Linux 2.4.15 is out; Linux 2.5.0 has also begun.

jbondjr writes: "It appears 2.4.15 is released. It's not quite updated on kernel.org's main page, but it is there if you dig through the tree. You can find 2.4.15 on one of your friendly Kernel.org Mirrors (note the 2.4.15 Changelog) From the 2.5 readme: "Linux-2.5.0 is exactly the same as 2.4.15, except for a version number change." So, enjoy the The 2.5 Tree

Comments

Release

[mroeder]

Is this Freshmeat or Slashdot ?

This isn't a new question - I know. But really does /. have to act as an advertising board for EVERY kernel release increment ?

2c

Re:Release

[aliebrah]

But really does /. have to act as an advertising board for EVERY kernel release increment ?

This isn't just any kernel release, this is the release of 2.5.0, the new kernel development branch.

Re:Release

[Mercury]

This is the release where Linus hands over the stable tree to someone else. AND the announcement of the first kernel of the new devel tree. Thus, it is news.
Yeesh.
Zephaniah E. Hull.

Re:Release

[atheos]

yea, but slashdot posts releases, regardless.
For the complainer, you can always go to your preferences, and select Linux articles to be excluded from your frontpage.
I noticed that the article wasn't available to me, until I logged in anyway.
Maybe that's a slashdot tactic to rid some of the AC posts.

Re:Release

[bclz]

Yeah~ It's Big News for OS

Re:Release

[gabba_gabba_hey]

Bleh, whatever... This comes up every time a kernel release is announced on /. I for one like to see these releases announced on the front page. If you don't like it, edit your preferences. Perhaps I should just check kernel.org every day and see whats up, or freshmeat. The fact is that I've grown accustomed to coming here as my first news spot of the day and generally just get new kernel release info from here - as it's always here. Personally, I don't mind a bit.

Re:Release

[MartinG]

Is this Freshmeat or Slashdot ?

It's slashdot. You can tell it is because it says "Slashdot - News for nerds" at the top.

You can easily recognise freshmeat when you see it as well - it has a logo at the top with "FM" written in it.

I'm surprised you have difficulty telling them apart. They really are quite different.

Re:Release

[szo]

IMO he's right. /. should have a separate topic for software release announcements, so people can exclude them. Now you can't just exclude linux, because a lot of other stuff is running under the same topic.
Or: /. could have sub-topic, or topic-modifier flags, so one could exclude linux-announcement, gimp-announcement, and so on...

Szo

Re:Release

[Colin Bayer]

My guess is that the Slashdot editors do this to answer the eternal philosophical question:

"How many people who can't read 'use the mirrors' can fit over a 100MB pipe?"

:)

Re:Release

[42forty-two42]

Ahem...

Yes.

Re:Release

[42forty-two42]

Slashdot has mirrors?!!

So who else is downloading 2.5

[Chuck Chunder]

so they can be cool and trendy and be on the development tree while it's still stable?

Re:So who else is downloading 2.5

[krorvik]

Thought about it, but then I realized that I would have to go *back* to 2.4.16 to keep on the stable side. So I'll just stick with 2.4.15 for now ;)

Re:So who else is downloading 2.5

[entrigant]

ME!! =D

Re:So who else is downloading 2.5

[ddent]

Funny you say that because thats *exactly* what I was thinking at the time. Hey, if I run 2.5.0, I can tell people I'm running the 2.5 branch. Heh.

Re:So who else is downloading 2.5

[KjetilK]

Me three! However, I failed to get it running. Can't get my network running. I have no idea what's going on, and I had two hackers working on it for me for two hours without getting anywhere. So, I'm back to 2.4.7-10 (came with RH7.2)

Re:So who else is downloading 2.5

[Cro Magnon]

I plan to, after I get home from work.

Re:So who else is downloading 2.5

[42forty-two42]

*ahem* Ever heard of up2date?

Re:So who else is downloading 2.5

[Sabalon]

did you compile in netlink (and routing messages under that) that RH7.2 needs? It's under the networking menu - and if you search on Deja you can find that you need this for RH7.2 due to the way the ifup script now works.

Re:So who else is downloading 2.5

[KjetilK]

Jihaaaaa! It worked, thanks a lot! :-) I compiled in netlink, on the advice of my hacker friends, but they didn't think I would need routing messages. When I enabled that too, it worked!

Re:So who else is downloading 2.5

[Sabalon]

Glad I could help...this is the way /. used to be - discussions and helping people. Too many political based articles lately just pissing people off left and right!

Re:So who else is downloading 2.5

[poundincludegeek]

From the README:

Linux-2.5.0 is exactly the same as 2.4.15, except for a version number change.

So you could say you're being cool and trendy and using 2.5.0, although it doesn't really matter. I'm downloading it now!

Re:So who else is downloading 2.5

[Cro Magnon]

Only problem is, 2.4.15 will eat your filesystem and burp loudly. You'd think it was a 2.5 kernel - oh wait, it is!

patch for XFree86 4.0.3 dri?

[san]

Is there a way to use a new(er) kernel with XFree86 4.0.3 DRI? It seems my mb agpgart (VIA KT266) is not supported by 2.4.7 but it may be by 2.4.14.

Re:patch for XFree86 4.0.3 dri?

[LinuxGeek8]

You can use an ac kernel.
Alan Cox has updated his dri modules in a way so you can still use XFree 4.0.3.

Re:patch for XFree86 4.0.3 dri?

[leine]

Linus has a little package in the v2.4 directory. It is called drm-4.0.x.tar.gz. You can use it to get DRM for XFreee86 4.0.x.

Re:patch for XFree86 4.0.3 dri?

[webcrafter]

So, finally Linux will incorporate Digital Rights Management?
Interesting... it's GPL'd I suppose...
However, is it possible to circumvent it by not using XFree? You seem to imply that

(Note for the humor impaired: yeah I'm just joking. Go ahead and mod me offtopic, you know you need it! indulge yourself)

Re:patch for XFree86 4.0.3 dri?

[Anonymous Coward]

DON'T open the link in the above post if you value your lunch.

Re:patch for XFree86 4.0.3 dri?

[Anonymous Coward]

You have to find the patch by Robert Love for
"Via vt8336"... it patches the agp.h file and some other files under drivers/char/agp/... and it is strange it hasn't entered the kernel yet.

Then you should use the -ac kernel series for 4.0.3 drivers. If you have a Radeon card you should go for XFree 4.1.0 and standard linus/marcelo kernel.

I thought I found it on http://www.linuxhq.com, even if now it says it is under reorganization

Turkey!

[iworm]

Build 2.4.15 with some modules. Look where it installs them... (Clue: Think when it was released)

Linus is a little joker, isn't he? ;-))

Re:Turkey!

[chabotc]

Ps, if anyone else, like me, likes his /lib/modules tree without holidays that are not celibrated here in europe; or is a religious vegitarian, simply edit the toplevel Makefile and remove the 'extraversion' (-greased-turkey) from it. Then you will have a 'plain' 2.4.15 ;-)

Re:Turkey!

[webcrafter]

Oh! you spoiled the fun for me!

Re:Turkey!

[imrdkl]

Turkey-Day is a great tradition. I guess this is Linus' second or third one now, which means that he's also feeling anticipation leading up to the day. (at least if he's celebrating with a good cook) Otoh, if he spent the day making a release, maybe he ordered take out.

I bet he still goes back home for Christmas, tho.

Re:Turkey!

[Martin Maciaszek]

or do as others suggested on lkml. Change it to:
-tasteful-salad
-sitting-duck
or
-speedful-penguin

Re:Turkey!

[Anonymous Coward]

Speaking of kernel oddities, I was thinking about editing my own kernel source to include odd lines.
boot messages are a good target: I already know what they're going to say (fine, if I update my hardware Ill go back to main-stream kernel) and it gives a great oppurtinity to make someone booting the machine thinking "did I really see that?"
("Sound blaster loaded. given severe spanking")
I have other kernel messages on my target, too..
So I now proclaim this to be the kernel-message-joke-thread. Feel free to post

("Power down"
1 sec delay
"No, wait, now it's down")

Re:Turkey!

[PM4RK5]

Yes, but its nice to see developers having fun with their work.

Gobble, gobbble!

Re:Turkey!

[Anonymous Coward]

>>> if anyone else, like me, likes his /lib/modules tree without holidays that are
>>> not celibrated here in europe; or is a religious vegitarian

Ummm, I personally think that when Linux said 'Greased Turkey', he was talking about Bill Gates, not Thanksgiving... :-)

And, sorry, but we Yanks have a lot to be thankful for, so please don't begrudge us our holidays, and we won't begrudge you yours. (BTW, it's 'Europe', with a capital 'E').

Re:Turkey!

[Anonymous Coward]

Linus is a little joker, isn't he? ;-))

Yes, he is. And I bet that our friends at MS will waste no time pointing out this kind of thing to businesses and asking them, "Do you want to bet your company's IT foundation on an OS written by people who do stupid things like this?"

No, it's not a fair thing to do, but who said the computer business was fair? When you hand someone a stick that big, it makes little sense to complain or even be surprised when they hit you over the head with it.

Re:Turkey!

[Anonymous Coward]

Ye, well, is US with U and capital S? I thought so...So better when talk put it as us...so we know to whom we should address...bugger...

Re:Turkey!

[Anonymous Coward]

yeah... if only we could see the little snide comments abut competitors they include in THEIR code over there in Redmond. And all the immaturity.

Preemptible kernel

[ThatComputerGuy]

Don't forget the Preemptible Kernel patch. This will reduce latencies and overall make a desktop machine feel snappier.

From what I hear, however, it's not recommended for servers where bandwidth is more important. But then again, you wouldn't be putting a bleeding edge kernel on there anyway, now would you?

Re:Preemptible kernel

[chabotc]

Also, as we speak, the rml preempt patches for 2.4.15 have not yet been released, and the older ones (2.4.15-pre7) do not apply cleanly to the 2.4.15 tree. If you want to use that patch, wait a few hours/days. (they are very much worth using on a interactive desktop machine).

Re:Preemptible kernel

[Anonymous Coward]

There's some argument for it on servers too - On a lot of servers, it's not the absolute max throughput that matters, it's how quickly it give the appearance of responding to requests - it's often better if the users of a server can immediately connect and download at a few K a second, than if they can't connect in a reasonable time in the first place. - so, even on a server, latency can override throughput considerations.

Re:Preemptible kernel

[ukryule]

So, this would be a version which when a new kernel is released, pauses all the low-priority task, upgrades the kernel, then happily continues?

Kinda useful given the number of releases :-)

Re:Preemptible kernel

[WzDD]

I'm interested in trying this, but I'm running XFS on my / - does anyone know about how well the patches work with XFS, if at all?

Re:Preemptible kernel

[Colin Bayer]

Actually, there was a thread or seven on the linux-kernel mailing list about this a couple months ago. However, it didn't go much of anywhere, as to upgrade a kernel on the fly would take some major code-fu (chucking variables into reserved spots in memory and hoping that you could keep track of them while you're running without a kernel)...

Re:Preemptible kernel

[Spoing]

I like Robert M. Love's Preempt patch...I've used it for the past month and it works well. Unfortunately for me and some others there are two issues that will make it hard to add into the main kernel tree (maybe 2.5 as an option?);

1. * Kernel modules that aren't SMP safe aren't preempt kernel safe.

2. * Specifically, the binary part of the Alcatel/Thompson Speedtouch ADSL modems aren't SMP safe.

There is probably nothing that RML can do anything about that. Alcatel needs to look into it, and consider releasing a fully open module, so that thier modems are better supported under Linux and other non-Linux OSs.

Re:Preemptible kernel

[mpe]

From what I hear, however, it's not recommended for servers where bandwidth is more important. But then again, you wouldn't be putting a bleeding edge kernel on there anyway, now would you?

It depends what kind of "server" is involved. Where one is mainly running interactive clients (or X terms) then such a patch might be very useful.
But for a file server, web server (serving mostly static content), email server, web proxy, etc. You probably don't want to apply it.
Hopefully soon this will become a compile option, rather than a patch.

Re:Preemptible kernel

[areguly]

I use it flawlessy for a few weeks now. A few lines are pushed down in the source, but that doesn't hurd. I recommend it.

I apply first xfs then preempt patch.

Re:Preemptible kernel

[evil_one]

The preempt patch for pre9 applies cleanly to greased-turkey. I'm running it right now.
Linux dwpws 2.4.15 #3 Fri Nov 23 07:49:28 EST 2001 i586 unknown

Re:Preemptible kernel

[Anonymous Coward]

I use bleeding edge kernels on my DEVELOPMENT servers all the time. In fact, I am compiling 2.4.15 for them right now...

Re:Preemptible kernel

[42forty-two42]

And what ha advertising ens if you have three versions: x, y, and z y has an extra variable in struct a from x z has an extra variable in struct a from x You upgrade from x -> z What happens?

Re:Preemptible kernel

[_Knots]

FYI, the patch *makes* it a compile option under "Processor Type and Features." It's a Y/N called "Preemptible Kernel."

--_Knots

Re:Preemptible kernel

[pwagland]

There is already a 2.4.15pre9 patch here. As always, please use the mirrors!

This patch is not 100% clean, but the only reject is on an already patched file.

patching file CREDITS Reversed (or previously applied) patch detected! Assume -R? [n] Apply anyway? [n] Skipping patch.

Linus is SOOOooo unprofessional

[barneyfoo]

Just look at the kernel name, for 2.4.15

You'd think it would be 2.4.15, consistent with all other kernels.

but NoooOOooo...

he has to name it:

2.4.15-greased-turkey

Re:Linus is SOOOooo unprofessional

[Scooby Snacks]

Just edit the Makefile and remove "-greased-turkey" following EXTRAVERSION, if it bothers you. Can't miss it, it's the fourth line.

Re:Linus is SOOOooo unprofessional

[barneyfoo]

Yes I know this. You couldn't detect the sarcasm?

Re:Linus is SOOOooo unprofessional

[Anonymous Coward]

Perhaps he`s american? they have a big problem with jokes that dont aren`t labelled:

/*
Joke pending
Dont take the next bit seriously - its just a
bit of fun.
*/

The door is a jar // this is the joke

/*
ok, i`m returning to being serious again.
*/

Re:Linus is SOOOooo unprofessional

[Anonymous Coward]

Wow, moderators. This guy tries to be helpful and you mark him redundant. If you would've compared the timestamp on this and the other post, you'd see they were submitted at the same time. Hey, why wasn't the other one redundant instead? Lighten up.

Re:Linus is SOOOooo unprofessional

[Anonymous Coward]

Of course, let us not forget the simple fact that Linus can name his kernel version however he wants ;)

Re:Linus is SOOOooo unprofessional

[Anonymous Coward]

So I guess, to be prefessional in your opinion, He should have named it:
1) Chicago
2) Cairo
3) Netscape Engineers are weenies

Or perhaps, he should embed A Flight Sim in the Kernel, or perhaps, a NSA Security Patch, or perhaps as many security flaws as possible?
That would make you happy. no?

Todo list?

[LinuxGeek8]

I'm sure there isn't a Todo list. And I assume there never will be one.

I was just wondering (read karma-whoring) what the first big change in 2.5.0 will be.
I guess cml2, the new config/make system.

Some other changes that might be expected;
Jfs and Xfs in the main kernel,
Fixing the latency, and maybe make the kernel pre-emptive.
There will be a scsi layer rewrite and maybe a ide layer rewrite (that's correct?),

Re:Todo list?

[Dr.Dubious DDQ]

Don't forget the possibilities of (in no particular order):

• ALSA sound
• MOSIX
• BTTV2/V4L2

Re:Todo list?

[chabotc]

To get a good feel for what the posible new cool things will be in the 2.5 kernel, check out the video / subscripts from the 2.5 kernel summit a while back. They mention

Raw access to block devices (DB work)

SCTP (nice network tricks)

Block layer redesign (scsi, ide, etc)

High performance file systems (xfs, jfs, etc)

Re-worked network driver API

Hot-Plug Devices (pci/pcmcia)

Kernel build system

Intergration of some NSA secure linux idea's (mandatory ACL's, etc)

Async-IO addons (non posix conforment)

ACPI / Power management (needed for hot plug?)
And a lot of other topics. 'Specialy make sure to check out the video's. They provide some nice insights into the wo the file systems (xfs, etc), raw access to block devices (oracle was hot on this i think),rld of kernel hero's ;-)

check out this link to the lwn article for more details, descriptions and links.

Re:Todo list?

[slittle]

And what about the firewall code? I don't know about you guys, but I just love spending a large chunk of a weekend learning the new firewall every time a new kernel series arrives. It just wouldn't be the same if it didn't fsck up my firewall scripts...

Re:Todo list?

[rasactive]

Last I recall, there was an article on Slashdot where Linus described what he wanted in the 2.5 kernel. Some karma whore can post it, but I believe one of the goals was to eliminate human configuration, thus eliminating human error.
But I could be wrong.

Re:Todo list?

[Joff_NZ]

How hard would it be to set the kernel to be pre-emptable as a choice, when you're doing a "make menuconfig" or whatever... that way you could keep both the server and the desktop people happy....?

Re:Todo list?

[Syberghost]

raw access to block devices (oracle was hot on this i think)

Our DBAs tell me that under the latest versions of Oracle, there is no performance increase on raw versus cooked partitions, and that it's therefore not worth the trouble anymore.

Re:Todo list?

[Frank T. Lofaro Jr.]

Perhaps in your situation, but there are likely users for whom that is NOT the case.

Think large enterprises which have huge databases. Oracle might know better about how to organize and access and cache the data than the OS does. The OS buffer could make things LESS efficient. Think read-ahead in situations where it is not appropriate.

Sometimes you just need the OS to get out of the way.

Re:Todo list?

[Syberghost]

Perhaps in your situation, but there are likely users for whom that is NOT the case.

Think large enterprises which have huge databases.

Right, like us. These databases are multi-terabyte. There are only a hundred and change larger businesses than us in the world. We've got more IT employees than many large telecommunications firms, and more airplanes than all but one of the largest airlines in the world.

I'm talking about FedEx, and the DBAs in question work on the largest projects in the company, including the one that does all the billing for FedEx Express.

They're quite adamant about the performance of the latest Oracle on cooked vs. raw.

Re:Todo list?

[slamb]

And what about the firewall code? I don't know about you guys, but I just love spending a large chunk of a weekend learning the new firewall every time a new kernel series arrives. It just wouldn't be the same if it didn't fsck up my firewall scripts...

What are you talking about? You can still run 2.0.x ipfwadm stuff on the latest 2.4 kernel. Likewise 2.2.x ipchains stuff. iptables has backwards compatibility. If you want to use the newer interface, you can. No one's forcing you to.

From linux-2.4.14/Documentation/Configure.help:

pchains (2.2-style) support
CONFIG_IP_NF_COMPAT_IPCHAINS
This option places ipchains (with masquerading and redirection
support) back into the kernel, using the new netfilter
infrastructure. It is not recommended for new installations (see
`Packet filtering'). With this enabled, you should be able to use
the ipchains tool exactly as in 2.2 kernels.

If you want to compile it as a module, say M here and read
Documentation/modules.txt. If unsure, say `N'.

ipfwadm (2.0-style) support
CONFIG_IP_NF_COMPAT_IPFWADM
This option places ipfwadm (with masquerading and redirection
support) back into the kernel, using the new netfilter
infrastructure. It is not recommended for new installations (see
`Packet filtering'). With this enabled, you should be able to use
the ipfwadm tool exactly as in 2.0 kernels.

If you want to compile it as a module, say M here and read
Documentation/modules.txt. If unsure, say `N'.

Re:Todo list?

[JLouder]

Think large enterprises which have huge databases. Oracle might know better about how to organize and access and cache the data than the OS does. The OS buffer could make things LESS efficient. Think read-ahead in situations where it is not appropriate.

I've seen numbers from a Sun engineer where mounting cooked partitions with the Solaris 'directio' option (which disables buffering) makes cooked Oracle partitions perform better than raw.

Granted, that was Solaris, but it shows that raw partitions don't necessarily beat cooked partitions hands-down every time.

Use the mirrors

[jbondjr]

Looks like kernel.org has updated their main page and their 100Mbps connection looks pretty maxed.

http://www.kernel.org/mirrors/

Sorry, seldom post, so don't know how to make a link on /. Also surprised my first submission got accepted :)

Re:Use the mirrors

[gazbo]

Sorry, seldom post, so don't know how to make a link on /.

Well...Kinda like you'd make a link anywhere really. So assuming you're not posting as code or extrans, I suggest you buy a book on html.

Re:Use the mirrors

[entrigant]

Just use html

Ive always

[nervlord1]

Ive always loved the thrill of a development kernel, i bet linus is very excited, constant bug fixing does get quite annoying. But ive come to realise the ingenius nature of his development: odd and even numbers beeing stable and unstable respectively, its an ingenius little system, I know that 2.5 has some exciting developments like ALSA being our sound system in the works, but i wonder what else we can look forward to.
Coincidently, does anyone remember the site that popped up when atlast 2.4 kernel was released and we got to predict when we thought 2.6 would come out, id like to see that site again to see what i predicted. Cheers

Re:Ive always

[Anonymous Coward]

It's out there somewhere. Soon after 2.4 was out it got updated for 2.6 and reset. I think I voted for Q1 2002 hahahaha..

ext3

[ma2tias]

Now, for the people who use ext3:
You will not need to download ac/ext3 patches anymore to get your journaling running.

Re:ext3

[Cardhore]

One thing worth mentioning, that it seems isn't well known, is that ext3 support journaling of the real data on the disc as well as the metadata. The other journalling filesystems only do metadata I think, so you might want to consider ext3 for filesystems where you have really critical data.

Re:ext3

[josepha48]

in 2.4.15 you just can't umount your system. There is a workaround I found the other day.

How to shutdown 2.4.15

Bring your machine to init 1 on the command line. Make sure that there are no files in /var/lock and /tmp/.X11* if you are running X. Then umount -n /var and /tmp if you have them on seperate partitions. Then umount -n al other mount points. And then shutdown. This will help prevent the nasty fsck. However it is a real PIA.

There should also be noted that there is a nasty bug in ECP in 2.4.14 adn 2.4.15 and 2.5.0 that may cause problems with devices that need ecp like webcams.

Damn!

[zloppy303]

Damn! I just got 2.4.14 to compile using a part of the 2.4.15pre patch.

I could have had some decent sleep! :)

Kbuild sorted in 2.5 I hope !!

[johnjones]

I hope that all the kbuild stuff gets in soon !!!

regards

john jones

It will happen again?

[bockman]

My impression, as a linux user, is that with 2.4 we (well...they) have moved to a three stage development cycle:

• alpha : the development tree
• beta : developers realise that they have done all the tests they could and call 'stable' the development tree. More users starts using the new kernel (users which would not have used a 'development' release). New bug and problems come out; some of them may not be fixable with small patches, but may require some strong re-enginering.
• final : the developers and beta users are satisfied with the current performances of the stable kernel ( and maybe also tired of working on it ;^>). The stable kernel is given to the maitainers. A new development kernel begins.

This is not a bad model. Maybe the kernel is getting too big for the two-stage model. And maybe too many people are using Linux for critical business, starving the 'development' releases of the testers it needs to get production quality.

Re:It will happen again?

[nyseal]

Or....maybe just don't trust it enough.

fast development

[carpe_noctem]

Dammit! I just downloaded 2.4.14 and recompiled not more than a few days ago, and time to do the same thing all again. Seems like the kernels are coming out faster than one can keep up with them. =)

Re:fast development

[Anonymous Coward]

If you had followed the kernel development, you wouldn't have upgraded to 2.4.14 in the first place. 2.4.15 was expected "Real Soon Now" and it would merge the Alan fixes, be handed over to the new maintainer, and last but not least, be Really Really Stable.

Look at the changelogs. I think seeing all those fixes and an entry that says "Alan Cox: final merges" (not to mention that the -ac tree has haulted) is well enough to warrant saying, "I think I'll wait for this."

Re:fast development

[Anonymous Coward]

>>>Seems like the kernels are coming out faster than one can keep up with them.

If you think that's bad, try keeping up with the security patches for Microsoft's so-called "operating systems". :-)

Re:fast development

[Anonymous Coward]

Maybe you should use SGL instead.....

how to make a link Re:Use the mirrors

[leuk_he]

just code html to make a link. you should type:

<a href="http://www.kernel.org/mirrors/">http://www.k ernel.org/mirrors/</a>

This works in html formatted <B> and </B> plain old text. (i think code as well).

just a tip.

Stable or not?

[doorbot.com]

Since the kernel is now on the 2.4 series kernel, is it officially considered a "stable" kernel yet? With the recent discussions of the VM system and pre-emptable kernels, I wonder if the latest version is stable enough for my server. I'm still running 2.2.19 and it's working fine, but I'm looking forward to upgrading some time in the next couple weeks (maybe).

One big thing I'm looking for a fix for (hopefully with the 2.4.x series) is for the SNMP/netstat info to use more than 32 bit integers to store the number of bytes transferred. Mine keeps resetting after 4 GB transferred and over 6 months that would've added up, but right now it resets itself back to zero after it hits the 4 billion byte limit.

Also, the SMP is supposed to be better, but will that really make a difference on a server which does mild web serving, other miscellaneous duties, and who's primary purpose is serving Unreal Tournament games? As of now, it doesn't look like UT takes advantage of both processors, it just uses one...

Re:Stable or not?

[Anonymous Coward]

SMP will help somewhat, in that it would free up one CPU for use entirely by UT, while other tasks could run on the second.

Re:Stable or not?

[Ami Ganguli]

Stable is totally relative. The 2.2.x tree has had a lot longer to mature. It'll probably be a year before 2.4.x is as solid. This is true with most operating systems. If you have a mission critical server that's been working flawlessly on 2.2.x, then you have no reason to upgrade to 2.4.x. Since the old series is still maintained with security fixes, you can stay with the old kernel for a long time.

On the other hand, if you're installing a new machine you might benefit from some of the 2.4.x features. As long as the machine isn't really critical, you might as well go for it. If you're starting a project that won't be deployed for another six months or so, then 2.4.x is definately the way to go. Likewise on a desktop machine.

In your particular case, you need to evaluate how stable your Unreal Tournament server needs to be. Will you go out of business if it crashes? If not then I'd go ahead and upgrade. Chances are it'll be rock solid, and if not you can always go back.

Re:Stable or not?

[anpe]

IIRC, SNMP integers are coded in ASN.1 standard, I wonder if your 4 bytes limitation doesn't come from the SNMP protocol itself ...

Re:Stable or not?

[TheGratefulNet]

snmpv2c and v3 (most people just use v2c) DOES support counter64.

I suspect its the kernel that's only keeping the 'ifconfig' counters to 32bits. the ucd-snmp pkg (the most popular) supports 64bit counters for the o/s's that have real native support for that.

being an snmp guy myself, I'll take a look at this. if, in fact, the linux kernel DOES support 64bit ifconfig counters, then its trivial to wrap that support back into the ucd pkg.

Re:Stable or not?

[Billly Gates]

The stable version is right here . You never have to worry about poor, unplanned, and non cvs development with this one.

Re:Stable or not?

[yem]

IMHO its stable. I've used 2.4 exclusively since 2.4.0 at home and since 2.4.4 (WITH AC PATCHES) on servers. Have had absolutely ZERO problems.

File servers, web servers, database servers, firewalls, you name it. I have *not* used the plain unpatched "linus" kernels much at all so I can't speak to the stability/performance of those. The -ac ones are good though.

I'm on 2.4.13-ac7 right now (the last of the stable -ac kernels). I might stick with this for a wee while.

I will *not* be using 2.4.15+ on any servers until some serious time/testing has taken place.

Re:Stable or not?

[Anonymous Coward]

So run two separate UT server processes on different ports!

hard packed versions

[ma2tias]

The readme file in 2.5.0 says:

Linux-2.5.0 is exactly the same as 2.4.15,
except for a version number change.
Subsequent releases diverge, with Marcelo Tosatti
maintaining the stable 2.4.x kernels, while the 2.5.x kernels are for development work.

and looking at the filesizes we find what we expect that

linux-2.5.0.tar. takes up 129 699 840 bytes
linux-2.4.15.tar takes up 129 699 840 bytes

are the same. But if we look at the packed versions

linux-2.5.0.tar.bz2 23 748 963
linux-2.4.15.tar.bz2 23 747 061

A difference of 1902 bytes

linux-2.5.0.tar.gz 29 404 635
linux-2.4.15.tar.gz 29 404 736

A difference of -101 bytes

How come?

Re:hard packed versions

[Cow4263]

I'd assume the difference lies within the only difference.. the file name. 15 is more then 0 so there would be a slight increase in the 4 series kernel.

Just a thought...

Re:hard packed versions

[iworm]

Maybe they backed out the Greased Turkey patch for 2.5.0?? Dunno, haven't looked. (Or maybe 2.5.0 should be Cold Turkey?)

Re:hard packed versions

[-brazil-]

gzip and bzip2 compression isn't completely deterministic, different implementations of the algoritm, or command options can yield more or less different (but correct) output.

Re:hard packed versions

[w00d00]

since the name is different, the overall entropy is different. IMHO the packaged size then depends on the algorithm..
just my .02

Re:hard packed versions

[Anonymous Coward]

They most certainly *are* deterministic; compress the same file with bzip2 or gzip and you'll get the same file every time. Of course, using different versions of gzip/bzip2 might not necessarily give you the same file; that's not being undeterministic, it's just the natural result of changes in the code.

Re:hard packed versions

[-brazil-]

Well, "not deterministic" may not be the best way of putting it, what I meant is that the format itself isn't so strict that the same input always has to result in the same output.

Re:hard packed versions

[42forty-two42]

a: the subversion joke in 2.4.15 b: compression is very picky. A change of one byte makes a big difference.

Re:hard packed versions

[jquirke]

What a shame there isn't a way to properly moderate those who ask stupid questions.

Re:Belch...

[Anonymous Coward]

That chick that keeps showing up in the Speigel ads on iWon drives me nuts. She ain't all that good looking, but damn I want to nail her every time I see her picture.

Upgrade

[vidarlo]

Damn. Downloaded 2.4.14 over a 56K link yesterday... Well.Patches suck but..
I suppose we realy need all those new kernels on 2 year old pcs?
Most normal pcs don't need this at all.
Unless you're a freak

Re:how to make a link

[jbondjr]

Thanx. I just didn't think it was so HTML, so I was thinking there must be some escape keys or something.

Informative

[beable]

There is a new technique called "Don't Click On The Link Of Stories You Don't Want To Read". Apparently it allows people to decide whether they want to read the story just by looking at the headline, and if it says "New Linux Kernel Out", it allows the user to not click the link, thus avoiding the story they didn't want to read. I can't see how this new-fangled "Don't Click The Link" technology could possibly work, but some people say they are very happy with it.

So... has anybody found any bugs on 2.4.15 yet?

Re:Informative

[Anonymous Coward]

Do you have a patent on that?

Re:Informative

[Anonymous Coward]

this is english humor?

Re:Informative

[Anonymous Coward]

Yeah.... you can tell because it's actually funny.

Re:Informative

[Johnny Vector]

There is a new technique called "Don't Click On The Link Of Stories You Don't Want To Read".

Well okay, but see I wacky-parsed the headline as

Linux 2.4.15 is out; Ringo's in

So naturally I clicked on it.

Re:Informative

[42forty-two42]

'Wacky-parsed'? What's that?

Re:Informative

[UtilMan]

Didn't Amazon patent this "Dont Click The Link" technology?

Re:Informative

[jumpfroggy]

Is this patented? *hmm*

Re:Belch...

[Anonymous Coward]

The ad and Here's the link to the original on the Speigel site.

At last

[javajerk]

Must have been a week since I installed 2.4.14 - to long a time :) This will be a nice weekend, I guess.

What I really want to know: how is ext3 doing? And why is Alan still patching .13?

Hmm, hopefully will be able to leave work early this evening...

Cheers, Lars

Re:At last

[Anonymous Coward]

Ext3 is in 2.4.15, and Alan has merged everything considered he considered stable with into that tree as well.

I don't think he plans on doing the -ac patches anymore, at least not the way he used to.

Request for 2.5.x

[redcliffe]

I want support for ATAPI/IDE CD Burners. The SCSI emulation solution isn't the best solution I don't think. That's the extent of my wishlist. I'll be happy with whatever else Linus gives us in his infinite wisdom. Anyone else got a wishlist for 2.5.x?

Re:Request for 2.5.x

[Doug Neal]

ATAPI is SCSI emulation. ATAPI is, essentially, the SCSI protocol over the IDE bus - Windows has the same kind of thing with it's ASPI layer which is used on SCSI and IDE drives. Linux's SCSI emulation is the perfect solution because programs such as cdrecord don't need to worry about whether the drive is SCSI or IDE, because they can speak the same language to any of them. cdrecord should work with your IDE CD burner unless you've got a particularly wierd one... but I always like to check Linux compatibility before buying new kit anyway...

Re:Request for 2.5.x

[chabotc]

Actualy, it would be quite simple to support ATAPI burners (since ATAPI is already supported). However the downside is in the userspace code. Most of the populair tools (read: cdrdao and cdrecord) are written with scsi interfaces in mind.. adding IDE writer support to those would be a lot more time consuming.

So scratch that of your linux 2.5.x wishlist, and put it on your cdrecord TODO list ;-)

Re:Request for 2.5.x

[Anonymous Coward]

Personally, my wish is that it will stop dying whenever my machine starts to use swap heavily. That'd be a big plus for me. :/

ALSA Re:Todo list?

[Laven]

Does anyone know the reason why they haven't included ALSA in the main kernel yet?

I finally switched my Red Hat 7.2 and recompiled KDE with Alsa support, and things are SO MUCH nicer than free-OSS. I can finally use Real Player or Quicktime Player (via CrossOver) seemlessly with KDE. No more fussing with artsdsp to get Real Player to work, or terminate artsd in order for Wine audio to play.

Re:ALSA Re:Todo list?

[Anonymous Coward]

Beats the fuck outta me. I am so tired of this.. every time I upgrade my kernel, I can't hear anything. Then I remember: oh yeah, I have to reinstall alsa.

Re:ALSA Re:Todo list?

[Anonymous Coward]

It does not have full OSS compatibility yet. The last time I tried it, the sound on my SB Live was full of static (the music played, but the quality sucked). A bug report had already existed for a while, and a year later, nobody had even tried to fix it. And recently Vorbis has had ALSA problems, because ALSA and OSS implemented the same function in different ways (something to do with blocking/non-blocking access to /dev/audio).

Re:ALSA Re:Todo list?

[J4]

I'm using the unstable version of ALSA with a SBLive and it sounds good. On some machines raising the value latency timer value (which I haven't had to do here) will get rid of the crackling.

Re:ALSA Re:Todo list?

[Laven]

OSS used to crackle horribly on both my SB Live and integrated AC97 audio chip. I installed the latest ALSA beta and it works very cleanly now. It works for me.

Stability

[sydb]

It's probably stable.

But why take my word for it (or anyone else's) - download, compile, install, stress test, use.

If you don't have a test machine - wait a while, read the mailing list, see if anyone reports bad uptime / errors which affect you.

If you're so worried about stability, I don't see any other real option for you. Certainly, it seems naive to consider an 'official' designation of 'stable' sufficient to jump in with both feet. Either test it yourself, or read other peoples reports - there's plenty of them.

My experience? I have had no trouble with the 2.4 series, except for VM probs in 2.4.9 which brought my desktop machine to a short lived pause once or twice. I installed 2.4.13 last night and am hoping Andrea's VM will cure this.

Loopback device fixed..and new pcmcia-cs 3.1.29

[buserror]

For some reason I couldn't get the loopback block device module building in 2.4.14 on my laptop (depmod chocked on undefined symbols)

works again in 2.4.15

There is also a new (at last! almost 4 months since .29) 3.1.30 pcmcia-cs package there

Re:Loopback device fixed..and new pcmcia-cs 3.1.29

[evildead]

loop.c is easy to fix in 2.4.14. Just remove the two lines that say 'deactivate_page'

Re:Loopback device fixed..and new pcmcia-cs 3.1.29

[JahToasted]

There is also a new (at last! almost 4 months since .29) 3.1.30 pcmcia-cs package

Wow... cool... thanks for letting me know, and according to the changelog the preemptable kernel patch will work now... Have a lot of patching to do this weekend.

Problem with ALSA and 2.4.15?

[Yomlogs]

Hmm, my very first /. post and it's a complaint. ALSA 0.5.12 (emu10k1 driver) seems to have been broken by this release - amixer returns this error:

amixer: Mixer 0/0 open error: Invalid argument

...and xmms segfaults immediately. :-(

direct links to kernel mirrors

[ddent]

Please avoid slashdoting the main server. Here is list of direct links to mirrors, by country.

Version 2.4.15, full tarball:
al
dz
as
ad
ao
ai
aq
ag
ar
am
aw
ac
au
at
az
av
bs
bh
bd
bb
by
be
bz
bj
bm
bt
bo
ba
bw
bv
br
io
bn
bg
bf
bi
kh
cm
ca
ic

Almost totally off-topic but...

[biglig2]

...reading this I went over to catch up on AC's diary. More human interest than the change log. Got to November 5th and read:

"Windows installation day one. Getting rid of the old windows was easy - they fell apart quite happily, and certainly wont be re-installable anywhere else. "

Blimey, I thought, until I remembered his house is getting done up...

Re:Almost totally off-topic but...

[Lennie]

Yes same here, I almost spit my thee over my laptop.

Re:Almost totally off-topic but...

[42forty-two42]

...and he hasn't updated it for a week...

please mod parent down, this is rediculous

[CaraCalla]

In a strict sense its true that the post is "Informative". As informative as if he wrote that Slashdot is usually green, that Rodger Rabbit is a Cartoon and that Bin Laden is evil.

Did I already mention that Bin Laden is evil?

--
Live on earth may be expensive, but it includes an annual free trip around the sun!

Re:please mod parent down, this is rediculous

[MartinG]

ask your mum to get you a sarcasm detector for xmas :)

my post was not supposed to be informative (although strictly it is as you point out)

Those who share my sense of humour might have said funny rather than informative.

Network downtime

[YellowSubRoutine]

Curiously trying this patch on a 2.4.13 machine a while back, I found myself in the very unpleasant situation of having about 50% packet loss towards that host (lan where I have absolutely 0% loss regulary)

No-one can explain this, but two very identical setups, only differing by this patch...

Preemptability has nothing to do with stability

[barneyfoo]

Preemptability is just a feature that you can add to your kernel to make kernel threads preemptable (using the in-place SMP infrastructure).

Preemptability is not:

-an attempt to fix a bug in linux or make linux more stable
-a way to make Linux slightly more advanced than MacOS
-Something that is desperately needed

Preemptability is:

-A way to make processes that have long execution threads inside the kernel not take away from higher priority user-space threads (priority is not the same as niceness).

So PLEASE, lets stop this bandying about of "preemptability" when you know not of what you speak.

Why?

[brunes69]

Quick question: I was wondering why there is so much fuss about this KBuild thing, and why there is going to be so much time / resouces spent on it in 2.5. I have never had a problem with the current kernel config, and find it quite intuitive (make xconfig especially). I can think of many more things which I would think deserve a higher priority than this. Someone care to enlighten me?

Re:Why?

[Crimson Midget]

The changes aren't so much for you the user, as they are for the developers and maintainers. The current system is a bit of a tangled mess and considered a hack. "make dep" is practically broken, configuration, compiling, pretty much everything takes longer than it should.

But the user should see some changes as well. The configuration will be streamlined, and some level of automatic configuration will be available. According to Eric the new system will make it impossible to create invalid configurations. Also, a single "make" will compile everything you need, including modules and dependencies. Parallel builds will be improved and it will be possible to build individual sections instead of recompiling the whole kernel if need be.
However, there is a big gripe some have with CML2; it's (currently) implemented in python.

Marcelo Tosatti

[rammer]

Why is Marcelo Tosatti maintaining the 2.4.x tree?
Is Alan Cox still tied up in 2.2.x maintenance or is he just tired of maintaining stable trees?

Anyone know more about this guy?
I know he's been a kernel hacker for a long time and has done some work with linux-ha project.

Re:Marcelo Tosatti

[Teukels]

Marcello is maintaining 2.4.15> because Alan Cox won't, it seems he made a personal choice.
Alan Cox is going to be in touch more with the customers of his employer, RedHat.

Marcello is a knowledgeable guy, he has proven to be sceptical and is imho able to discriminate well between a 'bad' patch and a Good(TM) patch. Alan said he would assist and advice Marcello if he needed help.

If you want to know more about Alan Cox, please search google for Alan's diary and website. (Or try Telsa's website, might be even more fun to read.)

Re:Marcelo Tosatti

[42forty-two42]

linux.org.uk

Unprofessionalism has a price

[Sits]

If you use make bzlilo then you will end up with the error:

Fatal: Label "2415-greased-turkey" is too long

While it's easy enough to fix (edit lilo.conf), I bet this is going to cause problems for someone.

I'm crying

[barneyfoo]

really, I am.

but anyway, last I checked "make install" put bzImage into /boot/vmlinuz and System.map into /boot/System.map, and move your old vmlinuz and System.map to vmlinuz.old and System.map.old, then reran lilo.

Never used make bzlilo myself. Sounds like one of those features of the kernel you really shouldn't use, but some people do, and if you do use it you should really know what you're doing, but I guess that should be true of anyone compiling their own kernel.

2.5 not fully open yet

[kanelephant]

It will probably be at least a week until 2.5 really gets moving. Linus on the lkml

...when I release 2.4.15, I'll at the same time release a 2.5.0 that is identical except for version number (that makes synchronization easier later on). And I'll probably _not_ start accepting all the big waiting patches immediately, I'd rather wait for at least a week or two to see that there aren't any other issues.

It's much easier doing some of the IO patches in particular knowing that the base you start out from is stable.

Linus

-K

Bad Timing

[Cro Magnon]

I had some free time after stuffing my face with Greased Turkey(tm), so I compiled 2.4.15-pre8 for ext3 support. Then, when I got to work this morning, I found out that Linus had not only released 2.4.15, but also 2.5! Damn, damn, damn!

Re:Bad Timing

[be-fan]

It's not that bad! I track the XFS tree, so I tend to compile the kernel every few pre-patches (this is a personal workstation, so it doesn't matter). It takes 15 minutes to do the compile, as long as you keep your config files around. Quite painless, really.

Oops.

[42forty-two42]

Oops...
Something weird happened...

Corrected version:

And what happens if you have three versions: x, y, and z

y has an extra variable in struct a from x
z has an extra variable in struct a from x

You upgrade from x -> z

What happens?

3 stage? no.

[evil_one]

In the case of the Linux kernel we have maintaince and development.

Major changes in structure and systems are made in development, with drivers added as well.
Maintaince is bugfixes and drivers.
That's it.
The reason that 2.5 didn't open sooner is that the 2.4 series was not stable enough for the maintainer's taste. Now that it is, active development (rather than maintaince) will continue.

Re:3 stage? no.

[Anonymous Coward]

no, I say he's right. early 2.4 has a whole string of changes that were definitely more than 'bugfixes and drivers' (VM espescially). Only now has it really settled down into pure maintenence, as kernel hackers move forward with all the cool stuff for 2.5 :-)

PPC woes

[vs]

Yeah, and as PenguinPPC mentions somebody spilled PPC support.

The advantage is that we probably get the most obscure unames ever: Linux issan 2.4.13-benh+xfs+ipsec-snap+irda-hotfix #1

Re:second!

[Anonymous Coward]

Think the average stormtrooper knows how to install a toilet main? All they know is killing and white uniforms.

Yes it does !!!

[Macka]

Because there are lots and lots and lots of people like *me* who are pleased that Slashdot keep track of the kernel releases.

What's your problem? If you don't see the point in it, don't open up the bloody article. If costs you what ... 2 seconds of your life to skip onto the next one!

You're in the minority here pal, accept it and move on. Better still, edit your preferences and screen out the stuff you don't like. That's why they are there!

SCSI Emulation?

[evil_one]

It's actually not emulation, but transport. ATAPI is actually SCSI, implemented on top of IDE. Remember those parallel port zip drives? They are SCSI over parallel port. SCSI "emulation" simply lets userspace apps talk to the devices in their "native" tounge.

talk about verison inflation

[Cardhore]

sheesh. 2.4 to 2.5 and not a single change.

Re:talk about verison inflation

[Anonymous Coward]

....dipshit

Version numbers suck

[Anonymous Coward]

Thats why microsoft called windows 4.0 95!

Two different things: Kbuild and CML2

[LunaticLeo]

Kbuild is the new Makefile system. As you said "make dep" is reliable, faster compiles (search google for "recursive makefiles harmful", more parallelism in the make (same recursive makefile problem), easier for developers to use.

CML2 is the rule system for features to be compiled into the kernel. The rule solver is built in python.I am a perl guy, but I see no problem with this. Unified text, curses, and GUI configurator. The problem it solves is it is possible to compile a kernel with features requiring subsystems like SysV IPC, but not actually having SysV IPC included in the kernel build.

These efforts are complementary. Makefiles are the low level rules to compile code and CML2 is the high level rules for whole features/subsystems.

Download Patch Here

[Laven]

It seems that a patch against 2.4.15-pre9 was released, but the web page wasn't updated. You can get it from here. http://www.kernel.org/pub/linux/kernel/people/rml/ preempt-kernel/2.4/

thanks

[Anonymous Coward]

thanks for that info, you saved my day :)

Warning: Troll - Re:Todo list?

[Anonymous Coward]

* Raw access to block devices (DB work)

Can anyone say "cache coherency". I knew you could. FreeBSD had this but removed it a year ago because it's impossible (and would defeat the purpose of raw access) to keep the buffer/cache aware of what's going on.

* Block layer redesign (scsi, ide, etc)

Try, oh try again. So this will be the third scsi layer in the 2.x series? Hopefully it will be done right this time.

ALSA and the 2.(4.15|5.0) kernel

[pwagland]

Hi all,

While we are talking about incompatible kernel patches, please be aware that ALSA 0.5.12 does not work under 2.4.15. You need to get the CVS version, as described here. ALSA 0.5.12 compiles, but does not work.

Re:ALSA and the 2.(4.15|5.0) kernel

[slashdot_commentator]

ALSA v0.9.0.beta9 works great on kernel 2.4.14. I can't imagine it breaking in the next kernel. I don't think anyone, other than a packages maintainer, would really care that they had ALSA 0.5.12 working on a kernel version.

Re:ALSA and the 2.(4.15|5.0) kernel

[pwagland]

Actually, it only appears to work great. The great conumdrum, is something broken if it appears to work? Read up more here. I.e. It does have the same problem, it is just masked....

For good details on what broke ALSA 0.5.12, see here.

Linux NFS

[cpeterso]

Don't forget that Linux NFS is perpetually broken and outdated. I believe there will be another NFS rewrite for Linux 2.5.

With so many compete rewrites planned (VM, SCSI, VFS, NFS, ...) for Linux 2.5, does that mean Linux is really great or just really sucky?

Re:Linux NFS

[Anonymous Coward]

or maybe it means 2.5->3.0 instead of 2.5->2.6 :-)

Warning: DO NOT use ext3 with recent -ac kernels

[archnerd]

Later versions of the -ac kernel contain an ext3 bug that causes frequent kernel panics. This bug is introduced into the linus kernel in 2.4.15-pre2 and fixed in 2.4.15-pre8. Bottom line: ext3 users should use 2.4.15/2.5.0, not 2.4.13-ac8.

Warning: likely file corruption bug in 2.4.15

[Anonymous Coward]

(dirty data does not get written onu nmount)

Copy from linux-kernel mail (by Alexander Viro)

Again, as a workaround - sync before umount (and don't boot
unpatched 2.4.15/2.4.15-pre9 again, obviously).

Breakage happens when you umount filesystem (_any_ local filesystem, be
it ext2, reiserfs, whatever) that still has dirty inodes.

IOW, if you are running 2.4.15 - build a patched kernel, install it and
do the following:
* switch to single-user
* sync
* umount everything non-buys
* remount the rest read-only
* turn the thing off
* boot with patched kernel or with anything before 2.4.15-pre9

How are preX patches applied?

[Sanity]

I have been trying to figure out how the pre-release kernel patches are applied. I have tried using the patch-kernel script to no avail. Are they patches against the previous prerelease, or the previous release?

Re:How are preX patches applied?

[AYEq]

the preX patches are applied against the last releases source. ex.

2.4.15-pre5 is applied against 2.4.142.4.15-pre6 is also applied against a virgin 2.4.14

I think that www.bzimage.org has some incremental patches though, but I have never used them.

Ext3 Utilities / Updates

[PM4RK5]

For those of you actually compiling in ext3 support, you're going to need the new ext2 / ext3 tools (1.25) which adds support for ext3 filesystems. Then you should run:

tune2fs -j (device)

for each device, to add ext3 journaling to existing ext2 partitions.

File system corruption under 2.4.15

[Anonymous Coward]

There is file system corruption bug in the 2.4.15
Here is a description a what to do if you are running 2.4.15 already. And here is the cure.

Unmount problem with 2.4.15-final (BPB Release)

[mikefe]

Hello,

It looks like this may be another BPB - Brown Paper Bag - 2.4 kernel release.

There is already a patch available though from Al Viro.

I believe Al can say it best. Here he is replying to war on LKML:

Damn this junk removal feature. I could've quoted if it didn't have trouble with ">"!! Grr.

http://marc.theaimsgroup.com/?l=linux-kernel&m=1 00 655627817897&w=2

References:
http://marc.theaimsgroup.com/?l=linux-kernel&m=1 00 655058402225&w=2

Patch:
http://marc.theaimsgroup.com/?l=linux-kernel&m=1 00 655322609664&w=2

Don't use this release: FS-corruption problem

[spektr]

A bug in the inode handling of 2.4.15 can leave stale inodes (unremovable files) when a filesystem is unmounted. The problem exists for all filesystem types. Do not use 2.4.15-pre9 and 2.4.15-final and wait for the patch.

Al Viro on linux-kernel: Sigh... Supposed fix to problems with stale inodes was completely broken.
What we need is "if we are doing last iput() on fs that is getting shut, sync it and don't leave it in cache". And yes, we have a similar path in iput(). Similar, but not quite the same.

Al Viro on linux-kernel:Breakage happens when you umount filesystem (_any_ local filesystem, be it ext2, reiserfs, whatever) that still has dirty inodes.

IOW, if you are running 2.4.15 - build a patched kernel, install it and do the following:
* switch to single-user
* sync
* umount everything non-busy
* remount the rest read-only
* turn the thing off
* boot with patched kernel or with anything before 2.4.15-pre9


Russel King on linux-kernel: I think 2.4.15-greased-turkey should be renamed to 2.4.15-dead-duck. 8)

Re:Don't use this release: FS-corruption problem

[spektr]

Here's another useful post from Russel King on linux-kernel for those who ran into troubles:

> What kind of breakage are we looking at here? I had a system that ran 2.4.15
> and got shut down without a sync. What kind of corruption will occur and is
> it something a simple fsck will fix?

fsck does seem to fix it, but it won't automatically detect the problem
(since the filesystem is marked clean).

It basically removes the inodes from the disk, but leaves the names in
the directory. On the next boot, init scripts which clear out certain
directories fail, and various daemons fail to start because of it.

It seems that the only solution is to force a fsck at boot:

shutdown -F -r now

should do the trick.

Re:Don't use this release: FS-corruption problem

[spektr]

...and by the way: kernel.org just crashed.

H. Peter Anvin wrote on linux-kernel:

zeus.kernel.org has suffered what appears to be a multiple disk RAID
failure, and no longer appears to boot. I can't do any further
diagnostic without physical access to the machine; however, there is a
good chance we have lost a RAID.

I will attempt to bring the machine back, but, in the meantime, I
would like to ask if there is anyone or any company who would be
willing to donate eight (8) 73 GB SCA SCSI disks on short notice. We
have been suffering from a shortage of space on the affected RAID for
quite a while, and if it is truly dead it might be reasonable to
replace the drives with higher-capacity ones instead.

Please email me if you would be able to help.

-hpa

Re:my god, slashdot sux

[Anonymous Coward]

also, why is galeon crashing everytime I open slashdot?? nothing changed about the galeon I had installed... perhaps it has to do with whoever kidnapped CT and put that "love the xbox' statement in his mouth.
yeah, their story posting is a sham, i think. I have sent stories only to have it posted under someone else's name, like that person had a better write up, or is better friends with the editors, or something.

And at last, the solar designer patch is ported...

[MattW]

According to openwall, the non-exec stack and other security patches so useful in 2.0.x and 2.2.x are finally on the way to 2.4.x, giving you that extra bit of protection. Of course, it looks like it will have its own beta period, but those patches protected my 2.0.x box for quite a while from 0-day exploits, and let it manage a full year of uptime at one point despite dozens of users and a bunch of services (including the ever-dangerous wu-ftpd).

2.4.15 corrupts data

[damian]

as most of you lkml reading people surely noticed: 2.4.15 corrupts data when you unmount a filesystem (for example when rebooting), it doesn't do a sync before the umount.

Back out to 2.4.15-pre8 to avoid it. Or find the patch on lkml.

FS corruption -- see link

[Anonymous Coward]

FYI :: http://marc.theaimsgroup.com/?l=linux-kernel&m=100 655058402225&w=2

Hope it helps to avoid problems for some.

What Linux needs is a Documentation czar

[Anonymous Coward]

Ever looked in the Documentation directory recently?

It's a MESS.

Ever noticed other documentation spread out through the Linux source tree? Yeah, it's a mess too.

I'd have a stab at cleaning the Documentation tree up so that it mirrors the layout of the rest of the source tree if only there was a hope in hell of Linus accepting the patch and integrating it in.

Patching again... getting XPish hein?

[CapnGrunge]

Heck, I didn't like it when I immediately had to patch 2.4.14 due to the problem of loopback. Now I wonder when I will finally be able to upgrade my HP NetServer with a Cyclades PC300 with *one* kernel.

Re:And at last, the solar designer patch is ported

[CTachyon]

According to openwall, the non-exec stack and other security patches so useful in 2.0.x and 2.2.x are finally on the way to 2.4.x, giving you that extra bit of protection.

These days, many writers of buffer overflow exploits (the *only* exploit that a non-exec stack could possibly save you from, in case you didn't know) assume that vulnerable software is running with a non-exec stack, and code around it. Non-exec stack patches may have saved you from the 'kiddies in the past, but they afford no protection in 2001, although it certainly doesn't hurt your security so long as Assumption #0 is that the 'kiddies won't even be slowed down by it.

Re:And at last, the solar designer patch is ported

[MattW]

No one, least of all me, said that a non-exec stack was all you needed to keep a box safe. However, it is still useful to have, and contains a variety of other useful patches, just as restricted access to /proc and some things to avoid games in /tmp, like symlinks to other system files, etc. Nonetheless, there's also such a plethora of buffer overruns available, and they are still common, probably because there is so much "example" code available, and smashing the stack is easier to code than trying to heap overrun malloc'd values or whatever a would-be exploiter might choose to do. Furthermore, many exploits that make it into the wild begin as proof-of-concepts, and therefore START as buffer overflows, since the writer has no intention of cracking anything. Several buffer exploits have been out in 2001. If you buy yourself any extra time, it is a patch well-applied, and the other features may be useful depending on the system.

Alternately, a lock on your house door won't protect you from a burglar (the *only* entry that a lock will stop is a standard check-for-open-door entry, in case you didn't know), so you can have a lock on your door, as long as Assumption #0 is that the lock won't even slow down a burglar.

Raw devices are good

[Frank T. Lofaro Jr.]

You aren't supposed to use block and character devices for the same device at the same time!

You have the raw device ignore the buffer cache. If you must switch between block and char devices (unlikely), use the block device, then flush it before any access to the character device (unless it already has been flushed).

Just because using raw char devices messes up the block device cache isn't a reason to remove the feature. It is a reason to NOT MISUSE the feature.

Do your light sockets have anything in them designed to keep your tougue out of them? ;)

Patch download for FS corruption problem

[DeeKayWon]

If you try to cut and paste from the mailing list archives, you'll probably have spaces instead of tabs in the patch, causing patch to choke on it. I've put a tabbed patch up here.

Re:And at last, the solar designer patch is ported

[CTachyon]

No one, least of all me, said that a non-exec stack was all you needed to keep a box safe.

Cool. It's just that there are quite a few people out there who think that having a non-exec stack is some sort of magic pixie dust that will guard against all buffer overflows, and I felt obliged to point it out in case you were one of them (which you obviously aren't).