Slashdot Mirror
Linux Kernel 2.4.4 Released
Marek Kupsta writes: "Linux 2.4.4 is out! Go get it! =)" Check the Changelog. As I write this, the bandwidth meter on kernel.org says 34.4 MBit/s.
← Back to Stories (view on slashdot.org)
← Back to Stories (view on slashdot.org)
It also includes the official IPtables FTP MASQ security fix, which is good. Time to upgrade... ;-)
Short answer: No
Slightly longer answer:
Adding ACL support to the kernel would require large-scale changes to the entire VFS, as well as the individual filesystems. This isn't going to happen in a stable kernel series. IIRC, this is one of the major goals for the 2.5 development cycle (along with generic extended atributes and maybe named streams), so the ACL support introduced in 2.5 may be backported to 2.4 (but that would be up to Alan Cox).
Although patches currently exist to add extended attributes and ACL support to the current VFS, I don't think that the kernel gods are entirely happy with its architecture. (This has been a frequent topic on the linux-fsdevel list.)
2.4.3 does have serious issues.
The reader-writer semaphore implementation is broken, resulting in processes getting stuck in the D state in down_semaphore. Heavily threaded programs (like Mozilla) are most likely to hit this bug, resulting in lots of stuck threads and an unusable program.
(Nothing actually used the rw-sems until fairly recently, which is why this bug went undetected for so long.)
Also fixed: the iptables FTP connection tracking security hole, some potential filesystem corrupting bugs and a bunch of other bugs that weren't likely to affect anybody.
And Dave Miller's zerocopy networking changes were merged in, which is pretty cool.
I've put up a Swarmcast mirror of the content here.
Swarmcast is BETA (it gots boogs) software that will soon be released under the GPL. The installation procedure sucks right now but the situation will quickly improve.
Swarmcast is peer-to-peer software that uses the bandwidth of the peers that are currently or recently downloaded a piece of content to help boost everyone elses download. So its basically like having a bunch of partial-horsepower mirrors.
If you're not into that "altruism" stuff then you can just ignore this post and hammer directly on kernel.org.
You can join the Swarmcast devel mailing list here. and let us know what you think.
TmpFS is an extension of shmfs able to swap off unused memory pages. This virtual filesystem *rocks* . Compiling new software in a TmpFS area is a lot faster than on a conventional filesystem. Since it was introduced, I can't live without it.
But why the hell is TmpFS only in -ac series and never in 'official' releases ? I will have to wait for 2.4.4-ac1 before upgrading.
TmpFS is now rock stable (it wasn't the case in the early times, I agree) . It's fast. It's easy to set up, and it's a very good enhancement to the Linux kernel.
When will TmpFS merged to the main tree ?
{{.sig}}
- mirrors organized by country
- UK
- US
In general, the form is www.country.kernel.org where country is the country's two letter code. For example, NZ is a link to New Zealand's mirror.Slashdot should maintain topic pages for each category with links like this, yes?
Too big to fail? Does that make me to small to succeed?
Of course, it all depends on what you're previously running - if you're running a previous 2.4 kernel from Linus.
Many file system corruption bugs were found by us during development of Red Hat Linux 7.1. The 2.4.2-2 kernel in Red Hat Linux 7.1 should be safe, any kernel released prior to that (our release, not 2.4.2 in general - our kernels has lots of fixes, some made after the official release of 2.4.3)
#!/bin/sh
/usr/src
$ 0.tar.gz
# Note: this is untested code written ad-hoc
# in a slashdot comment box. You decide
# how much you want to trust it.
#
# Usage: lspinstaller <version-of-kernel> <previous-version-of-kernel>
# Must run as root.
ARCH=i386
cd
wget ftp://ftp.kernel.org/pub/linux/kernel/v2.4/linux-
# make sure we only have a linux-oldversion dir
rm linux || mv linux linux-$1
tar xvfz linux-$0.tar.gz
mv linux linux-$0
ln -s linux-$0 linux
# recover our kernel configuration
cp linux-$1/.config linux-$0/.config
cd linux-$0
# answering "no" to new kernel features is not necessarily what you want, but it's a good place to start
yes n | make oldconfig
mv Makefile Makefile.configged
# I find it's always a good idea to tag kernels with your hostname so you can tell what they're meant for
sed s/EXTRAVERSION =/EXTRAVERSION = `hostname`/ Makefile
make dep clean bzImage modules modules_install
# if you use lilo, this automatically runs it.
# if you use grub, do it manually I guess.
installkernel $0-`hostname` arch/$ARCH/boot/bzImage System.map
echo Windo^H^H^H^H^HLinux has finished installing your new software. Press any key to reboot, but ctrl-alt-delete will probably be more effective.
--
It's rare that you're presented with a knob whose only two positions are Make History and Flee Your Glorious Destiny.
When this was posted, the bandwidth meter read 34.4 Mbps. Five minutes later it was at 40 Mbps. Five minutes after that it was at 50 Mbps. Now it is at 71.39 Mbps.
Can't you guys wait a while and download it from the mirrors?
Tarsnap: Online backups for the truly paranoid
Why exactly does using 71.39% cause a problem?
As usage tends towards 100%, speed tends towards zero and packet retransmission increases.
Tarsnap: Online backups for the truly paranoid
One misconception that most people have is that the BIOS POST routines actually test RAM. They don't.
The only thing BIOS routines do is check for the existance of RAM, and as a side benifit some failures are detected.
Other failures can't be found with the quick xor tests done in the BIOS because;
They are masked by a cache that is not defective.
Are heat/cold sensitive -- and at boot the system might be OK.
The failure of one bit only occurs after other bits are flipped.
Even a good memory test program like Memtest86 can't find all defects, and definately not on the first pass. As a reasonable sanity check, Memtest86 is highly recommended.
A firewall can not protect you from yourself. Turn off what you do not need. Do not use the firewall to do your work.
insert this into /usr/src/linux/include/linux/skbuff.h below the line /* Internal */ Each quotation is a new line (remove the quotes before you paste) "static inline atomic_t *skb_datarefp(struct sk_buff *skb)" "{" " return (atomic_t *)(skb->end);", "}".
from openprojects irc 'apt'
"don't fuck up. This is why we have... James M... say fuck enough... all the algo is pure shit and should be replaced (From tcp in IPv4 - makes you feel good, eh?)." The whole IP NAT for FTP thing is "eloquent like `fuck'." The irnet_irda module is made to "Give a kick in the ass of ppp_generic so that he sends us some data".
The people who put together te setup code for the i386 arch were curious to know "What lunatic came up with this shit?" The IA64 arch includes some "weird ass code needed for error injection". MIPS/cobalt setup is referred to as "Oh shit, this is so crappy ..."
SysIRIX has "2,191 lines of complete and utter shit coming up..." There's more of that for MIPS, though - "Big shit, we now may have two dirty primary cache lines for the same..." I don't even want to hear the end of that! In fact the MIPS/Irix people seem as if they just "Don't Care, a rats ass we couldn't give". That same file includes 10 iterations of "XXX AFS shit".
HyperSPARC has is "Verified, my ass..." The developer for a SPARC checksum code wants to merely "give up. I'm serious, I am going to kick the living shit..." The SPARC kernel process claims that somebody should "fuck me plenty". I agree :) At least they think SPARC64 is "One bad ass cpu". Then again "are two fold. Firstly, they cannot pair with jack shit,".
Everyone should "Remember: "Different name, same old buggy as shit hardware." Speaking of which, one of the SCSI drivers contains to lines of "shit". "... We don't want to fuck directly..." So we'll do it second-hand!
ASM has so much "More PROM shit. Probably has to do with VME RMW cycles???" They've been having to "having to fuck around with the syscall interface themselfes." Ultimately, everyone agrees that "If you don't see why, please stay the fuck away from my code."
/.ers only have a karma of 49...
REAL
SIG: HUP
- Alan Cox: more resyncs (ARM down, but more to go) - Alan Cox: more merging (S/390 down, ARM to go).
Oh my GOD!!! Alan Cox is being merged into the Kernel!!! They have his ARM merged in now, what next, WHAT NEXT?????
--Volrath50
The latest kernel still can't pass the Purity Test. You can see by running:
egrep -r -e "(\bfuck\b|\bass\b|\bshit\b)" /usr/src/linux
In fact, I found an increase in degree of profanity in the new release:
egrep -r -e "(\bfuck\b|\bass\b|\bshit\b)" /usr/src/linux-2.2.18 | wc -l ---> 48
egrep -r -e "(\bfuck\b|\bass\b|\bshit\b)" /usr/src/linux | wc -l ---> 52
That's too bad, my boss wouldn't like it.
Question is - did AMD drop their NDA silliness for sharing the bug workaround or did the kernel developers just hack it (by reading NDP twice)
I expect the latter but I'm psyched anyway!
As always - thank you kernel developers!
--
Top Most Bizarre/Disturbing Error Messages