1367366578 Q * FireEgl Remote host closed the connection 1367375567 J * Romster ~romster@202.168.100.149.dynamic.rev.eftel.com 1367376847 Q * imachine Read error: Connection reset by peer 1367380821 J * imachine ~imachine@robot.greenhost24.pl 1367381283 N * Bertl_zZ Bertl 1367381287 M * Bertl morning folks! 1367384433 Q * jrayhawk Remote host closed the connection 1367388692 J * jrayhawk ~jrayhawk@nursie.omgwallhack.org 1367390791 M * Bertl off for now ... bbl 1367390796 N * Bertl Bertl_oO 1367395230 J * nkukard ~nkukard@197.87.148.190 1367398559 J * bonbons ~bonbons@2001:a18:209:3a01:f8f5:c385:3381:9b02 1367398885 J * ser ~ser@host1.tldp.ibiblio.org 1367399382 J * Ghislain ~aqueos@adsl1.aqueos.com 1367406647 Q * ircuser-1 Ping timeout: 480 seconds 1367407304 M * alex3 hi, I need a method to identify the host on that a vserver is running out of a vserver, what could I do? 1367407448 M * alex3 oh thing that i could do is matching the mac address if the network interfaces agains a collected list of all mac addresses of all hosts, but I am looking for something more clever 1367410109 J * ircuser-1 ~ircuser-1@35.222-62-69.ftth.swbr.surewest.net 1367410358 J * nlm_ ~nlm@host178.186-124-177.telecom.net.ar 1367410462 Q * nlm_ Read error: Connection reset by peer 1367410574 J * nlm_ ~nlm@host178.186-124-177.telecom.net.ar 1367410712 Q * nlm Ping timeout: 480 seconds 1367412380 J * nlm ~nlm@host178.186-124-177.telecom.net.ar 1367412485 Q * nlm_ Ping timeout: 480 seconds 1367412904 M * ptitoliv_ hello Bertl_oO 1367412923 M * ptitoliv_ i finally had time to generate the archive for namespace issue : send the link in private 1367413655 J * melbar ~me@177.17.100.121 1367413772 M * melbar Hi. When I enable network namespaces (/etc/vserver//spaces/net) and create some interfaces in the namespaces, almost everything works fine except /proc/net comes out empty, which breaks netstat. 1367413787 M * melbar Anyone knows a fix for that? 1367413829 M * ptitoliv_ melbar: hey 1367413831 M * ptitoliv_ i have the same issue 1367413849 M * melbar Actually, most classic unix programs (ifconfig, route, etc) also fail because of the same reason: /proc/net is empty 1367413854 M * ptitoliv_ i reported it to chan here 1367413855 M * melbar the 'ip' utility works fine though 1367413860 M * ptitoliv_ we try to investiage 1367413869 M * ptitoliv_ melbar: you are exactly in the same case than me 1367413881 M * melbar any progress on this, ptitoliv_? 1367413898 M * melbar The only thing I miss is netstat 1367413912 M * melbar The 'ip' utility works fine and does almost everything I need 1367413928 M * ptitoliv_ melbar: not really progress 1367413932 M * ptitoliv_ but it"s my fault 1367413939 M * melbar how so? 1367413952 M * ptitoliv_ i didn't had time to send my vserver configuration for furthner investigation 1367413958 M * ptitoliv_ i just did it today 1367414029 M * melbar I'm trying to read the code to see if there's an obvious 'if' preventing it from working 1367414051 M * melbar but I guess I am far too inexperienced about the kernel code (: 1367414083 M * melbar I guess the underlying reason is buried deep in some obscure data structure 1367414132 M * melbar ptitoliv_: how can I help? 1367414213 M * ptitoliv_ melbar: i don't know i'm like you 1367414220 M * ptitoliv_ too inexperimented in the kernel code 1367414342 M * melbar well 1367414363 M * melbar If anyone knows this and would like to help me fix it, contact me! 1367414412 M * ptitoliv_ for know the only way is to use chbind if you want to use /proc/net 1367414417 M * ptitoliv_ melbar: i have an additional question 1367414446 M * ptitoliv_ did you find an elegant way to attach a network namespace usable with netns to a vserver ? 1367414472 M * melbar depends on what you consider 'elegant' (: 1367414510 M * melbar I wrote a script that locates a process in the vserver context to get to the namespace 1367414527 M * melbar then I create the interface and migrate it to the ns 1367414538 M * ptitoliv_ of so you do it after the vserver start 1367414546 M * ptitoliv_ me i would want to do the opposite 1367414559 M * melbar yep, at post-start.d 1367414570 M * ptitoliv_ me i wanted to do before 1367414584 M * melbar But I guess it is fine, because the namespace starts only with the loopback 1367414587 M * ptitoliv_ but in that case if you have services on your vserver that you xwant to bind only in a specific card 1367414594 M * melbar before would be nice 1367414598 M * ptitoliv_ you need to start the service after 1367414666 M * melbar yep. 1367414675 M * melbar So I guess it is far from elegant 1367414840 M * ptitoliv_ well it works 1367414852 M * ptitoliv_ but my dream would be a parameter wich would bne 1367414869 M * ptitoliv_ use /var/run/netns/ 1367414879 M * ptitoliv_ using setns 1367415004 M * melbar can't you set up the namespace beforehand? 1367415031 M * melbar spaces/net can contain the namespace ID if I am not mistaken 1367415069 M * ptitoliv_ yes but it doesn"t seems to work 1367415087 M * ptitoliv_ when i try i have a no_such_name process returned by the vc_enter_namespace function 1367415108 M * melbar yeah. the same I got here. 1367415128 M * melbar oh boy. seems like I have some dense code reading ahead of me. 1367415293 M * ptitoliv_ :) 1367416484 M * Bertl_oO alex3: what about putting a file in each guest with the host id? 1367416548 M * Bertl_oO melbar, ptitoliv_ for the /proc/net issue, this seems to be a problem with the 'net' entry being created within the network namespace 1367416660 M * melbar Bertl_oO: guessed as much. It's still unclear to me where to fix that though. 1367416736 M * Bertl_oO well, the best way to fix it is probably to set the hide/unhide attribute correctly for those entries, assuming (and the debug so far confirms that) it is actually the hide mechanism which removes those entries 1367416760 M * Bertl_oO if there is a different (mainline) mechanism at work, it will get complicated 1367416847 M * melbar you mean something like setattr --~hide /proc/net, but within that namespace? 1367416896 M * Bertl_oO well, /proc/net points to /proc/self/net and /proc/self points to /proc/, so it's not so simple, but yeah, the idea is the same 1367416974 M * melbar yeah, I followed the symlinks 1367417022 M * melbar and with the varying pid number, seems like we've got a moving target 1367417112 M * Bertl_oO it is part of the per process dynamic values 1367417139 M * Bertl_oO so, by default, it should be visible, which makes this a little strange 1367417158 M * melbar I felt the strangeness as well (: 1367417174 M * Bertl_oO but I have to do some debugging to figure out what's going on 1367417181 M * melbar my gut tells it should work yet it doesn't 1367417190 M * melbar can I help in any way? 1367417230 M * Bertl_oO it would be beneficial to figure out a 'setup sequence' to create a minimal namespace/context environment which exposes the issue 1367417270 M * Bertl_oO i.e. a sequence/combination of vcontex/vnamespace commands which should show /proc/net but do not because of the issue at hand 1367417306 M * Bertl_oO (similar to a guest startup, but without the unrelated stuff) 1367417330 M * melbar sure. isolate the issue, remove the crap. 1367417353 M * melbar to find out the exact spot where the issue shows up 1367417367 M * melbar i'm trying to do this as we speak. 1367417857 M * melbar here's what I found so far... tell me if I'm the right direction 1367417891 M * melbar I created a vserver named 'nsvs' with network namespaces and I have a nc -l -p 5000 running there so netstat -tanp should show it 1367417904 M * melbar however 1367417909 M * melbar vspace -e nsvs --net -- vcontext --migrate --xid nsvs -- netstat -tanp 1367417913 M * melbar triggers the problem 1367417921 M * melbar : 1367417922 M * melbar netstat: no support for `AF INET (tcp)' on this system. 1367417924 M * melbar yet 1367417940 M * melbar vspace -e nsvs --net -- netstat -tanp 1367417942 M * melbar works just fine 1367417984 M * melbar I'm guessing the context isolation might be conflicting with the network namespaces thing 1367418016 M * melbar by making /proc/ on that context not show because it is unaware of the netns 1367418025 M * Bertl_oO that's the interesting part, because the context isolation does not really care about network namespaces at all 1367418046 M * Bertl_oO so, the question is more: 1367418062 M * Bertl_oO why does it work with vcontext --migrate --xid nsvs -- ... 1367418085 M * Bertl_oO after all, the host network namespace is a network namespace as well 1367418098 M * melbar if I do 1367418101 M * melbar vcontext --migrate --xid nsvs -- netstat -tanp 1367418110 M * melbar by itself, without the vspace 1367418117 M * melbar I get the sockets on the 'root vserver' 1367418137 M * Bertl_oO yeah, so why isn't /proc/net hidden there? 1367418145 M * melbar which makes sense to me because I haven't switched to the network namespace 1367418153 M * melbar yeah... here's the mystery 1367418203 M * melbar my (wild) guess is that the data needed to make /proc/net work is in some _other_ data structure 1367418246 M * melbar notice that within the namespace, /proc/net exists... it is the files within it that do not 1367418280 M * Bertl_oO would make sense, but the attributes (including the hide flag) are in the dentry for proc entries :) 1367418345 M * Bertl_oO so it really is a kind of mystery, but we'll figure it out ... what I meant with sequence to recreate was something which doesn't actually need a running guest 1367418376 M * Bertl_oO i.e. something creating the spaces/contexts on the fly in the command sequence 1367418823 M * melbar you mean something like 1367418825 M * melbar vspace --new --net -- vcontext --create --xid 41234 -- cat /proc/net/tcp 1367418826 M * melbar ? 1367418833 M * melbar this one triggers the problem 1367418864 M * melbar it says 1367418865 M * melbar cat: /proc/net/tcp: No such file or directory 1367419017 M * Bertl_oO that's probably because /proc is missing 1367419051 M * melbar oot@melbar:/usr/lib/util-vserver# vspace --new --net -- vcontext --create --xid 41234 --namespace -- cat /proc/ 1367419052 M * melbar New security context is 41234 1367419052 M * melbar cat: /proc/: Is a directory 1367419065 M * Bertl_oO yeah, but I guess it's empty 1367419080 M * melbar vspace --new --net -- vcontext --create --xid 41234 --namespace -- ls /proc/ 1367419080 M * melbar New security context is 41234 1367419080 M * melbar 11546 execdomains ioports locks mounts stat tty 1367419080 M * melbar cpuinfo filesystems kcore meminfo net swaps uptime 1367419080 M * melbar crypto interrupts kmsg misc self sys version 1367419081 M * melbar devices iomem loadavg modules slabinfo sysvipc 1367419461 M * melbar the --namespace is redundant in this case, the same happens 1367419522 M * melbar so if I understand correctly, merely combining context isolation with namespaces seem to trigger the problem 1367420425 J * BenG ~bengreen@bmex-gw.bristolwireless.net 1367421984 M * alex3 Bertl_oO: a file in each vserver could work... maybe something with a bind mount 1367421989 M * alex3 Bertl_oO: thanks 1367423106 Q * melbar Remote host closed the connection 1367424347 Q * BenG Quit: I Leave 1367424725 N * ensc Guest4010 1367424735 J * ensc ~irc-ensc@p54ADEADA.dip0.t-ipconnect.de 1367425147 Q * Guest4010 Ping timeout: 480 seconds 1367425273 M * fback Bertl_oO: with 3.4.42-vs2.3.3.9 it's not possible to write to routing table from the guest anymore 1367426341 J * benl ~benl@dockoffice.sonassihosting.com 1367426452 J * nlm_ ~nlm@host178.186-124-177.telecom.net.ar 1367426767 Q * nlm Ping timeout: 480 seconds 1367426783 M * benl Hi all 1367426824 M * benl Is there any reason a Vserver would be slow to "enter" (vserver GUEST enter) 1367426848 M * benl Some guests are taking a few seconds to "enter" (despite the load being 0.00) 1367427268 M * benl Doh. I found out why. 1367427284 M * benl Something being slow to execute in either /etc/profile, .bashrc or .bash_profile 1367427610 Q * benl Quit: HydraIRC -> http://www.hydrairc.com <- Like it? Visit #hydrairc on EFNet 1367433885 J * melbar ~me@177.17.100.121 1367433968 M * melbar hi again. Bertl_oO, did you have a chance to look into the missing /proc/net/ entries issue? 1367437395 M * melbar Bertl_oO: disabling vx_hide_check on fs/proc/generic.c:proc_lookup_de makes the missing /proc/net/tcp entry show up even with network namespaces 1367437405 M * melbar but reading it gives 'permission denied' 1367438941 J * BenG ~bengreen@cpc35-aztw23-2-0-cust207.18-1.cable.virginmedia.com 1367440552 Q * BenG Remote host closed the connection 1367441864 Q * bonbons Quit: Leaving 1367446202 Q * Ghislain Quit: Leaving. 1367449312 Q * puck Ping timeout: 480 seconds 1367449468 J * puck ~puck@2404:130:0:1000::23:10 1367449619 M * melbar ptitoliv_, are you there?