1171065657 M * daniel_hozac because mainline does choose_new_parent(p, vx_child_reaper(p)); 1171065682 M * daniel_hozac which will do the right thing. 1171065726 M * daniel_hozac (maybe not with regard to threads, but for this case) 1171065767 M * Bertl but the code, except for the threads is not different, or did I miss something? 1171065777 M * daniel_hozac hmm? 1171065782 M * Bertl both start with reaper = father 1171065792 M * daniel_hozac mainline's forget_original_parent is entirely different. 1171065802 M * Bertl both check the siblings 1171065817 M * Bertl reaper = next_thread(reaper); 1171065823 M * daniel_hozac but mainline only uses the reaper for the ptrace children. 1171065826 M * Bertl if (reaper == father) reaper = vx_child_reaper(father); 1171065868 M * Bertl but we agree that so far the same things happen in mainline and fc? 1171065882 M * daniel_hozac yep, so far. 1171065897 M * daniel_hozac but this is where things deviate wildly. 1171065934 M * Bertl so now the ptrace checks happen in mainline 1171065941 M * Bertl (which are missing in fc) 1171066063 M * Bertl but yes, doing choose_new_parent(p, vx_child_reaper(p)) would be an option too 1171066074 M * Bertl at least for the non ptrace processes 1171066094 M * daniel_hozac so we never want to use reaper? 1171066116 M * Bertl oh yes, for the ptrace children 1171066127 M * daniel_hozac right, but for regular processes. 1171066155 M * Bertl well, mainline does the exception based on 1171066161 M * Bertl father == p->real_parent 1171066189 M * daniel_hozac yeah. 1171066241 M * Bertl IMHO you are doing a 'workaround' because FC doesn't differenciate between ptrace and non ptrace 1171066261 M * Bertl so you could as well put that workaround into choose_new_parent() 1171066311 M * daniel_hozac i put it there for now. 1171066343 M * daniel_hozac i'm still not entirely convinced that we do the right thing in mainline w.r.t. reaper for regular processes. 1171066367 M * Bertl okay, let's walk that trough shall we? 1171066378 M * daniel_hozac or does the reaper have threads only when running under ptrace? 1171066384 M * daniel_hozac ok, thanks. 1171066565 M * Bertl http://www.13thfloor.at/~herbert/delta-exit.diff.hl 1171066619 M * daniel_hozac hunk 7 is what's bothering me. 1171066626 M * Bertl yes, I see 1171066698 M * Bertl okay, let's take a step back and take a different approach 1171066733 M * Bertl let's assume we use reaper there, which IMHO looks better 1171066795 M * Bertl then we should be par with fc, well at least from the special case 1171066840 M * daniel_hozac right. 1171066856 M * Bertl now, that would then mean we hit choose_new_parent() with a condition which_will_BUG_ON() us 1171066914 M * Bertl the Make sure we're not reparenting to ourselves 1171066923 M * Bertl is not added by us, it is in Mainline 1171066929 M * daniel_hozac yep. 1171066966 M * Bertl so, I seetwo options, either we bail out with vs2.2.0 (after changing hunk 7 1171066990 M * Bertl or fc is doing something wrong, which results in this 1171067003 M * Bertl do you agree? 1171067022 M * daniel_hozac yeah, sounds correct. 1171067037 M * Bertl okay, can you verify that easily? 1171067089 M * daniel_hozac shouldn't be a problem. 1171067134 M * Bertl okay, so iif we hit the same with mainline, we have to think about what actually causes this special case, and deal with it properly (or fix another bug) 1171067147 M * Bertl btw, I agree that hunk 7 looks wrong 1171067222 M * daniel_hozac IMHO it's caused by father == vcontext waiting for execve in child (starting init) to succeed, with the child setting itself up as the init of the guest. 1171067257 M * Bertl okay, but that is something which can happen on a normal system too 1171067258 M * daniel_hozac yep, got it when reverting hunk 7. 1171067270 M * daniel_hozac (on plain 2.6.19.3-vs2.2.0-rc12) 1171067272 M * Bertl i.e. we are not changing the child/parent behaviour 1171067285 M * daniel_hozac no, but the reaper is then incorrect. 1171067315 M * Bertl yes, but we special case the init process 1171067319 M * daniel_hozac the reaper is correct for the father (the guest's init), and any children which are not init, but init should have the real child_reaper as the parent. 1171067347 M * daniel_hozac hmm, where? 1171067365 M * Bertl if (vxi->vx_reaper == p) 1171067377 M * daniel_hozac ah, we add that, okay. 1171067384 M * Bertl in vx_child_reaper() 1171067392 M * daniel_hozac oh. 1171067404 M * Bertl so, IMHO, we are hitting a case where 1171067406 M * daniel_hozac yeah, but the thing is that vx_child_reaper is just run once. 1171067410 M * daniel_hozac and that's for the parent. 1171067477 M * daniel_hozac it needs to run for each child. 1171067489 M * Bertl ah, i.c. we use the wrong process for determining the reaper 1171067541 M * daniel_hozac maybe the if (reaper == father) case should set reaper to NULL and if reaper == NULL, use vx_child_reaper(p)? 1171067577 M * Bertl let me think about the for a moment ... 1171067654 M * Bertl it's probably better to modify vx_child_reaper() in such a way 1171067660 P * stefani parting (is such sweet sorrow) 1171067670 M * Bertl that it takes two arguments, a child (p) and a 'suggested' reaper 1171067715 M * Bertl we then simply check for the special cases, and replace that then, no? 1171067731 M * Bertl of course, the same will be used with a NULL reaper at the beginning 1171067779 M * daniel_hozac yeah, makes sense. 1171067795 M * Bertl actually, we can pass the child_reaper in the beginning instead 1171067834 M * Bertl let me prepare a patch for that 1171067990 J * bronson ~bronson@adsl-75-36-145-145.dsl.pltn13.sbcglobal.net 1171068157 M * Bertl welcome bronson! 1171068267 J * Daniel15 ~dansoftau@server.daniel15.com 1171068291 M * Bertl wb Daniel15! 1171068300 M * Daniel15 Hello :) 1171068395 M * Bertl daniel_hozac: btw, can we add a test for that to testme.sh or could you give me a vcmd sequence for that? 1171068419 M * daniel_hozac does vcmd ever fork? 1171068437 M * daniel_hozac i guess that's the essential part. 1171068438 M * Bertl sure, the 0.08 version can fork, clone and unshare :) 1171068451 M * daniel_hozac wow :) 1171068599 M * daniel_hozac vcontext --create --xid ... -- vcontext --migrate-self --endsetup --initpid --disconnect -- true should do it. 1171068632 M * Bertl ah, nice, seems to work 1171069286 M * daniel_hozac hmm, seems true doesn't work, sleep 5 seems do to the trick though. 1171070180 Q * meandtheshel1 Quit: Leaving. 1171070243 M * Bertl true worked here ... 1171070288 M * daniel_hozac guess it depends. as long as it doesn't exit before the parent... 1171070370 J * Aiken ~james@ppp111-64.lns2.bne4.internode.on.net 1171070412 M * Bertl morning Aiken! 1171070416 M * Aiken Bertl how does the current vserver handle being on a machine using briding 1171070427 M * Bertl should work 1171070442 M * Aiken with my old vs1.2.10 machine I am trying to go from eth0 eth1 to br0 and vserver is giving me grief 1171070465 M * Bertl interesting .. how so? 1171070493 M * Aiken normally I specify eth0 or eth1 in the conf files 1171070507 M * Aiken when using bridging I have use br0 1171070510 M * Aiken which is fair enough 1171070529 M * Aiken to shut down the guests I have to modify the .conf to use eth0 or eth1 1171070541 M * Aiken then to restart the guest set the .conf back to br0 1171070550 M * daniel_hozac why?! 1171070592 M * Aiken when br0 is active the guests won't start usingng the normal ethX devices 1171070608 M * Aiken and shutting down they refuse to shutdown properly when using br0 1171070612 M * Bertl ahem 1171070625 M * Bertl when you use bridging, you should not use the eth devices at all 1171070632 M * Aiken I know 1171070639 M * Bertl i.e. you want to change the guests to use br0 instead 1171070641 M * Aiken and the guests will start with br0 1171070656 M * Aiken BUT the guests won't stop with br0 in the conf file 1171070678 M * Aiken and when the networking is setup for the guests they still create eth0:XXX 1171070696 M * Aiken which is why I am wondering if the current code is better behaved :) 1171070711 M * daniel_hozac sounds like a userspace thing to me. 1171070751 M * daniel_hozac (or are bridges unable to cope with aliases, forwarding them to the real interface?) 1171070773 M * Aiken once my raid 5 set is rebuilt I will be able to try again 1171070859 M * Bertl yeah, maybe update userspace ... 1171070986 Q * DreamerC Quit: leaving 1171071009 J * DreamerC ~dreamerc@125-225-101-221.dynamic.hinet.net 1171071030 M * Aiken they can with 2.6 1171071088 M * Bertl hmm? 1171071103 M * xp_prg2 there is a vserver-copy, why didn't you guys tell me about that? :( 1171071110 M * Aiken also with 2.4 1171071118 M * daniel_hozac because it's legacy and doesn't work with new-style configurations? 1171071158 M * Aiken didn't think 1.2.10 worked with the new configurations 1171071184 M * daniel_hozac (my comment was to xp_prg2) 1171071191 M * Aiken oh 1171071207 M * daniel_hozac but actually, the new-style configuration should work with them all. 1171071242 M * xp_prg2 so vserver-copy doesn't work or it does? 1171071271 M * daniel_hozac if you for whatever reason are using legacy configurations, it might work. 1171071296 M * xp_prg2 I am not using legacy configurations 1171071528 M * Bertl daniel_hozac: do we still have those in util-vserver? 1171071541 M * Bertl if so, maybe we should rename them to *.legacy or so? 1171071577 M * Daniel15 Hmmm, I just noticed vserver-copy now... So you're saying it won't work with the new-style configurations? 1171071602 M * Daniel15 Isn't there another command to duplicate a vserver 1171071605 M * Daniel15 ? 1171071633 M * Bertl yep, vserver * clone should be there 1171071726 M * Daniel15 Looks like verver-debiantools also has a "dupvserver" script :) 1171071731 M * Daniel15 vserver-debiantools* 1171071751 M * Bertl those should be depreciated even by debian by now :) 1171071753 M * Daniel15 I've never tried it before, though 1171071762 M * Daniel15 Oh, OK :P 1171071771 Q * Aiken Read error: Connection reset by peer 1171071792 M * Bertl iirc, those are the addons done by the deian developer we do not speak about anymore :) 1171071906 M * Bertl daniel_hozac: http://vserver.13thfloor.at/Experimental/delta-reaper-test01.diff 1171071929 M * Bertl it's only considered a test, because it shows unexpected behaviour here 1171071958 M * Bertl I had a version with extended vx_child_reaper() too, but after thinking that through, it ended up with the same result 1171071961 M * xp_prg2 http://www.micropp.se/vserver/ 1171071965 M * xp_prg2 Bertl check this out 1171071973 M * xp_prg2 tell me what you think 1171072108 M * Bertl looks complicated, but if it works ... 1171072268 A * xp_prg2 is about to give up his effort to copy a vserver and just start a new vserver from scratch :( 1171072330 M * Daniel15 Can't you just make a new barebones vserver, copy the stuff in /etc/vservers/[old vserver name] and /vservers/[old vserver name] over, and then edit it? 1171072354 M * xp_prg2 I don't know 1171072367 M * xp_prg2 this is my first time trying to copy a vserver 1171072411 M * daniel_hozac xp_prg2: and what do you find lacking in the rsync or clone build methods? 1171072481 M * Bertl probably the documentation on the wiki, as usual ... 1171072504 J * nigel ~nigel@leibniz.catalyst.net.nz 1171072508 M * Bertl daniel_hozac: strange thing is, in my test setup, the patch ends up killing init :) 1171072521 M * daniel_hozac killing? 1171072545 M * Bertl I use bash -c 'echo $$' as command 1171072550 M * Bertl [root@(none) /]$ 30aper: 79302090[30,#100] replaced with 79157510[1,#0][root@(none) /]$ 1171072553 M * Bertl <0>Kernel panic - not syncing: Attempted to kill init! 1171072560 M * daniel_hozac humm. 1171072569 M * Bertl note: the executing bash _is_ the init :) 1171072590 M * Bertl so, yes, it becomes the parent, but I have no idea why it is killed ... 1171072607 M * Bertl welcome nigel! 1171072609 M * daniel_hozac isn't the panic only if task == child_reaper? 1171072630 M * nigel hello. I am trying to compile vserver support for 2.6.20. But it bombs out when compiling kernel/timer.c, saying that all these vx_* symbols are undefined (in particular VXF_VIRT_UPTIME) 1171072643 M * nigel anyone know why? :) 1171072645 M * Bertl daniel_hozac: yes, as I said, the bash I'm starting the sequence from _is_ the init process 1171072661 J * Aiken ~james@ppp111-64.lns2.bne4.internode.on.net 1171072662 M * nigel Bertl: hiya 1171072664 M * Bertl nigel: which patch and what config? 1171072668 M * daniel_hozac ah, _the_ init process, okay. 1171072676 M * Aiken it seems to be working now 1171072679 M * daniel_hozac i thought you meant in the guest. 1171072680 M * Aiken a reboot was involved 1171072693 M * Bertl daniel_hozac: nah, _the_ init :) 1171072694 M * nigel patch: 2.3.0.9 1171072708 M * nigel what do you mean, "what config"? 1171072718 M * Bertl nigel: okay, please upload your .config file somewhere 1171072722 M * nigel righto 1171072743 M * Daniel15 Hey, the messages you get on the screen when booting a Linux system, what are they called? 1171072761 M * Daniel15 I just call it "Output when booting", but I'd like to know the proper name for it :P 1171072761 M * Bertl bootup messages? 1171072762 M * daniel_hozac "scrolling text thingies"? :) 1171072784 M * Bertl yeah, I like that one .. lol 1171072789 M * Daniel15 Hehe :P 1171072799 M * nigel Bertl: http://pastebin.ca/348310 1171072910 J * almak ~almak@willers.employees.org 1171072911 M * Bertl nigel: k, will take a moment 1171072915 M * Bertl welcome almak! 1171072918 M * nigel :) 1171072919 M * almak hi there 1171072928 M * almak trying to pick your brain 1171072973 M * daniel_hozac Bertl: it's not due to bash doing weird stuff either? 1171072992 Q * thunder1 Ping timeout: 480 seconds 1171073011 M * Bertl daniel_hozac: could be, but it's surspising that it gets send a signal besides SIG_CHILD from the command sequence, no? 1171073030 M * daniel_hozac yeah, anything else would be odd. 1171073046 M * Bertl I guess I have to enable some signal logging for that 1171073095 M * daniel_hozac IMHO the test patch shouldn't actually change anything. 1171073133 M * Bertl I thought so too, funny thing is, without the patch it doesn't do that 1171073163 M * daniel_hozac it seems to work fine here, the vcontext reproducer doesn't do anything now. 1171073185 M * Bertl yeah, I actually expect the patch to do the proper thing ... 1171073209 M * Bertl i.e. you should get exactly one such message (the replace one) for the guest 1171073223 M * Bertl when the actually init process is reparented 1171073233 M * daniel_hozac yeah. 1171073258 M * Bertl and the same should work for fc too 1171073318 M * Bertl could you try if starting such a test chain directly from init is something your system would survive? 1171073340 M * daniel_hozac i.e. booting with init=/bin/bash? 1171073353 M * Bertl either that or adding it to inittab 1171073372 M * Bertl probably the former is less harmfull :) 1171073459 M * nigel Bertl: also: http://pastebin.ca/348317 (the error messages I'm getting) 1171073475 M * Bertl nigel: it's currently compiling here 1171073485 M * nigel :( 1171073485 M * Bertl i.e. we will soon see what's up 1171073491 M * nigel oh 1171073492 M * nigel heh 1171073494 M * Bertl (not finished yet) 1171073500 M * nigel oarsum :) 1171073556 M * almak I have a number of guestos eaching having an interface bind to it. Is it possible to configure the routing table such that each guestos thinks it only has one interface and all it's traffic will go through that interface? 1171073621 M * Bertl yes 1171073642 M * almak can you show me how? 1171073645 M * daniel_hozac Bertl: the vcontext sequence seems to work fine when executed from init too. 1171073685 M * Bertl okay, good, then we remove the log message and that's it, yes? 1171073698 M * Bertl almak: basically you assign a single ip to each guest 1171073714 M * Bertl almak: and you set up a table for each guest ip 1171073723 M * daniel_hozac i think so. 1171073740 M * Bertl almak: i.e. rule table x from 1171073750 M * Bertl almak: and route add ... table x 1171073789 N * DoberMann DoberMann[ZZZzzz] 1171073916 M * almak so there is no need to config external net ip 1171073988 M * Bertl how is that related to the table/routing separation? 1171074140 M * almak rephrasing: for each table, is there a need to add the external net IP? I am trying to avoid such configuration. 1171074305 M * Bertl hmm, so you have N external ips, but want to use private ips for the guests or what? 1171074317 M * Bertl nigel: compiled perfectly fine here 1171074369 Q * FireEgl Quit: ... 1171074401 M * nigel grrrRRR 1171074412 A * nigel sadly pokes linux config 1171074469 M * nigel maybe I could add an include to kernel/timer.c to "fix" it? 1171074637 M * Bertl nigel: either you have a different kernel 1171074646 M * Bertl or the patch is broken somehow ... 1171074650 M * nigel hmm 1171074656 M * nigel well it's been patched with a few things 1171074670 M * nigel one of them is suspend2 1171074683 M * nigel that's the only one I can think of that would patch kernel/timer.c as well... 1171074714 M * nigel hmm, yes it does 1171074755 A * nigel makes vserver patch first 1171074843 M * almak bertl: So each guest has external ip. What I want to do is without 1171074852 M * nigel oooh 1171074854 M * nigel hunk failed 1171074859 M * almak bertl: configuring extra routes to the rest of the world in the guest 1171074866 M * almak bertl:it will use the ip bind to it send traffic. 1171074891 M * Bertl almak: yep, that will work out of the box 1171074912 M * almak ok. Let me try it. Thx 1171074958 M * Bertl simple rule is, if ping -I www.google.com works on the host, the guest should be fine 1171074983 M * Bertl (just assign the to the guest and it should work 1171075116 M * nigel Bertl: it works now 1171075132 M * nigel both suspend2 and vserver were trying to add includes to the same place in timer.c 1171075147 M * nigel because vserver was last, it was being rejected, and I didn't spot that 1171075177 M * nigel thanks for your help :_ 1171075181 M * nigel err, :) 1171075227 M * Bertl you're welcome! 1171075281 M * almak Bertl: in the guest I still have to ping -I ... www.google.com, I cannot just ping www.google.com 1171075308 M * Bertl then your guest setup is wrong, or was that a question? 1171075350 M * almak This is my guest's behavior. 1171075359 M * almak So I have a setup problem. 1171075385 M * Bertl what does /proc/virtnet//* show? (upload to paste.linux-vserver.org) 1171075538 M * almak just sent 1171075574 Q * pflanze Quit: night 1171075589 M * Bertl so your public ip is 11.11.10.2/24 then? 1171075599 M * almak yes 1171075603 M * Bertl and what ip do you use in the ping to make it work? 1171075640 M * Bertl ah, I guess your problem is in /etc/hosts 1171075728 M * almak I have empty hosts 1171075749 M * Bertl could you show me the output of 1171075771 M * Bertl ping -c 1 72.14.203.104 1171075772 M * Bertl and 1171075784 M * Bertl ping -I 11.11.10.2 -c 1 72.14.203.104 1171075790 M * Bertl inside the guest, please? 1171075822 M * almak hum.. this is a private network but I can use one of my internal IP 1171075833 M * Bertl okay, do that 1171075927 Q * Piet_ Quit: Piet_ 1171075953 M * almak it will be a min. I have routes setup. 1171075980 J * Daniel15-2 ~dansoftau@server.daniel15.com 1171075989 Q * Daniel15 Killed (NickServ (GHOST command used by Daniel15-2)) 1171075997 N * Daniel15-2 Daniel15 1171076238 M * almak Just sent the output to pastbin 1171076364 M * Bertl seems you have a somewhat strange routing setup and maybe? the guest ip assigned to two interfaces? 1171076417 M * almak there is only one in the guest 1171076449 M * Bertl could you upload the output of 1171076472 M * Bertl 'ip route ls' and 'ip addr ls' (on the host and guest) too, please? 1171076482 M * Bertl btw, which kernel/patch is this? 1171076499 M * almak vserver patch 2.0.1 1171076614 M * almak sent 1171076738 M * Bertl ah, okay, that explains it 1171076898 M * Bertl the 11.11.11.1 is on a different network, you know? 1171076950 M * almak right. this only works if I setup a route or specify a -I in ping. 1171076972 M * Bertl yes, and then it just work 'by accident' 1171076978 M * Bertl *works 1171077008 M * Bertl is there a reason for the strange vlans? 1171077029 M * Bertl and more important, any reason for the even stranger ips? 1171077047 M * Bertl (which, btw, are public ips, at least some of them :) 1171077048 M * almak the IP, no, just my test network. 1171077111 M * Bertl in a real world setup, you would put the guests on a single network with a common gateway 1171077133 M * Bertl then they would reach that gateway quite fine, without any additional routing 1171077154 M * almak The strange 11.11.11.1 is an address in my router. My application talks to different VLANS 1171077163 M * almak per vserver 1171077197 M * Bertl okay, thing is, the 11.11.11.1 request will be sent to 1.100.50.126 1171077232 M * Bertl I guess you can verify that with tcpdump -vvnei if1 1171077250 M * Bertl now it really depends on how the router will answer that 1171077272 M * Bertl I'd expect that the answer differs between the two pings 1171077299 M * Bertl the game is probably a completely different for actual (tcp/udp) traffic 1171077392 M * almak interesting, so the 11.11.11.1 will be sent to 1.100.50.126 even when it is not present in the guest 1171077422 M * Bertl it is a 'remote' ip 1171077431 M * Bertl so it will take the default route 1171077526 M * almak So if I setup a default route in a table for 11.11.10.2 will the 11.11.11.1 request use it? 1171077549 M * Bertl yep 1171077573 M * Bertl you need the rule/route/table setup to do that though 1171077598 M * almak How about without setting up the 'remote' network ip 1171077602 M * almak in the table 1171077662 M * Bertl hmm? 1171077726 M * almak I think it is not possible. I want to mimic each guest just like a machine with a single interface. All traffic defaults through it. No route needs to be configured. 1171077750 M * Bertl that is what happens without all the vlans and routing stuff 1171077769 M * Bertl except for 'local' traffic, which will always stay local 1171077783 M * Bertl i.e. traffic between the guests and traffic to the host 1171077817 M * almak what is special about vlan? 1171077847 M * Bertl nothing, just that you do not use the vlans on the host 1171077861 M * Bertl i.e. you want _different_ setups between guest and host 1171077869 M * Bertl so you have to have different routing tables 1171077882 M * Bertl in a default setup, regardless of via vlan or not 1171077900 M * Bertl the host will be on the same network as the guests, and use the same default gateway 1171077984 M * almak ok. In this case, we have different default gateways. So routing is required. 1171078087 M * Bertl the best way, as I suggested in the beginning, is to have a completely separate routing table for the guest(s) 1171078117 M * Bertl put your routes, including the default route there, and use the ip rule ... from ... to map the guest ips to the table 1171078128 M * baldy my perl scripts working 1171078136 M * baldy sing 1171078141 M * Bertl almak: check that on the host, with ping and traceroute 1171078141 M * baldy dance 1171078145 M * baldy soooo 1171078145 M * Bertl baldy: congrats! 1171078151 M * baldy gn8 guys 1171078176 M * almak Cool thanks Bertl. 1171078276 M * Bertl you're welcome! 1171081785 J * olivierk ~olivier@olivierk.org 1171081892 Q * olivierk_ Ping timeout: 480 seconds 1171083077 Q * bronson Quit: Ex-Chat 1171084822 J * bronson ~bronson@adsl-75-36-145-145.dsl.pltn13.sbcglobal.net 1171084827 Q * bronson Read error: Connection reset by peer 1171085162 J * DoberMann_ ~james@AToulouse-156-1-64-247.w90-16.abo.wanadoo.fr 1171085269 Q * DoberMann[ZZZzzz] Ping timeout: 480 seconds 1171086292 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171086782 Q * infowolfe Quit: Leaving 1171086790 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171087187 Q * ZLinux Ping timeout: 480 seconds 1171087683 Q * Aiken Ping timeout: 480 seconds 1171088728 J * dna ~naucki@126-204-dsl.kielnet.net 1171091716 Q * dna Quit: Verlassend 1171093364 J * rgl ~Rui@84.90.10.107 1171093368 M * rgl hello 1171093454 M * rgl anyone running named inside vserver? the latest centos update borks with -"u with Linux threads not supported: no capabilities support or capabilities disa 1171093454 M * rgl bled at build time", even thou I've recompiled the rpm with --disable-linux-caps like in previous rpm :| 1171093576 M * Daniel15 Nope, sorry, I haven't tried named in a while 1171093828 J * Aiken ~james@ppp111-64.lns2.bne4.internode.on.net 1171094469 M * Bertl rgl: what patch/kernel version? 1171094500 M * rgl Bertl, I'm still on 2.6.17.13-vs2.0.2.1 1171094517 M * Bertl okay, this is worked around with 2.2.x 1171094533 M * Bertl on 2.0.x you have to recompile it with --disable-linuxcaps or so 1171094544 M * rgl I have to use --disable-threads when building named ? I'l almost sure I didn't do that on the previous build :| 1171094562 M * Bertl the threads are not important 1171094584 M * Bertl the linux capability thing is broken in bind 1171094586 M * rgl hat do you mean? 1171094599 M * rgl wee s,hat,what, :D 1171094618 M * Bertl http://linux-vserver.org/Problematic_Programs 1171094625 M * rgl I did rebuild it with --disable-linuxcaps 1171094657 M * rgl Bertl, yeah, read that too. but honestly, I don't remeber building it with --disable-threads (nor I have that on my notes) 1171094669 M * Bertl didn't know about the thread brokenness, must be a new feature :) 1171094704 M * Bertl okay, if the thread issue is valid on 2.2.x too, I'll investigate that further 1171095234 M * rgl Bertl, did you build bind/named on centos? 1171095246 M * Bertl nope, why should I? 1171095267 M * rgl to see the error :D 1171095281 M * Bertl it errors out on build? 1171095286 M * rgl its a FAQ thing on the bind sources: Configuring BIND9 with the --disable-threads option (the default) causes a 1171095286 M * rgl non-threaded version to be built, which will allow -u to be used. 1171095317 M * Bertl hmm, you could probably just remove -u then? 1171095319 M * rgl I wonder why the default rpm build .spec uses --enable-threads :| 1171095338 M * Bertl probably some 'evil' centos thing :) 1171095342 M * rgl I don't want to run it under root :D 1171095367 M * Bertl put it in a separate guest, and you can run it as root too 1171095429 M * rgl humm, when a binary uses threads, the /lib/libpthread.so.* should appear on ldd, right? 1171095509 M * Bertl not necessarily, but usually 1171095522 M * Bertl note, it is often part of glibc 1171095537 M * Bertl (i.e. glibc already pulls that in) 1171095547 M * rgl thats when glibc is nptl? 1171095570 M * Bertl both, nptl and linux threads support threading 1171095589 M * rgl yes. but they cannot be used at the same time, right? 1171095603 M * Bertl no, usually not 1171095605 M * rgl I mean, its an exclusive thing. use one or the other 1171095919 M * rgl oh, there are two libc version, one with linux threads at /lib/ and other with nptl at /lib/tls/ 1171096264 M * rgl Bertl, oh, I can now see that I didn't build the older bind with threads. because strings /usr/sbin/named|grep thread does produce any output *G* 1171096280 M * rgl err does *not* 1171096296 M * Bertl okay, so it was somehow disabled back then 1171096380 M * rgl I must have disable them, but didn't update my notes :( 1171096473 M * Bertl np, didn't take long to find ... 1171097369 M * waldi hmm? 1171097385 M * waldi which bind version? 1171097418 M * Bertl bind9/centos on 2.0.x as it seems 1171097485 M * waldi bind9/debian sarge works flawless with -u and threads 1171097879 J * jmcaricand ~kvirc@d83-179-149-229.cust.tele2.fr 1171097925 M * rgl waldi, its 9.2.4-24 1171098160 M * Bertl welcome jmcaricand! 1171098174 M * Bertl okay folks, I'm off to bed now ... have a good one everyone! cya! 1171098180 N * Bertl Bertl_zZ 1171098189 M * jmcaricand Bert: Hi 1171098195 M * rgl bye Bertl_zZ 1171098210 M * nigel must be from nz like me with a sleeptime like this 1171098311 M * Daniel15 I believe that Bertl is from... Actually, I have no idea what country he is from :P 1171098322 M * nigel :) 1171098334 A * Aiken thought it was Austria 1171098353 M * Daniel15 Probably is 1171098353 M * nigel austria? He'd be going to sleep at about 11am then... :) 1171098367 M * nigel australia? 1171098374 M * Daniel15 I'm from that big country next to New Zealand :D 1171098386 M * Daniel15 Bertl is not from Australia... I am, though :D 1171098392 M * nigel heh 1171098393 M * Aiken so am I 1171098402 M * nigel oo er 1171098406 A * nigel feels outnumbered 1171098415 M * Aiken Daniel15 where about? 1171098422 M * Daniel15 Melbourne 1171098450 M * Aiken I am about 1300 km north of you 1171098465 M * Aiken near Toowoomba 1171098499 M * Daniel15 :) 1171098578 M * Daniel15 http://linux-vserver.org/Developers 1171098587 M * Daniel15 Look at the map at the bottom :) 1171098614 M * Daniel15 One of those markers is where Bertl lives 1171098623 M * Daniel15 (I guess, anyways :P) 1171098627 M * nigel heh, not aussie then 1171098674 M * jmcaricand I want to create a configuration for a guest without to create root filesystem for it. What's the good method ? 1171098694 M * jmcaricand A simple methode. 1171098796 M * Daniel15 So, you want to create a vserver, with no files in the root filesystem? 1171098811 M * Aiken vserver build -m skeleton will create a very basic configuration and also create /vserver// with a populated dev 1171098813 M * Daniel15 vserver [name] build -m skeleton .... I guess 1171098830 M * Daniel15 Heh, Aiken replied one second faster than me :P 1171098892 M * nigel damnit, now I'm all distracted by the googlemap 1171098938 M * jmcaricand Thank. I tested. That works !!! 1171098956 M * Daniel15 No problem :) 1171099797 J * bonbons ~bonbons@83.222.37.103 1171100101 Q * rgl Quit: Fui embora 1171100563 Q * Daniel15 Quit: ( www.nnscript.de :: NoNameScript 4.02 :: www.XLhost.de ) 1171101013 J * id23 ~id@213.155.87.73 1171101024 M * id23 morning has broken... 1171101720 N * DoberMann_ DoberMann 1171102232 M * nebuchadnezzar Hello, I have a bug when booting 2.6.20-vs2.3.0.9 on my sparc system 1171102234 M * nebuchadnezzar http://paste.linux-vserver.org/1131 1171102253 M * nebuchadnezzar Bertl_zZ: when you will be up :-) 1171102384 J * dlezcano ~dlezcano@AToulouse-252-1-69-72.w81-250.abo.wanadoo.fr 1171102498 Q * jmcaricand Ping timeout: 480 seconds 1171103266 J * jmcaricand ~kvirc@d83-179-158-135.cust.tele2.fr 1171103650 Q * infowolfe Quit: Leaving 1171104958 Q * nigel Killed (NickServ (GHOST command used by Nigel__)) 1171105170 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171106073 J * meandtheshell ~markus@85-124-175-211.dynamic.xdsl-line.inode.at 1171107706 J * ZLinux ~ZLinux@88.213.62.142 1171108396 Q * ZLinux Remote host closed the connection 1171109886 M * phedny has anyone ever tryed to use vmware server and linux-vserver on the same host server? 1171110037 M * bonbons phedny: I tried once but results were rather bad on the VMWare side (vmware-player), NUT, I have not check if it would have worked better with a mainline kernel 1171110072 J * pflanze ~chris@84-73-56-44.dclient.hispeed.ch 1171110084 M * pflanze Hello. 1171110114 M * pflanze Since I'm making use of COW links, I'm seeing those in my kernel logs: ReiserFS: dm-1: warning: vs-8115: get_num_ver: not directory or indirect item 1171110140 M * eyck phedny: I'm using it in such scenario on multiple hosts without problems. 1171110262 M * phedny okay, then I'll give it a try 1171110355 M * bonbons the issues I had were freezing vmware guests when it wanted to switch video mode 1171110384 M * eyck no such problems here, 1171110420 M * phedny the server won't have a display connected, vmware will only be controlled remote 1171110436 M * eyck and I've got one of machines configured this way is in semi-production since ~5 months 1171110441 Q * Aiken Quit: Leaving 1171110469 M * pflanze Problem: /etc/vservers/*/fstab contains the usual column with a flag for fs checking; but e2fsck is being run in interactive mode, and fails when starting the vserver on boot time, so this feature is of no use. Right? 1171110485 M * phedny eyck: that's nice 1171110502 M * pflanze (s/no/not much/) 1171110567 M * eyck pflanze: hmm, e2fsck on boot usually is being run non-interactively, right? 1171110588 M * eyck phedny: definitely worth a try. 1171110588 M * pflanze eyck: no, it complains upon host booting. 1171110620 M * pflanze eyck: it says "e2fsck: need terminal for interactive repairs" and fails, thus the vserver doesn't start. 1171110637 M * pflanze It did take me quite some time to figure this out. 1171110647 M * eyck I don't get it, why do you have a filesystem access in guest at all? 1171110649 J * duckx ~Duck@tox.dyndns.org 1171110668 M * pflanze eyck: I don't, I have a mount in the fstab of a vserver. 1171110697 M * pflanze well, I *do* have *filesystem* access in the vserver, but not raw partition access, to be precise. 1171110711 M * pflanze e2fsck is run by the vserver tools. 1171110715 M * daniel_hozac pflanze: fsck shouldn't even run. 1171110748 M * pflanze daniel_hozac: why not? I have this line in the fstab: /dev/mirrorvg/e5_mysql_log /mnt/mysql_log ext2 defaults,noatime 1 1 1171110773 J * |jmcaricand| ~kvirc@d83-179-133-205.cust.tele2.fr 1171110789 M * daniel_hozac because grep fsck src/secure-mount.c doesn't return anything. 1171110808 M * pflanze Well.. for some reason it is being run. 1171110823 M * daniel_hozac nor does grep -Hr scripts return anything relevant. 1171110943 M * pflanze I've found out by running 'vserver foo start > `tempfile -p star-` 2>&1' separately from a system boot script. 1171110952 M * pflanze In the output I'm seeing the above message. 1171111007 M * pflanze I even *like* fsck to run. The problem is only that it would either have to run with the -n or -y options, or attach itself to a console. 1171111057 M * daniel_hozac are you sure it's not the guest that's running fsck? 1171111068 M * pflanze (The latter being the safer variant, but in my case even -y would be ok.) 1171111130 M * pflanze daniel_hozac: yep. The host doesn't have the device, and it's fstab contains only the line '/dev/hdv1 / ext2 defaults 1 1' 1171111138 M * pflanze ehr s/host/vserver/ 1171111183 M * daniel_hozac so changing those 1s to 0s doesn't do anything? 1171111198 M * pflanze Running "vserver foo start" manually from a terminal runs e2fsck sucessfully. 1171111208 Q * jmcaricand Ping timeout: 480 seconds 1171111254 M * pflanze hm I think changing them to 0 will prevent fsck from being run, 1171111264 M * pflanze but I'd have to check (by rebooting the host, after all) 1171111285 M * daniel_hozac given that the guest doesn't have the device, why does it matter? 1171111328 M * pflanze ?, the guest is getting the filesystem, and the filesystem should be checked if it's not using journaling.. 1171111354 M * pflanze I could mount the filesystem from the host's /etc/fstab, of course, 1171111365 M * pflanze but that's a bit ugly (and does not use namespaces) 1171111385 M * daniel_hozac or run fsck in an initialize script. 1171111397 M * pflanze Yes, that's what I've been planning to do. 1171111417 M * pflanze But I thought I'd say here, maybe investing into a cleaner solution would be worthwhile. 1171111527 M * pflanze Actually it should be enough to run the vserver startup init script tied to a console 1171111605 M * daniel_hozac given that fsck won't run anyhow, i'm not sure i see the point. 1171111617 M * pflanze why won't run anyhow? 1171111627 M * pflanze it *does* run if "1 1" is in the fstab 1171111634 M * daniel_hozac which fstab? 1171111643 M * pflanze /etc/vservers/foo/fstab 1171111651 M * daniel_hozac and you have verified that? 1171111666 M * daniel_hozac because there's _nothing_ running fsck based on that fstab. 1171111677 M * pflanze yes, as I said, "vserver foo start" shows messages from e2fsck 1171111682 M * pflanze hm 1171111687 A * pflanze double checks his scripts 1171111706 M * daniel_hozac and changing the 1s to 0s in _that_ fstab makes the message go away? 1171111708 M * pflanze *sorry* 1171111714 M * pflanze scripts/prepre-start contains the fsck call. 1171111721 M * pflanze Forgot about this somehow. 1171111738 M * pflanze ok. 1171111804 M * pflanze (to my defense, those "0 0" in the fstab did delude me) 1171111814 M * pflanze (dunno how those got into my fstab's though) 1171111955 Q * phreak`` Quit: leaving 1171111965 J * phreak`` ~phreak``@deimos.barfoo.org 1171112277 M * pflanze What is stdin/stdout/stderr when running lib/util-vserver/vserver-wrapper from init.d scripts on host boot? 1171112324 M * pflanze (more precisely, what are stdin/out/err when the scripts/prepre-start is being run by vserver-wrapper) 1171112538 A * pflanze is preparing a wiki entry 1171112548 M * daniel_hozac the defaulttty. 1171112588 M * daniel_hozac /etc/vservers/guest/apps/init/tty, /etc/vservers/.defaults/apps/init/tty or if neither is set, /dev/null. 1171112604 M * pflanze ah, thanks 1171112655 M * pflanze Is it correct that columns 5&6 (when counting from 1) are ignored by the utils? 1171112784 M * daniel_hozac the dump and fsck columns? yes. 1171113247 M * pflanze daniel_hozac: do you accept http://paste.linux-vserver.org/1132 ? 1171113248 M * |jmcaricand| Does somebody use selinux with vserver ? 1171113351 M * waldi it explicitely uses the capability security module 1171113389 M * daniel_hozac pflanze: i'm not sure how i feel about putting documentation in configuration files. 1171113415 M * pflanze daniel_hozac: you could just remove the "0 0" as alternative 1171113430 M * daniel_hozac pflanze: and running fsck from the initialize script is better. 1171113454 M * daniel_hozac during prepre, the root fs has already been mounted. 1171113463 M * pflanze ah. 1171113477 M * daniel_hozac pflanze: wouldn't it be better to actually use the dump field? 1171113501 M * pflanze how? 1171113517 M * daniel_hozac s/dump/fsck/, of course. 1171113545 M * pflanze I mean, for the future it may be a good thing to have those fields there. 1171113553 M * pflanze If someone wants to make use of them. 1171113563 M * pflanze But until then, it's confusing. 1171113579 M * pflanze I don't see problems with comments in that file. 1171113590 M * pflanze Since /etc/fstab can contain comments as well. 1171113607 M * pflanze But that may be a question of whether all /etc/vserver/* files accept comments. 1171113630 M * daniel_hozac the comments are not the problem, documentation in it is the problem. 1171113651 M * pflanze ah. Well so just add the line with the link to the wiki? 1171113654 M * daniel_hozac i'd rather accept a patch that runs fsck on the filesystems. 1171113664 M * pflanze hm 1171113669 M * pflanze okay.. 1171113693 M * pflanze But there's still the trap of not having a tty tied to the thing. 1171113714 M * pflanze But maybe that should be "dealt" with in some other way. 1171113727 M * pflanze What about a warning tool? 1171113737 M * pflanze Which checks the /etc/vservers contents? 1171113744 M * daniel_hozac so, it runs with -n. if the filesystem is corrupt, starting the guest should fail. 1171113748 M * pflanze vserver-config-check or something 1171113773 M * pflanze yeah, that may be a good way. 1171113857 M * pflanze So where is fstab being handled? 1171113887 M * pflanze (And what should "dump" do? I've never ever used this) 1171113944 J * DavidS ~david@chello062178045213.16.11.tuwien.teleweb.at 1171113961 M * daniel_hozac i guess it's used by dump 1171113978 M * daniel_hozac implementing that would probably be overkill though. 1171113981 M * pflanze yeah, I just wouldn't know what to implement in the vserver scripts. 1171114002 M * daniel_hozac it's not something that should happen on boot, IIUC. 1171114025 M * daniel_hozac the fstab is handled in secure-mount. 1171114038 M * daniel_hozac but fsck should run before that. 1171114062 M * daniel_hozac before mountRootFS in scripts/vserver.start. 1171114385 M * pflanze well what with other partitions than the root fs? 1171114412 M * pflanze is everything being handled at once by secure-mount? 1171114637 A * pflanze starts editing secure-mount.c 1171115453 M * pflanze Hm, specifying -n is problematic, after man fsck. 1171115481 M * pflanze It's not guaranteed that the checkers report problems with this option. 1171115522 M * pflanze So it's probably a bad idea to special case a missing tty stdin to "-n". 1171115536 M * pflanze And instead failing right away is probably better. 1171115577 M * pflanze The only issue I have with this is that it's non-obvious why an operation fails on host boot but not when issued from a terminal. 1171115603 M * pflanze But that problem may not be limited to fsck'ing. 1171115619 M * pflanze So better try to "fix" this otherwise. 1171115650 M * pflanze Maybe use a log file instead of /dev/null when init/tty is missing? 1171115658 M * pflanze Or why not syslog? 1171116061 M * pflanze I still don't know how init scripts are being started on boot (normally, e.g. on host). One can see script output on the console, so one would expect that stdout/err is tied to the console? 1171116087 M * pflanze Or is it going to syslog and some proxy is outputting syslog messages to the console? 1171116149 M * pflanze My question basically is: why is vserver-wrapper opening to /dev/null ? 1171116210 M * pflanze or who is opening to /dev/null, daniel_hozac? 1171116368 M * pflanze Hm, looks like ./scripts/vserver.start is redirecting to /dev/null 1171116416 M * pflanze (or is it just in cleanup() ) 1171116448 A * pflanze is stopping further development for the moment and going to other work 1171117476 M * daniel_hozac pflanze: when --defaulttty is specified, and no tty has been configured, /dev/null is used. 1171117486 M * daniel_hozac vserver-wrapper uses --defaulttty. 1171117533 M * daniel_hozac and IMHO, it's better to check all the filesystems at once, before much of anything has happened, and output a "fsck returned X for filesystem Y" sort of message and fail. 1171117799 M * pflanze But what if no output can be seen by the user? 1171117814 M * pflanze What's the reason to open /dev/null? 1171117840 M * pflanze pipe to a syslogger would make it be seen. 1171117881 M * pflanze What I'm wondering why do reopen stdout/err *at all* unless a defaulttty setting is configured? 1171117909 M * pflanze Since I suspect if no reopen is being done, the output at least goes the same route as any other system boot output, right? 1171117993 M * daniel_hozac because cluttering the host's boot process with messages from multiple guests starting in parallell would be extremely confusing, at best? 1171118027 M * pflanze hm 1171118073 M * pflanze use a separate syslog name for each vserver? 1171118147 M * pflanze or just let it clutter, that at least forces the user to set up logging correctly. 1171118166 M * daniel_hozac piping the output to syslog doesn't make much sense, IMHO. 1171118179 M * pflanze yeah, a separate logfile per vserver would be more practical 1171118217 M * pflanze so either configure a tty, or a logfile. If neither is done, just let it clutter the boot console. 1171118241 M * daniel_hozac i prefer a Just Works(tm) methodology. 1171118279 M * pflanze It didn't work for me, I didn't find log output and had to wrap scripts to make output seen. 1171118286 M * daniel_hozac to be perfectly honest, i don't see what harm it does if the output is sent to /dev/null by default. 1171118327 M * daniel_hozac so, you had a broken script, and util-vserver just ought to deal with that? 1171118349 M * pflanze I didn't know how to see output. 1171118365 M * daniel_hozac vserver ... start? 1171118375 M * pflanze yeah, but then my problem didn't happen! 1171118387 M * pflanze it only happened on boot, because there was no tty. 1171118393 M * daniel_hozac because you had a broken script. 1171118403 M * pflanze "yeah" but see above: 1171118413 M * pflanze my point is, I'd rather not try to fall back to "-n" to fsck. 1171118434 M * pflanze No more not-really-working special cases. 1171118447 M * pflanze That makes it more difficult to track down problems. 1171118454 M * daniel_hozac fsck should always be called with -n. 1171118473 M * pflanze hm 1171118478 M * pflanze is it called like that by the boot process? 1171118488 M * pflanze I just read the warning in the man page. 1171118519 M * pflanze Maybe if only checkers don't report failure where it is not important, then ok. 1171118544 M * pflanze How are commands being run by the system boot process? Is there a tty on stdin? 1171118548 M * pflanze normally, that is. 1171118570 M * daniel_hozac depends on the OS, i guess. 1171118592 M * daniel_hozac IMHO we just want to run fsck to check for problems. 1171118596 M * daniel_hozac if there are any, bail out. 1171118607 M * daniel_hozac we don't want to interactively repair the filesystem. 1171118610 M * pflanze yes, if -n is suitable for this, fine. 1171118637 M * pflanze but then I'd always run with -n, not only if a tty is missing. 1171118648 M * daniel_hozac _exactly_. 1171118658 M * nebuchadnezzar I have a strange error, I do not understand what happens http://paste.linux-vserver.org/1133 1171118686 M * daniel_hozac nebuchadnezzar: looks like the network context doesn't exist anymore? 1171118693 M * daniel_hozac nebuchadnezzar: what does ls -l /proc/virtnet show? 1171118715 M * daniel_hozac pflanze: note that i have no idea what option is suitable for that. 1171118731 M * daniel_hozac nebuchadnezzar: oh, that's during start, hmm... 1171118740 M * matti daniel_hozac: Hi there ;] 1171118744 M * daniel_hozac hey matti. 1171118751 M * pflanze Can I configure a logfile as output (not a tty) for vservers with the current tools? 1171118766 M * nebuchadnezzar daniel_hozac: I found the problem 1171118766 M * daniel_hozac somewhat. 1171118780 M * nebuchadnezzar I use 65535 for the context 1171118784 M * nebuchadnezzar to high 1171118802 M * daniel_hozac nebuchadnezzar: ah. 1171118840 M * daniel_hozac pflanze: if you point the apps/init/tty link at a file, stdout/err will go there, and stdin will be read from it. 1171118842 M * nebuchadnezzar strange, on a 2.6.16-17-vs2.1.1-rc21 it works 1171118859 M * daniel_hozac nebuchadnezzar: with 65535? 1171118865 M * nebuchadnezzar yes 1171118875 M * nebuchadnezzar maybe now there is a XID-1 somewhere :-) 1171118978 M * nebuchadnezzar daniel_hozac: do you have any clue about http://paste.linux-vserver.org/1131 ? 1171119007 M * daniel_hozac no, not really... 1171119025 M * daniel_hozac i think someone else reported a similar issue. 1171119080 M * nebuchadnezzar I there a way to see what's wrong in my /proc after that ? 1171119175 M * daniel_hozac setattr -R --~hide /proc/tty gets you the same error? 1171119223 M * nebuchadnezzar I reboot on that kernel 1171119324 M * daniel_hozac nebuchadnezzar: have you disabled dynamic contexts on that kernel? (where you use xid 65535) 1171119354 M * nebuchadnezzar 2.3.0.9 do not have dynamic anymore, no ? 1171119464 M * daniel_hozac ah, you're right. 1171119514 M * daniel_hozac does vserver-info -q 65535 XIDTYPE static && echo yes output yes? 1171119723 M * nebuchadnezzar daniel_hozac: the setattr -R --~hide /proc/tty give me the Bad address error 1171119742 M * daniel_hozac okay, i guess an strace of that might be interesting. 1171119748 M * nebuchadnezzar and the vserver-info output yes 1171119763 M * daniel_hozac and you still can't start a guest with xid 65535? 1171119764 M * nebuchadnezzar strace -Ff ? 1171119767 M * daniel_hozac yeah 1171119788 M * nebuchadnezzar yes, I can't 1171119824 M * daniel_hozac hmm, you're sure that command outputs yes? 1171119855 M * daniel_hozac because i think there's a bug in the utils that would cause it to always output nothing. 1171119882 M * nebuchadnezzar http://www.asgardr.info/~nebu/vserver/setattr-R--hide-proc-tty.strace 1171119909 M * nebuchadnezzar arf, no sorry 1171119912 M * nebuchadnezzar I made a typo 1171119919 M * nebuchadnezzar it does not output yes 1171119939 M * daniel_hozac okay. 1171119986 M * daniel_hozac http://svn.linux-vserver.org/projects/util-vserver/changeset/2493?format=diff&new=2493 should fix it. 1171120022 M * nebuchadnezzar ok 1171120063 M * nebuchadnezzar is the 0.30.213 released ? 1171121588 Q * |jmcaricand| Quit: KVIrc 3.2.4 Anomalies http://www.kvirc.net/ 1171121967 M * daniel_hozac no. 1171122377 M * Guy- can xserver-xorg be run in a vserver? 1171122412 M * Guy- (can it do its fancy low-lever hardware woodoo?) 1171122422 M * Guy- s/er/el/ 1171122442 M * Guy- and voodoo 1171122459 M * Guy- this is not my day as far as spelling is concerned :) 1171122487 M * daniel_hozac if you give it the device nodes and capabilities required, sure. 1171122520 M * daniel_hozac http://oldwiki.linux-vserver.org/MoreUbuntu is AFAIK the only documented Xorg usage. 1171123730 Q * shedi Quit: Leaving 1171124129 M * pflanze I'm trying to hook a "screen" session to apps/init/tty. Not sure how to start a screen session w/o any program but only to generate tty's, though. 1171124190 M * pflanze (It might turn out being a good solution to a standard output setup. Does not take up disk space, and allows interaction if necessary.) 1171124269 J * ema ~ema@lart.galliera.it 1171124354 M * pflanze (s/allows interaction/allows remote interaction/, and does not need a multitude of virtual console instances) 1171125065 N * Bertl_zZ Bertl 1171125069 M * Bertl morning folks! 1171125184 M * Bertl pflanze: hmm, what's the idea behind that? 1171125232 J * olivierk_ ~olivier@olivierk.org 1171125292 M * pflanze Bertl: sometimes vserver setup is broken in subtle ways and doesn't startup through vserver-wrapper on host boot. 1171125337 M * pflanze It would be nice to have access to the error messages then. Or even interactive access. 1171125344 Q * olivierk Ping timeout: 480 seconds 1171125365 M * pflanze Linking /etc/vservers//apps/init/tty to a file may be problematic. 1171125374 M * Bertl pflanze: hmm, why not redirect that to a file then? or have it write to a cnsole/terminal? 1171125391 M * pflanze Since init/tty also connects stdin to read from that file. 1171125426 M * pflanze So while thinking about a better approach, "screen" has come to my mind. 1171125445 M * Bertl but you are interested in the startup messages _on_ the host or _inside_ the guest? 1171125466 M * pflanze For now only on the host. 1171125480 M * pflanze It may be useful for vserver owners to get acess to them also, though. 1171125487 M * Bertl so those should not do any fancy I/O IMHO 1171125487 M * pflanze without them having access to the host. 1171125507 M * Bertl i.e. it should be fine to redirect output and error to a file 1171125555 M * Bertl (guest init should never get in touch with the host side) 1171125565 M * pflanze yes 1171125603 M * pflanze For me currently just a file would be ok, I can always run "vserver foo start" from the host cmdline. 1171125616 M * pflanze Not sure what vserver owners should do, though. 1171125636 M * pflanze But yeah, probably they never should get access to the startup sequence anyway 1171125653 M * pflanze (My example is fsck running, asking for y/n style input) 1171125668 M * pflanze (But users shouldn't be allowed to decide what fsck is fixing and what not.) 1171125686 M * Bertl ahem? fsck on the host for guest startup? 1171125709 M * pflanze yes, if you put an ext2 partition into /etc/vservers//fstab 1171125735 M * Bertl I don't think that will be checked ... 1171125747 M * pflanze Currently, secure_mount does not honor the fsck flag. 1171125757 M * pflanze I've been talking with daniel about adding this. 1171125776 M * Bertl hmm, not sure that is a good idea ... 1171125777 M * pflanze Currently I'm running fsck from a scripts/prepre-start (which is wrong anyway) 1171125782 N * olivierk_ olivierk 1171125806 M * Bertl why not run that on host startup with the host scripts/fstab? 1171125816 M * pflanze That ties the vserver to the host. 1171125834 M * pflanze Or: ties it more to the host than otherwise. 1171125859 M * Bertl but the filesystem will be tied to the host anyway, after all, it's host devices, no? 1171125874 M * pflanze It could be a network block device, for example. 1171125894 M * Bertl you have to setup that first 1171125896 M * pflanze In some ways it will be tied to the host anyway, yes. 1171125919 M * Bertl note: I'm not against the fsck stuff at all 1171125939 M * Bertl I'm just not convinced that it is such a good idea to add more host side stuff to guests 1171125959 M * pflanze I thought it was nice. 1171125966 M * pflanze But this leads to my first problem: 1171125979 M * pflanze I was mislead by the "0 0" columns in the example fstabs. 1171126004 M * Bertl that means, do not check and make full backup :) 1171126005 M * pflanze I did change them to "1 1". What I didn't remember is that this didn't work and I then added the fsck to the prepre-start script. 1171126047 M * Bertl why not to the initialize script? 1171126050 M * pflanze Today I have told daniel at length that vserver scripts are calling fsck while he said they didn't. 1171126064 M * pflanze Until I found out that I did add that prepre-script line. 1171126090 M * pflanze I think a comment line warning about dump/fsck flags being ignored should be added to the example fstab 1171126120 M * pflanze Daniel didn't want to add this, saying that he'd prefer not adding documentation, but fixes to the code. 1171126140 M * pflanze Yes, he told me to add it to the initialize script instead. 1171126173 M * pflanze So I'm pretty much back at my suggestion to add the document patch. 1171126221 M * pflanze And move the fsck call to some host startup script (not /etc/fstab, btw, since I dont' want to *mount* it in the host context). 1171126296 M * Bertl okay, well, I will not interfere with daniel_hozac's decisions regarding the tools (that's his and ensc's domain), but I think there are much more usefull improvements which could be done, than adding a probably problematic (because it might require interaction or fail) fsck which should not be needed in almost every setup ... 1171126368 M * pflanze yes, imho just adding the documentation patch (with the link to the wiki fstab page) should be fine for now. 1171126376 M * Bertl I'm still wondering what the status of 'vserver reconfig/config/backup' is ... 1171126391 M * daniel_hozac waiting on the config library. 1171126403 M * daniel_hozac (which is on my TODO-list for 0.30.214) 1171126489 M * Bertl hey daniel_hozac! 1171126499 M * Bertl the nebuchadnezzar issue is resolved? 1171126505 M * daniel_hozac hmm, which one? 1171126513 M * Bertl http://paste.linux-vserver.org/1131 1171126531 M * daniel_hozac no, not yet. 1171126543 M * daniel_hozac (seems i even missed the strace) 1171126553 M * Bertl okay, any ideas? dietlibc? 1171126589 M * daniel_hozac i suppose, but that shouldn't be kernel dependent, should it? 1171126757 M * Bertl could be a tricky special case with the readdir return behaviour 1171126832 M * Bertl the libc function gives NULL on EOF and error 1171126850 M * daniel_hozac http://www.asgardr.info/~nebu/vserver/setattr-R--hide-proc-tty.strace shows getdents64 returning EFAULT though. 1171127017 M * Bertl did you google for getdents64? 1171127042 M * Bertl the first 100 pages contain 'problem' :) 1171127046 M * daniel_hozac lol 1171127185 J * lilalinux ~plasma@80.69.41.2 1171127214 M * Bertl wb lilalinux! 1171127246 M * Bertl daniel_hozac: but I guess I have a test program somewhere ... 1171127328 M * Bertl ah, no, that was for proc fs explicitely 1171127338 M * Bertl http://vserver.13thfloor.at/Stuff/penum.c 1171127353 M * Bertl thought I had one for opendir too ... 1171127408 N * DoberMann DoberMann[Course2caddie] 1171129159 M * nebuchadnezzar Bertl: hi 1171129167 M * Bertl hey 1171129183 M * Bertl the readdir() issue is not present with 2.2.x ? 1171129201 M * nebuchadnezzar I do not test 1171129219 M * nebuchadnezzar I believe I can try 1171129221 M * nebuchadnezzar :-) 1171129224 M * Bertl ah, daniel_hozac said something about depending on the kernel 1171129449 Q * fosco Remote host closed the connection 1171129452 J * fosco fosco@konoha.devnullteam.org 1171130601 Q * mnemoc Ping timeout: 480 seconds 1171130629 J * mnemoc ~amery@kilo105.server4you.de 1171130667 Q * dlezcano Ping timeout: 480 seconds 1171130851 Q * lilalinux Remote host closed the connection 1171131353 N * DoberMann[Course2caddie] DoberMann 1171131353 J * dlezcano ~dlezcano@AToulouse-252-1-9-35.w82-125.abo.wanadoo.fr 1171131868 M * Bertl wb dlezcano! 1171132213 Q * fosco Ping timeout: 480 seconds 1171132332 J * Piet hiddenserv@tor.noreply.org 1171132582 J * fosco fosco@konoha.devnullteam.org 1171132748 M * nebuchadnezzar Bertl: I have the bad address with vs2.2.0-pre3 too 1171133201 M * Bertl okay, so, when do you have it not? 1171133232 M * nebuchadnezzar when do I have it not ? sorry my english is not very great ;-) 1171133254 M * nebuchadnezzar It's ok with 2.6.19.2-vs2.3.0.8 1171133262 M * nebuchadnezzar if it's what you want 1171133265 M * Bertl yep 1171133373 M * Bertl so, switching from 2.6.19 to 2.6.20 causes that? 1171133391 M * Bertl or is the 2.3.0.9 on 2.6.19.3 ? 1171133408 M * nebuchadnezzar I do not test 2.3.0.9 on 2.6.19.3 1171133459 M * Bertl okay, then let's try that one too, if possible, just for the case 2.6.20 is causing this 1171133486 M * nebuchadnezzar I just have the unaligned access ont the 2.6.19.2 1171133491 M * nebuchadnezzar I'll try the 2.6.19.3 1171133518 M * Bertl okay, if you get unaligned access there too, please let me know 1171133537 M * nebuchadnezzar I have unaligned access from long time now 1171133557 M * Bertl but with 2.6.20 it is gone? 1171133612 M * nebuchadnezzar It's replaced by Bad address 1171133629 M * nebuchadnezzar when the util-vserver init script is launched 1171133642 M * Bertl ah, okay, that now makes sense 1171133675 M * Bertl daniel_hozac: guess the improper alignment we saw in readdir() is now replaced by a simple EFAULT 1171133698 M * Bertl (kernel side, so it isn't fixed up by the kernel, but just bails out) 1171133760 M * Bertl this would mean two things, a) it's a dietlibc/util-vserver issue, and b) fixing up the alignment will magically cure the fault 1171134020 J * lilalinux ~plasma@80.69.41.2 1171134199 N * DoberMann DoberMann[PullA] 1171134270 J * ZLinux ~ZLinux@88.213.62.142 1171134381 M * Bertl wb ZLinux! 1171134394 M * ZLinux thx Bertl :) 1171134496 Q * lilalinux Remote host closed the connection 1171135083 M * daniel_hozac Bertl: do we know a is true for sure? 1171135158 M * daniel_hozac and where would we align it? 1171135247 M * Bertl no, we do not know, but it is a good theory, no? 1171135269 M * Bertl and I know that alpha and sparc have special alignment requirements for 64bit types 1171135284 M * Bertl i.e. they have to be aligned on 64bit boundaries or worse 1171135351 M * daniel_hozac hmm, yeah, sounds possible. 1171135413 M * Bertl I'm somewhat confident that testing with 2.6.19 (and the 'old' config) will bring back the misaligned stuff, and make the readdir() work 1171135421 M * Bertl okay, off for now .. back later ... 1171135425 N * Bertl Bertl_oO 1171135428 M * daniel_hozac okay, cya! 1171135633 M * daniel_hozac nebuchadnezzar: could you test a hack-patch to dietlibc? 1171135727 M * daniel_hozac nebuchadnezzar: http://people.linux-vserver.org/~dhozac/p/m/delta-dietdirent-test01.diff 1171136428 J * Piet_ hiddenserv@tor.noreply.org 1171136794 Q * Piet Ping timeout: 480 seconds 1171136836 J * shedi ~siggi@ftth-237-144.hive.is 1171138208 M * nebuchadnezzar daniel_hozac: your patch seems to work 1171138236 M * nebuchadnezzar no misaligned message nor bad address 1171138252 M * nebuchadnezzar setattr -R --~hide /proc/tty seems to work 1171138453 M * daniel_hozac okay, thanks. 1171139387 Q * Piet_ Quit: Piet_ 1171139737 Q * infowolfe Ping timeout: 480 seconds 1171139973 Q * olivierk Read error: Operation timed out 1171140010 J * olivierk ~olivier@olivierk.org 1171140171 M * pflanze Strange. Somehow "vserver foo exec apt-get install --reinstall openssh-server" will not work, 1171140176 J * flewid ~flewid@CPE000024c43e1c-CM000f9fac6da2.cpe.net.cable.rogers.com 1171140183 M * flewid hi there 1171140190 M * pflanze whereas "vserver foo exec bash -c 'apt-get install --reinstall openssh-server'" will 1171140198 M * flewid i have a quick question - i've been using a gentoo vserver/host for about a year and a half now 1171140200 M * pflanze consistently. 1171140211 M * flewid yesterday i logged into my host system, issues a "vserver XXXX restart" 1171140216 M * flewid and none of the programs restarted 1171140223 M * flewid and now it's happening in all the vserver guests :( 1171140233 M * flewid i am able to vserver XXX enter, and then start the programs manually 1171140237 M * flewid but that's a bit of a pain in the ass 1171140240 M * flewid anyone have any ideas? :) 1171140303 J * FireEgl Proteus@2001:5c0:84dc:1:211:9ff:feca:b042 1171140311 M * daniel_hozac pflanze: "will not work"? 1171140364 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171140467 M * daniel_hozac flewid: any leftover run files or similar? 1171140492 M * flewid well, i've tried this with a brand new vserver that i just setup, that has one program (syslog-ng) 1171140495 M * flewid and it's still not starting 1171140500 M * flewid when i boot them, i see 1171140506 M * flewid root 24708 0.0 0.2 3052 1312 ? S 20:32 0:00 /bin/bash /sbin/rc default 1171140510 M * flewid running about 5-8 times 1171140533 M * pflanze daniel_hozac: here's an example: http://paste.linux-vserver.org/1134 , here's another one: http://paste.linux-vserver.org/1135 1171140564 M * pflanze daniel_hozac: it's always another weird error. But always an error. While running it through bash always worked so far. 1171140630 M * flewid should i try re-emerging baselayout or something within one of the guests? 1171140662 M * pflanze daniel_hozac: hmm, scratch that "always". Every now and then bash -c also fails. Investigating more. 1171140687 M * daniel_hozac flewid: your guests use baselayout? 1171140700 M * flewid i'm just grasping at straws :) 1171140704 M * daniel_hozac flewid: they're supposed to use baselayout-vserver. 1171140726 M * flewid well, that's what i meant - baselayout-vserver, should i try re-emerging that within the guest to see if that makes it boot? 1171140843 M * flewid interesting 1171140846 M * flewid that seems to have fixed it? 1171140850 M * daniel_hozac i guess. 1171140850 M * flewid well, sorta, sshd is started 1171140867 M * flewid localhost / # /etc/init.d/syslog-ng start 1171140868 M * flewid * Starting syslog-ng ... 1171140868 M * flewid Error opening file /proc/kmsg for reading (Operation not permitted) 1171140868 M * flewid Error initializing configuration, exiting. 1171140868 M * flewid * Failed to start syslog-ng [ !! ] 1171140870 M * flewid hmm 1171140918 M * daniel_hozac as expected. 1171140922 Q * ema Quit: leaving 1171140988 M * flewid sorry, it's been a while since i've played around with this :/ don't mean to seem newbish 1171141026 M * flewid sweet, well, re-emerging baselayout-vserver in the vservers 1171141030 M * flewid seems to have fixed the bootup 1171141041 M * flewid expected == with syslog-ng you mean? 1171141048 M * daniel_hozac yep. 1171141061 M * daniel_hozac you need to remove the /proc/kmsg part from the configuration. 1171141137 M * pflanze daniel_hozac: there is definitely never an error with "vserver fula enter" and then "apt-get install --reinstall openssh-server"; and there is also never an error with: 1171141142 M * pflanze root@elvis root# vserver fula exec bash -c 'sleep 1; apt-get install --reinstall openssh-server' 1171141150 M * pflanze The sleep 1 makes the difference. 1171141180 M * daniel_hozac pflanze: and what exaclty is it that's failing? 1171141192 M * pflanze always different, weird file not found and such. 1171141198 M * pflanze see: http://paste.linux-vserver.org/1137 1171141216 M * daniel_hozac and what kernel is this? 1171141225 M * pflanze 2.6.19.3-vs2.2.0-rc11 1171141234 M * pflanze 0.30.213-rc1 1171141410 M * pflanze daniel_hozac: you remember my error from yesterday?: 1171141413 M * pflanze Feb 08 23:53:24 daniel_hozac: what's this?: vserver imap hashify -> sh: /lib/tls/libc.so.6: version `GLIBC_2.3.4' not found (required by sh)\n failed to determine configfiles 1171141423 M * daniel_hozac yep. 1171141425 M * pflanze I suspect it's the same issue 1171141835 M * daniel_hozac if you use something simpler, like ls -l /, do you ever get the host's contents or similar? 1171141847 A * pflanze checks 1171141921 M * pflanze fun, http://paste.linux-vserver.org/1138 1171141935 A * pflanze thinking about how to do sub-second sleeps in shell 1171141968 M * pflanze btw /root/bin/ls is a wrappe by mine. 1171141982 M * pflanze root@elvis root# vserver fula exec bash -c '/bin/ls -l /' -> shows the contents of the vserver 1171142000 M * daniel_hozac what's line 53? 1171142024 J * infowolfe_ ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171142038 M * pflanze open PP," gives the same problems as without the sleep 1 1171142475 M * pflanze so the exec reverses the sleep benefit. 1171142744 M * flewid hmm 1171142747 M * flewid yah, within the vserver 1171142754 M * flewid the /proc/sys/kernel/sem gives the same values 1171142770 M * flewid it's weird, all the other vservers are working as expected again 1171142776 M * daniel_hozac pflanze: so, apt-get requires the parent to be within the guest? 1171142776 M * flewid but this one wont' even start apache 1171142792 M * pflanze some more data points: http://paste.linux-vserver.org/1140 1171142802 M * pflanze daniel_hozac: yes exactly, as is shown by the perl example. 1171142827 M * pflanze daniel_hozac: note however that sometimes apt works even with the parent in the host. 1171142847 M * pflanze it's just random 1171142856 M * flewid Filesystem Size Used Avail Use% Mounted on 1171142856 M * flewid /dev/hdv1 20G 32K 9.6G 1% / 1171142862 M * flewid yeah cause it's definately still has space left 1171142862 M * flewid :p 1171142886 M * daniel_hozac flewid: df -i? 1171142899 M * flewid Filesystem Inodes IUsed IFree IUse% Mounted on 1171142899 M * flewid /dev/hdv1 2000000 22 1999978 1% / 1171142899 M * flewid none 64280 3 64277 1% /tmp 1171142922 M * daniel_hozac flewid: i guess chcontext --xid 1 ipcs -s might shed some light on it. 1171142924 M * pflanze To be more explicit: this always works (note the fork): vserver fula exec perl -we 'my $pid=fork; defined $pid or die "fork: $!"; if($pid) { wait } else {exec qw(apt-get install --reinstall openssh-server) or die $!}' 1171142957 M * pflanze Without the fork it's the random business as always. 1171142962 M * daniel_hozac pflanze: well, this could be fixed in -rc12. 1171142980 M * flewid vserver 1021 # chcontext --xid 1 ipcs -s 1171142981 M * flewid ------ Semaphore Arrays -------- 1171142981 M * flewid key semid owner perms nsems 1171143005 M * flewid samething if i change --xid to 1021 (the vserverid) 1171143009 M * daniel_hozac we did fix a bug with the refcounting on the fs_struct's (which store the root), so it's possible you're getting one which has incorrect values. 1171143010 M * flewid brb, i think i need a coffee 1171143027 M * pflanze daniel_hozac: rc12 of the kernel patch, right? ok I'll build a new kernel. 1171143121 M * flewid ok back 1171143194 M * flewid boom fixed it 1171143201 M * flewid had to do it from the vserver host, not within the vserver 1171143241 M * flewid now just to fix ejabberd on my vserver and we'll be back in business :) 1171143270 P * dm8tbr User started wasting time elsewhere! 1171143598 J * _Medivh ck@paradise.by.the.dashboardlight.de 1171143598 Q * Medivh Read error: Connection reset by peer 1171144503 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171144539 Q * infowolfe_ Ping timeout: 480 seconds 1171144652 N * DoberMann[PullA] DoberMann 1171144772 J * chris_ ~chris@84-73-58-227.dclient.hispeed.ch 1171145085 Q * pflanze Read error: Operation timed out 1171145430 Q * DavidS Quit: Leaving. 1171145682 N * chris_ pflanze 1171146758 J * infowolfe_ ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171147148 Q * infowolfe Ping timeout: 480 seconds 1171147331 M * pflanze daniel_hozac: now that is fun: 1171147346 M * pflanze My tests did hose the ssh installation *on the host* 1171147364 M * pflanze It *did* execute the stuff in the host context, half way. 1171147459 M * daniel_hozac nah, just the chroot would've been affected. 1171147468 N * DoberMann DoberMann[ZZZzzz] 1171147472 M * daniel_hozac is -rc12 okay? 1171147475 M * pflanze I have sarge on the host, which is calling the package "ssh". I did run the tests against an etch vserver, which is calling the package "openssh-server". 1171147504 M * pflanze It did uninstall "ssh" on the host and halfway install the etch openssh-server package on the sarge host. 1171147515 M * pflanze Of course this failed, it was left in iU status. 1171147537 M * pflanze With an ssh binary that cannot be started because of missing shared libs. 1171147561 M * pflanze And the package system complained about missing package dependencies. 1171147579 M * pflanze "apt-get -f install" did remove the openssh-server package. "apt-get install ssh" did restore the sarge package. 1171147594 M * pflanze This is what I had to do to gain ssh access to the host again. 1171147617 M * pflanze I'm running rc12 now. Going to do my tests again now. 1171147774 M * pflanze BTW I've double checked: I do *not* have etch sources in the souces.list on the host. 1171147800 M * pflanze So the *only way* how an etch package could have landed on the host was through a program on the etch vserver. 1171147927 M * daniel_hozac well, if you're hitting the fs_struct issue, there's really no telling what would use what root. 1171148048 M * daniel_hozac assuming there's not a single other process on the system forking, some heuristics could probably be devised, but i don't think it's worth it. 1171148075 M * pflanze heuristics for what? 1171148089 M * flewid daniel: thanks again for your help earlier :) i think all is working again 1171148097 M * flewid minus ejabberd , but i think that's ejabberd's fault, not the vserver 1171148124 M * daniel_hozac heuristics for which process would get which fs_struct, and when it might change to one of the host's rather than the guest's, or the other way around. 1171148153 M * pflanze Why is it difficult to choose the right fs_struct? 1171148216 M * daniel_hozac because the accounting was wrong. 1171148225 M * daniel_hozac which would free it prematurely. 1171148238 M * daniel_hozac leaving the previous address up for grabs for new fs_structs. 1171148311 M * pflanze why would this need heuristics? 1171148341 M * pflanze refcounting should work precisely as long as there are no cycles involved. 1171148403 J * mire ~mire@199-166-222-85.adsl.verat.net 1171148499 M * pflanze Up to now all my non-working exec calls are working now. 1171148508 M * pflanze But with memory issues one never knows, of course. 1171148554 M * pflanze Maybe I should try later, after the machine has been running for some time. 1171148569 M * daniel_hozac every fork should allocate a new fs_struct. 1171148735 M * pflanze You say that it was a premature refcount decrement and thus free of an fs_struct, and subsequent reuse of the memory for a new structure, if I understand right. If my machine at the moment will not have the same memory pattern, reuse stuff could be different than before the reboot. 1171148739 M * pflanze That's what I mean. 1171148804 M * daniel_hozac so, run a forkbomb in the host while trying to vserver ... exec stuff. 1171148925 M * transacid i have a dedicated server with 2 publix ip adresses. one is assigned on the host system (eth0). the vservers are in a 10.1.1.0 net. now i want to assign the second public ip adress to one vserver but it's not reachable from outside 1171149063 M * daniel_hozac how did you assign it to the guest? 1171149095 M * transacid in the host in /etc/vserver/f00/interfaces/* 1171149170 M * daniel_hozac by doing creating a new directory and adding the ip, dev and prefix files in it? 1171149181 M * daniel_hozac s/doing// 1171149226 M * transacid i created a new guest and set the stuff in the interface dir 1171149292 M * daniel_hozac yes, but what i'm getting is, what's "stuff"? how does it not work? is the address assigned? 1171149321 M * transacid in the guest ifconfig says it's assigned 1171149369 M * daniel_hozac to the correct interface? 1171149404 M * pflanze (Heh, my fork bomb suddenly vanished while I've been cloning a new vserver. How could be? I've got no cpu limit set, only ulimit -u 20. Sneaky fork bomb killer? But no messages in the kernel log.) 1171149414 M * transacid what do you mean by correct interface? the server (hardware) has only one interface 1171149502 Q * shedi Remote host closed the connection 1171149564 M * transacid perhaps iptables has a problem with that 1171149594 M * daniel_hozac so, how are you testing it, and how is it not working? 1171149623 M * transacid i tried to ssh to the public ip i assigned 1171149642 M * daniel_hozac and you mentioned iptables, are you sure that's not getting in the way? 1171149696 M * transacid i'm not sure, maybe, because i dunno how to do the DNATing 1171149715 M * transacid because the host has only eth0 1171149769 M * daniel_hozac and what happens if you just reset iptables? can you access it then? 1171149794 M * daniel_hozac have you checked with tcpdump to see the traffic? 1171149848 M * transacid mom i'll try that 1171149864 Q * ensc Killed (NickServ (GHOST command used by ensc_)) 1171149874 J * ensc ~irc-ensc@p54B4DA4B.dip.t-dialin.net 1171149907 M * transacid ah ok it's definately a iptables prob 1171149911 M * transacid sory for bothering 1171149925 M * transacid hmm now i need to find out how to work around 1171149938 M * transacid *reading iptables book again* 1171149942 M * daniel_hozac work around what? 1171149975 M * transacid no not "work around", i just need new rules for input chain 1171150006 Q * meandtheshell Quit: Leaving. 1171150200 M * pflanze Ok, the new kernel seems to behave reliably now. 1171150287 M * pflanze The processes still get the pid of the host parent as ppid, though, but I guess this is what you expect. 1171150308 M * pflanze But I wonder, why do you return that pid and not 1? 1171150333 M * daniel_hozac because the parent _is_ the host process. 1171150340 M * pflanze The process should have no way of contacting the real pid, so what use is it to give it? 1171150358 M * pflanze For the client it should look like the process had been forked right from init, right? 1171150379 M * daniel_hozac if we reassign the parent of the child to the guest's child_reaper, the host process will not get SIGCHLD for it. 1171150408 M * pflanze hm, I see. 1171150460 M * pflanze well if it's only my own program having difficulties with that then no problem :) 1171150569 M * daniel_hozac i suppose it might be possible to just fake the parent pid, but that'd require a new wrapper for the pid stuff. 1171150616 M * pflanze Yeah, and usually it's just init or the rc startup script being called, so I don't see a real problem with it. 1171150640 M * pflanze As long as there's no sneaky way in the kernel to get permissions to the ppid or something. 1171150674 M * daniel_hozac not that we know of. 1171150722 M * pflanze I think this fs_struct problem deserves being mentioned on the list. 1171150731 M * pflanze It could really be security relevant. 1171150789 M * daniel_hozac it was. 1171150799 M * daniel_hozac that's how it was brought to our attention. 1171150806 M * pflanze ah 1171150819 M * pflanze And people know they should upgrade? 1171150823 M * daniel_hozac http://archives.linux-vserver.org/200702/0033.html 1171150901 M * daniel_hozac if people are running rcs, upgrading to the latest one is implied... 1171150955 M * daniel_hozac it's not like Bertl_oO's putting out new rcs just because it's fun :) 1171150963 M * daniel_hozac we actually _do_ fix bugs. 1171151033 Q * m`m`h Ping timeout: 480 seconds 1171151098 M * pflanze So if a final version comes out, and there's a serious bug discovered, how do you publish a new version? 1171151111 M * daniel_hozac 2.2.0.1. 1171151115 M * daniel_hozac just like we did with 2.0.2. 1171151132 M * daniel_hozac (2.0.2 had a break-out-of-chroot bug) 1171151187 M * pflanze If a precise (automatically checkable) formula for finding out whether an upgrade is needed exists then that's ok. 1171151236 M * pflanze Currently I'm checking http://vserver.13thfloor.at/Experimental/?C=M&O=D for this regular expression: 1171151242 M * pflanze qr/patch-(2\.6\.[\d.]+-vs2\.2\.[^<\s-]*)\.diff/ 1171151253 M * pflanze and the first which hits is being reported to me if it changes. 1171151259 J * m`m`h ~simba@deb30.mgts.by 1171151266 M * daniel_hozac checking Experimental is a bad idea for releases. 1171151276 M * pflanze ok, which is the right idea? 1171151283 M * daniel_hozac as, well, those aren't Experimental. 1171151330 M * daniel_hozac i guess subscribing to the wiki might work. 1171151340 M * daniel_hozac if you don't need it the second Bertl_oO uploads it. 1171151355 M * pflanze Which wiki page? The front page? 1171151370 M * daniel_hozac http://linux-vserver.org/Template:CurrentPatchTable