1150243600 J * ag- ag@caladan.roxor.cx
1150243706 Q * nebuchadnezzar Ping timeout: 480 seconds
1150243913 J * itguru ~itguru@host86-141-48-66.range86-141.btcentralplus.com
1150243946 P * itguru Leaving
1150244023 J * itguru ~itguru@host86-141-48-66.range86-141.btcentralplus.com
1150244166 M * itguru I am likin the sound of the Vserver thingy
1150245273 J * shedi ~siggi@130.208.221.254
1150246401 J * blake- blake@ip24-250-20-40.ri.ri.cox.net
1150248653 N * sarnold sars
1150248950 J * DarthVader ~Aniken@203.177.212.165
1150249304 P * blake- 
1150250287 N * sars sarnold
1150250872 Q * mountie Remote host closed the connection
1150250885 J * mountie ~mountie@CPEdeaddeaddead-CM000a739acaa4.cpe.net.cable.rogers.com
1150251251 Q * sb Quit: 
1150251309 Q * DarthVader Quit: Leaving
1150251986 Q * s0undt3c1 Ping timeout: 480 seconds
1150252181 N * sarnold sars
1150257083 J * s0undt3ch ~s0undt3ch@bl7-244-104.dsl.telepac.pt
1150257182 J * fwl ~f_@83-215-237-1.seek.stat.salzburg-online.at
1150257243 Q * fwl Quit: 
1150261825 N * Guest285 otaku42
1150262281 J * s0undt3c1 ~s0undt3ch@bl7-242-26.dsl.telepac.pt
1150262480 Q * s0undt3ch Ping timeout: 480 seconds
1150262735 J * fwl ~f_@83-215-237-1.seek.stat.salzburg-online.at
1150264837 Q * fwl Quit: This computer has gone to sleep
1150265311 J * fwl ~f_@83-215-237-1.seek.stat.salzburg-online.at
1150266708 J * s0undt3ch ~s0undt3ch@bl7-240-252.dsl.telepac.pt
1150266800 Q * s0undt3c1 Ping timeout: 480 seconds
1150269294 J * dna ~naucki@dialer-167-28.kielnet.net
1150269380 J * ||Cobra|| ~cob@pc-csa01.science.uva.nl
1150269672 J * _coocoon_ ~coocoon@84.160.83.151
1150269679 M * _coocoon_ morning
1150269742 M * phedny mornin' _coocoon_ 
1150271018 J * yelifu ~hongdanst@202.38.114.129
1150273042 J * DarthVader ~Aniken@203.177.212.163
1150273207 Q * jkl Ping timeout: 480 seconds
1150273419 J * Viper0482 ~Viper0482@p54975BC2.dip.t-dialin.net
1150274465 J * Fushi Rawr@69-172-154-250.atlsfl.adelphia.net
1150276513 J * CoBoLt ~countcobo@84.196.135.76
1150276518 M * CoBoLt oioi
1150276524 M * CoBoLt how's everyone doing?
1150276685 J * jhaig ~jhaig@mantis.office.netline.net.uk
1150276708 M * CoBoLt I'm getting this error on alpha
1150276717 M * CoBoLt rpm-fake-resolver: vc_set_cflags(): No such process
1150276717 M * CoBoLt rpm-fake.so: failed to initialize communication with resolver
1150276723 M * CoBoLt anyone seen this before?
1150276835 M * jhaig I have set up nss_vserver so that I can log on to guests directly via ssh.  When I try 'ssh vserver-user@host' I can log in but I get an error "Could not chdir to home directory /etc/vservers/vserver/vdir/./home/user: Permission denied".  Any idea what may be wrong?
1150277188 M * jhaig Also, I can use 'ssh vserver-user@host' but 'ssh user@vserver' doesn't work (NB, I am actually specifying @host and @vserver as ip addresses).  http://linux-vserver.org/HowtoHostAuth says that I should be able to use the latter form.
1150277445 J * Milf ~Miranda@141.12.9.154
1150278360 Q * jhaig Remote host closed the connection
1150278372 Q * shedi Quit: Leaving
1150279089 Q * DarthVader Quit: Leaving
1150279175 J * jhaig ~jhaig@mantis.office.netline.net.uk
1150279699 N * Bertl_zZ Bertl
1150279703 M * Bertl morning folks!
1150279713 M * phedny Bertl: had a good sleep?
1150279714 M * _coocoon_ hello bertl
1150279725 M * Bertl phedny: short but good, tx :)
1150279729 M * Bertl hey _coocoon_!
1150279773 M * phedny I have this question about x86_64
1150279801 M * phedny as I read things, it is possible to run 32-bit binaries on a 64-bit kernel by using compat-libraries
1150279822 M * phedny but I wonder whether it is possible to run an entire 32-bit guest on a 64-bit kernel?
1150279837 M * Bertl well, yes, and it is done quite often
1150279854 M * Bertl only thing you need is the ia32 vompatibility/emulation in the kernel
1150279892 M * phedny and then I don't need these compat libraries?
1150279931 M * Bertl they ae already present in the guest (as normal ia32 libts)
1150279936 M * Bertl *libs even
1150279950 M * phedny point is, my computer died and I'm thinking about buying AMD64-based system and I'd like to transfer vservers from backup to it
1150279961 M * Bertl will work without issues
1150279966 M * phedny okay, nice :)
1150280043 M * daniel_hozac CoBoLt: that looks weird. does testme succeed?
1150282214 J * lilalinux ~plasma@dslb-084-058-228-103.pools.arcor-ip.net
1150282222 M * Bertl wb lilalinux!
1150282386 Q * Viper0482 Remote host closed the connection
1150282491 M * cehteh phedny: i made a recipe on the wiki for a 32 bit env on a 64 bit system
1150282587 M * cehteh echo linux_32bit > /etc/vservers/$NAME/personality
1150282587 M * cehteh echo i686 > /etc/vservers/$NAME/uts/machine
1150283006 Q * Nam Ping timeout: 480 seconds
1150283063 M * derjohn cehteh, is that really necessary? i do a export ARCH=i386 ; vserver build .... and there is a 32bit guest on a 64bit env. but beware to complile ia32 emulation in the 64 bit kernel! (else you will get a execve error ...
1150283069 M * derjohn phedny, too ...
1150283134 M * cehteh derjohn: i think thats mostly optional normal apps will work without it .. 
1150283155 M * derjohn what? the ia32 emulation? no, it's needed ...
1150283172 M * cehteh the things above
1150283193 M * cehteh and ia32 is not emulated on x86_64 ...
1150283321 J * Nam ~nam@70.78.64.62
1150283347 M * derjohn well I empirically tested what I need: [*] Kernel support for ELF binaries  <M> Kernel support for MISC binaries  [*] IA32 Emulation <M>   IA32 a.out support
1150283369 M * derjohn maybe it's only the misc binaries?
1150283390 M * Bertl ia32 emulation should be essential for having 32bit guest
1150283394 M * Bertl +s
1150283408 M * Bertl without that, the entire 32bit API is not present
1150283508 M * Bertl derjohn: regarding ubuntu in paris, if somebody pays the trip/accomodation I'm fine with it, but I guess it's easier if somebody from france shows up there ...
1150283601 M * derjohn Bertl, did you look at the 'sponsoring' page? interesting that I found  Paul Sladen (-> sladen ) on that list. Is he still in vserver business?
1150283783 M * Bertl sladen: ping? are you?
1150283934 M * derjohn Bertl, I think you would be the right one there. if ben collins ask me about kernel details i would have to talk about the weather instead.
1150283964 M * derjohn (not true for packaging or promotion .... ;))
1150284004 M * Bertl well, if they ask me about ubuntu stuff, I'd have to switch to the weather too :)
1150284020 M * derjohn Bertl, in africa there is always good weather ;)
1150284045 M * Bertl hmm, tell this to the people there ...
1150284074 J * Viper0482 ~Viper0482@p54975BC2.dip.t-dialin.net
1150284077 M * derjohn phedny, cehteh: http://linux-vserver.org/some_hints_from_john <-- '64 bit' FAQ added. would you mind to check?
1150284160 M * doener derjohn: so you got your bullshi^Wkernel running? ;)
1150284164 M * doener SCNR
1150284215 M * derjohn doener, i am afarid, yes, my bullshit serves a large companies webserver now ;)
1150284220 M * phedny derjohn: it's what has just been told in this channel and assuming that is correct, you text would be :)
1150284227 M * derjohn *company's
1150284248 M * phedny derjohn: but as soon as I've bought my AMD64 system, I'll verify it and let you know
1150284249 M * derjohn probably more soon if Bertl doesnt release rc23 .... :)
1150284263 M * derjohn phedny, *lol* ok ....
1150284513 J * jldalla ~jld@200.5.207.233
1150284532 P * jldalla 
1150284877 J * liquid3649_ ~Viper0482@p54975BC2.dip.t-dialin.net
1150284934 Q * Viper0482 Ping timeout: 481 seconds
1150285380 M * daniel_hozac derjohn: is the uts/machine necessary?
1150285396 M * daniel_hozac vserver x86-builder exec uname -a
1150285397 M * daniel_hozac Linux x86.builders.hozac.com 2.6.16-1.2123_FC5.vs2.0.2.0.rc21.1 #1 SMP Mon May 22 16:54:21 EDT 2006 i686 athlon i386 GNU/Linux
1150285414 M * doener not in general, but IIRC eg some Java stuff needs that
1150285438 M * daniel_hozac it seems to be set by the linux32 personality though.
1150285457 M * daniel_hozac (i have no uts/machine, just linux32 in personality)
1150285636 M * derjohn daniel_hozac, doener i took over those lines from cehteh. i dont use them myself, but it may make sense to set it ti i686 instead of i386? 
1150285674 M * daniel_hozac isn't i386 the hardware platform?
1150285692 M * daniel_hozac i get that on my regular i686 boxes as well.
1150285875 M * doener daniel_hozac: uname -mpi should tell...
1150285944 M * doener hm, well I cannot tell in which order the output is, except for "machine" being the first one, the others are unknown here...
1150285980 M * derjohn daniel_hozac, uhm, no idea. I only saw in Debian a libc6-i686 (optimized) ... but thats a compiler optimisation only, not a differnt $ARCH.  I think mandriva offers i586, that not an $ARCH, too ?
1150286055 M * Bertl strictly speaking the hardware platform is x86
1150286061 M * doener derjohn: hm, is libc6-i686 just optimized? IIRC it also provides (or used to) NPTL support while the plain libc6 does not
1150286089 M * Bertl but out of tradition, _most_ distros use i386, some i586 and a few other variants like k6/k7/i686
1150286100 M * Bertl +something
1150286189 M * derjohn doener, a standard libc has no native posix threads? well, I dont code c as you may remember, so i never really cared ; .. besides that it nice weather outside today! 
1150286246 M * Milf ... when you're not sitting in front of the AC exhaust shaft
1150286248 M * Milf  :)
1150286329 M * Milf can I once again ask about my localhost issue that I also posted to the list?
1150286338 M * doener derjohn: well, the debian libc6 package, not glibc in general...
1150286360 M * daniel_hozac are you sure?
1150286382 M * daniel_hozac i thought NPTL required tsc or some instruction only present on 586 or 686.
1150286446 M * doener daniel_hozac: thus the libc6-i686 package, but distros compiled for i686 will probably have it in their plain libc packages
1150286499 M * doener I just wanted to say that it is not like that "in general", but depends on your distro
1150286533 M * phedny but the output of uname should not be influenced by any distro specific stuff
1150286537 M * Bertl Milf: sure ...
1150286563 M * derjohn the nice thing with standards is that you can choose from a large set of of them :/
1150286584 M * derjohn Milf, dont ask to ask, just ask ;)
1150286602 M * Milf I can run the ported OX inside a vserver as long as I map lo into it.
1150286633 M * Milf Just mapping localhost to the guests IP in /etc/hosts inside the guest doesn't get me any positive result.
1150286638 M * derjohn Milf, OX != macos 10 ?
1150286646 M * derjohn :)
1150286653 Q * liquid3649_ Remote host closed the connection
1150286657 M * Milf OpenExchange. Commercial version. Will install on SLES9
1150286700 M * daniel_hozac have you found out why it doesn't work with the rewriting?
1150286733 M * doener Milf: did you try to disable hiding of network interfaces?
1150286746 M * Milf The part that doesn't work without loopback interface is written in java and I would have to spend a week finding and interpreting the source code.
1150286761 M * Milf doener: How do I find out if I am hiding any interfaces?
1150286762 M * daniel_hozac eww, Java.
1150286785 M * doener Milf: does ifconfig show lo? if not, hiding is enabled...
1150286809 M * Milf me shrugs its shoulders. It just wants to use the OX, whether it be written in Java, Intercal or Basic.
1150286812 M * derjohn Milf, as i still dont know the problem .... I think you cannot bind to 127.0.0.1 in a guest
1150286849 M * derjohn Milf, forgot to mention Eiffel ;)
1150286850 M * Milf I have a proprietary application written in Java that seems to want 127.0.0.1
1150286857 M * daniel_hozac Milf: also, have you tried the new CONFIG_VSERVER_REMAP_SADDR? IIRC you have to disable legacy to get it to show, but it might do the trick.
1150286857 A * Milf nods to derJohn
1150286901 M * derjohn Milf, so it wants to to bind 127.0.0.1 ... I assuem you set 'localhost' to your public ip in /etc/hosts ?
1150286905 J * Viper0482 ~Viper0482@p54975BC2.dip.t-dialin.net
1150286925 M * derjohn if not, grep -ril 127.0.0.1 /path/to/java/src/* ...
1150286956 M * Milf Ok, let's do this one by one. Doener: Will disabling interface hiding do anything different compared to setting up lo for the guest? 
1150286965 M * Milf derjohn: I tried that. No cigar.
1150286976 M * daniel_hozac binds to 127.0.0.1 are remapped to the first IP address of the guest.
1150286980 M * doener it will show all interfaces but still disable direct access to 127.0.0.1
1150286996 M * doener ie. you see more, but cannot do more
1150287012 M * Milf doener: sounds like another thing to try. How do I disable hiding?
1150287041 M * daniel_hozac vattribute --xid ... --bcap ~0 --flag ~hide_netif
1150287065 M * doener 2fast4me ;)
1150287073 M * Milf daniel_hozac: Do I do this while the guest is running?
1150287082 M * daniel_hozac yes.
1150287102 M * daniel_hozac if you have the guest stopped, echo ~hide_netif >> /etc/vservers/.../flags
1150287107 M * derjohn daniel_hozac, do you really _know_ that the binding to a mapped ip works? or do you just assume it? I i ever run accross an example where it doesnt work, i will report.
1150287128 M * daniel_hozac derjohn: well, yes.
1150287165 M * daniel_hozac BIND's control channel is bound to 127.0.0.1.
1150287181 M * doener I assume that OX either wants to see lo, or that it checks that its socket is really bound to 127.0.0.1...
1150287194 M * daniel_hozac the latter would be quite silly, IMHO.
1150287204 M * doener the former would be solved by ~hide_netif, the latter would need the lo patches to be solved in a safe manner
1150287220 M * Milf Hmmm, no cigar. Admin-Interface gives me 'invalid credentials' on login.
1150287223 M * derjohn doener, maybe Java has such a security layer (-> sandbox ...)
1150287232 M * Bertl apropos lo patches *G* maybe we should have a network discussion today?
1150287238 M * daniel_hozac sure.
1150287263 M * doener I'll be gone soon, back at about 20:00 CEST
1150287269 M * Milf I can see lo inside the vserver and I have localhost mapped to first ip of the guest in /etc/hosts
1150287289 M * Bertl daniel_hozac: so around 2000 is fine for you too?
1150287292 M * daniel_hozac Milf: have you tried strace?
1150287298 M * daniel_hozac yeah, 2000 works fine.
1150287305 M * Bertl okay, sounds like a plan ...
1150287314 M * Milf strace I tried, but I don't know what to strace.
1150287323 M * daniel_hozac Milf: the process that fails :)
1150287330 M * Milf tomcat?
1150287352 M * daniel_hozac if you say so.
1150287395 M * Milf The part that fails is a java applet running on a tomcat. Which I know about as much of as the dark side of the monn. Which makes it hard for me to debug.
1150287404 M * Milf s/monn/moon/
1150287426 M * doener invalid credentials... that sounds like a security check for dst or src address..
1150287444 M * Bertl Milf: what does localhost resolve to?
1150287461 M * Milf My only hint on what to go on is that it works if I mkdir interface/lo;echo 127.0.0.1>ip;touch nodev;echo 255.0.0.0>mask 
1150287484 M * Milf bertl: you mean by running ifconfig inside the guest?
1150287624 M * doener Milf: where exactly does the error message show up? and is that the complete error message?
1150287671 M * Milf The error message shows up on the web interface.
1150287680 M * Milf But hold on, I may have found another problem.
1150287688 M * phedny Milf: is there something in the Tomcat logfiles that may help?
1150287707 M * Milf phedny: I checked there. Nope. No cigar to be won.
1150287755 M * Milf bertl: According to 'host' "localhost.ipsi.fraunhofer.de has address 127.0.0.1"
1150287776 M * CoBoLt daniel_hozac, testme runs fine, i'll be back in about 30 minutes
1150287797 M * CoBoLt just did some shopping :)
1150287806 M * phedny Milf: what does "grep host /etc/nsswitch.conf" say?
1150287833 M * Milf hosts: files dns
1150287860 M * Milf I just had to create resolv.conf, as I made the image of the real server while having booted from Knoppix
1150288034 M * phedny can you post all localhost of 127.0.0.1 related lines from /etc/hosts ?
1150288310 M * Milf #141.12.26.118  miketest-int1.ipsi.fraunhofer.de        miketest-int1
1150288310 M * Milf 141.12.26.118   miketest-int1.ipsi.fraunhofer.de        miketest-int1   localhost
1150288310 M * Milf #127.0.0.1       localhost
1150288326 M * sladen derjohn / Bertl: hell
1150288333 M * sladen derjohn / Bertl: ...o
1150288361 M * derjohn sladen, ehlo ! you will be on the ubuntu summit ?
1150288437 M * sladen derjohn: I'm unsure yet.  I'm paying my own way if I get there...
1150288462 M * derjohn sladen, so the sponsoring application was refused?
1150288475 M * Bertl sladen: so maybe you want to speak for linux-vserver there?
1150288484 M * sladen derjohn: however the logistics are fairly easy since it's just at the other end of the Eurostar to me
1150288508 M * Bertl sladen: we might have to bring you up-to-date, but I guess that wouldn't be a problem :)
1150288548 M * sladen are they specically wanting somebody to talk about vserver then?  Or is this in relation to vserver packagin in Ubuntu?
1150288585 M * Bertl well, honestly I don't know, maybe derjohn should bounce you the email?
1150288610 M * derjohn sladen, we talked to mark about vserver and ubuntu on linuxtag. we would love to see it included in edgy.
1150288632 M * sladen derjohn: you mean having the kernel patch included by default?
1150288657 M * Bertl I don't see a good reason against it, as debian has xen/vserver hybrid kernels which work quite fine
1150288685 M * Bertl i.e. you can build/install all variations and combinations
1150288688 M * sladen derjohn: generally the Ubuntu kernel follows the mainline one except for additional drivers and the like or laptop fixes and the like that are going upstream
1150288710 M * Bertl sladen: well, they seem to include xen support though
1150288713 M * sladen Bertl: ah right, so having a separate additional linux-image-vserver ?
1150288728 M * Bertl sladen: precisely, and a xen-domu-linux-vserver
1150288756 M * sladen okay, so that's too more variations
1150288773 M * Bertl yup
1150288801 M * sladen two more
1150288857 M * sladen was there any pariticular aspect that caught Mark's eye?  That's the way to get it into Ubuntu
1150288863 M * derjohn sladen, yes, "vserver and xen - just good friends" :)
1150288889 M * Bertl sladen: I guess he liked the moreubuntu multi seat desktop (tm) :)
1150288897 M * derjohn sladen, we showed him a multi-seat Xorg, each xorg in one vserver ;)
1150288986 M * derjohn sladen, (suse is also very interested in that feature, but Milf and zeng didn't finish the howto yet)
1150288986 M * sladen one of the first 'jobs' that Canonical had come in was the HP 444 (1 PC, 4 video cards, 4 usb keyboards, 4 usb sound thingies)
1150289015 M * Milf The Howto is as done as it will get. Bugger Zeng a little so he doesn't forget to post it.
1150289046 M * sladen they still all need /dev/mem don't they
1150289046 M * sladen for X
1150289050 M * Milf sladen: what about 4 mice?
1150289057 A * Milf nods
1150289058 M * derjohn Milf, zeng is still at computex, so not at this pc now :)
1150289070 M * Milf derjohn: Hmmm, he told me other things.
1150289092 M * Milf derjohn: But he's busy catching up at work after his leave
1150289092 M * sladen the.  lag.  here.  is.  unbleliveable.
1150289096 M * derjohn sladen, Bertl made first plans to virtualize /dev/mem ....
1150289141 M * sladen ooh, that's really interesting.  like  ioperm()  for it?
1150289176 M * Bertl yup, basically based on assigning PCI ranges to guests
1150289181 M * derjohn sladen, how is the waether in .uk ? (read: I dont know anythings about it ;))
1150289332 M * sladen derjohn: alternating between a heatwave and tropical rain
1150289334 J * id ~id@p50811B0F.dip0.t-ipconnect.de
1150289340 M * id Hi #vserver
1150289343 M * id hello Bertl 
1150289368 M * Bertl welcome id!
1150289371 M * sladen Bertl: if you can see that separate from the context it would have wider-ranging uses for locking down /dev/mem
1150289393 M * sladen derjohn: where are you based?
1150289404 M * derjohn sladen, as20755 :)
1150289411 M * derjohn sladen, .de :)
1150289419 M * sladen derjohn: is your plan that you'd like me to go to UFK and pimp vserver
1150289428 Q * fwl Quit: This computer has gone to sleep
1150289429 M * sladen derjohn: and is there a spec already on the wiki?
1150289432 Q * jhaig Quit: Download Gaim: http://gaim.sourceforge.net/
1150289454 M * derjohn sladen, what is yoz current e-mail ?
1150289471 M * sladen derjohn: vserver@paul.sladen.org
1150289480 M * derjohn sladen, spec? what kind of spec do you expect?
1150289506 M * sladen Bertl: BTW, I wasn't expecting to see you in a suit at LinuxTag :)
1150289534 M * Bertl sladen: hmm, why no?
1150289536 M * Bertl *not
1150289544 M * derjohn sladen, you wont believe it, it's his everyday clothing :)
1150289551 M * sladen derjohn: https://launchpad.net/distros/ubuntu/+specs
1150289571 M * sladen Bertl: I guess, I'd just never imagined it.
1150289576 M * sladen Bertl: very respectable
1150289593 M * derjohn sladen, have you been on LT ?
1150289593 M * CoBoLt bertl, have you read my question on the alpha release?
1150289634 M * Bertl CoBoLt: hmm, obviously not .. 
1150289639 M * derjohn sladen, the plan was to talk to ben collins about the integration of the kernel patch.
1150289641 M * CoBoLt :)
1150289655 J * cdrx ~legoater@cap31-3-82-227-199-249.fbx.proxad.net
1150289664 M * CoBoLt when I do a build of a vserver, 
1150289665 M * CoBoLt i get
1150289675 M * Bertl welcome cdrx!
1150289677 M * derjohn sladen, util-vserver should be a piece of cake (at least the etch packages a good!)
1150289692 M * cdrx hi all !
1150289702 M * sladen derjohn: yes, the utilties can be a straight sync (import) from Debian
1150289705 M * CoBoLt rpm-fake-resolver: vc_set_cflags(): No such process
1150289705 M * CoBoLt rpm-fake.so: failed to initialize communication with resolver
1150289726 M * CoBoLt now this is an alpha running centOS4.3
1150289734 M * CoBoLt testme gives me succeed on all parts
1150289748 M * sladen derjohn: and if you add the moreubuntu tools as a seperate package, depending on util-vserver then they're built in layers
1150289779 M * CoBoLt btw, LinuxTag, would that be interesting for now german speaking people?
1150289794 M * CoBoLt and second would it be interesting for non geeks too? As in my girlfriend?
1150289801 M * CoBoLt *non-german
1150289843 M * Milf Hmmm, first question: a definite yes; a lot of talks were in english
1150289861 M * Bertl CoBoLt: well, guess it was interesting for non german speakers, and I guess it was less interesting for non computer folks :)
1150289869 M * CoBoLt hehe
1150289873 M * CoBoLt she's computer minded
1150289878 M * CoBoLt but not yet linux minded :)
1150289881 M * Bertl CoBoLt: testme.sh works fine? could you upload that to paste.linux-vserver.org ?
1150289882 M * CoBoLt mind , the not yet :p
1150289887 M * Milf Second question: Ask your girlfriend if she likes fluffy penguins, but if you mostly go to booths and talk geek stuff, your GF will probably get bored like you do when she's trying on shoes :)
1150289928 M * Milf That's the third thing: LinuxTag was beginning of May. The next one is next year.
1150289930 M * Bertl Milf: maybe CoBoLt is a shoe fetishist?
1150289936 M * CoBoLt Bertl , no thank you
1150289938 M * CoBoLt :p
1150289942 M * sladen CoBoLt: LinuxTag is *very* interesting for German people.  It's the only conference I know of where they don't speak English :)
1150289958 M * Wonka lol, Milf 
1150290008 M * CoBoLt http://paste.linux-vserver.org/86
1150290013 M * CoBoLt bertl, that was for you :)
1150290039 M * CoBoLt for else I might be interested in coming over
1150290082 M * CoBoLt but I'll have to take holiday and oc my gf want a part of me :p
1150290101 M * CoBoLt sladen, are all the expo's in german?
1150290115 M * Bertl CoBoLt: I see glibc, but you compiled with diet, which version btw?
1150290116 M * CoBoLt of in general can we take in account that most speak some english
1150290122 M * CoBoLt hang on
1150290135 M * CoBoLt I used daniels srpms for fedora
1150290141 M * CoBoLt as the same for his util-vservers
1150290158 M * CoBoLt recompiled them on the alpha
1150290171 M * Bertl CoBoLt: so it is dietlibc 0.29 or so?
1150290179 M * CoBoLt 0.28-1
1150290285 M * sladen CoBoLt: only LinuxTag
1150290314 M * CoBoLt nono that was not what I meant sladen, I meant the general booths over there :)
1150290315 M * sladen CoBoLt: daniel stone?  (Mr. xorg?)
1150290336 M * Bertl CoBoLt: basically all germans there will speak english too
1150290341 M * sladen CoBoLt: yes, virtually all of them were
1150290341 M * CoBoLt as you guys speak english that should not be a rpoblem :)
1150290344 M * CoBoLt great
1150290357 M * CoBoLt we have fosdem overhere
1150290365 M * CoBoLt so that like my annual linux meeting
1150290371 M * CoBoLt but germany isn't that far :)
1150290383 M * CoBoLt and there is something like trains, cars airplanes :p
1150290444 M * sladen derjohn: recieved
1150290454 M * CoBoLt Bertl any ideas on what goes wrong?
1150290469 M * CoBoLt sladen, what do you mean with mr Xoorg?
1150290470 M * Bertl CoBoLt: you should talk with Aiken once he is around, he is using an alpha too, I have one available too, but this one runs debian, and the guest building work fine there
1150290481 M * CoBoLt hmmm
1150290488 M * CoBoLt I'll check with running a legacy build
1150290492 M * CoBoLt maybe that works
1150290504 M * CoBoLt do you know which distro Aiken uses?
1150290514 M * Bertl CoBoLt: you might try installing the tools from sources to eliminate the remote possibility of rpm issues
1150290531 J * tmyneii ~localhost@161.53.107.100
1150290541 M * CoBoLt okay
1150290559 M * CoBoLt could I just take the sources from the srpms or do you think it is better to run the latest versions?
1150290574 M * CoBoLt and what would you prefer, the stable versions or not?
1150290795 M * Bertl welcome tmyneii!
1150290831 M * Bertl CoBoLt: I'd try latest dietlibc with 0.30.210
1150290837 M * Bertl CoBoLt: kernel should be fine
1150290843 M * CoBoLt ok
1150290856 M * CoBoLt I am building dietlibc 0.29 now
1150290917 M * CoBoLt damn an alpha compiles fast :)
1150290941 M * Bertl hmm, depends on the cpu :)
1150291026 Q * hallyn Quit: leaving
1150291077 Q * tmyneii Quit:  HydraIRC -> http://www.hydrairc.com <- IRC for those that like to be different
1150291140 M * Milf Ok, back to the localhost problem:
1150291154 M * Milf unhiding the interfaces didn't help
1150291220 M * CoBoLt 533 miata :)
1150291265 M * Bertl Milf: could you just change the 'localhost' to the first guest ip?
1150291278 M * Bertl (as it is suggested everywhere)
1150291282 M * Milf where? in /etc/hosts?
1150291294 M * Bertl yep, just add a line there with <ip> localhost
1150291316 M * Milf Like this: 141.12.26.118  miketest-int1.ipsi.fraunhofer.de        miketest-int1   localhost
1150291322 M * Milf Tried it, no cigar.
1150291336 M * Bertl what does the strace/error show in this case?
1150291350 M * Milf That was the other problem: What to strace?
1150291362 M * Milf It's a java applet running on tomcat, displaying a web page.
1150291388 M * Bertl okay, so what about the error?
1150291417 M * Milf The error is the  string 'invalid credentials' displayed on the login page that doesn't let me into the admin app
1150291448 M * Bertl I thought there was something like interface not found or so?
1150291479 M * Milf I wish there was.
1150291503 M * Bertl probably I mixed that up now, but why the hell are you trying to unhide lo then?
1150291529 M * Milf 'cause doener said to try that approach.
1150291535 M * Bertl what you actually want to try is to disable the localhost rewriting (as available in rc22)
1150291550 M * sid3windr cool
1150291567 M * Milf Argl, you mean to tell me recompile my kernel?
1150291572 M * Bertl which will give you 127.0.0.1 for the connection, which in turn might work around the brokenness in that applet
1150291600 M * Bertl JFYI: any application which 'assumes' that localhost == 127.0.0.1 is broken
1150291653 M * Milf Yes, I know that. Try telling that to the openexchange people. I couldn't even find their mailing list, even though it's based on an open source project.
1150291674 A * Milf sings 'No need no subset'
1150291681 M * Bertl should there be the 'source code' somewhere?
1150291703 M * Milf If I look hard enough I probably will find the source code.
1150291705 J * pagano ~pagano@131.154.5.20
1150291737 M * Milf So, next thing to try is patch up my kernel to rc22 and then do what?
1150291764 M * Bertl well, if you _find_ the source, I'd grep for a hardcoded 127.0.0.1 and replace that
1150291774 M * Bertl then send a patch to the openexchange folks :)
1150291797 M * Milf ... and wait for the next commercial release :)
1150291826 M * Bertl well, for proprietary applications, you might try disabling the address rewrite stuff
1150291839 M * Bertl (it's a kernel option and we need some testing/feedback anyways :)
1150291868 M * Milf Ok, what will that do? Can I do that with my rc20 kernel?
1150291906 M * Bertl if you apply certain patches, yes :)
1150291928 A * Milf shivers with anticipation at building yet another kernel.
1150291949 M * CoBoLt Bertl custom build gives me the same error
1150291953 M * CoBoLt I'll try legacy now
1150291978 M * Bertl very interesting ... seems like centos is kind of different nowadays
1150291979 M * CoBoLt this works which I expected as I am not using rpm atm
1150291985 M * CoBoLt as in?
1150291992 M * Bertl ah, which reminds me, could you try with a sane gcc too?
1150292003 M * CoBoLt without dietlibc?
1150292006 M * Bertl maybe force downgrade to 3.3.something
1150292010 M * Milf Ok, Bertl, I'll give it a try. Gimme them patches and I'll build me an alternate kernel.
1150292010 M * CoBoLt hmmm
1150292021 M * Bertl Milf: better get rc22
1150292052 M * doener Milf: why shiver? should only take a few seconds..
1150292060 M * CoBoLt Bertl why downgrade to 3.3?
1150292081 M * Bertl CoBoLt: just to eliminate the possibility of a gcc/diet issue
1150292084 M * CoBoLt I ran vserver on x86 with slackware with 3.4 as gcc
1150292097 M * CoBoLt what would happen if I uninstalled dietlibc?
1150292104 M * CoBoLt how many issues would I run into?
1150292107 M * Bertl yes, but you don't know what centos did put into the gcc
1150292112 M * Milf doener: remember my problems with my last kernel build and the issues with initrd?
1150292112 M * CoBoLt nope
1150292167 M * doener Milf: you know how to do it now, right? ;)
1150292260 M * Milf doener: Yes, but I get these flashbacks :)
1150292391 M * doener btw I just got the OX source and there appears 127.0.0.1 in ldap_addressbooks.xml in conf/groupware/ and conf/webmail/ ... and in their examples, ldap.conf in PREFIX/etc/{groupware,webmail}/ also has 127.0.0.1 in it
1150292407 M * doener maybe you also have 127.0.0.1 in these configuration files in your install
1150292447 M * Milf Hmmm, indeed I haven't grepd for 127.0.0.1 in the whole server, like I did with the originla IP of the server that provided the image
1150292460 M * Bertl doener: uh-oh, you mean a problem between keyboard and chair?
1150292529 M * doener if you like to call it like that... I'd prefer to say "stupid default config combined with unexpected security checks" (if my assumptions hold)
1150292530 M * Milf bertl: Not nice. It would be a problem between user and uaotmatic setup :)
1150292542 M * doener at least the config could use "localhost" instead of "127.0.0.1"
1150292548 A * Milf likes doener's explanation much better :)
1150292564 M * Bertl hehe, I'm fine with that too :)
1150292568 M * Milf or just put the guest's ip in there
1150292585 J * fwl ~f_@83-215-237-1.seek.stat.salzburg-online.at
1150292627 M * Milf and until that find is done, I'll delete some users :)
1150292699 M * CoBoLt guys didn't we have a legacy to new config converter?
1150292699 M * Bertl wb fwl!
1150292710 M * CoBoLt got it :)
1150292712 M * CoBoLt me blind
1150292712 M * CoBoLt :p
1150292744 M * Milf cobolt: post the URL please. I lost my bookmark to that one it seems
1150292757 M * CoBoLt http://linux-vserver.org/Legacy-To-Newstyle-Config
1150292758 M * CoBoLt :p
1150292766 M * CoBoLt couldn't be easier :D
1150292772 A * CoBoLt definately blind
1150292779 M * Milf thanx
1150293473 M * Milf Ok, greped the whole tree for 127.0.0.1 and exchanged all the relevant findings with guests IP. Doesn't help :(
1150293542 Q * brc Ping timeout: 480 seconds
1150293801 J * orionpanda_bbl orionpanda@netblock-66-245-252-180.dslextreme.com
1150293880 M * orionpanda_bbl bertl: I've got delta-ocfs2 running on two nodes (with 2x aoe shared disks). Thanks for the patch!
1150294358 M * Bertl orionpanda_bbl: great, there will be proper support (for testing) soon
1150294376 N * orionpanda_bbl orionpanda
1150294417 M * harry hahaaaaaaaa... tomorrow... /me will make a new grsec + vserver patch!!!!
1150294421 M * orionpanda I don't know if you've run ocfs2 over aoe yet. However, there are some performance issues. Namely, copying over OCFS2 is verrrry sloooow.
1150294427 A * harry has definitive hardware :)
1150294497 M * orionpanda For instance: I can copy a 1gig vserver from one aoe device to another in ~4min (no ocfs). Over OCFS2, I had to kill the rsync after 1 hour.
1150294506 M * Bertl orionpanda: well, personally I would not start with aoe, but a shared disk system
1150294534 M * Bertl but the ocfs2 folks ar very cooperative, you might ask regarding those obvious issues
1150294541 M * orionpanda yes, you're right. I probably jumped the gun. I got excited when I saw the patch.
1150294616 M * id cya laters
1150294618 Q * id Quit: Leaving
1150294786 M * orionpanda Question: Are there any limitations to running vservers from an ext3/resier loopback filesystem? Does all of the functionality (barrier flag, cpu limits, dentry limits, etc) work?
1150294813 M * Bertl sure, when the filesystem is there, the features are there too 
1150294859 M * orionpanda excellent. just wanted to be sure.
1150295001 M * CoBoLt Bertl what is the difference between copy and legacy build?
1150295005 M * CoBoLt except for the config scheme?
1150295030 M * Bertl not much I guess, but you have to ask the tool folks about details
1150295046 M * CoBoLt who is doing that?
1150295063 M * Bertl personally I'm using the 'normal' build functionality i.e. distro and skeleton
1150295086 M * Bertl Enrico and recenttly daniel_hozac :)
1150295098 M * CoBoLt well the problem is that rpm / yum don't work
1150295103 M * CoBoLt never used skeleton before
1150295116 M * CoBoLt so would use legacy
1150295123 M * CoBoLt but don't have legacy API in my kernel :)
1150295139 M * Bertl well, it's going away anyways ...
1150295165 M * Bertl IMHO you should work with enrico and/or daniel_hozac to figure _what_ fails in the resolver
1150295184 M * CoBoLt I will
1150295188 M * CoBoLt but daniel was off atm
1150295189 M * Bertl maybe a simple strace might be sufficient to narrow it down
1150295196 M * CoBoLt hmmm
1150295199 M * CoBoLt haven't tried that yet
1150295210 M * Bertl strace -fF -o my.trace ...
1150295294 Q * yelifu Quit: 
1150295411 M * CoBoLt Bertl this is further down the road to go
1150295427 M * CoBoLt remember osf_syscall which gave me errors during compilation of my kernel?
1150295439 M * CoBoLt osf_syscall(0x3, 0x11fb91890, 0, 0, 0, 0x11fb91850) = -1 ERRNO_352 (Unknown error 352)
1150295440 M * CoBoLt osf_syscall(0x2, 0x11fb91560, 0x11fb915f0, 0, 0x20000093790, 0x11fb91850) = 54
1150295445 M * CoBoLt see the previous errors?
1150295461 M * Bertl hmm, who is doing thos syscalls?
1150295465 M * Bertl *those
1150295511 M * CoBoLt vserver build does 
1150295523 M * Bertl then something is definitely wrong
1150295538 M * CoBoLt I'll output my strace see what I can do ok?
1150295562 M * Bertl upload the stuff on paste.linux-vserver.org
1150295577 M * Bertl btw, what gcc is currently used for your tool builds?
1150295587 M * CoBoLt regular centOS 4.3
1150295590 M * CoBoLt so still 3.4
1150295605 M * Bertl 3.4 really?
1150295606 M * CoBoLt because I _really_ don't want to go a long road away from regular distro
1150295612 M * CoBoLt 3.4.5 I believe
1150295631 M * Bertl isn't there a 3.3 gcc package for centos available too?
1150295639 M * Bertl I doubt that they are 3.4 only
1150295644 M * CoBoLt not for 4.3 as I am aware
1150295704 J * kir ~kir@swsoft-mipt-nat.sw.ru
1150295720 M * CoBoLt Bertl I now have regular gcc (which is 3.4.5 in packages) and there is also gcc4
1150295724 M * CoBoLt but that's it
1150295732 M * CoBoLt do you want the whole trace?
1150295735 M * CoBoLt as it is huge...
1150295763 M * Bertl yes please
1150295768 M * CoBoLt all in pastebin?
1150295780 M * Bertl or if you have some other way to uplaod it, then there
1150295791 M * CoBoLt i'll upload it :)
1150295835 M * CoBoLt http://www.c2root.be/my.trace
1150295838 M * CoBoLt there you go
1150295849 M * Bertl tx
1150295864 Q * ||Cobra|| Remote host closed the connection
1150295965 M * CoBoLt told you is was long :)
1150295974 M * CoBoLt ok
1150295975 M * CoBoLt back in a few
1150295983 M * Bertl k,cya
1150295986 M * CoBoLt need to put new windowwasher on my car :)
1150295989 M * CoBoLt gimme 10 :p
1150295994 M * Bertl np
1150296312 M * CoBoLt back
1150296314 M * CoBoLt that was easy :)
1150296374 M * Bertl okay, I still would like you to try with a different gcc, do you think you could recompile and use an srpm for that?
1150296409 Q * Viper0482 Remote host closed the connection
1150296421 M * CoBoLt for upgrading gcc?
1150296425 M * CoBoLt or downgrading?
1150296428 M * Bertl actually for downgrading it
1150296435 M * CoBoLt would I need to recompile my kernel too :s
1150296443 M * Bertl no, leave it as is
1150296447 M * CoBoLt oef
1150296453 M * CoBoLt I can try
1150296510 M * CoBoLt I'll use the source of centOS 3
1150296532 M * Bertl try with this one, it should work fine:
1150296602 M * CoBoLt which one?
1150296606 M * Bertl sec
1150296610 M * CoBoLt I am downloading 3.2.3 atm
1150296681 M * Bertl http://vserver.13thfloor.at/Stuff/Cross/MDK/gcc-cross-3.3.6-10mdk.src.rpm
1150296745 M * CoBoLt downloading 
1150296751 M * CoBoLt although these are mandrake packages
1150296758 M * CoBoLt so it might need MDK based deps...
1150296766 M * CoBoLt btw 13thfloor from the movies?
1150296768 M * CoBoLt great one
1150296771 M * CoBoLt have it over here
1150296773 M * Bertl yep, indeed :)
1150296779 M * CoBoLt f##k matrix
1150296781 M * CoBoLt this is better
1150296786 M * CoBoLt the ending is great
1150296786 M * Bertl well, try to build it with --nodeps 
1150296790 M * CoBoLt I will
1150296817 M * CoBoLt the way they made "the end of the world" was just fascination
1150296819 M * CoBoLt great image
1150296837 M * CoBoLt saw it a long time ago
1150296845 M * CoBoLt nut only recently bought the dvd
1150296849 M * CoBoLt 10 euro  :)
1150296852 M * CoBoLt couldn't resist
1150296876 M * CoBoLt I appears that gcc might build fine
1150296963 M * CoBoLt nope give me an error
1150297001 M * CoBoLt grmbl
1150297003 M * Bertl which one?
1150297013 M * CoBoLt %{CROSS_ARCH} error
1150297020 M * CoBoLt how do I set the arch type?
1150297023 M * Bertl ah, k, forgot about that
1150297030 M * CoBoLt Invalid configuration `%{CROSS_ARCH}-linux': machine `%{CROSS_ARCH}' not recognized
1150297033 M * CoBoLt uhu
1150297036 M * CoBoLt same here :p
1150297046 M * Bertl maybe try this one instead: http://vserver.13thfloor.at/Stuff/gcc-3.3.6-1mdk.src.rpm
1150297057 M * CoBoLt different one?
1150297066 M * Bertl it's probably better in your case anyway
1150297078 M * Bertl the former one is my cross compiling source package
1150297085 M * CoBoLt aha
1150297094 M * Bertl but as you already figured, you need to specify the target arch :)
1150297121 M * CoBoLt uhu
1150297123 M * CoBoLt but how?
1150297152 M * Bertl sec, btw, the latter one ist still uploading
1150297159 M * CoBoLt woeps
1150297163 M * CoBoLt was already donloading :p
1150297262 M * Bertl you can do --define="CROSS_ARCH alpha"
1150297286 M * CoBoLt wokee
1150297286 M * Bertl rpm -b* should understand that
1150297301 M * CoBoLt rpmbuild on centos
1150297312 M * CoBoLt normally I would do rpmbuild --rebuild package
1150297319 M * Bertl rpm -b* usually calls rpmbuild :)
1150297322 M * CoBoLt lets see :)
1150297341 M * Bertl so, yes, I assume rpmbuild will understand the --define :)
1150297346 M * CoBoLt hehe
1150297375 J * Viper0482 ~Viper0482@p54975BC2.dip.t-dialin.net
1150297610 M * CoBoLt nope doesn't work
1150297617 M * CoBoLt /var/tmp/rpm-tmp.61707: line 42: fg: no job control
1150297618 M * CoBoLt :s
1150297641 M * Bertl gee, such an old rpm system but a new compiler :)
1150297665 M * Bertl you can easily fix that, but I'd try the other rpm first (which was uploaded by now)
1150297711 M * CoBoLt lol
1150297843 M * CoBoLt are you sure the upload is finished?
1150297853 M * Bertl well, it says so here
1150297864 M * CoBoLt ok
1150297869 M * CoBoLt was giving me an error
1150297884 M * CoBoLt now If this doesn't work, I'll try the centOS 3 srpm
1150297897 M * Bertl 4bf2de9bbb812198ad07aed5512dd91c  /usr/src/RPM/SRPMS/gcc-3.3.6-1mdk.src.rpm
1150297916 M * Bertl I'd prefer a gcc3.3.something
1150297922 M * CoBoLt uhu
1150297928 M * Bertl IIRC, the older centos had a 3.2.x
1150297938 M * CoBoLt 3.2.3-54 version
1150297942 M * CoBoLt but remember it is redhat
1150297956 M * Bertl maybe a redhat gcc3.3.x would work?
1150297963 M * CoBoLt so this means that it would be 3.3.x but they just patched it to keep version dependencies
1150297974 M * CoBoLt they always do that with EL versions 
1150297982 M * Bertl ah, maybe, you can try ...
1150298001 M * CoBoLt lol doesn't work as it wan an mandrake-release :p
1150298022 M * Bertl you can probably fix that up easily if you like to 
1150298026 M * CoBoLt just did 
1150298026 M * CoBoLt :)
1150298030 M * CoBoLt ln -s :)
1150298101 M * CoBoLt I do hope this is something fixable in the utils...
1150298118 M * CoBoLt so I can stay with the mainstream versions of centOS
1150298138 M * Bertl well, once we know _what_ it is, we might be able to fix it, unless it's a bug in gcc :)
1150298159 M * CoBoLt hehe
1150298165 M * CoBoLt hey I didn't use 2.95 :p
1150298222 M * CoBoLt lol
1150298226 M * CoBoLt dies on the same error :p
1150298243 M * Bertl okay, you know how to build an rpm from the spec?
1150298251 M * CoBoLt rpmbuild -bb 
1150298266 M * Bertl okay, just do rpm -i gcc-3.3.6-1mdk.src.rpm
1150298283 M * Bertl edit the spec file and look for '% ' commands in the file section
1150298296 M * CoBoLt no was the forground error
1150298300 M * CoBoLt not the arch error
1150298306 M * CoBoLt that was fixed with --define
1150298333 M * Bertl yes, I though of %arch or something like this
1150298350 M * Bertl that's something which requires newer rpm
1150298356 M * CoBoLt nope no job control thing :)
1150298363 M * CoBoLt is latest rpm provided by redhat
1150298375 Q * itguru Quit: Leaving
1150298375 M * Bertl which is often caused by rpm commands which are unknown to rpm
1150298408 M * Bertl but let's try with the older centos gcc
1150298486 M * daniel_hozac strace works over context creations/enters on alpha?
1150298535 M * Bertl hmm, do we have context stuff here?
1150298543 M * daniel_hozac we certainly should.
1150298551 M * Bertl for building a guest?
1150298557 M * daniel_hozac yeah.
1150298562 M * Bertl aha, hmm?
1150298566 M * daniel_hozac rpm-fake-resolver is meant to be inside the guest.
1150298586 M * Bertl yeah, but I thought without the context stuff
1150298607 M * CoBoLt hi daniel
1150298630 M * CoBoLt I have upload my strace 
1150298636 M * CoBoLt www.c2root.be/my.trace
1150298698 M * CoBoLt this only occurs when I try o use yum/rpm to install a guest os
1150298733 M * CoBoLt legacy build works
1150298823 M * CoBoLt second: -m copy say that there is no vserver-build.copy (which is correct)
1150298916 Q * Milf Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org
1150298963 M * daniel_hozac indeed... 
1150299005 M * CoBoLt daniel_hozac, do you think this has something to do with centOS?
1150299009 M * CoBoLt of within vserver?
1150299039 M * daniel_hozac the rpm-fake stuff?
1150299041 M * CoBoLt uhu
1150299045 M * CoBoLt that is my whole issue
1150299056 M * CoBoLt that is why I can't install guests with yum
1150299083 M * daniel_hozac i'm more inclined to think it's the alpha part that's ruining it.
1150299102 M * Bertl I wonder where those many, many osf syscalls come from
1150299115 M * CoBoLt damn
1150299116 M * daniel_hozac me too. i thought that was basically unused.
1150299123 M * CoBoLt :s :'(
1150299133 M * Bertl yes, but something does call it repeatedly ...
1150299229 J * yarihm ~yarihm@vpn-global-dhcp1-32.ethz.ch
1150299241 M * daniel_hozac CoBoLt: on your legacy guest, does vrpm work? (i guess you might need vserver ... pkgmgmt internalize first)
1150299266 M * CoBoLt daniel_hozac I am now creating one
1150299269 M * CoBoLt so might take a while
1150299280 M * CoBoLt did already a million things and same amount reinstalls
1150299289 M * CoBoLt so creating a clean vserver
1150299367 M * daniel_hozac you should be able to yum a new guest manually, using something like yum --installroot=/vservers/blah install glibc filesystem coreutils ...
1150299449 M * CoBoLt uhu
1150299459 M * CoBoLt take all the packages in the minimal OS install?
1150299469 J * zkbrsnie ~zkbrsnie@83-64-146-226.klosterneuburg.xdsl-line.inode.at
1150299475 M * CoBoLt but I am still tackling all the configuration thingies :p
1150299483 M * daniel_hozac i guess you could do that... i'm more of a minimalistic guests type of person :)
1150299484 M * Bertl okay, first, when I do your command on my alpha I get this one
1150299495 M * CoBoLt because the flower page is ... destracting :p
1150299504 M * Bertl /usr/lib/util-vserver/vserver-build: line 206: -n: command not found
1150299514 M * daniel_hozac Bertl: shame on you, ancient utils :P
1150299516 M * Bertl which might be related to the fact that it is 0.30.209
1150299527 M * CoBoLt it 0.30.210 overhere
1150299527 M * Bertl so I will update right now :)
1150299531 M * CoBoLt dietlibc 0.29
1150299533 M * CoBoLt hehe
1150299536 J * vasko ~vasko@unreal.rainside.sk
1150299547 M * Bertl daniel_hozac: do we have a tar.gz with your patches too?
1150299550 M * daniel_hozac that means something in the build failed though.
1150299553 M * Bertl welcome vasko!
1150299557 M * vasko hi :)
1150299561 M * daniel_hozac no, not that i'm aware of.
1150299571 M * daniel_hozac i guess i could make one.
1150299572 M * Bertl okay, should this work with mainline?
1150299575 M * vasko can i ask a question?
1150299582 M * daniel_hozac yeah, it should.
1150299583 M * Bertl vasko: you already did!
1150299596 M * vasko heh, well...
1150299606 M * Bertl vasko: want to try with another?
1150299617 M * vasko is there a way to  to prevent host to access loopback?
1150299635 M * Bertl hum, loopback as in network lo or as in loop device?
1150299651 M * vasko as network lo
1150299670 M * Bertl no, not really, as the host has all the networking
1150299684 M * Bertl why would you want to block lo there?
1150299694 M * Bertl (just curious)
1150299701 M * vasko i would like to have my guest unable to connect to any (maybe accidently run) service on 127...
1150299729 M * Bertl what about a simple iptables rule?
1150299736 M * daniel_hozac are guests able to do that now?
1150299742 M * daniel_hozac shouldn't the rewriting catch that one?
1150299751 M * Bertl daniel_hozac: yes, it should
1150299770 M * vasko i've expermented a bit with iptables but was unable to do much with lo
1150299784 M * vasko sec plz
1150299789 M * Bertl np
1150299791 N * otaku42 otaku42_away
1150299942 M * vasko hmmm, sorry for the noise, i must have been overlooking something in last try, it is actually possible to do it with iptables :)
1150299981 M * Bertl good to hear .. np, feel free to hang around and/or ask more questions ...
1150300087 M * CoBoLt daniel_hozac, I'll need to go onward first with conversion from legacy to new style
1150300096 M * CoBoLt I'll let you know wether it works or not
1150300097 M * CoBoLt ok?
1150300132 M * vasko Bertl: atm everything works fine for me, but i'll stay anyway
1150300157 M * Bertl even better, what arch/version do you use?
1150300173 M * vasko 2.6.16.12-vs2.0.2-rc18
1150300199 M * vasko and 2.6.16.16-vs2.0.2-rc20 on amd64
1150300221 M * Bertl with 32bit guests on x86_64?
1150300238 M * vasko yes
1150300322 Q * lilo2 Read error: Connection reset by peer
1150300425 M * vasko actually both 32bit and 64bit guests
1150300443 J * lilo2 ~0710AAD4@tor-irc.dnsbl.oftc.net
1150300653 M * Bertl ah, nice to hear that somebody is actually using 64bit guests
1150300666 M * Bertl wb lilo2!
1150300919 M * vasko 32bit ones i do use in production also, even with oracle, the 64bit is under testing, but so far there was no single problem
1150301070 M * daniel_hozac i'm using 64-bit guests too ;)
1150301082 M * Bertl daniel_hozac: why didn't you tell me :)
1150301130 M * daniel_hozac i haven't noticed any problems either.
1150301275 J * bonbons ~bonbons@83.222.39.166
1150301359 M * Bertl welcome bonbons!
1150301490 M * bonbons Hi Bertl!
1150301559 M * bonbons Bertl: is there a major difference between some ifconfig version (that would cause some addresses to show up at some times, some to not show up)?
1150301573 M * Bertl bonbons: you might be interested in the scheduled network talk @ 2000
1150301617 M * Bertl bonbons: it should not happen, but I remember some older ifconfig (debian?) doing strange things ...
1150301669 M * bonbons ok, that may be useful information! I have one IPv6 tester whose IPv6 does not show up in the guest with ifconfig, but does with ip
1150301752 M * bonbons Bertl: what's the subject of conversation @ 20:00 (timezone? UTC+2 or UTC)
1150301808 M * Bertl BUT (i.e. CEST atm :)
1150301953 M * bonbons would be easier if it was UTC ;)
1150302001 M * Bertl topic is some general 'future' network steps ...
1150302134 M * bonbons that's certainly interesting, should be present
1150302542 M * Bertl daniel_hozac: okay now I get:
1150302545 M * Bertl ERROR: Can not find configuration for the distribution 'centos4';
1150302574 M * daniel_hozac when doing vserver ... build -m yum?
1150302593 M * Bertl obviously lost a '2' there
1150302602 M * Bertl # vserver pollux build -m yum -- -d centos42
1150302602 M * Bertl mount: mount point /etc/rpm does not exist
1150302617 M * daniel_hozac do you have rpm installed?
1150302618 M * Bertl no idea what this one means though ...
1150302631 M * Bertl ah, no probably not on debian :)
1150302668 M * CoBoLt Bertl CentOS4 config was built from what was on the site regarding centOS install
1150302686 M * CoBoLt you can install a centos42 but I don't know about ports on alpha
1150302695 M * CoBoLt didn't loose a 2
1150302696 M * CoBoLt :p
1150302697 M * Bertl CoBoLt: could you try with the centos42 ?
1150302702 M * CoBoLt 4.3
1150302710 M * CoBoLt and the tutorial deals with 4.X :p
1150302711 M * CoBoLt :D
1150302715 M * CoBoLt wrong character
1150302723 M * CoBoLt I run centOS 43 on this machine
1150302725 M * Bertl there is a centos42 in the tools
1150302737 M * CoBoLt I know but will it take my alpha port?
1150302739 M * Bertl could you just for a test try with this one?
1150302762 M * Bertl I'm just curious if that gives the same results
1150302824 M * CoBoLt okay
1150302837 M * Bertl hmm, there is a yum and rpm in debian available ...
1150302844 M * CoBoLt same error bertl
1150302848 M * Bertl okay, tx
1150302848 M * CoBoLt no another issue
1150302853 M * CoBoLt *now
1150302859 M * CoBoLt I have a legacy install
1150302866 M * CoBoLt when stopping I get this 
1150302875 M * CoBoLt /usr/lib/util-vserver/vserver.stop: line 100: 29719 Segmentation fault      $_LOCKFILE "$1" $tmp $2
1150302910 M * CoBoLt also with vserver.start
1150302950 M * CoBoLt /usr/lib/util-vserver/vserver.start: line 113: 30524 Segmentation fault 
1150303011 M * CoBoLt I am on and off computer (fixing dinner atm)
1150303022 M * CoBoLt I'll have an eye on this channel though
1150303156 M * Bertl okay, np
1150303512 M * CoBoLt vserver-stat seems to report ok
1150304131 J * brc bruce@20151197230.user.veloxzone.com.br
1150304215 N * sars sarnold
1150304513 M * Bertl daniel_hozac: okay, the mount seems to be something different
1150304519 M * Bertl # rpm --version
1150304519 M * Bertl RPM version 4.4.1
1150304531 M * Bertl yum --version
1150304535 M * Bertl 2.4.0
1150304540 M * Bertl # vserver pollux build -m yum -- -d centos42
1150304540 M * Bertl mount: mount point /etc/rpm does not exist
1150304576 M * Bertl mkdir seems to help here
1150304687 M * Bertl wow, it seems it is working now!
1150304701 M * Bertl well, I get:
1150304707 M * Bertl No Repositories Available to Set Up
1150304710 M * Bertl No Match for argument: glibc
1150304755 M * Bertl CoBoLt: could you provide a tar of your Centos specs/defaults in /etc/vservers ?
1150304902 M * CoBoLt http://www.c2root.be/vservers.tgz (here you go, kinda large meal I'm prepping:) )
1150305076 Q * yarihm Quit: Leaving
1150305518 M * Bertl tx
1150305620 M * Bertl CoBoLt: hum, as far as I can tell, this does not include any distro defaults for a 'centos4' no?
1150305639 M * Bertl i.e. for me that results in 
1150305642 M * Bertl ERROR: Can not find configuration for the distribution 'centos4';
1150305761 M * CoBoLt Bertl
1150305762 M * CoBoLt hang on
1150305770 M * CoBoLt I'll give you the centos4 specs
1150305805 M * Bertl don't worry, and please do not let me keep you from the cooking :)
1150305835 M * CoBoLt trout :)
1150305845 M * CoBoLt with tomatoes and mashed potatoes
1150305853 M * CoBoLt with a mix of salads :)
1150305855 M * CoBoLt anyone?
1150305866 M * CoBoLt http://www.c2root.be/distri.tgz
1150305872 M * Bertl salads? isn't that too healthy?
1150305879 M * CoBoLt this is /usr/lib/util-vserver/distributions
1150305887 M * CoBoLt who says geeks have to be unhealthy?
1150305887 M * CoBoLt :p
1150305890 M * CoBoLt gonna continue now
1150305895 M * CoBoLt tty in a few
1150305900 M * Bertl k,cya
1150306070 M * Bertl hmm, well, kind of 'probably would work' here
1150306349 M * Bertl http://paste.linux-vserver.org/88
1150307124 M * daniel_hozac hmm, what yum version is that? a patched one?
1150307268 M * Bertl debian, unpatched yum, v2.40
1150307274 M * Bertl 2.4.0 actually
1150307448 M * daniel_hozac why don't the utils complain?
1150307473 M * Bertl no idea, but if you like you can have a look yourself ...
1150307473 J * doener_ ~doener@i5387ECAC.versanet.de
1150307491 M * daniel_hozac did you update to 0.30.210 already?
1150307496 M * Bertl yes
1150307518 M * Bertl although not the debian tools, maybe I should check them out too
1150307885 Q * doener Ping timeout: 480 seconds
1150307924 M * daniel_hozac does yum --version output several lines? i don't see what else would be failing...
1150307947 M * Bertl yes, it issues a warning and the version number
1150307950 M * Bertl Warning, could not load sqlite, falling back to pickle
1150307950 M * Bertl 2.4.0
1150308028 M * daniel_hozac ah, i guess that's the problem.
1150308081 M * daniel_hozac /usr/lib*/util-vserver/vyum-worker:42 add | tail -n 1
1150308440 M * Bertl hmm, I have a tail -n 1 there
1150308475 M * Bertl (did install the debian tools in the meantime)
1150308479 M * Bertl retrying now
1150308565 M * Bertl gtting dirty hack now
1150308568 M * Bertl *getting
1150308576 M * Bertl installation is running ...
1150308614 M * Bertl time for the network discussion, I'd say? everybody around?
1150308627 M * bonbons I'm present
1150308638 M * Bertl doener_: ?
1150308697 M * Bertl okay, let's start, doener_ will join later I hope/guess :)
1150308701 A * doener_ just arrived
1150308706 M * Bertl excellent!
1150308710 M * doener_ but needs some food first
1150308730 M * Bertl good, get something to eat, I'll fix the topics :)
1150308755 M * Bertl from my side the following points are of interest:
1150308807 M * Bertl  - do we want to keep the single ip special case? if yes, do we want to make it special, i.e. require a special flag or something like that?
1150308856 M * Bertl - when we remove the #16 ip limit, does that affect any of the existing interfaces? what issues/solutions do we have/need for legacy here?
1150308908 Q * cdrx Ping timeout: 480 seconds
1150308923 M * Bertl - what kind of ip/address/whatever matching does make sense, and how shall we implement it? should the implementation of the logic happen in userspace (as with iptables), or in the kernel, and be transparent to userspace
1150308952 M * Bertl - with ipv6 and lo soon being part of the networking, what special casing do we need for them?
1150309003 M * Bertl that's it for now, probably will add some others later
1150309030 M * Bertl ah, yes, forgot, required API changes
1150309037 M * daniel_hozac lo implies no for 1, right?
1150309053 M * daniel_hozac as every guest will then have a minimum of two IP addresses.
1150309068 M * Bertl depends, there are two exceptions here
1150309077 M * Bertl a) guests without a separate ip
1150309090 M * Bertl b) guests without lo isolation active and a single ip
1150309107 M * daniel_hozac ah, the lo isolation is meant to be voluntary?
1150309159 M * CoBoLt hi guys
1150309161 M * CoBoLt I'm back
1150309162 M * CoBoLt :)
1150309172 M * CoBoLt finished a healthy meal ... Don't shoot me :p
1150309177 M * bonbons Bertl: could you give a small overview of how lo isolation works? (high-level)
1150309182 M * Bertl CoBoLt: good, please have a desert too ...
1150309222 M * Bertl daniel_hozac: well, for service isolation, folks might want to use chbind with a single ip, so yes, I'd say so
1150309240 M * Bertl bonbons: sure, basically it works like this:
1150309280 M * daniel_hozac Bertl: this is not implemented right now, is it? IIRC the patch doesn't require you to assign 127.0.0.1 to the guest.
1150309309 M * Bertl daniel_hoza: no, will become a flag or cap
1150309320 M * daniel_hozac ok.
1150309371 M * Bertl bonbons: packets are handled in so called skb (socket buffers)
1150309392 M * Bertl the major modification is that we network-id tag those skbs
1150309411 M * Bertl this allows to handle the lo traffic independantly for each network context
1150309427 M * bonbons so for host-local communication both endpoint network contexts are known at any time?
1150309447 M * Bertl i.e. we already know which socket belong to what context, now we also know that for the packets
1150309474 M * Bertl precisely, so we basically block traffic going on 'lo' across contexts
1150309502 M * Bertl and permit traffic between contexts (using lo too, but with non-lo ips)
1150309542 M * Bertl of course, this also requires a few changes (which are quite tricky) to make it work as expected
1150309585 M * Bertl doener_: did you find some time to look at the network context patches posted to lkml?
1150309586 M * bonbons so if someome wants there is context-tagging available for iptables (for any potential IPTables guy that would be around) it's then feasible as well
1150309617 M * Bertl yes, that was actually the first ngnet approach
1150309668 M * Bertl and of course, it can be used/extended for generic iptables matching/tagging
1150309711 M * bonbons ok . While you said lo with lo address: are dummy and similar interfaces seen as lo at skb level?
1150309727 M * Bertl regarding dropping support for the single ip special casing, maybe derjohn will have some numbers soon if it is worth the efford ...
1150309767 M * Bertl bonbons: I'm always talking about the actually used interfaces, not the interfaces which accidentially carry the ips :)
1150309826 M * bonbons yes, my question is wether the kernel keeps some information about the interface carrying the IP
1150309859 M * Bertl yes, it does, but it's not the one sending or receiving the packet
1150309884 M * Bertl i.e. put 192.168.0.1 on dumm0 and 192.168.0.2 on eth0
1150309899 M * Bertl then ping one from the other, and the only involved interface will be lo
1150309955 M * bonbons to this regard we may also need to be careful with IPv6 link-local addresses (as those are interface-local)
1150310008 M * Bertl means? they behave differently from v4 addresses?
1150310048 M * bonbons they can exist on multiple interfaces without conflict and MUST be used with specified interface
1150310087 M * bonbons so fe80::1 on dummy0 can never communicate with fe80::2 on eth0 by nature
1150310095 M * Bertl hmm, what happens if you ping one from the other?
1150310102 M * Bertl (on the same host?)
1150310129 M * bonbons this is the kind of address used for network autoconfiguration, neighbour detection, ICMPv6 (as replacement of ARP)
1150310257 M * bonbons if you try to ping the link-local address of eth0 from dummy0 you will never get an answer
1150310365 M * bonbons the same thing applies for a part of multicast addresses (ip6-allnodes, ip6-allrouters and such)
1150310383 J * cdrx ~legoater@cap31-3-82-227-199-249.fbx.proxad.net
1150310392 M * Bertl wb cdrx!
1150310484 M * cdrx yep fast diner 
1150310520 M * CoBoLt ok, when I ssh to my vserver I end up in my host :s
1150310531 M * CoBoLt also the hostname of my vserver stays the same as the host...
1150310540 M * Bertl CoBoLt: see restricting the host's sshd :)
1150310556 M * CoBoLt only listening on it's own ip?
1150310562 M * Bertl yup
1150310594 M * Bertl CoBoLt: btw, centos4 installation on my debain alpha is almost done
1150310602 M * CoBoLt wtf?
1150310606 M * CoBoLt wokee
1150310609 M * Bertl (68/204): tcp_wrappers-7. 100% |=========================| 115 kB    00:00     
1150310616 M * CoBoLt grmbl
1150310625 M * CoBoLt but under centos itself you didn't try it?
1150310636 M * Bertl no, I have no centos for alpha installed
1150310646 M * Bertl but this clarifies it is a centos issue :)
1150310660 M * CoBoLt yup that's a fact
1150310661 M * CoBoLt :)
1150310665 M * bonbons Bertl: so with lo isolation the single-IP case exists if guest has no lo or just 127.0.0.1?
1150310669 M * CoBoLt but centOS is so nice :(
1150310670 M * Bertl CoBoLt: (or at least an issue triggered by centos)
1150310679 M * CoBoLt I mean the maintainers think they are gods
1150310684 M * CoBoLt so telling them will end up in
1150310690 M * CoBoLt is that a custom kernel?
1150310696 M * CoBoLt no, well then we don't support it
1150310698 M * CoBoLt lovely guys
1150310702 M * Bertl bonbons: yep, iif we want to keep that
1150310704 M * CoBoLt got banned several times
1150310718 M * CoBoLt just for asking things which weren't compiled by default :s
1150310731 M * CoBoLt well Bertl, guess I'll stick to the legacy method for now
1150310755 M * CoBoLt if you could just tell me how I need to change my hostname, so I see it in my vserver..
1150310756 M * Bertl would be interesting to try binaries compiled with gcc33 nevertheless
1150310766 M * bonbons I guess single-IP case is mostly useful if the guest can have kind-of-exclusive access to the IP (aka guest binding on ANY won't take away port for the guest)
1150310789 M * Bertl CoBoLt: somehow I still suspect the gcc/binutils to be responsible for this ...
1150310809 M * CoBoLt Bertl possible
1150310824 M * CoBoLt the fact is if it is gcc 3.4.5 that gives you issues
1150310830 M * Bertl bonbons: the advantage of the single ip special casing is that we do not even touch the network context's structures on a lookup
1150310831 M * CoBoLt what's gonna happen if other distro's upgrade?
1150310850 M * Bertl bonbons: we just replace 0.0.0.0 by that specific ip
1150310869 M * Bertl CoBoLt: tools were compiled with 4.0 gcc here :)
1150310883 M * Bertl CoBoLt: so I assume once Centos upgrades it will be fine too :)
1150310887 M * bonbons yep, but if that's enabled we must ensure that the guest is not able to obtain more addresses (so that the "ANY" remains consistent)
1150310911 M * Bertl bonbons: yes, that's the drawback, but we could have that in two flavors
1150310927 M * Bertl a) we tell the folks that old sockets/connections will not catch up
1150310938 M * Bertl b) we do not permit remove/add there 
1150310961 M * bonbons I would prefer b for that case
1150310986 M * Bertl CoBoLt: ah, not enough space on my alpha for centos guest :)
1150310993 M * CoBoLt damn
1150311005 M * daniel_hozac Bertl: b where?
1150311017 M * daniel_hozac in the single IP case?
1150311027 M * Bertl daniel_hozac: in this case for the single ip case in the kernel
1150311042 M * Bertl i.e. you decide you want single ip, you get it but modification does not work
1150311059 M * daniel_hozac of course, that makes sense.
1150311065 M * Bertl alternatively we could do a c)
1150311078 M * Bertl patch up existing sockets to the new ip
1150311096 M * Bertl (which is something I'd rather avoid)
1150311102 M * daniel_hozac yeah, that sounds really ugly.
1150311116 M * bonbons that looks dangerous, especially if there are active connections
1150311128 M * Bertl IMHO the not being able to change is a reasonable tradeoff
1150311140 M * bonbons yep
1150311142 M * Bertl given that it gives a real performance advantage
1150311154 M * Bertl (which still has to be proven)
1150311205 M * bonbons probably the advantage is for repetitive short-lived connections (and UDP)
1150311205 M * Bertl I fixed up the nc_server/client checks to handle all relevant cases and to output an address agnostic result
1150311243 M * Bertl and the princeton folks are going to provide some test blades with a good network connection
1150311259 M * Bertl so I think we will get some numbers soon there
1150311311 M * Bertl for now, I'd suggest we avoid the single ip special case, and add it back later if it really gives an advantage .. your opinions?
1150311366 M * Bertl orionpanda: btw, want to test the latest iattr patch for ocfs2?
1150311398 M * bonbons that's fine (anyhow, single-ip case exists seprately for each protocol - except with the special ANY that is IPv6+IPv4)
1150311426 M * Bertl it's the ANY case we are talking about
1150311430 M * doener_ phone call from my mom... took some time ;)
1150311437 M * doener_ Bertl: no, didn't take a look yet
1150311438 M * Bertl (i.e. replacinf ANY by single IP)
1150311446 M * Bertl *replacing
1150311486 M * bonbons yep, but there are two ANY: IPv4 only any, IPv6 only any and IPv4+IPv6 any.
1150311512 M * Bertl hmm, 1+1 = 3:)
1150311518 M * bonbons the first two are normal single-IP cases, the third one is a more exclusive SINGLE
1150311552 M * Bertl but actually that's a good point
1150311569 M * Bertl regarding ipv4 and ipv6, do we want to mix them at all?
1150311588 M * bonbons so for 1. and 2. single-IP are IP-proto local, for 3. it's way harder (but IPv4+IPv6 sockets can be disabled completely)
1150311595 M * Bertl wouldn't it be sufficient to have an ipv4 and an ipv6 'personality' for a guest=
1150311600 M * Bertl s/=/?
1150311660 M * bonbons that's disable the 3. case, which I'm really fine with (it's quite complex to handle with collisions!)
1150311731 M * Bertl can ipv4 addresses collide with ipv6 ones? or vice versa?
1150311743 M * bonbons yep
1150311760 M * bonbons there are those IPv6 mapped IPv4 addresses which collide
1150311762 M * Bertl for anything except the ANY case?
1150311810 M * Bertl okay, so would it be reasonable for the userspace side to have ipv4 or ipv6 only guests?
1150311827 M * bonbons don't remember the range for IPv4 mapped in IPv6, but there you have IPv6 addresses which are equivalent to IPv4 addresses
1150311873 M * bonbons no, mixed IPv4/IPv6 guests are fine as long as IPv6 sockets are always IPv6 only
1150311897 M * Bertl i.c. and this can be selected in the kernel? or how?
1150311944 M * bonbons that is mixed sockets are disabled in /proc/sys/net/ipv6/bindv6only
1150311973 M * Bertl ah, okay, sow e could force that on
1150311979 M * bonbons or for each created IPv6 socket the IPv6only flag is set
1150312005 M * bonbons we can disable it easily
1150312035 M * Bertl okay, then let's move on for now to the 'matching' question
1150312058 M * Bertl i.e. what kind of address matching do we want, and where should we put the logic
1150312105 M * bonbons matching for "new" connections?
1150312131 M * Bertl general address in nx_info() matches
1150312152 M * bonbons I think we need matching based either in individual address, address range or subnet/mask
1150312174 M * Bertl be it for interface visibility or for collision detection
1150312183 Q * phedny Ping timeout: 480 seconds
1150312204 M * Bertl well, the collision case is a little trickier for ranges
1150312212 M * bonbons individual address is for few addresses (would say less than 5 to be reasonable), range is for non-power-of-two
1150312233 M * bonbons why so?
1150312257 M * Bertl first, how to define ranges, then how to check one range against the other
1150312303 M * bonbons ranges compared to ranges are relatively easy, defined by border addresses addr1-addrN
1150312338 M * Bertl but okay, let's assume we have single ips, ranges (ip + len), and networks (ip + mask)
1150312370 M * Bertl how do we a) combine them, and b) keep the decision/checking as fast/simple as possible
1150312392 M * bonbons so conflict is if range1min < range2max and range1max > range2min (would need to think in details if I'm not missing something but that should be the comparison)
1150312403 M * daniel_hozac what does it mean to assign a range/network to a guest?
1150312416 M * daniel_hozac just an easier way to give it tons of IP addresses?
1150312432 M * Bertl IMHO yes
1150312438 M * bonbons The most work will have to be done when configuring the addresses for the guests (performance there should not be really important)
1150312470 M * Bertl agreed the setup (guest start/address change) can be dog slow
1150312485 M * bonbons yesp, that same style as giving a subnet, just that you dont need to giver 2^n addresses but can assign a free number
1150312485 M * Bertl the important part is the matching which happens on each and every connect
1150312520 M * daniel_hozac that should be fairly simple, no?
1150312522 M * Bertl well, for udp on every packet and several times actually
1150312586 M * Bertl the thing is, with the current approach we always have to check all entries
1150312587 M * bonbons if users can reduce address "container" (individual addr/range/subnet) count to small number then either setup would be effective
1150312617 M * Bertl I'd prefer something which makes decisions like this:
1150312638 M * bonbons if the list of "containers" is sorted, we can check them in a dichotomic way -> log(n)
1150312651 M * Bertl first entry/rule: if outside range x/y, return no, otherwise continue
1150312679 M * Bertl second entry: if within network a/b, return yes, otherwise continue
1150312700 M * Bertl third entry: if ip == z, return yes, otherwise no
1150312733 M * Bertl that would drastically reduce the checks while enhancing the options
1150312741 M * daniel_hozac why the negated check for ranges?
1150312760 Q * derjohn2 Ping timeout: 480 seconds
1150312766 M * Bertl was just an example, all checks could be positive or negative
1150312776 J * derjohn2 ~aj@dslb-084-058-200-014.pools.arcor-ip.net
1150312782 M * bonbons why not keep a list of non-overlapping items in sorted order?
1150312793 M * CoBoLt ok guys
1150312794 M * daniel_hozac i guess returning no doesn't really work at all.
1150312795 M * CoBoLt I'm off
1150312797 M * CoBoLt ttytm
1150312806 M * Bertl CoBoLt: k, cya!
1150312811 Q * CoBoLt Quit: Leaving
1150312819 M * bonbons this way we can start in the middle of list and on each test consider just the half of interest
1150312864 M * Bertl well, doesn't buy you much on many entries
1150312881 M * Bertl a balanced tree will give you log(n) while you have n/2
1150312943 M * bonbons is the gain important for low count of nodes?
1150312945 M * Bertl daniel_hozac: why wouldn't the 'no' work?
1150312966 M * daniel_hozac what if the IP address is within a network or one of the assigned singles?
1150312969 M * bonbons with ranges or subnets I guess having many entries in the list will be very rare
1150312992 M * Bertl daniel_hozac: ah, well, the rules would have to be 'prepared' by userspace
1150313007 M * Bertl daniel_hozac: i.e. they would not necessarily match the input from the user
1150313038 M * Bertl for example, a very simple optimization IMHO would be to 'mask' the given IPs into one single mask or range
1150313061 M * bonbons yep, adding 1.0.0.1 and then 1.0.0.2 and then 1.0.0.3 should result in just range 1.0.0.1 to 1.0.0.3
1150313062 M * Bertl with the sole purpose of producing that 'no' on first access
1150313098 M * Bertl (this could even be stored in the nx_info without external data structures
1150313116 M * bonbons the idea of a enclosing range is good as well
1150313144 M * Bertl for 90% of the check (maybe even more) it would yield the no and be done
1150313171 M * bonbons so context would have either no IP at all, or something in a range/complete range
1150313183 M * Bertl (again assumed that the guests do not overlap, and special ips like 127.0.0.1 are handled differently
1150313186 M * bonbons the more guests the more true that 90% is
1150313186 M * Bertl )
1150313236 M * Bertl also operations like inclusion or exclusion (union or intersection) might be interesting here
1150313257 M * Bertl sidenote: my centos guest is now installing :)
1150313327 M * doener_ Bertl: special casing for 127.0.0.1 should be easy, just keep the tag for skb that gets into the receive chain, and for non-127.0.0.1 destined ones remove it... anything I miss there?
1150313333 M * bonbons yep, but we should not make it too complex either (that will probably cost much performance) Better confvince users to keep their addresses in a (few) ranges/subnets
1150313368 M * doener_ then receival of tagged skb's would immediately know which socket the skb goes to
1150313371 M * Bertl agreed, the simpler the better
1150313441 M * Bertl doener_: I had some crosstalk restrictions in mind/planned, which would not work with that
1150313478 M * doener_ hm, what exactly?
1150313489 M * Radiance hiya all :)
1150313531 M * Bertl doener_: something like guestX can produce crosstalk, guestY can receive crosstalk, if both is given, they will be allowed to talk to echother
1150313540 M * Radiance perhaps some one knows, i see on a remote server using "top" the init process saying init [2]   , why would it be 2 ? normally you'd see only [init]
1150313567 M * Bertl init records the runlevel (current) if there is enough space in the name
1150313593 M * Bertl there is even a special init option ro reserve more space, see init man page
1150313600 M * Radiance ah thanks mate
1150313601 M * Bertl s/ro/to/
1150313608 M * Radiance i thought for a second it had to do with a zombie issue or so
1150313662 Q * brc Quit: [BX] Reserve your copy of BitchX-1.1-final for the Sony Playstation today!
1150313801 M * Bertl okay, I guess we need at least a nigh of sleep to continue this network discussion, what do you think?
1150313806 M * Bertl *night
1150313807 M * bonbons the existance of subnets/ranges should be visible to userspace (e.g. allowing addition of range/subnet at once)
1150313864 M * bonbons there are two of the point whe could talk about before night, to make them take profit of it
1150313865 J * brc_ bruce@20151197230.user.veloxzone.com.br
1150313875 M * bonbons that's mostly the interface to userspace part
1150313903 M * Bertl okay
1150313940 M * bonbons the limit on address/item count is certainly not of any importance to userspace (too many addresses you be rejected with OOM)
1150313952 M * Bertl my last check of the userspace API gave me the impression that we do not change much there for recent API funtions
1150313993 M * Bertl *need
1150313999 M * bonbons if we do dynamic allocation of mem memory footprint of guest is not an issue either
1150314023 M * Bertl yes, we definitely move away from the in kernel static #n ips
1150314055 M * daniel_hozac only the legacy interface would have to be modified.
1150314066 M * bonbons that's transparent to userspace (should be!)
1150314066 M * Bertl we might once again put the first 4 or 5 rules inline, to avoid memory fragmentation
1150314100 M * bonbons daniel_hozac interface, only implementation
1150314103 M * Bertl (but those are implementation details)
1150314123 M * daniel_hozac yeah, that's what i meant.
1150314161 M * Bertl so don't worry about this, the 16ips are history, which will not affect/change userspace
1150314182 M * bonbons current API is just a special case of adding of "items" (addr/range/subnet)
1150314191 M * daniel_hozac util-vserver will have to be changed, IIRC it uses a static number too.
1150314210 M * Bertl daniel_hozac: that's fine for now, but yes, the changes will be required
1150314233 M * Bertl (not required for max 16 ips and legacy though)
1150314261 J * phedny ~mark@volcano.p-bierman.nl
1150314285 M * Bertl wb phedny!
1150314292 M * bonbons I would propose some API which allows adding subnet/range/ip for IPv4 or IPv6, one item at a time; having the kernel to the merging of that new infor with existing one
1150314333 M * Bertl while I would propose to do the merging in userspace :)
1150314334 M * bonbons this has the major advantage that we can change data management inside of kernel transparently to userspace
1150314361 M * Bertl and the major disadvantage that the complicated stuff has to be done in the kernel :)
1150314395 M * Bertl (doesn't mean that we won't do it this way :)
1150314407 M * bonbons yes, but kernel is sure to have the information the way it wants (otherwise it must validate averything that's equavalent to recalculating)
1150314436 M * Bertl it could also trust userspace data, as for example iptables does
1150314455 M * phedny 21:44:16 -!- Irssi: critical g_io_error_get_from_g_error: assertion `err != NULL' failed
1150314458 M * phedny Bertl: thankz :)
1150314508 M * bonbons just that it binds userspace and kernel implementation
1150314560 M * bonbons but the iptables does not build the whole ruleset in userspace either, as it has important costs when manipulating rules (not just a atomic mem-copy)
1150314569 J * m0Zzg ~m0zg@80.71.242.130
1150314573 M * m0Zzg hello
1150314757 M * bonbons Bertl: we will also have to to some locking on those new dynamic structures (as those I use for modular IPv6, exclusive just for write)
1150314764 J * CoBoLt ~countcobo@84.196.135.76
1150314767 M * CoBoLt hi bertl
1150314772 M * CoBoLt I am downloading debian netinstall atm
1150314773 M * CoBoLt :)
1150314782 M * CoBoLt hoping that this will work in the end :)
1150314797 M * m0Zzg did you heard about Ruslan Ermilov?
1150314823 M * m0Zzg he is russian developer of freeBSD
1150314833 M * CoBoLt uhu
1150314837 M * CoBoLt jst googled on him
1150314877 M * m0Zzg :) sorry, my english is bad. did you heard about his troubles?
1150314881 M * CoBoLt nope
1150314901 M * CoBoLt as I just google on him, I didn't even hear off him before the last 4 minutes
1150314908 M * CoBoLt wassup with the guy?
1150314951 M * m0Zzg his wife needs an operation... oncology
1150314965 M * CoBoLt awtch
1150314967 M * CoBoLt not good
1150314976 M * m0Zzg he told to all open source comunity
1150314999 M * m0Zzg he need very much money
1150315002 M * CoBoLt urls?
1150315023 M * m0Zzg http://people.freebsd.org/~ru/help/ru
1150315031 M * m0Zzg this is his open-letter
1150315039 M * m0Zzg russian language
1150315048 M * CoBoLt can't read the char settings
1150315060 M * m0Zzg koi8-ru
1150315074 M * CoBoLt don't have that installed
1150315078 M * CoBoLt and I don't speak russian
1150315120 M * m0Zzg save with another charset?
1150315148 M * CoBoLt I won't be able to read it still, won't I?
1150315153 M * CoBoLt I don't understand russian
1150315160 M * CoBoLt is it in english somewhere?
1150315188 M * m0Zzg i didn'n saw, but may be
1150315189 M * mnemoc babelfish translates russian 
1150315200 Q * cdrx Quit: Leaving
1150315275 M * m0Zzg try to translate
1150315527 M * m0Zzg so, if you want to help him
1150315564 M * m0Zzg you must write to: imp@FreeBSD.org or rwatson@FreeBSD.org
1150315569 M * Bertl btw, there are millions of people dying of hunger ...
1150315570 M * m0Zzg Warner
1150315570 M * m0Zzg Losh
1150315582 M * m0Zzg and Robert Watson
1150315602 M * Hunger of me?
1150315630 M * bonbons Bertl, daniel_hozac, doener_: it's time to have a night over the network subjet! (at least for me)
1150315663 M * doener_ good night then!
1150315672 M * m0Zzg they are coordinating donates in England and Amerika
1150315696 M * Bertl Hunger: well, of your fellow namesake ...
1150315731 M * Hunger :)
1150315770 M * CoBoLt night
1150315777 M * bonbons thanks! will back in about 20 hours
1150315811 Q * bonbons Quit: Leaving
1150315967 P * m0Zzg 
1150315975 Q * Zaki Ping timeout: 480 seconds
1150317451 M * Bertl okay, tx folks, I'm off for now .. probably back later!
1150317460 N * Bertl Bertl_oO
1150317935 M * CoBoLt are you guys running etch or sarge?
1150318712 M * sid3windr sarge here
1150318775 M * CoBoLt ok
1150318778 M * CoBoLt just installed 3.1
1150319041 Q * lilo2 Quit: brb
1150319074 Q * zkbrsnie Quit: 
1150319108 Q * m4z Ping timeout: 480 seconds
1150319220 Q * lilalinux Ping timeout: 480 seconds
1150319550 J * lilo2 ~0710AAD4@tor-irc.dnsbl.oftc.net
1150319601 Q * Viper0482 Remote host closed the connection
1150319732 J * lilalinux ~plasma@dslb-084-058-226-179.pools.arcor-ip.net
1150320037 J * lilo2_ ~0710AAD4@tor-irc.dnsbl.oftc.net
1150320115 Q * lilo2 Ping timeout: 480 seconds
1150321634 Q * Fushi Read error: Connection reset by peer
1150322773 J * m4z m4z@bastard-operator.from-hell.net
1150322841 Q * dna Quit: Verlassend
1150322972 J * afta ~a@chello085216135073.chello.sk
1150323103 M * afta hello. is it sound to have a vserver for each separate service (dns,web,mail) on a server with more ips, where all those services should be present on each ip - separating services role-wise but not ip-wise?
1150323270 M * doener_ sure
1150323295 J * shedi ~siggi@130.208.221.254
1150323333 M * doener_ ip address sharing is quite possible, you'll just have to restrict the ssh daemons to a single ip address or assign different ports to them
1150323361 M * doener_ s/quite//
1150323388 M * afta hm, i was thinking leaving the ssh just for the host
1150323401 M * afta which would have a dedicated ip
1150323431 M * doener_ well, AFAIK using ssh to enter the vservers is still preferred over "vserver foo enter"
1150323457 M * doener_ and as usual I don't remember the exact reasons :/
1150323466 M * afta hmm...
1150323500 M * doener_ daniel_hozac: do you remember why? (and if it is still true...)
1150323775 M * afta that forces me to remember the service-ip or service-port 8) but if you say so ;P
1150323927 M * cehteh anyone of you attending to froscon?
1150323939 M * doener_ afta: well, you can easily store that in your ssh config
1150323981 M * doener_ and if you intended to run sshd on the host only, you would probably need that config on the host only anyway. connect there first and then to the vserver
1150324070 J * Aiken ~james@tooax6-178.dialup.optusnet.com.au
1150324096 J * sdsaas ~asdsdadas@node-51-188.jambolnet.com
1150324113 M * sdsaas freeshell?
1150324126 M * sdsaas freeshell for ircd?
1150324195 M * bon :))
1150324200 M * bon better look somewhere else
1150324251 Q * sdsaas Quit: 
1150324271 M * doener_ afta: http://paste.linux-vserver.org/89
1150324294 M * doener_ here's an example for a ssh config
1150324384 M * bon mysql.vserver/
1150324389 M * bon what's that good for in sshd?
1150324403 Q * sladen Ping timeout: 480 seconds
1150324414 M * doener_ if I, for example, have a sshd on my.host.tld:22 and one on my.host.tld:5000
1150324432 M * doener_ sshd would store the HostKey for both sshd's as "my.host.tld"
1150324453 M * doener_ and obviously get bitchy because it is not the same in case of 2 different sshd's
1150324484 M * doener_ the HostKeyAlias tells ssh to store it as "mysql.vserver"
1150324485 M * bon ah
1150324486 M * bon i see
1150324487 M * bon that's nice
1150324490 M * bon but fortunately
1150324493 M * bon i have plenty of ips
1150324503 M * bon so i can have each vserver on different one if needed
1150324532 M * doener_ (don't ask why I wrote "mysql" once and then "mysql.vserver", it doesn't matter that much, but for example ssh bash completion will break)
1150324557 N * lilo2_ lilo2
1150324557 M * doener_ so using the same string for "Host" and "HostKeyAlias" is a better idea
1150324628 M * doener_ bon: it's already sufficient to have 2 different domain names, so it's mostly a problem for dyndns users (like me ;)
1150324665 M * CoBoLt guys I'm really off now
1150324673 M * doener_ I included it for completeness. The CheckHostIP thing is more important, to avoid that, you actually need one IP address per sshd
1150324674 M * CoBoLt still compiling my kernel on the alpha
1150324683 M * CoBoLt hopefully tomorrow I can run vservers on it :)
1150324685 M * CoBoLt ttytm
1150324688 M * CoBoLt night all
1150324693 M * doener_ good night!
1150324697 Q * CoBoLt Quit: Leaving
1150325380 J * mef ~mef@72-254-14-32.client.stsn.net
1150325390 Q * mef Remote host closed the connection
1150325662 N * sarnold sars
1150325680 N * sars sarnold
1150326382 J * sladen paul@starsky.19inch.net
1150327243 J * aphetadus ~joaopaulo@201.64.40.31
1150327252 P * aphetadus 
1150328579 Q * fwl Quit: This computer has gone to sleep
1150329500 J * fwl ~f_@83-215-237-1.seek.stat.salzburg-online.at