiserFS merged. (Score:3)

    by CocaCola ( 30016 ) on Saturday June 24, 2000 @11:54PM (#979266)
    Such a merge is very very easy:

    1) Reiserfs folks should start getting involved in everyday Linux-fs development, instead of sitting in their cathedral.

    2) Reiserfs folks should start posting useful patches to the linux-kernel mailing list. Patches that benefit all filesystems and the generic architecture of Linux.

    3) just post the patch in two parts: generic changes and lowlevel FS changes. Such patches are posted and merged on an everyday basis, eg. Linux now has a shared-memory filesystem!

    It's not at all up to Linus to do the merge. It's the *Reiserfs folks* who should get more involved with the Linux kernel and should learn how to merge things. There were similar or bigger projects merged lately, for example USB, RAID, LVM, framebuffer subsystem. So a merge is easy: JUST DO IT, and stop whining, please.

  • This reminds me of the the now famous 1.3.51 [iu.edu] and 2.1.129 [lwn.net] kernel announcements, aka the "Greased Weasel" releases.

    Not to be confused with 2.2.2pre4, the "Almost-valentines day", aka "horny greased weasel", aka "Presidents Day" [linuxtoday.com] release.

  • I have been runing test1 since it came out I has been perfectly stable. The only problem with my use is that the APM isn't working quite right, and USB is worth it. I'll have to try this, and hope that is addresses the APM. Oh yeah, And don't mix Maxtor and WD on the same IDE chain with this kernel. They have enabled UDMA, and some Westsern Digital drives chatter with Maxtors when that is on. You should be able to UDMA off in your BIOS. I had this problem, and it starts to errode your fs. Nothing I couldn't fix with apt of course. Right on Linus et al.

  • compile != bug-free (Score:3)

    by MrChucho ( 23875 ) on Saturday June 24, 2000 @05:22AM (#979281)
    The headline seemed to indicate that this release was "BugFree" and that "if it doesn't compile for you, you must be doing something wrong". Anyone with CS 101 experience knows that just because it compiles, doesn't mean that it's "BugFree".

Slashdot Top Deals

A Fortran compiler is the hobgoblin of little minis.

Close