1179360378 Q * onox Quit: zzzz 1179362919 J * zLinux ~zLinux@88.213.56.155 1179364163 J * ntrs ntrs@68-188-55-120.dhcp.stls.mo.charter.com 1179366096 Q * bzed Quit: Leaving 1179367656 Q * meandtheshell Quit: Leaving. 1179369294 N * ensc Guest158 1179369304 J * ensc ~irc-ensc@p54B4D617.dip.t-dialin.net 1179369409 Q * Guest158 Ping timeout: 480 seconds 1179369473 Q * baldy Remote host closed the connection 1179371393 Q * djbclark Remote host closed the connection 1179371423 J * djbclark dclark@opensysadmin.com 1179374361 Q * ahuman Quit: Leaving 1179376616 M * Bertl okay, off to bed now .. have a good one everyone! 1179376620 N * Bertl Bertl_zZ 1179378964 J * djzhan ~zhan@58.62.76.208 1179380494 Q * djzhan Quit: Leaving 1179381593 J * ravid ~fooz@ecstasy.openbsd.org.il 1179382637 Q * FireEgl Ping timeout: 480 seconds 1179384814 J * dna ~naucki@34-192-dsl.kielnet.net 1179387759 J * FireEgl FireEgl@4.0.0.0.1.0.0.0.c.d.4.8.0.c.5.0.1.0.0.2.ip6.arpa 1179388957 Q * besonen_mobile Quit: Leaving 1179388993 J * besonen_mobile ~besonen_m@71-220-227-185.eugn.qwest.net 1179389245 J * DavidS ~david@p57A4A547.dip0.t-ipconnect.de 1179390640 J * bonbons ~bonbons@ppp-111-132.adsl.restena.lu 1179391754 Q * FireEgl Read error: Connection reset by peer 1179392720 J * FireEgl FireEgl@Sebastian.Atlantica.DollarDNS.Net 1179395394 Q * cehteh Ping timeout: 480 seconds 1179395494 J * cehteh ~ct@pipapo.org 1179396429 M * CHTEKK back, yo all! 1179397175 Q * FloodServ resistance.oftc.net synthon.oftc.net 1179397175 Q * djbclark resistance.oftc.net synthon.oftc.net 1179397175 Q * ntrs resistance.oftc.net synthon.oftc.net 1179397175 Q * ruskie resistance.oftc.net synthon.oftc.net 1179397175 Q * besonen resistance.oftc.net synthon.oftc.net 1179397175 Q * mattzerah resistance.oftc.net synthon.oftc.net 1179397175 Q * mugwump resistance.oftc.net synthon.oftc.net 1179397175 Q * bragon resistance.oftc.net synthon.oftc.net 1179397175 Q * weasel resistance.oftc.net synthon.oftc.net 1179397175 Q * besonen_mobile resistance.oftc.net synthon.oftc.net 1179397175 Q * Aiken resistance.oftc.net synthon.oftc.net 1179397175 Q * phreak`` resistance.oftc.net synthon.oftc.net 1179397175 Q * hardwire resistance.oftc.net synthon.oftc.net 1179397175 Q * lylix resistance.oftc.net synthon.oftc.net 1179397175 Q * brcc_ resistance.oftc.net synthon.oftc.net 1179397175 Q * soltesz resistance.oftc.net synthon.oftc.net 1179397271 J * weasel weasel@weasel.noc.oftc.net 1179397288 J * besonen_mobile ~besonen_m@71-220-227-185.eugn.qwest.net 1179397288 J * Aiken ~james@121.45.222.137 1179397288 J * phreak`` ~phreak``@deimos.barfoo.org 1179397288 J * hardwire ~bip@rdbck-5851.wasilla.mtaonline.net 1179397288 J * lylix ~eric@dynamic-acs-24-154-33-109.zoominternet.net 1179397288 J * soltesz ~soltesz@aegis.CS.Princeton.EDU 1179397288 J * brcc_ bruce@i.am.someasshole.com 1179397294 J * djbclark dclark@opensysadmin.com 1179397294 J * ntrs ntrs@68-188-55-120.dhcp.stls.mo.charter.com 1179397294 J * ruskie ruskie@ruskie.user.oftc.net 1179397294 J * besonen ~besonen@dsl-db.pacinfo.com 1179397475 J * mugwump ~samv@watts.utsl.gen.nz 1179397475 J * mattzerah ~matt@121.50.222.55 1179397475 J * bragon ~bragon@sam.geeknode.org 1179398649 J * bzed ~bzed@dslb-084-059-110-069.pools.arcor-ip.net 1179400842 J * meandtheshell ~markus@85-125-192-122.dynamic.xdsl-line.inode.at 1179401330 J * Piet hiddenserv@tor.noreply.org 1179402483 Q * Piet Ping timeout: 480 seconds 1179402991 J * FloodServ services@services.oftc.net 1179403054 J * avi avi@193.30.161.200 1179403063 M * avi hey everybody 1179403077 M * avi lastest patch for kernel 1179403082 M * avi what is it ?~?/1/1?! /rn/rn/r/rnr/n 1179403104 M * daniel_hozac see topic or the front page of linux-vserver.org... 1179403117 Q * Aiken Quit: Leaving 1179403236 M * avi daniel_hozac, <3<3<#<# 1179403239 M * avi dude i love u ok !?!?! 1179403271 M * Hollow we love you too honey 1179403299 A * Hollow yawns 1179403303 M * Hollow morning :) 1179403355 M * CHTEKK yo Hollow 1179403372 M * CHTEKK calling it morning is stretching it for you... :) 1179403392 M * CHTEKK (I'm up since some 2 hours too :P yay holidays! :D) 1179403413 M * Hollow yay :) 1179403425 M * Hollow well, i'm up since 11am 1179403449 M * phreak`` wtf 1179403463 M * CHTEKK :D hi phedny 1179403466 M * CHTEKK err phreak`` 1179403478 M * phreak`` hey Luca ;) 1179403514 A * Hollow stabs phreak 1179403526 M * phreak`` Hollow: what ? 1179403534 M * phreak`` Hollow: remember, I'm NAT'ed ;) 1179403535 M * Hollow see jabber :P 1179403544 M * Hollow yeah, me too, but i configured port forwarding 1179405759 J * ema ~ema@rtfm.galliera.it 1179405768 Q * wibble Quit: Trillian (http://www.ceruleanstudios.com 1179406374 J * cruser ~chatzilla@72.242.194.162 1179406432 M * cruser daniel_hozac: ping 1179408426 N * Bertl_zZ Bertl 1179408430 M * Bertl morrning folks! 1179408812 M * daniel_hozac cruser: pong 1179408816 M * daniel_hozac morning Bertl! 1179408916 M * ravid morning Bertl <3 1179409085 M * cruser daniel_hozac: I compiled a vanilla kernel with the non-default memory split a few weeks ago. I getting ready to post to lkml about the issue. I am hesitant to post the 2,000 line configure lines in the .config file. Or should I post my meminfo and cpuinfo and wait from someone to ask to post the .config? 1179409119 M * daniel_hozac post at least the results of grep VMSPLIT .config and grep HIGHMEM .config 1179409128 Q * FireEgl Ping timeout: 480 seconds 1179409354 M * cruser Do you want to be cc'ed or bcc'ed in on the emails? 1179409368 M * daniel_hozac sure, cc me. 1179409383 M * cruser okay 1179409553 M * cruser daniel_hozac: what is you email address? 1179409578 M * daniel_hozac daniel@hozac.com 1179410296 J * FireEgl FireEgl@4.0.0.0.1.0.0.0.c.d.4.8.0.c.5.0.1.0.0.2.ip6.arpa 1179410774 M * cruser daniel_hozac: Could you look over the email before I send it to the kernel list? I sent it to your email address. 1179410806 J * Fercho ~Fercho@201.249.73.14 1179410816 M * Fercho hi 1179410880 M * daniel_hozac hello 1179410882 M * daniel_hozac cruser: sure. 1179411244 M * daniel_hozac cruser: looks fine to me. 1179411270 M * cruser okay I will send it and cc you in. Thanks. 1179411394 M * daniel_hozac thank you! 1179412257 Q * matti Ping timeout: 480 seconds 1179412384 J * matti matti@acrux.romke.net 1179412553 M * Fercho hello 1179412559 M * Fercho excuse e 1179412561 M * Fercho me 1179412586 M * Bertl Fercho: you are excused :) 1179412799 M * mjt hmm. If I bind-mount some directories on host to guest(s) - such as /var/lib/dhcp/ stuff - looks like i'm opening chroot restrictions, right? 1179412828 M * daniel_hozac hmm? 1179412856 M * mjt that "secure chroot barrier" thing - it doesn't work anymore 1179412866 M * mjt or does it? 1179412883 M * daniel_hozac it does. 1179413019 M * mjt ok 1179413111 M * CHTEKK back... hi Bertl! did you read wrt the scheduler? 1179413202 M * daniel_hozac (and reset the princeton machine ;)) 1179413216 M * Bertl CHTEKK: nope, care to give me a short overview? 1179413224 M * Bertl daniel_hozac: k, np ... 1179413306 M * CHTEKK Bertl, sure, well I was basically trying to set fillrate to 0, disable idle-time, and set the required min tokens to insane high values, on a CPU, so that in theory nothing would get scheduled on that particular CPU for that vserver (the other three I left at default settings) 1179413322 M * CHTEKK now according to /proc/virtual/x/sched, the settings were applied, but the CPU used anyway 1179413337 M * mjt blah.. util-vserver.org doesn't exist... ;) 1179413366 M * mjt (i typed util-vserver instead of linux-vserver in my browser ;) 1179413370 M * CHTEKK Bertl, http://paste.linux-vserver.org/1866 shows the problem 1179413392 M * CHTEKK the first cat was done with the vserver started correctly, I then just logged into it and started cpuhog -n 10 1179413417 M * CHTEKK as you can see in the subsequent cats, it looks like CPUID 3 is still being as heavily used as the others, when in theory it shouldn't be used at all 1179413450 M * daniel_hozac (this is what i was experimenting with on the princeton machine, see the sarge01 guest) 1179413481 J * shedi ~siggi@ftth-237-144.hive.is 1179413534 Q * Fercho Ping timeout: 480 seconds 1179413914 M * mjt as far as i understand... network is set up by "transferring" an IP address into guest's "addressspace", but preserving interface (name). Right? 1179413936 M * daniel_hozac no. 1179413943 M * mjt in other words, i basically can only "transfer" (grant) an existing host address 1179413954 M * daniel_hozac the guest is simply limited to some of the IP addresses. 1179413989 M * daniel_hozac and yes, the IP address has to be set up on the host. 1179414007 M * mjt that's obvious (set up on the host) 1179414063 M * mjt in other words, i basically can only "transfer" (grant) an existing host address -- in this case the guest "sees" an interface from host but with a single address only 1179414100 M * mjt (sure i can grant/transfer several addresses on the same interface - doesn't really matter) 1179414148 M * mjt the reason for this question is that /etc/vservers/$foo/interfaces/ stuff is somewhat confusing, esp. wrt interface naming 1179414168 M * daniel_hozac how so? 1179414206 M * mjt from the docs one may assume he is free to use whatever names he wants there. but in practice the names should be names of host interfaces. 1179414266 M * mjt lol. Hooray to vserver. 1179414330 M * mjt finally i managed to set up ntpd on this machine. It has some 500 dynamic tunnels, and ntpd had.. troubles with that mess (as it doesn't have an option 'listen address' - what a shame). Now i simple run it in a vserver with a single IP and everything's neat ;) 1179414361 M * mjt implementing 'listen address' externally ;) 1179414384 M * Bertl mjt: which names should be host interface names? 1179414414 M * mjt /etc/vservers/$foo/interfaces/$n/dev 1179414467 M * Bertl right ... I thought you meant the $n .... 1179414506 M * Bertl daniel_hozac: it locked up somehow or what? 1179414532 M * daniel_hozac yeah. 1179414546 M * mjt and ../interfaces/$n/name doesn't really do much - it becomes $dev:$name really 1179414572 M * Bertl yep, that is called alias 1179414579 M * mjt (nothing prevents it to be just $name) 1179414596 M * mjt "alias" term is long obsolete ;) 1179414608 M * daniel_hozac ip prevents it to be just $name. 1179414612 M * mjt no 1179414614 M * daniel_hozac yes. 1179414641 M * Bertl CHTEKK: did you enable the hardcpu scheduler, both kernel config wise as for that guest? 1179414644 M * mjt so i wonder how the .. i managed to have some 50 "aliases" with names like isp, intnet etc... 1179414664 M * mjt ifconfig dislikes them - that's right, but not ip. 1179414683 M * Bertl yes, you can name any interface arbitrarily 1179414694 M * mjt and any address too 1179414704 M * daniel_hozac # ip a a 192.168.99.3/24 dev wlan0 label test 1179414704 M * daniel_hozac "dev" (wlan0) must match "label" (test). 1179414770 M * mjt 6: eth-dmz: mtu 1500 qdisc pfifo_fast qlen 1000 1179414770 M * mjt link/ether 00:80:48:e6:68:7e brd ff:ff:ff:ff:ff:ff 1179414770 M * mjt inet 81.13.94.1/28 scope global eth-dmz 1179414770 M * mjt inet 192.168.2.1/29 scope global pvt 1179414770 M * mjt inet 81.13.94.2/28 scope global secondary nat 1179414773 M * mjt inet 81.13.94.3/28 scope global secondary cserv 1179414802 M * mjt i use `ip a add 81.13.94.2/28 dev eth-dmz label nat' for that 1179414811 M * Bertl daniel_hozac: sounds like a 'sane' userspace check to me 1179414880 M * mjt heh. And on another host, with more recent iproute package, that doesn't work anymore ;) 1179414893 M * mjt with exactly that error message 1179414938 M * mjt the one that works is from 2002, this is 2006 version 1179414952 M * CHTEKK Bertl, damn sorry, I was missing that one, gah didn't remember vcd doesn't auto-add it when setting the scheduler limits up 1179414994 M * Bertl CHTEKK: anyway, the hang daniel_hozac triggered sounds interesting, that might be a bug 1179415006 M * CHTEKK now it did this: before cpuhog: cpu 3: 10 13 0 0 0 R- 27 10000 12000 0/1 1/8 0 0, then I started it and it went: cpu 3: 37 13 0 0 0 H- 0 10000 12000 0/1 1/8 0 0 and remained stationary 1179415016 M * mjt thanks for the heads-up (unintentinal - but thanks anyway) -- you saved me quite..some troubles, as i planned to upgrade iproute 1179415021 M * daniel_hozac i think it's because i accidentally had idle-time enabled for the CPU, with a fill-rate2 of 0. 1179415025 M * CHTEKK so it increased minimally still, but I'm not sure if that's just expected by how things are done (little delay?) or whatever 1179415069 M * Bertl mjt: hehe, you're welcome! 1179415282 M * Bertl CHTEKK: if you didn't changethe 'token' settings, by defalt a context gets a calculated amount to start with 1179415299 M * CHTEKK yeah... and so it was using that up, ok got it, thanks! 1179415314 M * CHTEKK and yeah, vcd doesn't set the initial tokens, we let that be "default" 1179415328 M * CHTEKK so that explains it, great, it works then :) 1179415332 M * Bertl yes, should be fine, gives it a startup bonus :) 1179415429 M * CHTEKK hrmmm can't kill cpuhog anymore, that thing just went zombie... and still sucks resources... why :S 1179415455 M * CHTEKK I could always kill it with vx -kill... could it be because when it started it used some of cpuid 3, and then that one went on permanent hold? 1179415484 M * Bertl yep, could be 1179415489 M * daniel_hozac should get migrated to another CPU then, shouldn't it? 1179415505 M * Bertl not sure we have everyting in place for that atm 1179415525 M * Bertl i.e. in theory it _could_ get on the hold queue and stay there 1179415541 M * Bertl but that should be easy to detect from the proc info 1179415555 M * Bertl (and yes, if that is true, then it is a bug :) 1179415575 M * CHTEKK from which proc info? 1179415595 M * Bertl /proc/virtual/*/sched 1179415603 M * Bertl it should keep incrementing the hold count 1179415629 M * CHTEKK na, nothing changes there 1179415654 M * Bertl then the best explanation is, that the process was stopped (SIGSTOP) and thus cannot be killed atm :) 1179415663 M * CHTEKK http://paste.linux-vserver.org/1894 1179415667 M * Bertl try to send it a resume signla (SIGCONT) 1179415716 M * Bertl the last one, if that is your concern, is a zombie 1179415737 M * CHTEKK yeah I know, but it's the first time it goes zombie 1179415746 M * Bertl it should have got reaped by init 1179415762 M * Bertl well, maybe it still will be at some point 1179415762 M * CHTEKK usually a sigkil killed it, this time it went zombie and still sucks resources... so I stopped the vserver, and that then just hanged 1179415775 M * Bertl any messages in dmesg? 1179415849 M * CHTEKK Bertl, indeed: http://paste.linux-vserver.org/1895 1179415923 Q * FireEgl Ping timeout: 480 seconds 1179415950 M * Bertl CHTEKK: tx! will look into it shortly ... 1179415977 M * CHTEKK ok, thank you for your help! :) 1179415997 M * Bertl np 1179416054 Q * mugwump Ping timeout: 480 seconds 1179416130 J * stefani ~stefani@flute.radonc.washington.edu 1179416231 J * Punkie ~Punkie@home.pekelny.net 1179416253 M * Punkie hello 1179416348 M * Punkie i need some advise with sharing files between more vservers, please 1179416453 M * Punkie I am using mount --bind ..but when I create file in first vserver, I cant read and edit it in second vserver, i can only delete it 1179416509 M * Punkie Is this a question of any ccaps or bcaps options? 1179416531 M * DavidS do uids/gids match between guests? 1179416555 M * Punkie yes 1179416571 M * DavidS i ask because deleting is operation on directory, not the file 1179416615 M * daniel_hozac are you using tagxid? 1179416630 M * Punkie in second vserver I so that file with "?" front the name of file 1179416649 M * Punkie yes I use taxid, I dont know how to disable it 1179416683 M * Punkie so=see* 1179416699 M * Punkie sorry for my english :( 1179416745 M * daniel_hozac well, since you're using tagxid, you can't access files belonging to another context. 1179416750 M * daniel_hozac so that's perfectly expected. 1179416777 M * Punkie so I cant have parametr tagxid in /etc/fstab? 1179416836 M * CHTEKK daniel_hozac, when binding a service to 0.0.0.0:143 (all, port 143) inside a vserver which has assigned an IP, this should not influence at all another vserver with another IP, which trieds to bind too to 0.0.0.0:143, right? 1179416879 M * Punkie it is not posiible to disable it only for few vserver on the same partitions ..with other vservers, which needs to use tagxid? 1179416909 M * CHTEKK cause a friend of mine using util-vserver gets "make_sock failed, address already in use" inside vserver A with IP x.x.x.123 when starting an imap server, and the same imap server is also installed&started on vserver B with another IP, on the same host 1179417021 M * Bertl CHTEKK: definitely not, unless the ip subsets (between guests) overlap 1179417079 M * Bertl Punkie: what you can do is the following: 1179417106 M * Bertl Punkie: get another partition (could even be loopback on your existing one) and mount that somewhere without tagxid 1179417117 M * Bertl Punkie: then --bind mount it into the guests 1179417131 J * FireEgl FireEgl@Sebastian.Atlantica.DollarDNS.Net 1179417137 J * Guy- beeuc0H976@chardonnay.math.bme.hu 1179417163 M * Punkie Bertl: thanks a lot 1179417206 M * CHTEKK Bertl, http://rafb.net/p/RskSkp65.html that's his network config... probably wrong netmask? 1179417426 M * Bertl CHTEKK: yep, looks like :) 1179417627 M * CHTEKK what's the correct netmask for vservers usually? 255.255.255.255? 1179417755 M * Bertl usually the one coresponding to the assigned network :) 1179417758 M * DavidS usually the same as without vservers, isn't it? 1179417761 M * DavidS thanks 1179417763 M * DavidS :) 1179417815 M * Bertl it looks like /25 in your case 1179417869 M * CHTEKK yeah 1179418220 M * CHTEKK ok works now, was some problem with ip assignement and netmask it seems, thanks for the help :) 1179418265 M * CHTEKK otoh /me is asking myself... one the server here I have on the host IP 192.168.10.210 with netmask 255.255.255.0, ip addr ls on the host confirms it 1179418281 M * CHTEKK then I assigned that IP to a vserver while specifying a netmask of 255.255.255.255 1179418302 M * CHTEKK in the vserver, it tells me the netmask is 255.255.255.0... which is kinda expected cause that's what the host tells 1179418319 M * CHTEKK and afaik vserver just limits what ips and network stuff the vserver sees, but the config is all up to the host 1179418336 M * CHTEKK so... why can I actually specify a netmask when assigning an ip to a vserver? does that make sense? 1179419010 Q * DavidS Quit: Leaving. 1179419148 Q * ema Ping timeout: 480 seconds 1179420621 Q * Punkie Read error: Connection reset by peer 1179420649 J * Punkie ~Punkie@home.pekelny.net 1179420655 M * CHTEKK Bertl, well it seems cpuhog going zombie anyway isn't (fully) related to that kernel bug... I rebooted and redid the whole experience, this time without trying to stop the vserver... so I still have cpuhog running as zombie and being unkillable, using resources, and nothing in dmesg this time 1179420910 M * Bertl yep, was probably a corner case 1179420980 M * CHTEKK that still can hopefully be fixed ;) but any ideas wrt what's happening atm? maybe really something to do with the fact it started when all 4 cpuids were available, and now on hold on that one? 1179421245 M * CHTEKK Bertl, ok I can confirm that was it 1179421291 M * CHTEKK I just did vx -sched-set FILL_RATE=10 FORCE to reset the fillrate for that vserver, and the zombie disappeared immediately 1179421336 M * Guy- what does "vserver build -m clone" use the distribution argument for? 1179421385 M * daniel_hozac initpost/initpre scripts, e.g. 1179421413 M * Guy- do I need to specify it for Debian? 1179421633 M * daniel_hozac probably not. 1179422101 Q * Punkie Read error: Connection reset by peer 1179424036 P * cruser 1179424042 M * Guy- hashify is skipping files like usr/lib/perl/5.8.8/auto/ByteLoader/ByteLoader.bs - any idea why? 1179424120 M * Bertl ll usr/lib/perl/5.8.8/auto/ByteLoader/ByteLoader.bs 1179424322 M * Guy- ll? 1179424336 M * Bertl ls -la 1179424351 M * Guy- ah, it's a zero byte file :) 1179424368 M * Bertl not worth hashifying those :) 1179424417 M * waldi but it still not skips suid files and makes dpkg choke on them 1179424441 J * Johnnie ~jdlewis@c-67-163-247-109.hsd1.pa.comcast.net 1179424636 M * Guy- waldi: how so? 1179424663 M * waldi it does some weird things and fails with EPERM 1179424725 M * daniel_hozac not if you use a recent kernel 1179425046 M * CHTEKK Bertl, I have another one for you: http://paste.linux-vserver.org/1899 that happens when calling sched-info with some weird cpuid like -1 1179425135 M * daniel_hozac hmm, that should be fine. 1179425148 M * daniel_hozac what does addr2line say? 1179425220 M * CHTEKK daniel_hozac, you talking to me ? 1179425224 M * daniel_hozac yes. 1179425264 M * CHTEKK hmm what should I pass addr2line? 1179425387 M * daniel_hozac addr2line -e /source/tree/vmlinux ffffffff802a4e2b 1179425428 M * CHTEKK that means I need to have /vxmlinux of the compiled kernel around... damn 1179425454 M * CHTEKK well lemme recompile the kernel, keep those files this time, and reboot, and reproduce it 1179425607 M * CHTEKK ok /me off to eat while it compiles, cya in a bit 1179425792 M * waldi hmm, is there a tool which allows me to switch to a different network context? 1179425844 M * Bertl ncontext? 1179425891 M * waldi yep 1179426254 J * cruser ~chatzilla@72.242.194.162 1179426263 M * Bertl welcome cruser! 1179426279 M * cruser Thank you! 1179426326 M * cruser daniel_hozac: ping 1179426343 M * daniel_hozac pong 1179426405 M * cruser daniel_hozac: Should I expect any communication from the post to kernel.org? 1179426420 M * daniel_hozac how do you mean? 1179426438 M * Bertl probably when the kernel folks wake up, and read it ... yes :) 1179426446 M * cruser Will anyone reply to the email? 1179426457 M * cruser Should I be looking to answer? 1179426510 M * daniel_hozac yeah, hopefully ;) 1179426557 M * cruser Okay. I do not know what I am going to tell them. 1179426592 M * cruser Compiling kernels is not my area :) 1179426625 M * cruser Might they want me to try more compiles? 1179426638 M * daniel_hozac probably, if they get a fix. 1179426781 M * cruser daniel_hozac: how long should I wait? A week, month etc. 1179426816 M * daniel_hozac well, usually doesn't take more than a day to get the first reply... 1179426842 M * cruser Okay. Is it alright if I keep you posted? 1179426857 M * daniel_hozac well, you cc:ed me, right? 1179426876 M * cruser so they would post to you as well? 1179426882 M * daniel_hozac yeah. 1179426893 M * cruser okay, thanks. 1179426902 Q * ravid 1179426903 M * CHTEKK daniel_hozac, http://paste.linux-vserver.org/1902 1179426989 M * daniel_hozac CHTEKK: you need CONFIG_DEBUG_INFO too. 1179427071 M * CHTEKK which I can set where in the kernel config? 1179427118 M * CHTEKK found it 1179427796 M * CHTEKK daniel_hozac, ok this time it worked: http://paste.linux-vserver.org/1904 1179427816 M * CHTEKK seems like line 203 of kernel/vserver/sched.c 1179428483 Q * ruskie Ping timeout: 480 seconds 1179428574 J * ktwilight_ ~ktwilight@217.120-66-87.adsl-dyn.isp.belgacom.be 1179428982 Q * ktwilight Ping timeout: 480 seconds 1179429884 J * ema ~ema@rtfm.galliera.it 1179430543 Q * ema Quit: leaving 1179430904 J * ruskie ruskie@ruskie.user.oftc.net 1179431093 M * Guy- vserver-stat says: 1179431094 M * Guy- 22 197 15.9+ 269.1M 1m14s82 0m29s00 1h29m37 havocnotes 1179431104 M * Guy- what doest 15.9+ mean? 1179431107 M * Guy- does 1179431260 M * daniel_hozac IIRC, that it overflowed. 1179431399 M * Guy- where does it overflow? what's the limit? 1179431445 M * daniel_hozac well, most likely it's just lying anyway. 1179431471 M * daniel_hozac you have an old kernel/util-vserver, yes? 1179431474 M * Guy- no 1179431484 M * Guy- 2.6.20.6-vs2.2.0 1179431491 M * daniel_hozac and util-vserver 0.30.213? 1179431492 M * Guy- util-vserver_0.30.213-1_amd64.deb 1179431502 M * daniel_hozac that's peculiar. 1179431504 M * Bertl interesting ... 1179431513 M * Guy- this guest is running lotus domino 1179431527 M * Guy- it's perfectly believable that it allocates a LOT of RAM 1179431543 M * daniel_hozac what does grep VM /proc/virtual//limit say? 1179431595 M * Guy- VM: 593352 0/ 640677 -1/ -1 0 1179431595 M * Guy- VML: 0 0/ 0 -1/ -1 0 1179431657 M * Guy- when I started this vserver, I still had util-vserver 0.30.213-rc1, in case that matters 1179431734 M * daniel_hozac >2 GiB of address space? 1179431786 P * juuva 1179432088 M * Bertl daniel_hozac: not that hard, actually 1179432108 M * Guy- daniel_hozac: 2GB physical RAM plus some swap, so I guess yes 1179432116 M * Bertl daniel_hozac: at least if you mean the sum of allocated address spaces 1179432122 Q * nebuchadnezzar Quit: I'll be back 1179432332 M * daniel_hozac well, i'm more surprised that util-vserver doesn't handle more than that. 1179432340 J * nebuchadnezzar ~nebu@zion.asgardr.info 1179432810 M * CHTEKK Bertl, daniel_hozac: any ideas wrt the two issues I posted? /me just curious :) 1179433548 J * Aiken ~james@121.45.222.137 1179433664 M * daniel_hozac you used -1? 1179433676 Q * ruskie Read error: Connection reset by peer 1179433955 M * daniel_hozac i assume it works as expected for any other value? 1179434052 J * ruskie ruskie@ruskie.user.oftc.net 1179434524 M * CHTEKK daniel_hozac, in that failure I used -1, lemme try with other invalid values 1179434564 M * daniel_hozac i guess we need to check for >= 0 too. 1179434662 Q * bonbons Quit: Leaving 1179434732 M * CHTEKK interesting, it gets killed and dmesgs' with -1 -2 -5 for example 1179434744 M * CHTEKK but not with something like -22 1179434761 M * CHTEKK and for positive values, there are no problems it seems 1179434789 M * CHTEKK it just returns invalid argument then 1179434792 M * daniel_hozac okay, that's what i thought. 1179434820 M * matti :) 1179434836 M * CHTEKK -1 to -12 die, while -13 and onwards seem to work 1179434853 M * CHTEKK still any cpuid < 0 is invalid, always, I think 1179434873 M * daniel_hozac yep. 1179435086 M * matti ;] 1179435101 M * CHTEKK hi matti 1179435121 M * matti Hi CHTEKK :) 1179435149 M * CHTEKK daniel_hozac, also I'm not sure if sched-get works, at least this is what I see: http://paste.linux-vserver.org/1912 as you can see for the four cpuids, it always shows the same values, while they should differ 1179435198 M * CHTEKK http://svn.linux-vserver.org/projects/libvserver/browser/trunk/src/tools/vx.c#L520 the code for schedget looks ok... so maybe kernel issue? 1179435421 M * daniel_hozac where's the vx_sched_get implementation? 1179435581 M * daniel_hozac http://people.linux-vserver.org/~dhozac/p/k/delta-cpu-fix01.diff should fix it. 1179435592 M * daniel_hozac Bertl: or is there something wrong with that? 1179435645 M * Bertl I wonder why cpu_possible() should not check for negative values (somehow) 1179435656 M * daniel_hozac well, it basically expands to a bit test. 1179435682 M * daniel_hozac i guess they leave negative values to the caller. 1179435697 M * CHTEKK daniel_hozac, http://svn.linux-vserver.org/projects/libvserver/browser/trunk/src/lib/sched.c#L25 1179435702 M * Bertl and that works/succeeds with a negative value? 1179435735 M * Bertl daniel_hozac: we simply could cast it to unsigned, no? 1179435753 M * Bertl but the fix is fine, just funny that it isn't already covered 1179435795 M * daniel_hozac but isn't the problem that -1 == 0xffffffff, which i assume will always make the bit test true? 1179435826 M * Bertl why should that give a true? 1179435843 M * Bertl I mean, assumed that cpu_possible() would work 1179435859 M * daniel_hozac well, i don't know how the bit test is implemented. 1179435862 M * Bertl then both, -1 and 0xFFFFFFFF should yield false 1179435874 M * daniel_hozac sure. 1179435892 M * daniel_hozac but how would a cast to unsigned take care of the problem then? 1179435915 M * Bertl now the question is, does cpu_possible() do the right thing for large positive values? 1179435928 M * daniel_hozac CHTEKK: i don't see you setting the mask anywhere? 1179435943 M * Bertl or maybe negative values have a special meaning? 1179435955 M * CHTEKK Hollow, ^^ libvserver bug 1179435970 M * daniel_hozac CHTEKK: i'd say the bug is in vx. 1179435977 M * daniel_hozac oh, vx is also libvserver. 1179436009 M * CHTEKK yup.. Bertl good question, one I can answer with NO 1179436017 M * CHTEKK vx -sched-info 3 999999999999999999 1179436017 M * CHTEKK Killed 1179436077 M * CHTEKK hmm this time addr2line gives a different value 1179436078 M * CHTEKK include/asm/bitops.h:243 1179436107 M * CHTEKK http://paste.linux-vserver.org/1913 for the full error msg 1179436165 M * daniel_hozac so we have to check for that too... 1179436196 M * CHTEKK what's the upper limit in processors for linux, anyone knows that? 1179436208 M * daniel_hozac it's configurable. 1179436225 M * CHTEKK k 1179436483 M * CHTEKK Bertl, is there anything that can be done to fix the case where if a process when it started was also on a processor, which then went on hold cause of scheduler settings, and then you try to kill the proc, it just doesn't die? happens consistenly here with those scheduler settings :) 1179436500 M * CHTEKK can it even be fixed? 1179436598 P * cruser 1179437133 Q * dna Quit: Verlassend 1179437954 M * matti Hi Bertl. 1179438007 M * Bertl CHTEKK: I'd suggest a workaround or indirect approach 1179438042 M * Bertl CHTEKK: you could for example use cpu-sets to restrict the guest to certain cpus (an thus avoid running _any_ processes on that cpu) 1179438075 M * Bertl CHTEKK: or, you could configure the TB scheduler to grant a token every now and then (only minimal amount, e.g. 1 per second or so) 1179438095 M * Bertl that will take the process from the hold queue and reschedule it 1179438119 M * CHTEKK k, yeah, that way works... and yeah the real solution are cpusets 1179438141 M * CHTEKK and it seems you can only modify those through proc... that sucks... 1179438157 J * mugwump ~samv@watts.utsl.gen.nz 1179438162 M * CHTEKK probably can also be done using some C stuff and kernel, but the docs on any of that are < 0 1179438172 M * Bertl util-vserver has support for cpu-sets :) 1179438188 M * CHTEKK yup, using bash and the /proc/cpuset fs 1179438210 M * daniel_hozac /dev/cpuset. 1179438217 M * CHTEKK dev, right 1179438239 M * CHTEKK still if that's the only way, I'll probably write up something for vcd to modify the stuff in /dev/cpuset... 1179438260 M * Bertl shouldn't be too hard ... 1179438265 M * daniel_hozac that is the only way. 1179438273 M * CHTEKK ok then, that seals it :) 1179438320 M * CHTEKK it's not hard, that wasn't what worried me, it's just I'd have preferred to find a way to do it directly using C and syscalls etc, as anything else in vcd is basically done like that 1179438344 M * CHTEKK still, not too tragic... I'll work on it during the next days then 1179438356 M * CHTEKK thanks for your help! :) 1179438752 Q * nebuchadnezzar Remote host closed the connection 1179438752 M * CHTEKK Bertl, daniel_hozac: any hope of getting http://www.bullopensource.org/cpuset/source.html 's virtualization patch into vs-sources? that would make cpusets for vservers much more useful imo 1179438797 M * mugwump those patches are against 2.6.9! 1179438804 M * CHTEKK (basically makes it as such that you can tell a cpuset to be the only thing the vserver can "see" from inside, ie calling /proc/cpuinfo would show only the assigned cpus) 1179438838 M * CHTEKK mugwump, I saw that too, but if there's anyone taht can tell if they can be ported to actual kernels, it's the vserver sources maintainers ;) 1179438857 M * CHTEKK maybe that stuff didn't change much, no idea myself, never used cpusets myself yet 1179438862 M * mugwump http://linux-vserver.org/util-vserver:Documentation#.2Fetc.2Fvservers.2Fvserver-name.2Fcpuset 1179438921 M * mugwump 2.6.9 is literally prehistoric - linux.git stops at 2.6.12 1179438935 M * mugwump it's like version 1 CE :) 1179438951 M * mugwump (2.6.12 = 1 CE, that is) 1179438959 M * CHTEKK I very much know it's damn old, still it's the only patch I could find that does that, and I ask if it's possible to port/integrate that functionality to new vserver-sources :) 1179438996 M * daniel_hozac have you tried it? 1179439016 J * nebuchadnezzar ~nebu@zion.asgardr.info 1179439049 M * CHTEKK ehmmm no, I was just asking :P cause I saw it in the patch of the guy who did it for util-vserver 1179439123 M * daniel_hozac so why don't you try it? 1179439219 M * CHTEKK I will, first wanted to checkout cpuset support and make it work for vcd, then adding that feature, if the patch still works, is easy 1179439226 M * CHTEKK I'll report back on it then 1179440860 Q * ruskie Read error: Connection reset by peer 1179441428 M * Bertl okay, off to bed now ... have a good one everyone! cya! 1179441434 N * Bertl Bertl_zZ 1179442555 P * stefani I'm Parting (the water) 1179446201 Q * nou Ping timeout: 480 seconds