1160784039 M * doener Bertl: proc_pid_lookup->pid_revalidate ... that check?
1160784063 M * Bertl seems to happen earlier
1160784164 M * Bertl but yeah, it would be the logical cause
1160784217 M * doener hm, ls shows 1 though...
1160784224 M * doener funny stuff
1160784238 A * doener just dared to upgrade...
1160784240 M * Bertl yeah, we know lookup != readdir
1160784311 M * Bertl it's the find_task_by_pid()
1160784363 M * Bertl at least I think so ...
1160784442 M * doener in pid_task?
1160784503 M * Bertl check dmesg, you should have a dump :)
1160784526 M * doener one? a few dozens ;)
1160784643 M * doener WARN_ON(current_xid) looks pretty bogus, given that we do a VX_WATCH|VX_IDENT check above, doesn't it?
1160784670 M * Bertl why?
1160784702 M * doener ah, the whole thing is basically a "warn on", nvm...
1160784708 M * Bertl current->xid=0 task->xid=666 ?
1160784756 M * Bertl (would not warn atm)
1160784782 M * doener no, I don't get the check... what is it supposed to check?
1160784805 M * Bertl the 'first' check is for context mismatch (the if)
1160784853 M * Bertl the warn/vxw check only complains about mismatches found in a non host context
1160784931 M * Bertl the init case is the one I obviously forgot about
1160785016 M * Bertl the real question is, shall we extend the pid_task() to handle the blend through init
1160785033 M * Bertl (which I consider a bad idea ...
1160785056 M * Bertl or better work around that issue by handling that in procfs
1160785109 J * ScoobyD00 ~foo@82-44-80-99.cable.ubr08.newm.blueyonder.co.uk
1160785117 M * ScoobyD00 hello
1160785119 M * Bertl welcome ScoobyD00!
1160785125 M * ScoobyD00 hi berlt :O)
1160785130 M * ScoobyD00 *tl
1160785135 M * Bertl :)
1160785155 M * ScoobyD00 most times you help me solve problems so i am glad to see you!
1160785169 M * Bertl means, you have a new one? :)
1160785175 M * ScoobyD00 yeah :o)
1160785190 M * ScoobyD00 i have a very strange issue with 1 of my 2 servers (hosts)
1160785201 M * ScoobyD00 they both run vserver guests
1160785243 M * ScoobyD00 one of the machines is consistently failing - it stops listening on any of its IPs to external traffic (host and guests)
1160785265 M * ScoobyD00 although when i log into my second machine, i can ssh into it fine
1160785271 M * Bertl when you do what?
1160785284 M * ScoobyD00 nothing - it just happens at random
1160785300 M * Bertl hmm, that indeed sounds strange ...
1160785313 M * Bertl i.e. it would point to a hardware failure
1160785318 M * ScoobyD00 so i have to login to my seconf machine, and when i restart any of the vserver guests, IP starts responding properly again
1160785338 M * ScoobyD00 nothing else fixes it besides restarting a vserver
1160785351 M * ScoobyD00 (any one will do)
1160785366 M * Bertl so, what happens is that the network connectivity goes away somehow?
1160785379 M * ScoobyD00 yes, to everything except the machine sitting right next to it
1160785400 M * Bertl that sounds like a router/switch/arp issue then
1160785415 M * ScoobyD00 do you know why restarting any of its vservers would make it work again?
1160785416 M * Bertl do you have some kind of firewall on that machine?
1160785420 M * ScoobyD00 no
1160785436 M * Bertl do you have more than one interface on that machine?
1160785439 M * ScoobyD00 no
1160785462 M * ScoobyD00 (btw - i too thought it was router issue - just wanted to make sure there was no known bugs as vserver seems to fix it on restart)
1160785493 M * Bertl no, I just assume that the restart (as it probably assignes the guest ips) will send out some arp requests
1160785517 M * ScoobyD00 ok, i go pester my datacenter support ;o)
1160785522 M * Bertl and those packets will probably make the routing/switching work again
1160785588 M * ScoobyD00 while i am here - have you ever ran vservers from oracles OCFS2? :o)
1160785610 M * Bertl when you reach this state, it might be interesting to try a tracepath out and to the guest/host
1160785624 M * Bertl well, I did the necessary changes for OCFS2 :)
1160785633 M * ScoobyD00 cool - what version was that for?
1160785636 M * Bertl xattr support and that
1160785648 M * ScoobyD00 oh right - to the FS?
1160785664 M * ScoobyD00 or to vserver?
1160785667 M * Bertl yep, so it is supported on the devel branch
1160785682 M * Bertl the changes are for mainline (included by now) and vserver
1160785731 M * Bertl so it simply should work as expected :)
1160785753 M * Bertl if not, here is the place for bugreports ...
1160785809 M * ScoobyD00 i am running debain unstables vs2.0.2 with kenrel2.6.16 - is that ok?
1160785866 Q * FCOJ Quit: Leaving
1160786137 M * Bertl 2.02 is stable, it requires a devel kernel
1160786155 M * Bertl (at least I think so, let me double check that)
1160786802 M * Bertl yep, requires a devel branch
1160786904 M * ScoobyD00 what denoted a devel branch in linux (am from a freebsd background originally)?
1160786913 M * ScoobyD00 odd number?
1160786926 M * doener until 2.6
1160786945 M * doener now devel is basically just everything between 2.6.x and 2.6.x+1
1160786960 M * ScoobyD00 so my 2.6.16 should be ok?
1160786960 M * doener (well, some say that 2.6 is a huge pile of devel)
1160786961 M * Bertl in this case you need a devel branch vserver patch :)
1160786982 M * Bertl means 2.1.x instead of 2.0.x
1160787020 M * ScoobyD00 oh ok - so not much chance of that being in debian *stable then :o)
1160787065 M * ScoobyD00 was it just xattr - and what does that do?
1160787093 M * Bertl basically support for the ext2 attributes (including barrier and iunlink) and xid tagging
1160787139 M * Bertl there should be devel kernels in debian too (ask waldi for details)
1160787190 M * ScoobyD00 yes, the problem is i already regret moving from stable -> unstable, and any further move into development packages will flip me out even more :o)
1160787396 M * Bertl why OCFS2 then?
1160787421 M * Bertl I mean, it's a very new addition to the kernel, with features getting in one by one
1160787480 M * ScoobyD00 but it is in the kernel
1160787513 M * ScoobyD00 and so will be maintained better by debian (when the next release of debian hits soon, it will be maintained as stable)
1160787513 M * doener ext4 is in as well, I'd not touch it with a 5m pole though ;)
1160787532 M * ScoobyD00 :o)
1160787547 M * Bertl lol
1160787555 M * ScoobyD00 i like the idea of a cluster fs, and GFS didnt feel right
1160787576 M * Bertl yeah, well, you can use it without the vserver support too
1160787596 M * Bertl but be careful, it will be somewhat insecure and not all features are there
1160787607 M * ScoobyD00 like the chroot exploit thing?
1160787620 M * Bertl doener: what if we just mimic the init pid (as dir)?
1160787626 M * Bertl ScoobyD00: yep, precisely
1160787653 M * doener Bertl: what if we skip the pid_revalidate for blend through init?
1160787667 M * doener if that dies, you don't get anything from proc anymore anyway
1160787681 M * Bertl I'd prefer to skip the entire task get there if possible
1160787704 M * Bertl and just add the directory inode (and ignore any further checks if possible)
1160787721 M * doener ehrm, my thoughts hung somewhere in time, right before you found the actual cause of the problem... oops :)
1160787929 M * Bertl the question IMHO is, how much is required to make userspace tools happy
1160787979 M * Bertl i.e. is a dir enough? how many entries?
1160788086 J * yarihm ~yarihm@84-75-123-221.dclient.hispeed.ch
1160788548 Q * Johnnie Remote host closed the connection
1160788626 Q * ScoobyD00 
1160789043 J * neuralis ~krstic@solarsail.hcs.harvard.edu
1160789063 M * neuralis Bertl: hi there. around?
1160789297 M * Bertl yep, hey!
1160789321 J * Johnnie ~jdlewis@66.199.231.180
1160789370 Q * yarihm Quit: Leaving
1160789466 M * neuralis Bertl: do you have a few minutes now, or are you going to sleep? i could catch you tomorrow, if that's better.
1160789480 M * Bertl nope, no problem, still awake :)
1160789554 M * neuralis okay. do you prefer to talk here, or privmsg?
1160789572 M * Bertl I guess it is vserver related, so the channel should be fine
1160789580 M * Bertl (note that it is logged in realtime though)
1160789598 M * neuralis that's fine. i have just a few questions that i couldn't answer easily by browsing through the site, and i didn't have time to dig into the actual patches yet.
1160789634 M * Bertl let's hear then ...
1160789636 M * neuralis first, marc mentioned you're doing a "new" copy on write implementation. how does it work?
1160789662 M * Bertl the 'new' CoW is quite old now (not for marc though :)
1160789681 M * Bertl you know how 'unification' works?
1160789704 M * neuralis e.g. unionfs?
1160789713 M * Bertl nah, okay, I'll explain
1160789724 M * neuralis please do.
1160789757 M * Bertl I'll try to explain the historical path here, so you know why we do some things this way
1160789772 M * neuralis sure
1160789792 M * Bertl in the beginning, there was the idea to cut down on resources (mainly disk space) by using hard links between guests
1160789820 M * Bertl the resulting problem is obvious -- guests could modify those links too easily
1160789824 M * neuralis right.
1160789846 M * Bertl but, there is a flag/feature present on most filesystem -- immutability
1160789861 M * neuralis right. so you introduced a way to break the hardlinks on modification, if i read it correctly.
1160789865 M * Bertl so the next step was to add the immutable flag to those files, and remove the capability
1160789890 M * Bertl this is quite fine, as long as you don't want to change the files
1160789913 M * Bertl so one step further, we introduced a second flag, called immutable linkage invert
1160789938 M * Bertl which basically makes the 'immutable' links 'unlinkable' (read removeable)
1160789978 M * Bertl this was (and is to some extend) more than sufficient for most package systems
1160790000 M * Bertl they can remove the old file (link) and replace it with a new one
1160790030 M * Bertl the latest addition (a year ago) was the CoW functionality
1160790063 M * Bertl instead of simply 'blocking' access to such special immutable links
1160790086 M * Bertl we make a copy on the fly, and let the user work with that
1160790093 M * neuralis yep.
1160790108 M * Bertl that's how the CoW works ...
1160790139 M * neuralis sounds like the fs overhead is relatively minimal
1160790157 M * Bertl on read access, there is no measurable overhead
1160790168 M * Bertl after all it is identical to hard links
1160790185 M * neuralis right, i meant in terms of accounting (the extra hard links for each vserver)
1160790188 M * Bertl on write access, the kernel need sto make the copy
1160790203 M * Bertl nah, that is insignificant
1160790228 M * Bertl also there are no _additional_ hard links
1160790240 M * Bertl so you can have, let's say, two guests
1160790247 M * Bertl and unify them with eachother
1160790258 M * Bertl which will not use any additional inodes
1160790286 M * Bertl and typically you can share 90% of the distro (not user data) between guests
1160790301 M * neuralis that's very interesting.
1160790310 M * Bertl with the additional advantage that 'shared' files will only show up once in the inode cache :)
1160790335 M * neuralis right. so you have them all use the same chroot, but give each guest its own scratch space for files it modifies (where the links are broken)?
1160790341 M * Bertl and for shared libraries and executable mappings, the benefit is even larger, as the mappings do not happen more than once
1160790369 M * Bertl neuralis: nope, they use different chroots, the links are beneath that
1160790393 M * Bertl /path/to/guest/A/bin/bash hardlinked to /path/to/guest/B/bin/bash
1160790415 M * Bertl not a problem, as long as they are on the same filesystem
1160790440 M * Bertl so each of the guests has it's own 'file space'
1160790457 M * neuralis how does that not introduce additional hardlinks for each guest? 
1160790474 M * neuralis (i'm not worried about the overhead, i'm just trying to understand how this is built.)
1160790478 M * Bertl let's assume guest A consists of 10 executables
1160790492 M * neuralis okay
1160790494 M * Bertl let's further assume guest B is basically identical
1160790515 M * Bertl that would give roughly 12 inodes for guest A and 12 for guest B, right?
1160790528 M * neuralis right
1160790531 M * Bertl i.e. 2 dir inodes and 10 'normal' inodes
1160790536 M * neuralis yep
1160790550 M * Bertl and it would also consume, let's say, 10M of diskspace
1160790557 M * Bertl (static glibc binaries :)
1160790565 M * neuralis heh, yeah
1160790569 M * Bertl again, for each guest
1160790579 M * Hurga both guests actually use the file system of the root server, in which they are chrooted, so the (hardlinked) binary is only mapped to memory once.
1160790599 M * Bertl now, the unified case would use 14 inodes and 10M of space :)
1160790624 M * Bertl 2 dir inodes for A, two for B, and 10 for the files
1160790629 M * neuralis okay, got it. 
1160790641 M * Bertl so you saved 10 inodes here :)
1160790649 M * Bertl and 10M of disk space :)
1160790673 M * Bertl Hurga: could be shared with the host too, but usually we avoid that for security reasons
1160790682 M * neuralis yep, makes total sense. next question i had -- are there any strange semantics of shared pages across guests that one wouldn't expect? 
1160790686 A * Hurga needs quota, unfortunately, so he has to use different file systems...
1160790709 M * Bertl Hurga: yes and no, we already did context quota on a shared filesystem
1160790730 M * Hurga Bertl: I know, but that's not in the current version
1160790733 M * Bertl neuralis: what kind of shared pages do you refer to?
1160790750 M * Bertl Hurga: that's correct ... too little demand yet
1160790761 M * Bertl Hurga: all testers vanished sooner or later :(
1160790766 M * Hurga Bertl: I'd be happy to test :)
1160790792 M * Bertl we can start anytime, but you should know, it isn't easy to test
1160790807 M * Bertl Hurga: i.e. many, many corner cases have to be checked
1160790816 M * neuralis Bertl: actually, i guess you answered that. shared libs will only end up in memory once.
1160790838 M * Bertl yep, that is true for all file based (read only) mappings
1160790864 M * Bertl write (able) pages are separate by default
1160790896 M * neuralis makes sense. last thing i was wondering about -- can i pass around fds between guests?
1160790944 M * neuralis actually, let me rephrase that. can i pass *in* fds to guests when i'm starting them? 
1160790972 M * Bertl yes, but you have to be careful regarding security
1160790977 M * neuralis yes, of course
1160790986 M * neuralis but it's physically permitted by vserver?
1160790992 M * Bertl a guest is nothing more than a combination of several isolation technologies
1160791024 M * Bertl (which means, you can cherry pick what you want to isolate)
1160791053 M * Bertl here an example:
1160791055 M * Bertl ls | chcontext --xid 42 -- wc
1160791075 M * Bertl this lists the current dir, creates a context (42) which executes the wc
1160791088 M * Bertl it is connected via filedescriptors here
1160791113 M * Bertl does that answer your question?
1160791157 M * neuralis yes, it does.
1160791158 A * Hurga usually describes vserver as "chroot on steroids" :)
1160791203 M * neuralis so what i don't understand is how i haven't heard of vserver before. it sounds like an awesomely designed system; why isn't there more traction?
1160791265 M * Bertl because we focus on development not PR :)
1160791293 M * neuralis well, i bet i can help with that. :)
1160791301 M * Hurga It's getting more and more popular, but it isn't in the official kernel, and everything which needs kernel patching isn't popular by default.
1160791313 M * neuralis bertl: thanks a lot for the answers. i'll follow up with a few things in privmsg.
1160791329 M * Bertl neuralis: you're welcome! and please do so ...
1160792372 J * bronson ~bronson@c-71-198-75-160.hsd1.ca.comcast.net
1160792490 M * Bertl wb bronson!
1160792501 Q * Hurga Remote host closed the connection
1160794763 Q * spTim Quit: ChatZilla 0.9.61 [Mozilla rv:1.7.6/20050319]
1160797606 Q * sp Ping timeout: 480 seconds
1160797816 J * sp sp@ip-87-97.members.virt-ix.net
1160800611 J * bronson_ ~bronson@c-71-198-75-160.hsd1.ca.comcast.net
1160800611 Q * bronson Quit: Ex-Chat
1160800845 Q * Johnnie Quit: G'bye!
1160803020 J * dna_ ~naucki@p54BCF20B.dip.t-dialin.net
1160803036 M * Bertl okay, finally off to bed now ... back later!
1160803043 M * Bertl have a good one everyone! cya!
1160803048 N * Bertl Bertl_zZ
1160803933 Q * dna_ Quit: Verlassend
1160804780 Q * matti Ping timeout: 480 seconds
1160808654 J * Piet hiddenserv@tor.noreply.org
1160810956 J * meandtheshell ~markus@85-124-232-103.work.xdsl-line.inode.at
1160814754 J * Johnnie ~jdlewis@66.199.231.180
1160815411 J * bonbons ~bonbons@83.222.36.111
1160817281 Q * jkl Quit: Lost terminal
1160821280 M * bonbons Did anyone try out nbd for vserver guest partitions? I get kernel panic when I try to login to a guest with its / on a nbd device...
1160821464 M * bonbons kernel: 2.6.17.11 + vserver 2.1.1-rc31 + IPv6 patch, nbd client&server 2.8.6; panic with recursive die() when handling pagefault
1160822927 N * nammie Nam
1160823479 J * Piet_ hiddenserv@tor.noreply.org
1160823900 Q * Piet Ping timeout: 480 seconds
1160824219 J * ruskie_ ~ruskie@84.20.228.4
1160824280 Q * ruskie Killed (NickServ (GHOST command used by ruskie_))
1160824313 N * ruskie_ ruskie
1160824340 Q * Piet_ Remote host closed the connection
1160824391 J * Piet_ hiddenserv@tor.noreply.org
1160828590 Q * ensc Killed (NickServ (GHOST command used by ensc_))
1160828600 J * ensc ~irc-ensc@p54B4FB3F.dip.t-dialin.net
1160832964 M * bonbons regarding my panic from 3 hours ago, more details available here: http://paste.linux-vserver.org/489
1160833157 Q * starlein Ping timeout: 480 seconds
1160833610 M * derjohn bonbons, _are_ does not nbd, but drbd as guest partitions. And that runs very fine.
1160833662 J * Floh ~Floh@pD9E4710D.dip.t-dialin.net
1160833681 Q * Floh 
1160833716 M * bonbons derjohn: drbd, that's the replication-device, no?
1160833740 M * daniel_hozac bonbons: does nbd work in vanilla?
1160833781 M * bonbons I works when used from host (with guest running nerby) but crashes my box when used as root partition for a guest
1160833855 M * derjohn bonbons, drbd is a distributed block device .. in that sense you could call it a "replication device". 
1160833858 M * bonbons I nerver tried a vanilla host with nbd as root partition ;)
1160834014 M * bonbons regarding the output from kernel, not so sure what is causing the trouble (or where exactly it crashes), there are multiple traces flying over the screen, and only last one remains visible
1160834126 M * bonbons when I connected to the guest via ssh, the screen (not ssh session) just showed [xxxxxxx.xxx] ============= scrolling by the screen (constant time, last line flickering so I assume scrolling)
1160834238 M * daniel_hozac serial/net console?
1160834254 M * daniel_hozac bonbons: does it work ok on the host if you just use chbind?
1160834278 M * bonbons when I tried netconsole I couldn't get it to work, serial console is not possible as there is no serial connector on that laptop
1160834350 M * daniel_hozac USB-to-serial? i suppose that might be a bit overkill though.
1160834359 M * bonbons the nbd-client is run from host (so normally not affected), for the guest it *should* just be some part of vfs
1160834427 M * bonbons that I could try, but I don't have a "crossover serial cable"...
1160834436 M * bonbons aka null modem cable
1160834510 M * daniel_hozac what was the problem with netconsole?
1160834548 M * bonbons I never saw any packet on the target machine, not even using tcpdump
1160834583 M * daniel_hozac is it on the same segment?
1160834595 Q * Vudumen Remote host closed the connection
1160834622 M * daniel_hozac i don't know how netconsole works, but from the options i got the impression it required a host on the same network.
1160834624 M * bonbons all connected to same switch
1160834653 J * Vudumen 072544468a@perverz.hu
1160834659 M * bonbons and ip addresses also in same subnet
1160834694 M * daniel_hozac what options did you give netconsole? did you get the startup successful messages?
1160834716 M * bonbons don't remember...
1160834754 M * daniel_hozac and you did have netconsole support, right? :)
1160834810 M * bonbons I did compile it, once as module, once built in, none did help
1160834844 M * daniel_hozac humm.
1160834894 M * daniel_hozac i've got it as module, and just do modprobe netconsole netconsole=6665@<client IP>/<client dev>,6666@<server IP>/
1160834899 M * bonbons for the module version I tried loading it with different parameters
1160834950 M * bonbons I can try again, I have it compiled as module on the "problematic" laptop
1160835195 Q * glut Ping timeout: 480 seconds
1160835213 M * bonbons got 1 new line in dmesg after modprobing netconsole, never arrived at the other end...
1160835527 M * bonbons daniel_hozac: from the led on the switch, looks like no packet is ever leaving my laptop, in dmesg I get 7 lines indicating netconsole configuration...
1160835648 M * bonbons I do "modprobe netconsole netconsole=10000@192.168.0.123/eth0,10000@192.168.0.124/00:11:22:33:44:55" (using correct mac/ip addresses), and those values also show up in kernel log
1160835700 M * bonbons might it be that my broadcom network interface is not supported by netconsole?
1160835730 Q * ntrs Remote host closed the connection
1160836113 M * daniel_hozac i would've thought it was independent of the network card.
1160836182 M * daniel_hozac or well, the driver i mean.
1160836401 M * bonbons daniel_hozac: As far as I read it looked like kernel pushes packets directly to the card and has some requirements for this... (no interrupts or such) does it work for you?
1160836444 M * daniel_hozac indeed, i've used it in the past to capture some panics.
1160836487 M * daniel_hozac using the forcedeth driver.
1160836665 M * bonbons wonder why it doesn't work for me!
1160837408 M * bonbons and again panic...
1160837456 M * bonbons an I got something sent over network, looks like only messages with sufficiently high severity get sent over netconsole
1160837467 M * daniel_hozac ah
1160837515 M * bonbons look here: http://paste.linux-vserver.org/490
1160837629 N * Bertl_zZ Bertl
1160837636 M * Bertl morning folks!
1160837648 M * Bertl reg. netconsole: no it is not driver independant
1160837662 M * Bertl the driver needs to conform to the netpoll API
1160837667 M * bonbons Hey Bertl!
1160837677 M * daniel_hozac ah, ok!
1160837694 M * Bertl I know that because I adapted the netconsole stuff for 2.4 back then :)
1160837720 M * bonbons good reason :)
1160837773 M * Bertl interesting oops
1160837802 M * bonbons but it's not the same as the one I got a few hours earlier, but all somehow related to nbd
1160838132 M * Bertl it's a preemption bug
1160838147 J * ntrs ~ntrs@68-188-51-87.dhcp.stls.mo.charter.com
1160838242 M * Bertl welcome ntrs!
1160838272 M * bonbons not so good...
1160838514 M * bonbons Bertl: you mean preemption bug as in unsynchronized access to some data-structure?
1160838767 M * Bertl no, assymetry in pairing
1160838783 M * Bertl i.e. the preemption count is decremented more often than incremented
1160838797 M * Bertl (once it goes below zero, you hit this bug)
1160838823 M * bonbons ok... so question is where does that buggy decrementing come from!
1160838858 M * Bertl yeah, most likely it happened somewhere else, so I would not necessarily check in the code in the dump
1160838866 M * bonbons I think that know I know how to use netconsole I will reproduce the "first" crash, to check if the initial cause is the same
1160838884 M * Bertl yep, that could help
1160838902 M * Bertl if it happens in the same place (backtrace) then you want to look there
1160838914 M * bonbons lets make the kernel panic ;)
1160839421 M * bonbons Bertl: here is the 'original' one: http://paste.linux-vserver.org/491
1160840049 M * bonbons what is in common is the task: idle task
1160841000 M * Bertl nah, IMHO the trace doesn't show what really happens
1160841075 M * bonbons my impression is that it's linked to nbd
1160841094 M * Bertl could be, double check the preemption stuff there
1160841105 M * Bertl btw, is the preemption and mutex debug stuff enabled?
1160841117 M * Bertl (if not, I'd do so :)
1160841224 M * bonbons will check, right now I'm reading through nbc.c, not found mention of preemtion (or smp) yet, but there are semaphors, mutexes and friends
1160841359 M * bonbons cool, I have both preemtion and mutex debugging compiled in :)
1160841615 M * bonbons anything to tell the kernel at boot/run-time regarding both options?
1160841630 M * Bertl no, should work out of the box, IIRC
1160842482 M * bonbons I think I will issue a few more panics, they seem to happen at a random location, maybe we can get a better picture then (interpolating all locations will certainly not be easy)
1160842728 M * Bertl as I said, I supsect the real issue to happen somewhere else
1160842737 M * Bertl i.e. 1 increment, two decrement
1160842762 M * Bertl once the last pair decrements *bang*
1160842903 M * bonbons yep, looks like that, what's the best way to proceed?
1160843272 M * Bertl source code review, eliminating all drivers/stuff you do not need
1160843301 M * Bertl etc., etc. if it is reproducible, try upgrading/downgrading the kernel
1160843405 M * bonbons eliminating unneeded stuff is at least partially done, I guess upgrading/downgrading would be the next step
1160843469 M * bonbons checking with vanilla(+squashfs?) should be possible as well...
1160843494 M * bonbons let's try and see :)
1160846538 Q * derjohn2 Ping timeout: 480 seconds
1160846546 J * derjohn2 ~aj@dslb-084-058-241-103.pools.arcor-ip.net
1160847074 A * bonbons eating while make vanilla-2.6.17.14; make vanilla-2.6.18.1 
1160847147 M * phreak`` .14 ? hrmmmm
1160847155 M * phreak`` morning everyone btw :)
1160847185 M * Bertl good morning!
1160847230 M * phreak`` Bertl: the morning isn't that good :) anyway morning is faar away :)
1160847235 M * phreak`` Bertl: heya :)
1160847354 M * phreak`` guess I missed rc32-38 up till now :)
1160847364 M * Bertl *39
1160847375 M * phreak`` eh :) rc39 then :)
1160847375 M * mnemoc *g*
1160847513 M * phreak`` Bertl: where's _rc32 exactly ? :)
1160847522 M * daniel_hozac there was no rc32 :)
1160847529 M * phreak`` ah, ok :)
1160848905 Q * Borg- Quit: fuck this.. When the worlds collide...
1160848910 M * Bertl doener, daniel_hozac: here is some 'proof of concept' code for a 'new' fake init approach ... http://vserver.13thfloor.at/Experimental/delta-fakeinit-feat01.diff
1160848922 M * Bertl I'm off for now, but will be back later ...
1160848928 N * Bertl Bertl_oO
1160848986 Q * bronson_ Read error: Operation timed out
1160849000 J * bronson_ ~bronson@adsl-64-161-106-11.dsl.snfc21.pacbell.net
1160850039 M * doener Bertl_oO: !vx_flags(VXF_FAKE_INIT, VXF_FAKE_INIT) ? is the ! correct?
1160851580 J * shedii ~siggi@130.208.221.254
1160851744 Q * shedi Read error: Operation timed out
1160852835 M * bronson_ Is there an easy way to fake a localhost interface in my vservers?
1160853079 M * bonbons bronson_: fake in what respect? What you can do is map 'localhost' to your guest's IP address in /etc/hosts
1160853105 M * bronson_ bonbons: sounds good to me.  thanks!
1160853160 M * bronson_ Is it a linux-vserver thing to ensure that the hostname and the vm name are the same?  Or is that just an assumption made by the build scripts?
1160853195 M * bronson_ I'd like to name the vm 'rinspin' but give it the hostname of 'hydra.rinspin.com'... 
1160853214 M * bronson_ Would I cause myself grief if I were to change things?
1160853638 M * bonbons bronson_: both can be different, just issue 'hostname mynewname' to change the guest's hostname
1160853683 M * bronson_ I figured it was just a feature of the build scripts; just wanted to be sure.
1160853828 J * mire ~mire@79-167-222-85.adsl.verat.net
1160854627 M * bonbons Bertl_oO: I tried with vanilla 2.6.17.14 and 2.6.18.1, both crash the same way. For 2.6.18.1 (not tried with 2.6.17.14) when using ScrollLock I get emerge to freeze into disk-sleep state (and not kernel panic)
1160854700 M * bonbons any access to that partition ends up disk-sleep, so there's definitly some race-condition in nbd!
1160856072 M * waldi more than on cpu?
1160856212 M * bonbons non, single-processor, single core, second generation PentiumM, but with preemption enabled
1160856252 M * bonbons I guest multi-processor would have same kind of trouble (but no such box here to try out)
1160856347 M * bronson_ I get two errors when shutting down my vserver...   * Terminating any remaining processes...             cat: /proc/cmdline: No such file or directory
1160856366 M * bronson_ That looks like a harmless outcrop of Ubuntu's bootsplash stuff
1160856424 M * bronson_ (Though I don't know why /proc/cmdline isn't found; it's there when the machine is booted)
1160856441 M * bronson_ But this one...
1160856449 M * bronson_  * Unmounting remote filesystems...           mount: permission denied
1160856455 M * bronson_ Is that typical?
1160856566 M * bonbons looks your guest tries to unmount things and your guest does not have sufficient mount caps, normally you can just disable all unmounting inside guests
1160856636 M * bronson_ update-rc.d -f umountnfs.sh remove; update-rc.d -f umountnfs.sh stop 90 2 3 4 5 .
1160856656 M * bronson_ Yeah, that got rid of the unmount error.   Thanks bonbons.
1160856674 M * bronson_ That /proc/cmdline missing is odd but I can live with it.
1160856727 M * bronson_ So now I have a clean Dapper VM to start from.  Excellent.  Time to start cloning it...
1160856838 M * bonbons Bronson, /proc/cmdline is hidden for vserver-guests. A virtualised one would be cool, but that for the time when someone decides to do so
1160857670 J * dna_ ~naucki@p54BCD378.dip.t-dialin.net
1160858189 J * glut glut@no.suid.pl
1160858944 N * sladen mynickislongerthanyours
1160858997 N * mynickislongerthanyours sladen
1160859327 Q * duckx Remote host closed the connection
1160859674 J * duckx ~Duck@tox.dyndns.org
1160860353 Q * dna_ Quit: Verlassend
1160862222 Q * Piet_ Quit: Piet_
1160863203 Q * h01ger Quit: h01ger
1160863209 J * h01ger ~holger@socket.layer-acht.org
1160863973 Q * meandtheshell Quit: exit (0);
1160866309 J * _node node@c-69-143-90-233.hsd1.md.comcast.net
1160866384 M * bonbons Bertl_oO: I filed a bug regarding nbd on mainline (2.6.18.1 mostly locks up in disk-wait on ndb device, prior panic): http://bugzilla.kernel.org/show_bug.cgi?id=7364
1160866458 Q * bonbons Quit: Leaving
1160866597 N * Bertl_oO Bertl
1160866968 M * Bertl doener: nope that is basically DeMorgan
1160866997 M * Bertl doener: i.e. normally we have vx_flags(A|B,0)
1160867019 M * Bertl which is (A|B)^0
1160867038 Q * _node Ping timeout: 480 seconds
1160867065 M * Bertl now we have [(A|B) & c]^(A|B)
1160867107 M * Bertl which is a negation of the result, so we end uo as !(A&B)^0
1160867194 J * _node node@c-69-143-90-233.hsd1.md.comcast.net
1160868212 J * bronson__ ~bronson@c-71-198-75-160.hsd1.ca.comcast.net
1160868602 Q * bronson_ Read error: Operation timed out