1148083453 P * stefani I'm Parting (the water) 1148083868 Q * mountie Quit: LUNCK! 1148084045 J * mountie ~mountie@CPEdeaddeaddead-CM000a739acaa4.cpe.net.cable.rogers.com 1148084983 M * Skram how can i add an ip without restarting a VPS? 1148084985 M * Skram :| 1148087525 J * sarnold ~sarnold@sarnold.noc.oftc.net 1148088372 Q * s0undt3ch Remote host closed the connection 1148088578 J * s0undt3ch ~s0undt3ch@bl7-244-34.dsl.telepac.pt 1148088785 N * sarnold sars 1148089968 Q * softi42 Ping timeout: 480 seconds 1148090592 J * softi42 khpadm@p549D6D28.dip.t-dialin.net 1148090608 Q * pbryan Remote host closed the connection 1148091767 Q * Methos Quit: Methos 1148093161 J * Methos ~mjoconr@hacker.pineview.net 1148097160 Q * Methos Quit: Methos 1148097733 J * Methos ~mjoconr@hacker.pineview.net 1148098413 Q * Zaki Ping timeout: 480 seconds 1148100414 Q * Methos Quit: Methos 1148101586 J * dna ~naucki@dialer-148-138.kielnet.net 1148101834 J * Methos ~mjoconr@hacker.pineview.net 1148103127 Q * dna Quit: Verlassend 1148106915 M * Hollow morning folks! 1148106946 M * Skram hello 1148106952 M * Hollow hey Skram 1148106963 M * Hollow did you notice the reboot workaround? 1148107078 M * Skram where? 1148107082 M * Hollow http://daniel.hozac.com/vserver/util-vserver-0.30.210-vshelper.patch 1148107083 M * Skram (no) 1148107095 M * Skram oh cool 1148107108 M * Hollow it should work, but it's not a clean bugfix 1148107118 M * Skram I have 0.30.210-r13 right now 1148107138 M * Hollow yes, just apply that patch to /usr/lib/util-vserver/vshelper 1148107138 M * Skram i dont quite get it though, heh 1148107168 M * Skram okay 1148107177 M * Skram but the problem is from INSIDE the vps 1148107185 M * Hollow yep.. 1148107207 M * Skram ok 1148107228 M * Skram # patch /usr/lib/util-vserver/vshelper util-vserver-0.30.210-vshelper.patch 1148107233 M * Skram like that? :) 1148107234 M * Hollow if you call sys_reboot inside the kernel calls vshelper on the host, and vshelper then restarts your guest 1148107259 M * Skram patching file /usr/lib/util-vserver/vshelper 1148107261 M * Skram done :) 1148107277 M * Skram so.. nothing needs to be done from inside/to each vps, right? 1148107282 M * Hollow right 1148107301 M * Skram ill test 1148107304 M * Skram watch vserver-stat 1148107305 M * Skram woops 1148107386 M * Skram 01:43:00 up 0 min, 0 users, load average: 1.53, 1.23, 1.16 1148107388 M * Skram cool 1148107403 M * Hollow :) 1148107406 M * Skram so.. you say it isnt a stable fix? 1148107413 M * Skram and where can i signup to be notified of changes 1148107421 M * Skram i recently joined the mailing list 1148107434 M * Hollow well, there is no announcement of this pathc.. 1148107438 M * Skram ok 1148107446 M * Hollow but the final fix will be in the next ebuild i guess 1148107452 M * Skram cool, cool 1148107510 M * Hollow as of 5 minutes ago, vcd is now also able to start, stop and restart guests :) 1148107516 M * Skram nice 1148107522 M * Skram but not able to be run on 2.0.1, right? 1148107528 M * Hollow right 1148107536 M * Skram hopefully i will have a nice dev box up by next weekend 1148107540 M * Skram finals coming up, and oy vey 1148107545 M * Hollow i just noticed it yesterday because i had 2.0.1 running for testing 1148107548 M * Hollow and vcd didn't work :o 1148107551 M * Skram :) 1148107554 M * Skram No wonder :) 1148107556 M * Skram Rock on, Hollow 1148107936 M * Skram Hollow: possible to add an ip without restarting the vps? 1148107981 M * Hollow technically, yes.. 1148107997 M * Skram mind explaining? 1148108002 M * Hollow but i don't know exactly how (with util-vserver ;) 1148108008 M * Hollow wait, i'll look 1148108022 M * Skram ;) 1148108023 M * Skram okay 1148108057 M * Hollow try: chbind --nid --ip 1.2.3.4/24 1148108087 M * Skram after i already added it to /etc/vservers/name/interfaces? 1148108097 M * Skram or will that just add it for the duration of guest-run time 1148108111 M * Skram and when rebooted it will add it because of the configs 1148108120 M * Hollow right 1148108122 M * Hollow http://paste.linux-vserver.org/53 ;) 1148108220 M * Skram http://paste.linux-vserver.org/43 does that do "the job" for what i want 1148108222 M * Skram to count ram 1148108248 M * Hollow uff 1148108249 M * Hollow :) 1148108252 M * Skram ? 1148108260 M * Skram hercules masilverberg # chbind --nid 17657 --ip 70.86.176.55 1148108260 M * Skram No command given; try '--help' for more information 1148108263 M * Skram what kind of command? 1148108267 M * Skram Usage: chbind [--silent] [--nid ] [--ip [/]] [--bcast ] [--] * 1148108279 M * Hollow hm 1148108297 M * Hollow chbind --nid --ip 1.2.3.4/24 -- true 1148108298 M * Hollow :) 1148108348 M * Skram hercules masilverberg # chbind --nid 17657 --ip 70.86.176.55/24 -- true 1148108349 M * Skram ipv4root is now 70.86.176.55 1148108363 M * Skram so i should be able to enter 17657 and do an ifconfig 1148108365 M * Skram and see that 1148108368 M * Hollow try it ;) 1148108386 M * Skram sentien-shells / # ifconfig | grep 55 1148108389 M * Skram nothin 1148108389 M * Skram g 1148108399 M * Skram just get the two other ip's masks from that grep 1148108403 M * Hollow do you have legacy networking disabled? 1148108408 M * Skram oh shoot, i think so 1148108417 M * Hollow ok, good 1148108441 M * Skram no, i have it enabled 1148108446 M * Hollow not good :) 1148108448 M * Skram legacy config is disabled. i got confused, sorry 1148108454 M * Hollow yeah 1148108464 M * Hollow the kconfig options are quite iiritating 1148108465 M * Skram CONFIG_VSERVER_LEGACYNET=y 1148108465 M * Skram # CONFIG_VSERVER_LEGACY is not set 1148108473 M * Skram :( 1148108487 M * Hollow you need to restart your guest then i guess 1148108627 M * Skram yeah 1148108641 M * Skram its a shell 'server' so ill do it when my tech who uses it for irssi as well knows 1148108660 M * Skram can we help by hosting any files or anything for you all? 1148108675 M * Skram I dont want to curse our server, but we have plenty of bandwidth and a fair ammount of space. 1148108897 M * Hollow hm, i'd be more interesting in testing hardware besides x86/amd64, i have plenty of hosting atm ;) 1148108924 M * Skram hmm 1148109403 M * Skram Hollow, thanks again for the linlk 1148109415 M * Skram an daniel_hozac: power. 1148109416 M * Skram thanls 1148109420 M * Skram *thanks 1148109429 M * Skram im going to look at more of those patches just to analyze and learn 1148109441 M * Skram i think most of them have been implemnented into my current version (r13) 1148109498 M * Hollow indeed 1148109693 M * Skram a friend wants to know.. 1148109712 M * Skram is VServer a shared kernel environment 1148109719 M * Skram the user can "read" the kernel but not write 1148109732 M * Skram and of course no access to anything ring 0, right? 1148109821 M * Hollow it is a shared kernel, and ring 0 == host, then yes 1148109830 M * Hollow and you can neither read or write a kernel 1148109834 M * Hollow you can just issue syscalls 1148109868 M * Skram you == a vps user :) 1148109901 M * Skram (21:23:00) TProphet69: it means that they do have access to the host if they can issue syscalls because if they are not issuing syscalls to the host what are they issuing them to? 1148109903 M * Hollow well, the programs you use issue syscalls ;) 1148109908 M * Skram you would be better at answering 1148109935 M * Hollow why doesn't he join here? ;) 1148109945 M * Skram i dont knpow 1148109962 M * Hollow ask him to join 1148109969 M * Skram could u just tell me 1148110005 M * Skram (21:24:04) TProphet69: well in ring 0 localsystem is assumed but if ring 1 it would be a userland context of some sort 1148110011 M * Skram just answer that please 1148110020 M * Hollow 1) there is no ring in vserver 1148110023 M * Hollow we have contexts 1148110042 M * Hollow 2) there is one kernel responsible for _all_ processes of _all_ contexts 1148110062 M * Skram okay 1148110065 M * Skram makes sense 1148110068 M * Hollow 3) depending of the context/flags and such, the kernel decides what to do when syscall xyz is called 1148110093 M * Hollow i.e. it can deny the execution of the syscall, or modify it's behaviour 1148110126 M * Hollow so, if you remove the CAP_MKNOD capability, no process in that context can call sys_mknod anymore 1148110266 M * Skram Thanks 1148110280 M * Skram this enabled me to help him better 1148110280 M * Skram thanks 1148110296 M * Skram i think i turned another linux-vserver 1148110298 M * Skram :) 1148110309 M * Hollow you just have to think of it as an extended chroot 1148110338 M * Hollow limit some things, show a bit different information, and voila 1148110347 M * Hollow and that's why there is nearly no overhead 1148110411 M * Skram right 1148110612 M * Skram thanks for that Hollow 1148110620 M * Hollow you're welcome! 1148110632 M * Skram did you say you wanted any x86 to test with or 64 bit? 1148110741 M * Skram this is a guy that works for and with microsoft servers all day 1148110742 M * Skram :0 1148110752 M * Skram :0 1148110761 M * Skram "it's definitely inferior to ours for some purposes, but ours is decidedly inferior for performance and usage of system resources." 1148110789 M * Skram "they're two equally valid engineering approaches. I actually wish I had the choice of you all's approach on the windows platform though." 1148110792 M * Hollow i talked about everything beside x86/amd64 1148110798 M * Skram oh, okay 1148110798 M * Skram hmm 1148110811 M * Skram i think thats all we have, as far as current hardware 1148110813 M * Skram sorry, Hollow 1148112556 Q * danderson Ping timeout: 480 seconds 1148114058 J * bonbons ~bonbons@83.222.38.81 1148115302 M * Skram oy. 22 hours awake.. i do almost this every night. this cant be healthy. 1148118110 M * eyck yeah, we all know that only 20 hours is healthy 1148118217 J * Zaki ~Zaki@212.107.125.102 1148120841 J * dna ~naucki@dialer-164-139.kielnet.net 1148121673 J * Banu ~Sohbet@85.99.36.228 1148121676 M * Banu s 1148123023 Q * Zaki Ping timeout: 480 seconds 1148123508 J * harry_ ~harry@p54A8680F.dip.t-dialin.net 1148123635 Q * harry_ Quit: 1148124736 J * Acemi-Kiz ~Sohbet@85.99.166.134 1148125098 Q * Banu Ping timeout: 480 seconds 1148126406 Q * mnemoc Quit: leaving 1148126735 J * doener ~doener@i5387C30A.versanet.de 1148126798 J * mnemoc ~amery@216.241.24.10 1148128133 Q * s0undt3ch Ping timeout: 480 seconds 1148128572 J * s0undt3ch ~s0undt3ch@bl7-253-90.dsl.telepac.pt 1148128635 M * derjohn is there a known problem with unix socket in a guest? I am struggeling to get cyrus/lmtp running, postfix via tcp does it, but not via mux .... 1148128720 M * mugwump unix sockets are in their own namespace, if you want to connect to another guest's sockets you need bind mount trickery 1148128761 M * derjohn mugwump, in this case all is within one guest. i 1148128804 M * derjohn m doing pam/ldap auth (which works), but cyrus cant access the mailboxes (yes, they are created ;)) 1148128860 M * mugwump at a guess, something's being chrooted. are any chrooted in postfix/master.cf? 1148128885 M * derjohn what strikes me is that postfix has problems to deliver via lmtp, while is runs perfect via tcp. 1148128927 M * derjohn I run lmtp without chroot (and others stuff also !chroot due to sasl auth, which gets painfully in chroot) 1148128968 M * derjohn the really strange thing is that cyrus pop3d and postfix seem to have problems in file access ... (more or less) 1148128989 M * mugwump is strace showing anything? 1148129007 M * derjohn mugwump, what would you recommend to strace? pop3d ? 1148129063 M * mugwump well, whichever bit is failing to connect 1148129171 M * derjohn gnarf, cyrus pop3d is started rom the cyrus master, I dunno how to make it stay in the foreground :( 1148129365 M * mugwump use strace -p to attach to a process 1148129379 M * derjohn mugwump, besdies that, what is the current status of your "mainline projkect" ? (http://lkml.org/lkml/2006/5/18/271) 1148129386 M * derjohn *besides 1148129403 M * mugwump ok, well what happened was the general approach was shot down, 1148129403 M * derjohn do we have a change to get vserver in? at least parts? 1148129425 M * mugwump but there have been some really cool recent developments, so the planning work is being reused 1148129440 J * night_watch ~mturk4719@212.174.55.29 1148129447 M * derjohn you mean the biederman stuff? 1148129455 M * derjohn or more evil magic? 1148129479 M * doener serge's stuff actually 1148129491 M * derjohn was does it do? 1148129511 Q * night_watch Quit: 1148129512 M * mugwump have a look at http://lkml.org/lkml/2006/5/18/182 1148129527 M * doener I just remember the uts namespace part, but IIRC the patch series includes a few more bits 1148129569 M * mugwump yeah, I think we've got uts, maybe sysvipc and time just about ready to go 1148129580 M * mugwump sounds like AM wants something on the network front next 1148129626 M * derjohn well i read that article ... (me remembers 'openvs') 1148129650 M * mugwump (as mentioned in http://lkml.org/lkml/2006/5/19/89) 1148129653 M * doener yeah, but I'm not really sure if it is a good idea to start with networking 1148129663 M * derjohn mugwump, that get interesting. the vz people have virtualized nw, nor? 1148129695 M * doener the idea was to see if that works, and if not, scrap the whole thing... but that might make us miss other stuff that _would_ make sense, even if the networking stays out of mainline 1148129704 A * derjohn imagines a v12n-ed kernel with parts of this and that and fresh stuff ....) 1148129712 M * mugwump I think that was the approach they took initially 1148129756 M * mugwump (vz people virtualising the network entirely, that is) 1148129760 M * derjohn is akpm really behind the v12n stuff? or just reacting to what the kernelpeople want? 1148129782 M * derjohn read: the community 1148129788 M * mugwump he's acting as spokesperson for the core team at the moment I guess 1148129794 M * mugwump being the -mm maintainer 1148129827 M * doener derjohn: hm, you obviously missed part of his talk at LinuxTag ;) 1148129897 M * derjohn doener, yes, I had to changes places with zeng! 1148129902 M * mugwump I think it's funny that Jacques Gelinas' vhost is being considered for merging now. it's almost 10 years old! :) 1148129911 M * derjohn (it was 50/50 deal ... I got the first half) 1148129927 M * doener mugwump: vhost? 1148129931 M * mugwump ip binding 1148129944 M * mugwump that was what it was originally submitted as 1148129964 M * doener hm, I only know the announcement from 2001 1148129985 M * derjohn wtf? 10 year old patch? it jaques still active in vs? 1148130065 A * mugwump digs up an archive link 1148130075 M * doener derjohn: well, did you see him here lately? ;) 1148130115 M * derjohn doener, that doesnt mean anything. powerfox/muh isnt here and a very active member ... 1148130146 M * derjohn (but for ensc it may be a true correlation) 1148130245 M * derjohn s0undt3ch, FYI: I filed Debian bugs for the stuff we found out (nss/sasl/ldap) 1148130507 M * mugwump http://www.ussg.iu.edu/hypermail/linux/net/9712.1/0039.html # I was wrong, vhost was userland only 1148130601 M * derjohn does anyone know how similar vserver foo ... is to the solaris zones stuff ? 1148130683 M * mugwump zones have quite a different approach to the filesystem stuff, called views 1148130687 M * derjohn mugwump, that link from 1999 ? so, vserver didnt start 2001 as i mention my FAQ ... but that wrapper was a userland *only* one? I wonder if/how it jailed IP? 1148130689 M * mugwump I don't know many details though 1148130721 M * derjohn mugwump, ah, you assume a view is some kind of a virtual fs ? 1148130724 M * mugwump derjohn: I doubt it jailed it in kernel space, I remember that I never found an actual submission for that feature 1148130750 M * mugwump derjohn: sure, like a database view (or reiser4 view) 1148130981 M * doener http://www.cs.helsinki.fi/linux/linux-kernel/2001-40/1065.html - that's the first "vserver" announcement I know of 1148131054 M * derjohn doener, yes, I linked that on john's hint in the wiki :) 1148131277 J * hillct ~H@client200-5.dsl.intrex.net 1148131287 Q * shedi Quit: Leaving 1148131292 M * hillct Morning all 1148131678 M * gdm hia 1148132280 M * derjohn mugwump: "+OK Mailbox locked and read".... if i set "unixhierarchysep: yes", I can create mailboxes but not use them, since i get "-ERR [SYS/PERM] Unable to locate maildrop: Mailbox does not exist". with "unixhierarchysep: no", everything is fine. is there a known bug? The boxes are on XFS. I use the Debian package of Etch. just FYI, I doubt that this bug is vserver related. 1148132471 Q * michal` Ping timeout: 480 seconds 1148132620 J * vrwttnmtu ~eryktyktu@82-69-161-137.dsl.in-addr.zen.co.uk 1148132631 M * vrwttnmtu Guten Tag, Herr Bertl 1148132632 M * vrwttnmtu :) 1148132730 J * michal` ~michal@www.rsbac.org 1148132772 M * doener I'm not sure if Bertl is really around. he had some trouble with his internet connection lately 1148132777 M * vrwttnmtu Aaah, OK 1148132870 M * vrwttnmtu I just came in because I saw on the Debian security list that they're releasing a new version of their vserver stuff for fixing "privilege escalation", and their new patch ver is 2.0.1-4 1148132906 Q * Methos Quit: Methos 1148132914 M * vrwttnmtu My current patch is patch-2.6.12.4-vs2.0.diff - just wondering if I need/there is an new patch? 1148132966 M * doener any details on what privilege escalation they're referring to? 1148132991 M * vrwttnmtu Excuse the paste: 1148132994 M * vrwttnmtu "Jan Rekorajski discovered that the kernel patch for virtual private servers 1148132994 M * vrwttnmtu does not limit context capabilities to the root user within the virtual 1148132994 M * vrwttnmtu server, which might lead to privilege escalation for some virtual server 1148132994 M * vrwttnmtu specific operations." 1148133009 M * vrwttnmtu "For the unstable distribution (sid) this problem has been fixed in 1148133009 M * vrwttnmtu version 2:2.0.1-4. 1148133009 M * vrwttnmtu We recommend that you upgrade your kernel-patch-vserver package and 1148133009 M * vrwttnmtu rebuild your kernel immediately." 1148133029 M * vrwttnmtu CVE-2006-2110 1148133046 M * vrwttnmtu http://www.cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-2110 1148133073 M * doener ok, the ccaps stuff... yeah, AFAIK all 2.6 releases were affected... daniel_hozac, do you remember when this was fixed? 1148133143 M * derjohn vrwttnmtu, doener this was 2.6 stuff? They mention the kernel-package-vserver ... I thought this was sarge and the "legacy" patches (1.x) 1148133163 M * Bertl ah, seems I'm back again :) morning folks! 1148133176 M * derjohn Bertl: replugged ;) 1148133183 M * vrwttnmtu derjohn, Apparently it's 1.9, and 2.0 1148133193 M * vrwttnmtu Hello Bertl 1148133210 M * doener derjohn: 1.x didn't know any ccaps 1148133235 M * derjohn doener, derjohn doesnt know anything about VS < 2.x. So that fits :) 1148133257 M * Bertl hey derjohn! vrwttnmtu! 1148133307 M * Bertl vrwttnmtu: okay, to make it short and simple, all 2.6 versions were affected, no 2.4 version, and it is fixed in stable and devel in two different ways 1148133328 M * Bertl vrwttnmtu: the "stable" fix can basically be done/applied to any 2.6 version 1148133375 M * Bertl the "developer" fix is more a feature than a bugfix, it also makes bind9 work unmodified in guests 1148133444 M * vrwttnmtu Bertl, I'm looking at http://www.13thfloor.at/vserver/s_rel26/v2.01/ . Not sure what the "split" patch is, but the main patch - does that include the fix? 1148133480 M * Bertl no, 2.01 was not updated (yet) there was not much demand yet 1148133491 M * vrwttnmtu OK, only the development branch? 1148133499 M * mugwump ah, bind9. because DNS *needs* specific capabilities 1148133500 M * Bertl we a at a very late release candidate series for 2.02 1148133527 M * vrwttnmtu OK, I shall cross my fingers and wait for that :) 1148133613 M * vrwttnmtu I stopped using Gentoo's vserver-sources for some reason - and I can't remember why. Wonder if it is fixed in their latest? 1148133655 M * Bertl http://vserver.13thfloor.at/Experimental/delta-vxcapable-fix01a.diff 1148133669 M * Bertl this patch should fix it for stable, not tested on 2.01 though 1148133687 M * vrwttnmtu Stable isn't 2.01 then? 1148133692 M * Bertl you can get the latest rc, here: http://vserver.13thfloor.at/Experimental/patch-2.6.16.16-vs2.0.2-rc20.diff 1148133708 M * Bertl and the development version is here: 1148133713 M * Bertl http://vserver.13thfloor.at/Experimental/patch-2.6.16.16-vs2.1.1-rc20.diff 1148133739 M * vrwttnmtu OK - which do you recommend for maximum stabilty and security? :) 1148133742 M * Bertl 2.01 _is_ stable, but we didn't do a 'bugfix release' for that one 1148133758 M * Bertl I would suggest to go with the vs2.0.2-rc20 1148133783 M * Bertl it will (hopefully) be released as vs2.02 in a few days ... 1148133788 M * vrwttnmtu OK - I'll take your advice 1148133793 M * vrwttnmtu Time to get compiling. 1148133795 M * vrwttnmtu :) 1148133803 A * s0undt3ch pings derjohn 1148133822 M * derjohn s0undt3ch, vcmp echo reply 1148133833 M * s0undt3ch good afternoon derjohn 1148133839 M * trash vcmp? 1148133847 M * derjohn v = vserver :) 1148134016 M * derjohn Bertl, no more rc21 ? 1148134046 M * Bertl derjohn: any good reason for rc21? 1148134065 M * derjohn Bertl, I would say we should ask Bertl 1148134089 M * Bertl I still have to cleanup the patch, which I will hopefully do this weekend 1148134102 M * Bertl maybe we will do an rc21 for plm then ... 1148134112 M * derjohn Cleanuo in what way? beautify? 1148134118 M * derjohn plm? 1148134123 M * Bertl yes, beautify :) 1148134144 M * derjohn wtf plm tla ? 1148134265 M * gdm hey, i have a resource allocation question... (this might be me being stupid) 1148134286 M * gdm i know how to allocate and limit resources on teh vservers now (processes/cpu/memory etc) 1148134319 M * gdm but is it possible to limit the host as well, without impacting the vservers that is, in anyway, and if so, how? 1148134342 M * derjohn gdm, dont run anything on the host 1148134355 M * gdm derjohn: i am involved in a group 1148134357 M * derjohn but i think thats not the answer you wanted to hear :) 1148134403 M * gdm and we have 3 vservers: one production, one dev (both of a single site) and one backup of another site/box 1148134410 M * gdm and we send the logs to the host 1148134418 M * gdm so that if the vserver is compromised, we can tell why 1148134434 M * derjohn gdm, create a vserver guest 'loghost' 1148134450 M * gdm yeah, that is an idea 1148134486 M * Bertl gdm: it is not very practical to 'limit' the host beyond the 'normal' limits 1148134486 M * derjohn keep in mind syslog-ng (is used) has to remove the refernce to the klog AFAIR 1148134498 M * derjohn syslog should run fine - even with the hosts IP ! 1148134504 M * Bertl gdm: mainly because as host admin, you could easily remove/raise those limits :) 1148134519 M * derjohn (guest and host can share one ip -> nor reconfig trouble) 1148134537 M * vrwttnmtu I'm on a network where my traffic is measured and charged - if vserver1 sends its logs to vserver2 or the host, under vserver, the traffic won't leave the host, and hence won't be measured - is that right? 1148134560 M * gdm vrwttnmtu: yes 1148134569 M * gdm vrwttnmtu: as far as i understand it :-) 1148134587 M * vrwttnmtu gdm, Cos that's how it would be on a "normal" host - just wanted to check that vserver used the normal routing rules 1148134589 M * derjohn vrwttnmtu, of course you could measure with pmacct 1148134601 M * derjohn but not on the switchport 1148134622 A * derjohn is away now. 1148134642 M * gdm vrwttnmtu: i think you can check with iptables accounting rules as well 1148134647 M * vrwttnmtu derjohn, I don't have access to the switch, and I hope it's not reaching there if it's all on the same physical host 1148134650 A * gdm is not a maestro in iptables tho 1148134678 M * vrwttnmtu gdm, Yes you can - but that will just tell you the amount of traffic coming in - not if it was detected by the switch 1148134829 M * vrwttnmtu Regarding logging, I was wondering if it was possible to redirect vserver /dev/log to the host one so it would all happen "internally" 1148134830 M * vrwttnmtu ? 1148135079 M * gdm vrwttnmtu: http://docs.indymedia.org/view/Sysadmin/KompostInfo#Logging 1148135089 M * gdm vrwttnmtu: some tips on how we do it ;-) 1148136357 M * Bertl okay, translocating ... back later (I hope) 1148136363 N * Bertl Bertl_oO 1148136539 J * Viper0482 ~Viper0482@p54977EE2.dip.t-dialin.net 1148136708 M * weasel does CVE-2006-2110 also apply to vs2.1.1-rcX? 1148136717 M * daniel_hozac Bertl_oO: we probably want to fix anonc's COW issues in 2.1.1. 1148136720 M * daniel_hozac weasel: yes. 1148136727 M * weasel which RC fixed it? 1148136750 Q * vrwttnmtu Quit: Leaving 1148136775 M * daniel_hozac -rc18. 1148136783 M * weasel thanks 1148136824 M * weasel did I miss a security advisory on the vserver list, or was there none? 1148136902 M * daniel_hozac you did. 1148136923 M * daniel_hozac http://archives.linux-vserver.org/200604/0323.html 1148136941 M * weasel ah, there it is 1148136944 M * weasel thank you again 1148137534 M * daniel_hozac Bertl_oO, doener: are lists circle-like in structure? include/linux/list.h:__list_del looks dangerous if not. (vnamespace -c got me an oops there) 1148137544 Q * mnemoc Ping timeout: 480 seconds 1148137666 J * mnemoc ~amery@216.241.24.10 1148137853 J * Banu ~Sohbet@85.96.228.251 1148138184 Q * Acemi-Kiz Ping timeout: 481 seconds 1148138643 J * Dr4g ~Dr4g@82-40-202-184.stb.ubr06.uddi.blueyonder.co.uk 1148139077 Q * [PUPPETS]Gonzo Ping timeout: 480 seconds 1148139675 J * stefani ~stefani@c-24-19-46-211.hsd1.wa.comcast.net 1148140957 P * neo 1148141658 Q * Viper0482 Ping timeout: 480 seconds 1148141749 M * Skram morning all 1148141786 J * FaUl ~immo@user1.wlan.chaostreff-dortmund.de 1148141788 M * FaUl weha 1148141794 M * FaUl Bertl_oO: arround? 1148142069 M * Skram FaUl: 1148142070 M * Skram 09:45 < Bertl> okay, translocating ... back later (I hope) 1148142070 M * Skram 09:46 -!- Bertl is now known as Bertl_oO 1148142083 M * FaUl Skram: shit :-) 1148142096 M * Skram What's wrong? 1148142102 M * FaUl nothing 1148142113 M * FaUl i'd just want to offer some test-hardware :-) 1148142205 M * Skram Right on. 1148142220 M * Skram I hear they are interested in NON-x86/amd64 1148142233 M * FaUl Skram: sparc64 here 1148142240 M * Skram but cool. Our company is thinking about 'donating' a hosted box for a developer to play with 1148142253 M * Skram awesome. I regret to say I havent done much with those. 1148142276 M * Skram I need to jet though- gotta get a haircut :) 1148142296 J * Viper0482 ~Viper0482@p54975995.dip.t-dialin.net 1148142303 M * FaUl well - its our main server today, but i'm going to replace it with standard-x86-hardware because it is not stable enough with linux 1148143252 M * daniel_hozac i think that's the third or fourth sparc64 offer :) seems everyone is removing those from production now. 1148143275 M * daniel_hozac (third or fourth recently, i mean) 1148143597 M * Hollow i'd like to have one too :) 1148144341 N * nokoya nokoyaz 1148144348 N * nokoyaz nokoya 1148145299 M * hillct I wonder if folks could gime me a hand testing a network problem 1148145423 M * hillct please /msg me with the last 3 hops when you traceroute 66.193.195.145 1148145805 M * ray6 you know about traceroute.org? 1148145808 J * ub_lt ~ub@p5484AB8E.dip0.t-ipconnect.de 1148146712 Q * FaUl Ping timeout: 480 seconds 1148148511 M * Skram ray6: still need the info? 1148148548 M * Skram Im doing this from home, not a work server... 1148148579 M * Skram and being routed from my ISP => Level3 (Houston to LosAngeles) and then to TimeWarner Telecome, where it stops. 1148148585 M * Skram ray6: 11 unknown.Level3.net (209.245.88.110) 52.383 ms 54.542 ms 52.156 ms 1148148585 M * Skram 12 core-01-ge-3-1-1-1.lsag.twtelecom.net (66.192.251.32) 49.461 ms 51.160 ms 50.032 ms 1148148588 M * Skram 13 dist-01-so-2-0-0-0.frsn.twtelecom.net (66.192.250.21) 64.875 ms 68.328 ms 64.704 ms 1148148591 M * Skram 14 hagg-01-ge-2-3-0-510.frsn.twtelecom.net (66.192.251.197) 65.458 ms 64.619 ms 64.302 ms 1148148619 M * Skram From a work server with a BGP right to TWTtelegom. i get.. 1148148622 M * Skram 6 core-02-ge-3-1-3-503.dlfw.twtelecom.net (66.192.246.217) 0.989 ms 1.090 ms 1.085 ms 1148148625 M * Skram 7 dist-01-so-2-0-0-0.frsn.twtelecom.net (66.192.250.21) 42.219 ms 42.320 ms 42.333 ms 1148148628 M * Skram 8 hagg-01-ge-2-3-0-510.frsn.twtelecom.net (66.192.251.197) 43.789 ms 42.398 ms 42.455 ms 1148148631 M * Skram (dallas to wherever the server is) 1148148650 M * Skram Hope that helps.. I can also do a traceroute from a different company cage which is in Virginia, if needed. 1148150144 M * ray6 Skram: thanks but I wasn't asking :) 1148150172 M * ray6 I already msg-ed him a trace, and told him about traceroute.org which I usually use when I need such things :) 1148150367 M * Skram Woops 1148150379 M * Skram Sorry, I meant to address hillct 1148150381 M * Skram My appologies 1148150404 M * hillct Skram: thanks 1148150425 M * Skram Sure 1148150430 M * Skram Sorry about the confusion 1148150957 J * shedi ~siggi@inferno.lhi.is 1148151297 M * ub_lt asdfasdfasdfasdasfd 1148151320 M * eyck and dfasgdfasgdfasg to you too 1148151350 M * ub_lt Sorry, wrong window... 1148151703 Q * FireEgl Ping timeout: 480 seconds 1148151971 Q * Dr4g Read error: Connection reset by peer 1148152005 N * Bertl_oO Bertl 1148152009 M * Bertl evening folks! 1148152069 M * Bertl hey ub_lt! hey eyck! 1148152265 M * ub_lt Good evening :-) 1148152423 M * daniel_hozac evening! 1148152786 M * Skram wow. this sucks. grub cant find my kernel. This is on my laptop.. whats the best thing to do? use a gentoo live cd to boot, make a kernel, and move into the /boot, or what? 1148153231 M * eyck hey Bertl, evening folks 1148153242 M * eyck Skram: press 'tab' ;) 1148153272 M * eyck grub is veeery good at finding your lost kernel, try something like this -> go to prompt, type 'kernel /boot/vm' 1148153289 M * Skram okay 1148153290 M * Skram will try 1148153302 M * Skram i didnt know grub was that smart 1148153306 M * eyck oh, and you need to type root (hd0,0) or sth like that.. 1148153323 M * Bertl it's very similar to bash, you even get help and such :) 1148153326 M * Skram (brb.. laptop is upstairs) 1148153328 M * Skram <3 1148153329 M * eyck yeah, grub can read most filesystems 1148153380 M * eyck root (hd0, shoudl list all available filesystems together with their type etc... very cool when you've messed your system up;) 1148153413 Q * Viper0482 Remote host closed the connection 1148153706 Q * hillct Read error: Connection reset by peer 1148153711 M * Skram no luck.. it seemed like it saw the file, and i rentered it to make sure, but still gets file not found... will use livecd and mount boot 1148153830 M * Skram :( 1148154100 Q * blackfire helium.oftc.net xenon.oftc.net 1148154100 Q * Wonka helium.oftc.net xenon.oftc.net 1148154100 Q * alamar helium.oftc.net xenon.oftc.net 1148154100 Q * micah helium.oftc.net xenon.oftc.net 1148154114 J * alamar ~alamar@sol.diddens.de 1148154114 J * micah ~micah@208.99.202.72 1148154114 J * Wonka produziert@chaos.in-kiel.de 1148154114 J * blackfire blackfire@dp70.internetdsl.tpnet.pl 1148154403 M * Skram i dont think its the kernel thats the problem 1148154523 M * Skram this box is using a modified genkernel...what do i but for init=? 1148154562 M * Skram kernel /boot/kernel-genkernel-x86-2.6.12-gentoo-r10 root=/dev/ram0 init=/linuxrc ramdisk=8192 real_root=/dev/hda3 udev <<-- example but i dont have a /linuxrc 1148154587 M * Skram hmm 1148154605 M * Skram btw, i have never been a guru at setup or kernels 1148154750 M * Bertl looks like you're missing an initrd 1148154770 M * Skram thats on the next linere 1148154781 M * Bertl yes, but is it for that kernel? 1148154781 M * Skram and when i mounted /boot, it was there 1148154786 M * Skram yes 1148154803 M * Skram it gives me a file not found (error 15) for the kernel .... line.. but doesnt say what isnt found 1148154832 M * Bertl boot is a separate partition? 1148154851 J * NetAsh ~NetAsh@88.222.136.221 1148154857 M * Skram it is hda1 1148154860 M * NetAsh hello all 1148154862 M * Skram swap is hda2 1148154862 M * Bertl welcome NetAsh! 1148154867 M * Skram everything else is hda3 1148154879 M * Bertl Skram: then better remove the /boot from your path 1148154888 Q * ub_lt Quit: Konversation terminated! 1148154912 M * Skram oh i pasted that from a website 1148154932 M * NetAsh has anyone any experience with lwres ? 1148154932 M * Skram /boot isnt in my configs, ill go triple check though 1148154942 M * Bertl Skram: i.e. make that kernel (hd0,0)/kernel-genkernel-x86-2.6.12-gentoo-r10 1148154954 M * Bertl NetAsh: what's that? 1148154975 M * NetAsh part of bind 1148155037 M * NetAsh "lightweight resolver server" 1148155059 M * NetAsh I dont know how mature the one is, and is there any benefit on using one 1148155094 M * NetAsh insted of full fietured bind 1148155153 M * Bertl sounds interesting 1148155155 M * Skram VAndreas: right, i dont need the (hd0,0) as it is defined on the line before 1148155162 M * Skram right, sorry about that, i wonder how grub changed. 1148155189 M * daniel_hozac lwresd doesn't use DNS... 1148155200 M * daniel_hozac (to communicate answers to the clients) 1148155233 M * NetAsh why am I asking about lwres 1148155263 M * NetAsh imagine you run dedicated vhost for lwresd or full bind with enabled lwres 1148155264 Q * Banu Read error: Connection reset by peer 1148155304 M * NetAsh and all other vhosts comunicate with the one via libnss-lwres insted of clasical dns aproach 1148155330 M * NetAsh in theory lwres comunications shuld be lighter than normal dns 1148155368 M * NetAsh this might leed to beter resource utilization on big virtual host farms from the dns resolving perspective 1148155427 M * NetAsh as I am laizy a litle, I first asked: maby someone tested is there any benefit using lwres insted of real dns 1148155428 M * NetAsh :) 1148156316 M * Skram okay, im back. 1148157344 M * daniel_hozac Bertl: did you see my message about the vnamespace -c oops? 1148157347 J * FireEgl ~FireEgl@Atlantica.US.TO 1148157405 M * Bertl daniel_hozac: nope .. please elaborate! 1148157438 M * daniel_hozac < daniel_hozac> Bertl_oO, doener: are lists circle-like in structure? include/linux/list.h:__list_del looks dangerous if not. (vnamespace -c got me an oops there) 1148157570 J * FaUl ~immo@sun.do.bundessicherheitsministerium.de 1148157571 M * FaUl weha 1148157585 M * FaUl Bertl: arround? 1148157588 M * FaUl argh 1148157589 M * FaUl +spelling 1148157696 J * lubeznia_los ~c0l0ruSeR@223.Red-88-9-124.dynamicIP.rima-tde.net 1148157707 P * lubeznia_los 1148157717 J * lubeznia_los ~c0l0ruSeR@223.Red-88-9-124.dynamicIP.rima-tde.net 1148157730 M * lubeznia_los eiiiiiiiiiiiiiiiiiiiiiiii 1148157730 M * Bertl wb lubeznia_los! 1148157734 M * lubeznia_los hahahah 1148157739 M * lubeznia_los awe bg 1148157742 M * lubeznia_los tuka ima li> 1148157752 Q * lubeznia_los Quit: 1148157756 M * FaUl hey Bertl 1148157764 M * Bertl hey FaUl! LTNS! 1148157806 M * Bertl daniel_hozac: do we have the oops somewhere?# 1148157850 M * daniel_hozac http://paste.linux-vserver.org/54 1148157886 M * daniel_hozac addr2line says include/linux/list.h:153 1148158137 M * daniel_hozac got it from chcontext --xid 8005 -- vnamespace -n -- vnamespace -s -- vnamespace -c -- cat /proc/mounts 1148158314 M * Bertl ah, that's the cleanup code/call 1148158320 M * daniel_hozac indeed. 1148158388 J * hillct ~H@client200-5.dsl.intrex.net 1148158435 M * hillct this is odd 1148158452 M * daniel_hozac hmm, seems the oops must've caused some lock to be held and never released. 1148158481 M * hillct I'm seeing behavior where moth eth0 and eth1 are showing up as ifconfig output in a guest where only eth1 is configured 1148158497 M * hillct er both... 1148158514 M * daniel_hozac version? 1148158524 M * hillct this behavior is new since I upgraded from 2.0.1 to 2.0.2 1148158547 M * daniel_hozac hmm, it's possible the new ifconfig code misses that case. 1148158548 M * hillct and tools 208 to tools 210 1148158557 M * daniel_hozac what does ip a ls say in the guest? 1148158573 M * hillct both interfaces show up as having the same IP address 1148158578 M * hillct checking 1148158591 M * Bertl doener: ping? 1148158769 M * hillct VServer:www(root):/> ip a ls 1148158769 M * hillct 2: eth0: mtu 1500 qdisc pfifo_fast qlen 1000 1148158769 M * hillct link/ether 00:e0:81:34:d1:96 brd ff:ff:ff:ff:ff:ff 1148158769 M * hillct inet 66.195.193.148/24 brd 66.195.193.255 scope global secondary eth0 1148158769 M * hillct 3: eth1: mtu 1500 qdisc pfifo_fast qlen 1000 1148158770 M * hillct link/ether 00:e0:81:34:d1:97 brd ff:ff:ff:ff:ff:ff 1148158770 M * hillct inet 66.195.193.148/24 brd 66.195.193.255 scope global secondary eth1 1148158772 M * hillct VServer:www(root):/> 1148158785 M * daniel_hozac are you sure you don't have the IP address assigned to both interfaces? 1148158791 M * hillct yes 1148158800 M * daniel_hozac so ip a ls on the host doesn't show it as assigned to both? 1148158803 M * hillct eth0 doesn't appear anywhere in the guest config 1148158821 M * hillct I have eth1 and a nodev device 1148158824 M * Bertl could you please upload the output of 'ip addr ls' both inside and outside the guest to paste.linux-vserver.org 1148158902 M * Bertl daniel_hozac: do you intend to use the cleanup code? 1148158924 M * Bertl daniel_hozac: AFAIK it is currently unused (except for the -c option) 1148158944 M * daniel_hozac no, i was just testing it. 1148158958 M * Bertl okay, because I think we should basically disable it for now 1148158961 J * ourseven ~ourseven@cpe-24-167-94-120.houston.res.rr.com 1148158968 M * Bertl welcome ourseven! 1148158969 M * Skram a fellow texan! 1148158973 M * hillct http://paste.linux-vserver.org/55 1148158980 Q * ourseven Quit: 1148158994 M * Bertl daniel_hozac: because I think all this could be done in userspace (much easier) 1148159038 M * Skram so using ip and qdisk and all this, one could limit the b/w (bandwidth) of a vps to lets say.. a max of 1mps in/out? 1148159043 M * Skram or no? 1148159044 M * daniel_hozac Bertl: yeah, that's what waldi's patch does. 1148159051 M * Bertl hillct: hmm, your output shows the ip assigned to bioth interfaces? 1148159056 M * Bertl *both 1148159064 M * hillct inside 1148159070 M * Bertl no, on the host too 1148159079 M * Skram ill ask later, nevermidn 1148159092 M * daniel_hozac Skram: yes. 1148159112 M * waldi Bertl: i think you should just remove it 1148159124 M * Skram daniel_hozac: okay, what would the syntax be? 1148159125 M * hillct ah 1148159127 M * hillct I see 1148159127 M * Bertl hillct: so I consider this 'normal' and 'expected' behaviour :) 1148159133 M * hillct you're right 1148159140 M * hillct although 1148159157 M * Bertl waldi, daniel_hozac: okay so we simply disable that command for now, yes? 1148159160 M * doener Bertl: pong! 1148159166 M * daniel_hozac Bertl: sure. 1148159171 M * hillct the only place 193.148 appears on the host is in ifcfh-eth1 1148159185 M * hillct er ifcfg-eth1 1148159199 M * Bertl doener: regarding cleanup, what's your opinion? 1148159298 M * doener ? 1148159414 M * Bertl should we just disable the cleanup code in the kernel for now? 1148159431 M * Bertl or does it make sense to fix it up properly? 1148159681 M * Skram 16:04 < daniel_hozac> Skram: yes. 1148159683 M * Skram 16:05 < Skram> daniel_hozac: okay, what would the syntax be? 1148159721 M * Skram wpuld it be the same as any example or special for a vserver environmnet? 1148159852 M * Bertl Skram: basically the same as without 1148159869 Q * dna Quit: Verlassend 1148159873 M * Bertl accounting and/or limits based on IPs, the only difference imight be that those IPs are local 1148159905 M * Skram each VPS gets a PUBLIC IP of each server's /26 1148159914 M * Skram so it would be the same? 1148159925 M * Skram hmm 1148159951 M * Skram so i need to look up how to do that with "ip" anything else that could point me in the right direction? 1148160068 M * daniel_hozac http://lartc.org/ ? 1148160339 M * doener Bertl: hm, there aren't any real users of it AFAIK, so simply disabling it should be fine 1148160939 M * Bertl okay, then we do that ... 1148161385 Q * bonbons Quit: Leaving 1148163019 M * NetAsh by all 1148163030 Q * NetAsh Quit: 1148164245 M * daniel_hozac Bertl: i can reproduce anonc's oops, i think. 1148164278 M * daniel_hozac fs/namei.c:1731 1148164378 M * daniel_hozac hmm, that's one of the lines my patch adds. 1148164442 M * Bertl =) 1148164463 M * daniel_hozac does path_to_nameidata change path? 1148164497 M * Bertl let me take a look at the code, sec 1148164535 M * daniel_hozac i assumed the if (!path.dentry->d_inode) ~15 lines above would make it safe. 1148164796 M * Bertl daniel_hozac: okay, care to give me a short description how you trigger it? 1148164808 M * Bertl same way as anonc did? 1148164817 M * daniel_hozac find /vservers/gentoo -type f -print0 | xargs -0 setattr --iunlink 1148164825 M * daniel_hozac cp -al /vservers/gentoo /vservers/cow-test 1148164827 M * daniel_hozac vserver cow-test start 1148164884 M * Bertl is the guest start mandatory? 1148164912 M * daniel_hozac well, it's the easiest way to trigger it that i've come across so far. 1148164917 J * Methos ~mjoconr@hacker.pineview.net 1148164922 M * Bertl welcome Methos! 1148164933 M * Bertl daniel_hozac: okay, do we know where it happens? 1148164946 M * Bertl I mean, which dentry causes it? 1148165015 M * Skram whats the max # of interfaces a guest can have? 1148165062 M * daniel_hozac no, any easy ways to find out? 1148165213 M * Bertl Skram: by default 16 1148165234 M * Skram why do you say by default :) 1148165243 M * daniel_hozac because you can patch it. 1148165247 M * Bertl it can be changed, and certain aptches raise it to 64 1148165252 M * Skram hm 1148165257 M * Skram this server only has 64 ips 1148165258 M * Skram :) 1148165263 M * Skram 16 should be fine 1148165272 M * Bertl yes, that's what we thought too :) 1148165307 M * Bertl daniel_hozac: could you provide the line numbers of the last 5 stack entries please? 1148165366 M * daniel_hozac the call trace? 1148165381 M * Bertl yes, please (of your call trace, with the trace) 1148165405 J * harpagon ~harpagon@85.100.193.91 1148165413 M * Bertl welcome harpagon! 1148165413 M * harpagon arent people allowed to register projects on freshmeat anymore? 1148165429 M * Bertl don't know, but I guess you still can ... 1148165495 M * daniel_hozac http://paste.linux-vserver.org/56 1148165501 Q * harpagon Remote host closed the connection 1148165502 M * Bertl tx 1148165726 M * Bertl ah, the c017c6ae too please 1148165758 M * daniel_hozac 00:31 < daniel_hozac> fs/namei.c:1731 1148165771 M * Bertl ah, tx 1148165804 M * daniel_hozac (just to clarify, this is with http://daniel.hozac.com/vserver/delta-cow-fix01.diff) 1148165844 M * Bertl yep, saw that 1148165990 M * Bertl so we are missing the inode here 1148166042 M * Bertl please add a debug line like this: 1148166127 M * Bertl vxdprintk(!path.dentry->d_inode, "missing inode %s", pathname); 1148166130 M * daniel_hozac how can that be? shouldn't the check on line 1718 make that impossible? 1148166175 M * Bertl no, because we can call to ok 1148166184 M * Bertl 'label' ok: 1148166187 M * daniel_hozac ah, yes. 1148166224 M * Bertl we could avoid that by a check for d_inode, but I'd like to see _where_ that happens 1148166251 M * Bertl I'd assume it happens on a directory or missing/deleted entry 1148166702 M * daniel_hozac hmm, no output. 1148166743 M * Bertl but still an oops? 1148166774 M * doener there's another check for path.dentry->d_inode on line 1723, so it's probably changed in path_to_nameidata anyway 1148166784 M * daniel_hozac yep. 1148166798 M * Bertl okay, let me suggest to move the 'untagging' somewhere else :) 1148166808 M * daniel_hozac path_to_nameidata doesn't change path AFAICT. 1148166825 M * Bertl IMHO it would be more natural to do it where the link count gets decremented, no? 1148166883 M * daniel_hozac you mean remove it when the second link is broken? 1148166979 M * Bertl ext2_dec_count, e.g.? 1148166997 M * Hollow hey all! 1148167005 M * daniel_hozac well, i thought about that, but if you have a template which you hardlink your guests to, the template's file would get IUNLINK removed. 1148167012 M * daniel_hozac (if all guests update a file) 1148167023 M * daniel_hozac and guests created after that wouldn't get the COW behavior. 1148167058 M * Bertl hmm, good point, but that's probably not really resolveable 1148167082 M * daniel_hozac hmm? the current patch only removes IUNLINK when the template tries to write to the file. 1148167109 M * Bertl how would you tell wether the template is the 'other' link or another guest? 1148167127 M * daniel_hozac well, the template would be the file with IUNLINK and link count == 1. 1148167139 M * Bertl ahem ... 1148167147 M * Bertl let me give an example 1148167162 M * Bertl T-A-B all three unified, link count=3 1148167173 M * Bertl now we delete the 'template' 1148167181 M * Bertl A-B unified, link count=2 1148167301 M * Bertl so what am I missing? 1148167328 M * daniel_hozac i'm not sure i follow. 1148167357 M * Bertl we have a template T okay? 1148167361 M * daniel_hozac right. 1148167376 M * Bertl we clone guests A and B 1148167401 M * Bertl then we remove the template T 1148167486 M * Bertl when we break one of the remaining A-B links 1148167497 M * Bertl what do we do with the original? 1148167504 M * daniel_hozac nothing. 1148167530 M * daniel_hozac until it wants to be written to. 1148167568 M * Bertl and this is what I consider inconsistent 1148167591 M * Bertl we basically create a tagged file with special semantics 1148167612 M * Bertl if we really want to go that we 1148167614 M * Bertl *way 1148167628 M * Bertl we have to introduce the new semantics properly 1148167640 M * Bertl i.e. a file with immutable and unlink set 1148167652 M * Bertl reverts to a normal file on open for write 1148167663 M * Skram my life is boring. http://img164.imageshack.us/my.php?image=terminalsgalore6xp.jpg (screenshot of my desktop, how it is usually.. full of terminals) 1148167665 M * daniel_hozac exactly. 1148167694 M * daniel_hozac regardless of if it is the last link or the fifth. 1148167698 M * Skram heh, life of a programmer/linux systems admin 1148167713 M * Bertl daniel_hozac: no, this one _is_ differen 1148167717 M * Bertl +t 1148167732 M * Bertl because we do _not_ unlink/copy anything there 1148167782 M * Bertl a different approach could be to test for xid=0 1148167849 M * daniel_hozac hmm? 1148167858 M * daniel_hozac do we want to unlink/copy the file? 1148167886 M * daniel_hozac seems the inode pointer is invalid. 1148167904 M * daniel_hozac COW check, path /var/log/wtmp, dentry c073d368, inode caba30e8 1148167904 M * daniel_hozac COW check, path /var/log/wtmp, dentry d708e000, inode 2f006576 1148167951 M * daniel_hozac (why it would try to unlink the same file twice, i don't understand) 1148168196 M * Bertl another quite different approach could be to have a separate CoW flag 1148168219 M * Bertl (just brainstorming) 1148168239 M * Bertl or to simply ignore immutable when iunlink with nlink=1 1148168300 M * daniel_hozac and leave them set? 1148168305 Q * Methos Quit: Methos 1148168321 M * Bertl daniel_hozac: yes, would be an option 1148168326 M * daniel_hozac IMHO the removing of iunlink and immutable when the last link is written to is the most consistent option. 1148168438 M * daniel_hozac because then the write does the same thing to all files, except for the copy on the last one (as it's not needed). 1148168674 M * Bertl so, what about this approach: 1148168700 M * Bertl (let's assume unmodified vserver code without the cow fix) 1148168728 M * Bertl we have the COW test in line 1539 (fs/namei.c) 1148168752 M * Bertl we add another test there, which checks for nlink=1 1148168768 M * Bertl and in exactly this case, removes the flags 1148168791 M * Bertl (dirtying the inode on the way) 1148168806 M * daniel_hozac well, isn't that pretty much exactly what the current fix does, except in another function? 1148168814 M * daniel_hozac (right after the may_open call) 1148168835 M * Bertl but we _know_ that the inode is fine there, no? 1148168855 M * Bertl so, instead of returning the EMLINK and retrying 1148168870 M * Bertl we 'just' modify the inode on the fly 1148168986 M * daniel_hozac how does it differ from the cow-fix approach, other than the function where it happens? 1148169011 M * Bertl not at all, except that it should work, no? 1148169054 M * Bertl the other ideas all change the semantics 1148169077 M * Bertl we should consider them, but I guess that is post 2.1.1 stuff 1148169078 M * daniel_hozac does may_open modify the pointers? otherwise i don't see what the difference is. 1148169160 M * daniel_hozac although it's odd the may_open doesn't generate the oops. 1148169201 M * daniel_hozac or is it the fact that we use path.* rather than nd? 1148169247 M * Bertl yep, nd->dentry should work 1148169280 M * Bertl but I'd really like to move it into the may_open() 1148169295 M * Bertl and _not_ return the EMLINK at all 1148169310 M * daniel_hozac yeah, that makes sense. 1148169368 J * Methos ~mjoconr@hacker.pineview.net 1148169371 M * Bertl that should also work without changing the COWBL code in namei 1148169407 M * daniel_hozac shouldn't the COWBL code be nd->dentry in cow_break_link? 1148169453 M * Bertl I think there was a reason for using the path here, let me rethink that part