1109548889 M * Bertl but I guess that is obsolete now, what do you say? 1109549523 M * Doener probably... not sure though... 1109549658 Q * maharaja Remote host closed the connection 1109549710 M * Bertl well, IMHO this check was to block create calls which return an already existing context 1109549754 M * Bertl but a test with new tools and newstyle config should test/validate that 1109550010 M * Doener yeah, i _love_ confixx... on that virtuozzo server, i created some users myself (oh, i'm sooo bad) and confixx decided to remove them... (from passwd and shadow only of course) 1109550032 M * Bertl hmm? 1109550069 M * Doener confixx decided to remove all users created using useradd and just keep those created from within confixx 1109550078 M * Doener (except system users) 1109550099 M * Bertl a sane decision, why would you use other tools ;) 1109550122 Q * zimdog Quit: Leaving 1109550137 M * Doener because confixx-created users are completely broken ;) no write permissions in their own home directory 1109550182 M * Bertl that's the way to go! keeps disk usage low! 1109550208 M * Bertl btw: http://vserver.13thfloor.at/Stuff/SCRIPT/ctxtest.sh 1109550241 J * jd86 ~jim@ip68-9-97-23.ri.ri.cox.net 1109550252 M * jd86 i. hate. when. my. net. goes. down. 1109550262 M * Bertl who doesn't ;) 1109550294 M * Doener Bertl: hm, you got to select 'the right' chcontext 1109550310 M * Bertl yeah, it's just for qemu tests 1109550314 M * Doener alpha tools have a script called vcontext _and_ the legacy c binary 1109550318 M * jd86 /usr/sbin/vserver: line 647: pushd: /etc/vservers/gentoo-template/vdir: No such file or directory 1109550324 M * jd86 ooh 1109550325 M * jd86 i know why 1109550326 M * jd86 har har 1109551291 M * jd86 uh 1109551293 M * jd86 no i dont know wh 1109551345 M * Bertl hmm? 1109551349 M * jd86 i think i know why 1109551352 M * jd86 (sorry thinking out loud) 1109551397 M * Bertl ls -la /etc/vservers/gentoo-template/vdir 1109551404 M * Bertl jd86: very loud ;) 1109551415 M * jd86 i fixed 1109551421 M * jd86 ok i have a vserver working 1109551427 M * Bertl congrats! 1109551438 M * jd86 so 1109551441 M * Bertl Doener: http://vserver.13thfloor.at/Stuff/SCRIPT/ctxtest.sh (reload nicer version) 1109551457 M * jd86 (yea right eh?) can you help me with the quota stuff (xid ) we left off at the palce where i have to have a working vserver! 1109551467 M * Bertl sure ... 1109551478 M * Bertl vserver is starting and can be stopped quite fine? 1109551491 M * jd86 yea just did it twice 1109551513 M * Bertl okay, fs is mounted tagxid? 1109551531 M * jd86 /dev/hda4 /vservers ext3 rw,noatime,tagxid 0 0 1109551551 M * Bertl good, vdlimit tool is compiled an available? 1109551564 M * jd86 vserver-host vdirbase # vd 1109551564 M * jd86 vdir vdu 1109551569 M * jd86 no, where do i get it? 1109551595 M * Bertl http://vserver.13thfloor.at/Experimental/TOOLS/vdlimit-0.02.tar.bz2 1109551606 M * Bertl (or from the download page on 13thfloor) 1109551644 M * jd86 alright 1109551645 M * jd86 i have it 1109551647 M * jd86 i compiled it 1109551658 M * jd86 should i copy it to /bin or something? 1109551668 M * Bertl wherever you prefere it to have 1109551675 M * jd86 alright 1109551711 M * jd86 what next :-D 1109551737 M * Bertl breath in and out ;) 1109551742 M * jd86 ah yes lots of that 1109551779 M * Bertl okay, the basic concept is that the 'limit' is per context _and_ per device 1109551800 M * Bertl so you have to specify those two elements when you configure it ... 1109551809 M * jd86 and when you say context you mean "vserver" and device as like partition? 1109551828 M * Bertl for example 'adding' (the first step) a limit is done with -a 1109551837 M * Bertl (yes) 1109551845 M * jd86 ok 1109551860 M * Bertl vdlimit -a -x 1109551877 M * Bertl will add a limit for to the vserver with 1109551897 M * jd86 so vdlimit -a -x 2 /dev/hda4 right or no? 1109551905 M * Bertl yes for example 1109551929 M * jd86 ok 1109551931 M * jd86 am i doing this? 1109551932 M * jd86 or no 1109551938 M * Bertl yes do it 1109551956 M * jd86 /dev/hda4: 0,0,0,0,100 1109551958 M * jd86 it spat that out 1109551977 M * Bertl so it added a limit now (the values are default) 1109551982 M * jd86 ok 1109551988 M * Bertl you can always display the limit by doing 1109552010 M * Bertl vdlimit -x 2 /dev/hda4 1109552018 M * jd86 a yes 1109552022 M * jd86 s/a/ah 1109552029 M * Bertl and you can set them with 1109552048 M * Bertl vdlimit -x 2 -S 100,1000,20,40,100 /dev/hda4 1109552054 M * jd86 what are all thsoe number? 1109552073 M * Bertl they are current and maximum disk usage/inodes and the root reserve 1109552091 M * jd86 ah. so what units are they 1109552101 M * Bertl issue the command above and look at the output of: 1109552136 M * Bertl vcontext --migrate --xid 2 df -h 1109552142 M * Bertl vcontext --migrate --xid 2 -- df -h 1109552145 M * Bertl and 1109552149 M * Bertl vcontext --migrate --xid 2 -- df -i 1109552165 M * jd86 ut oh 1109552174 M * jd86 was i suppose to have had the vserver on when i did those vdlimit commands? 1109552183 M * Bertl yes, that's fine 1109552191 M * jd86 well i didnt have it on 1109552198 M * Bertl well, no problem either ... 1109552200 M * jd86 do i have to have it on to do those above vcontext commands? 1109552209 M * Bertl change the migrate to --create then to test 1109552251 M * jd86 vserver-host gentoo-template # vcontext --migrate --xid 2 df -h 1109552251 M * jd86 Filesystem Size Used Avail Use% Mounted on 1109552251 M * jd86 /dev/hda3 5.8G 817M 5.0G 14% / 1109552251 M * jd86 /dev/hda4 21G 1.5G 19G 8% /vservers 1109552251 M * jd86 none 252M 0 252M 0% /dev/shm 1109552253 M * jd86 df: `/proc/bus/usb': No such file or directory 1109552258 M * jd86 is that the way its supposed to be? 1109552275 M * Bertl ah, forgot a flag ... 1109552301 M * Bertl it's easier to do that from inside the context, so I'd suggest starting your vserver and entering it 1109552307 M * jd86 alright 1109552313 M * Bertl then just do 'df -h' and 'df -i' 1109552323 M * jd86 gentoo-template / # df -h 1109552323 M * jd86 Filesystem Size Used Avail Use% Mounted on 1109552323 M * jd86 /dev/hdv1 21G 1.5G 19G 8% / 1109552323 M * jd86 none 16M 16K 16M 1% /tmp 1109552333 M * Bertl (on another console reissue the vdlimit if required 1109552389 M * jd86 uhh 1109552391 M * jd86 i reissued it 1109552393 M * jd86 no go 1109552396 M * jd86 i mean it still says that 1109552441 Q * rs Quit: Lost terminal 1109552531 M * jd86 uhh ami supposed to see anything with the df -h? 1109552541 M * Bertl sec checking ... 1109552544 M * jd86 k 1109552662 A * jd86 is getting some water be rigt back 1109552710 M * jd86 backj 1109552924 M * Bertl interesting ... 1109552935 A * Bertl is now investigating ... 1109552939 M * jd86 alright 1109553060 M * Bertl ah, stupid me .. keep falling for that everytime .. 1109553086 M * jd86 oh?'' 1109553091 M * Bertl thing is I changed the arguments some (a long) time a go 1109553109 M * Bertl so you do not specify the 'evie' but instead a file or the mountpoint 1109553116 M * jd86 evie? 1109553121 M * Bertl s/evie/device/ 1109553128 M * jd86 ls 1109553129 M * jd86 sl 1109553130 M * jd86 ugh 1109553134 M * jd86 why did i type that 1109553135 M * jd86 so like 1109553136 M * Bertl so use 1109553153 M * jd86 vdlimit -x 2 -S 100,1000,20,40,100 /vservers/ 1109553155 M * Bertl vdlimit -x 2 -S 100,1000,20,40,100 /vservers 1109553159 M * Bertl precisely! 1109553181 M * Bertl but first the add (-a) 1109553187 M * jd86 ok 1109553210 M * jd86 sweet 1109553216 M * jd86 it has no space left on that one 1109553248 M * jd86 woah can you explain the inode thing to me? 1109553273 M * Bertl well, the files itself are stored in so called inodes 1109553286 M * Bertl so a directory or a file will use up some space, and one inode 1109553305 M * jd86 so no matter how big a file is it only uses one inode? 1109553313 M * Bertl df -i shows the available and used inodes 1109553333 M * jd86 gentoo-template / # df -i 1109553333 M * jd86 Filesystem Inodes IUsed IFree IUse% Mounted on 1109553333 M * jd86 /dev/hdv1 40 20 20 50% / 1109553333 M * jd86 none 64277 1 64276 1% /tmp 1109553340 M * jd86 so taht means...? 1109553360 M * Bertl you now have 'virtually' 40 inodes and 20 are already used 1109553372 M * jd86 ok but i still dont get what an inode is 1109553378 M * jd86 i have 40 'what' used 1109553397 M * Bertl linux filesystems are inode based ... 1109553403 M * Bertl (well, most are) 1109553422 M * jd86 yea but what exactly is an inode? 1109553424 M * Bertl this can be seen as some container to put stuff in ... 1109553426 M * jd86 does every file take one inode? 1109553475 M * Bertl on most filesystems, yes 1109553492 M * jd86 but i have _alot_ more then 20 files in that / tho 1109553496 M * daniel_hozac how about a directory? 1109553498 M * Bertl hardlinks do not take any inodes 1109553544 M * Bertl symlinks require an inode, and directories too 1109553562 M * Bertl but play around a little with your favorite fs and check the inode counts with df -i 1109553566 M * jd86 ^^ i dont get it tho i have tons of files how are only 20 inodes used 1109553575 M * daniel_hozac jd86: you set it. 1109553579 M * Bertl the dlimits are relative 1109553585 M * jd86 oh my 1109553589 M * jd86 i'm confuzeizzled 1109553624 M * Bertl the kernel can not know about the real values (or well, it would take a long time to evaluate them everytime you want to know) 1109553643 M * Bertl you ever done 'su -skx' or so? 1109553649 M * jd86 no 1109553649 M * Bertl *du 1109553653 M * jd86 still no 1109553653 M * Bertl du -skx 1109553662 M * Bertl well try it with some partition 1109553667 M * Bertl du -skx /vservers 1109553681 M * jd86 i just did it in a dir on my machine and its like slamming my hard drive. 1109553693 M * Bertl precisely 1109553694 M * jd86 jimboy root # du -skx 1109553694 M * jd86 560524 . 1109553706 M * jd86 so why do we bother setting inode limits? 1109553709 M * Bertl so nobody wants that to happen everytime you do df inside a vserver 1109553726 M * daniel_hozac but if you tell the kernel when the vserver starts, and then the kernel just does it normal accounting... 1109553744 M * Bertl jd86: it's not just inode limits it's also disk space limits 1109553759 M * jd86 what do you mean? 1109553775 M * Bertl the kernel will keep track of the changes to the 'current' values 1109553796 M * Bertl e.g. if you create a file inside the vserver, the values for inodes and blocks are updated 1109553819 M * jd86 so i do a du -skx /vservers/gentoo-template/ and then set it to that for used, and then think up a grand ol random number for max? 1109553830 M * Bertl exactly 1109553853 A * jd86 hopes xchat logs chats by default cuz i'll never remember this 1109553863 M * Bertl the channel is logged 1109553865 M * IrishKitty jd86: it doesn't 1109553870 M * IrishKitty try /set autolog on 1109553877 M * jd86 No such variable. 1109553885 M * IrishKitty hmm 1109553894 M * jd86 ok Bertl moving on, how do i set space limitations then? 1109553903 M * jd86 and what are each,of,those,comma,seperated,values? 1109553923 M * daniel_hozac space_used,space_total,inodes_used,inodes_total,reserved 1109553932 M * Bertl first two are disk space (1k blocks), the second two are inodes 1109553946 M * Bertl tx 1109553978 M * jd86 so do the first in regular ol kilobytes then do the node dance 1109553980 M * jd86 and what do i do with reserved? 1109554057 M * Bertl ever wondered why your df space doesn't sum up? 1109554073 M * jd86 nope not really 1109554081 M * jd86 you mean it doesn't? 1109554114 M * Bertl well, usually some space is reserved for root 1109554136 M * Bertl typically that is between 3 and 10% 1109554159 M * jd86 hrm 1109554217 M * jd86 what should i do for a inode limit? 1109554220 M * jd86 does it make a difference 1109554238 M * Bertl well, you can leave it unlimited 1109554247 M * jd86 what do i do to make it unlimited? 1109554247 M * jd86 0? 1109554252 M * daniel_hozac -1 1109554254 M * jd86 well does it make a difference? 1109554265 M * Bertl if somebody does: 1109554304 M * Bertl for n in `seq 1 100000`; do touch /tmp/$n; done 1109554322 M * Bertl you'll get 100000 used inodes, but no used up space 1109554357 M * Bertl if your filesystem/partition does just have 90000 inodes, it's some kind of denial of service/resource attack 1109554380 M * jd86 whats 1 gig in K? is it 1048576 1109554409 M * Bertl in my boke 1G in k is 1M ;) 1109554416 M * Bertl *book! 1109554430 M * jd86 what? 1109554449 M * Bertl k is *1024 1109554451 M * jd86 you trying to confuze me? if so your doign a good job 1109554454 M * daniel_hozac lol 1109554457 M * Bertl M is *1024*1024 1109554465 M * jd86 so its 1048576, no? 1109554468 M * Bertl G is *1024*1024*1024 1109554473 M * jd86 another one!?! 1109554476 M * jd86 numbers too long 1109554500 M * jd86 1.073741e+09 1109554514 M * Bertl but yes 1048576 is fine 1109554524 M * jd86 well i'm trying to have this thing set it to 1 gig 1109554590 M * jd86 actually the thing is 1 gig (how? ugh) 1109554607 M * jd86 1.4 gigs for a template hmm i'm going to have to trim that i think HMM 1109554626 M * daniel_hozac how does unification and disk limits work together? do they? 1109554636 M * Bertl yes, they do ... 1109554649 M * Bertl unified files have to belong to xid 0 anyway 1109554656 M * jd86 ok 1109554660 M * jd86 yea its not changing the thing any more 1109554673 M * jd86 do i ahve to restart the thing to have it update the size? 1109554678 M * daniel_hozac ah. so vservers can access files belonging to xid 0? 1109554685 M * Bertl yes 1109554696 M * Bertl jd86: no 1109554703 M * daniel_hozac ah, ok. i thought they could only access their own files. thanks. 1109554706 M * jd86 Bertl: it isnt working any more then.. 1109554732 M * Bertl check your command, what does it output? 1109554740 M * jd86 no i figured out why ti wasnt being nice 1109554740 M * jd86 but 1109554752 M * jd86 vserver-host vservers # vdlimit -x 2 -S 905720,1073741000,905720,100000,100 /vservers 1109554752 M * jd86 vc_set_dlimit: Invalid argument 1109554760 M * jd86 what happens when i want to give one like 20G 1109554771 M * Bertl well, the second one is a little exteme 1TB? 1109554773 M * daniel_hozac inodes_used > inodes_total 1109554777 M * jd86 I thought that was 1 gig 1109554790 M * Bertl that is a gig in bytes 1109554799 M * Bertl but you specify the amount in 1k blocks 1109554814 M * Bertl so you ahve to divide by 1024 ;) 1109554824 M * jd86 vserver-host vservers # vdlimit -x 2 -S 905720,1048576,905720,100000,100 /vservers 1109554824 M * jd86 vc_set_dlimit: Invalid argument 1109554832 M * jd86 whats wrong with that one? 1109554839 M * Bertl 905720,100000 1109554844 M * jd86 whats wrong with that? 1109554844 M * Bertl 905720 >= 100000 1109554849 M * jd86 shatizzle 1109554859 M * jd86 ahh 1109554860 M * jd86 all bettar 1109554869 M * jd86 teh w00t 1109554870 M * jd86 it works 1109554882 A * Bertl .o( that is the end of communication as we know it ;) 1109554894 M * jd86 ? 1109554895 M * daniel_hozac so your vserver uses as many inodes as it uses bytes? 1109554903 M * jd86 haha 1109554905 M * jd86 good question 1109554906 M * daniel_hozac s/bytes/kibibytes/ 1109554916 M * Bertl a very special setup, hard to achieve ;) 1109554925 M * jd86 i'm a master at those things 1109554960 M * jd86 905720 is a "du", and a du -skx is a 905720, i thought -skx would tell inode? 1109554985 M * Bertl du -skx shows the sum of bytes ... 1109554992 M * jd86 oh 1109554995 A * jd86 does a little stupid dance 1109554999 M * jd86 how do i tell the inode total? 1109555016 M * Bertl you can 'count' the inodes by doing something like 'find /vservers/gento... | wc -l' 1109555027 M * Bertl well maybe better 1109555029 M * jd86 i dont have that many fingers ;-) 1109555039 M * Bertl 'find /vservers/gento... --xdev | wc -l' 1109555061 M * jd86 says --xdev is invalid arg 1109555073 M * daniel_hozac du --inodes --xid 2 --one-file-system ;) 1109555074 M * Bertl remove one dash 1109555093 M * Bertl ah, that works? 1109555100 M * daniel_hozac well, with patches ;) 1109555115 M * Bertl aah ;) 1109555126 M * jd86 fixed 1109555136 M * jd86 163760 1109555192 M * jd86 voila 1109555194 M * jd86 how leet is that 1109555203 M * Bertl extremely! 1109555214 M * jd86 one more thing. 1109555221 M * jd86 when i do an ifconfig in the guest 1109555228 M * jd86 it shows the two things 1109555230 M * jd86 not just its own 1109555233 M * jd86 any way to change that? 1109555240 M * Bertl hide_netif 1109555251 M * jd86 where i put that? 1109555256 M * jd86 flags? 1109555258 M * Bertl yep 1109555269 M * daniel_hozac isn't that enabled by default? 1109555277 M * Bertl I thought so too .. 1109555309 M * Bertl but maybe he has a new ifconfig then it could be a netlink issue (which would be fixed in recent kernels) 1109555322 M * jd86 still does it 1109555344 M * jd86 vserver-host gentoo-template # ifconfig --version 1109555345 M * jd86 net-tools 1.60 1109555345 M * jd86 ifconfig 1.42 (2001-04-13) 1109555347 M * Bertl k, let's see the output (replace sensitive data with xxx) 1109555356 M * jd86 uh 1109555362 M * jd86 who really cares if u know my internal ips lol 1109555375 M * Bertl the world! the world! 1109555393 M * jd86 http://jim.psunit.com:989/~jim/ifout 1109555415 M * Bertl ah, okay, that's fine 1109555427 M * Bertl it's an ifconfig oddity 1109555443 M * Bertl if you look closely, you'll see that the first one has no ip 1109555482 M * jd86 o 1109555493 M * jd86 thats alright 1109555514 M * Bertl ifconfig just concludes that if there is eth0:gent, there has to be eth0 too 1109555524 M * jd86 ah 1109555528 M * jd86 smartass ifconfig 1109555540 M * Bertl yeah, ip addr ls is better there 1109555547 M * jd86 heh 1109555648 M * jd86 so i think that i should be able to write that tutorial after i setup these two servers :-D 1109555656 M * Bertl excellent! 1109555697 M * Bertl maybe have a look at existing ones (on other wikis and sites) too, they probably contain useful hints and such 1109555708 M * Bertl (but might be outdated of course) 1109555710 M * jd86 i think i've gained some knowledge from this whole vserver experience 1109555725 M * jd86 about inodes. and pages and random stuff 1109555734 M * jd86 and the rlimits and this bucket setup . 1109555750 M * Bertl linux-vserver educates! 1109555756 M * jd86 yea definatly 1109555768 M * jd86 almost as much as building my first computer :-D 1109555773 M * Bertl well, we have smart folks hanging around here ... 1109555783 M * jd86 i remember doing that, what a learning curve. 1109555789 M * jd86 well i'm not so smart :( 1109555827 M * daniel_hozac lol, putting together a computer was like those toys where you're supposed to put the cube in the square hole, the sphere in the round hole, etc ;) 1109555838 M * jd86 being 16 and having no formal computer 'classes' or anything.. i think all i need now is a big fat book that goes hardcore into this stuff so i can learn about all of it. 1109555843 M * jd86 daniel_hozac: yea well this was a while ago 1109555878 M * jd86 daniel_hozac: i basically went from knowing nothing to knowing sooo much in like a month, month and a half 1109555921 M * mugwump jd86: start with _The Pragmatic Programmer_, then - if you're thinking of going down the programming route 1109555953 M * Bertl Doener: okay, final? patch at http://vserver.13thfloor.at/Experimental/FOR-1.9.5/delta-create_info-feat04.diff 1109555955 M * daniel_hozac mugwump: is that a good book? 1109555968 M * jd86 mugwump: honestly i'm not very good at programming. i'm not a creative person and i think that effects my programming ability. 1109555991 M * mugwump hogwash, jd86, no-one is inherantly non-creative 1109556013 M * daniel_hozac everyone's creative, you just need to find your art. 1109556068 A * Bertl .o( hopefully not SunTzu ) 1109556125 M * Doener Bertl: does IS_ERR(NULL) evaluate to true? 1109556156 M * Doener hm, actually it doesn't matter... 1109556156 M * Bertl probably not, can we get that? 1109556189 M * Doener the "no dyn. xid avail." path 1109556204 M * mugwump daniel_hozac: there are few books that manage to express the algorithms involved in *being a programmer* as well as TPP... It's one of those books that I've scanned, seen lots of sage advice in, and thought, "I wish I had have read this when I first started programming" 1109556238 M * jd86 hrm 1109556261 M * Doener hm, missed the check in vc_ctx_create, so we should probably find some error value for the dyn. xid path 1109556271 M * Bertl Doener: hmm, right we should return some meaninful output 1109556313 M * jd86 i'm going to sleep i'm beat. 1109556321 M * Doener night jd86! 1109556324 M * Bertl night then! 1109556327 M * jd86 Bertl: thanks for your help you are excelent at this helping me thing 1109556344 M * Bertl you're welcome! 1109556485 M * Bertl Doener: EDEADLK? 1109556489 M * Doener Bertl: hm, don't we end up with -EINVAL for dynamic context again? 1109556503 M * Doener ah no, my fault 1109556519 M * Doener i'm bad at reading code today... 1109556542 M * Bertl np, or should we use EAGAIN? 1109556596 M * Doener yep, EAGAIN 1109556645 M * Bertl okay! 1109556778 M * Bertl http://vserver.13thfloor.at/Experimental/FOR-1.9.5/delta-create_info-feat05.diff 1109556792 M * Bertl enough for me for today ... I'm off to bed now! 1109556807 M * Bertl thanks for the support Doener! have a good night! 1109556822 M * Doener night Bertl! 1109556823 N * Bertl Bertl_zZ 1109557576 N * Bertl_zZ Bertl 1109557582 M * Bertl hmm, forgot something ... 1109557719 M * Bertl http://vserver.13thfloor.at/Experimental/patch-2.6.11-rc5-vs1.9.4.9.diff 1109557727 M * Doener hehe 1109557732 T * Bertl http://linux-vserver.org/ | latest stable 1.2.10, devel 1.9.4.9, ng9.0 -- He who asks a question is a fool for a minute; he who doesn't ask is a fool for a lifetime -- share the gained knowledge on the wiki, and we'll forget about the minute ;) 1109557745 M * Bertl so this might get some testing ... ;) 1109557751 M * Bertl okay, gone now for real! 1109557756 N * Bertl Bertl_zZ 1109561416 Q * eyck Ping timeout: 480 seconds 1109561709 N * Doener Doener|x_x 1109561933 J * eyck eyck@81.219.64.71 1109565993 J * nox- ~nox@213.39.207.212 1109566323 Q * nox Ping timeout: 480 seconds 1109566355 N * nox- nox 1109567936 J * eyck_ eyck@81.219.64.71 1109567936 Q * eyck Read error: Connection reset by peer 1109570627 Q * monrad Quit: Leaving 1109571281 M * lilo Bertl: thanks :) 1109576417 J * BWare ~bware@212.26.196.154 1109576467 M * BWare morning 1109578151 J * pulsar ~pulsar@82.209.246.234 1109578165 M * bro mornin` all 1109578467 M * pulsar Is rss in rlimits/rss.{hard,soft} in kb or in pages like in vserver-stat ? 1109579800 J * rs ~rs@193.155.140.132 1109579808 M * rs hi there 1109579912 M * BWare morning 1109580604 J * prae ~prae@ezoffice.mandrakesoft.com 1109580647 M * prae moorrrning :) 1109580657 M * prae (morning is too hard) 1109581115 J * jsambrook ~jsambrook@aelfric.plus.com 1109582419 Q * Doener|x_x Ping timeout: 480 seconds 1109583441 J * LinuXman ~LinuxMan@host54-54.pool80182.interbusiness.it 1109583443 M * LinuXman hi 1109583454 M * LinuXman i have a virtuozzo account on vserver.de 1109583464 M * LinuXman i want to know if i can load kernel modules for enable ipv6 support 1109583533 M * pulsar Is virtuozzo vserver based software ? 1109583556 M * LinuXman no 1109583564 M * LinuXman virtuozzo is a software of sw soft 1109583577 M * LinuXman server:/proc# cat version 1109583578 M * LinuXman Linux version 2.4.20-021stab028.3.777-enterprise (st@current.eng.sw.ru) (gcc version 2.96 20000731 (Red Hat Linux 7.3 2.96-110)) #1 SMP Wed Feb 2 21:08:59 MSK 2005 1109583578 M * LinuXman server:/proc# 1109583581 M * LinuXman this is the kernel 1109583708 M * LinuXman . 1109583921 J * erwan_taf ~erwan@81.80.43.68 1109584386 Q * LinuXman Quit: 1109584488 M * prae erwan_taf: hi :) 1109584512 M * erwan_taf lo prae 1109584578 J * nck ~nck@220.224.18.131 1109584591 M * nck how does process inside kernel knows that its running inside chroot environment 1109584733 M * erwan_taf first at all, hello nck 1109584758 M * nck hi 1109584873 M * nck ?? 1109585754 P * nck Leaving 1109589041 J * Duckx ~duckx@195.75.27.158 1109589279 Q * erwan_taf Remote host closed the connection 1109589885 J * erwan_taf ~erwan@81.80.43.68 1109589895 J * erwan__taf ~erwan@81.80.43.68 1109589912 Q * erwan__taf Quit: 1109590540 N * Bertl_zZ Bertl 1109590553 M * Bertl morning folks! 1109590570 M * erwan_taf lo Bertl 1109590587 M * Bertl taf? 1109590659 M * erwan_taf taf means 'at work' 1109590668 M * erwan_taf this is some french slang 1109590683 M * Bertl ah, and ho was something like home? 1109590711 M * erwan_taf ho means home :) 1109590720 M * SiD3WiNDR _maison? :P 1109590733 M * Bertl k, how are you? and morning SiD3WiNDR! 1109590820 M * SiD3WiNDR hi :) 1109590869 M * SiD3WiNDR inside a vserver, can you access serial ports? 1109590946 M * Bertl if you provide the device for the serial port, yes of course 1109590970 M * SiD3WiNDR provide the device? 1109590975 M * SiD3WiNDR as in, mknod bla bla? 1109590987 M * Bertl yep 1109591003 M * SiD3WiNDR but then, another root in another vserver can do the same, and mess with my serial stuff? 1109591032 M * Bertl if he got a serial device node too then yes 1109591044 M * SiD3WiNDR well he can just make one himself, can't he? 1109591052 M * Bertl no he can't! 1109591082 M * Bertl this needs a capability he usually doesn't have 1109591089 M * SiD3WiNDR aha! 1109591090 M * SiD3WiNDR :) 1109591122 M * SiD3WiNDR so I have to make the device node from the host, in the vserver's filesystem? 1109591131 M * Bertl precisely! 1109591135 M * SiD3WiNDR nice :) 1109591150 A * SiD3WiNDR will play with that later today then 1109591155 M * SiD3WiNDR argh 1109591159 M * SiD3WiNDR this will suck 1109591171 M * Bertl ? 1109591174 M * SiD3WiNDR in belgium, if you go over a certain data transfer limit, you get put on "smallband" 1109591180 M * SiD3WiNDR which in my case is 4 kB/sec up & down 1109591198 M * Bertl ah, and you plan to do that ;) 1109591209 M * SiD3WiNDR the connection where the server is where I want to make that new vserver to play with is on such a 4 kB/sec connection atm 1109591214 M * SiD3WiNDR doing a net install will take ages! 1109591248 A * SiD3WiNDR looks at the date 1109591259 M * SiD3WiNDR ooh, last day of the month. I'll just wait 12 hours then I'm on 6Mbit again ;) 1109591263 M * SiD3WiNDR btw, vserver rocks :) 1109591266 M * Bertl i.c., well CDs and DVDs have been invented to ease that, well not exclusively ;) 1109591273 M * SiD3WiNDR hehe 1109591281 M * Bertl thanks! (for the rocks) 1109591298 M * SiD3WiNDR here at work I have a local debian mirror (but no smallband anyway so..) 1109591305 M * SiD3WiNDR did you decide anything yet about the job? 1109591323 M * Bertl which job? 1109591362 M * SiD3WiNDR didnt you get a job offer from some company? :) 1109591376 M * Bertl ah, that, well doesn't really need a decision ;) 1109591398 M * SiD3WiNDR hehe 1109592310 J * maharaja maharaja@chello080109078225.4.15.vie.surfer.at 1109592312 M * maharaja hi there 1109592328 M * Bertl morning maharaja! 1109592340 M * maharaja i'm in chaos right now :) 1109592349 M * SiD3WiNDR isn't it afternoon ;) 1109592355 M * maharaja i though ill try 2.6.11-rc5-1.9.4 1109592365 M * Bertl try 1.9.4.9 ;) 1109592384 M * SiD3WiNDR when can we expect a linux-vserver talk at fosdem, Bertl ;) 1109592387 M * maharaja any link? 1109592428 J * Doener doener@193.24.208.125 1109592479 M * Bertl SiD3WiNDR: well, probably when I get there, right? 1109592495 M * Doener morning! 1109592500 M * Bertl maharaja: http://vserver.13thfloor.at/Experimental/patch-2.6.11-rc5-vs1.9.4.9.diff 1109592503 M * Bertl morning Doener! 1109592504 M * maharaja bertl: with try, you mean its better to use that one or not :) 1109592522 M * maharaja bertl: this is a whole diff, right? 1109592530 M * Bertl it has some new features but also some to-be-tested fixes 1109592540 M * Bertl it's a patch agains 2.6.11-rc5 1109592541 M * SiD3WiNDR Bertl: hehe 1109592545 M * maharaja mhm - well, basically i need some stable stuff right now *g* 1109592560 M * SiD3WiNDR Bertl: just find a nice sponsor ;) 1109592562 M * maharaja as i wanted to migrate some produktion maschines to 2.6 1109592568 M * maharaja with vserver enabled 1109592569 M * Bertl it has good chances to be released as 1.9.5 ... 1109592602 M * maharaja would you use it for stable maschines? 1109592611 M * maharaja for productive use? 1109592614 M * Bertl probably ... 1109592617 M * maharaja or would you stick with 2.4.x 1109592679 M * Bertl there is only one answer to that, use 2.4/1.2 when you do not need the features of 2.6, it's well tested and it's definitely stable ... 1109592710 M * Bertl if you need or want features of 2.6/1.9 it's usually best to go for rc versions 1109592717 M * maharaja the only major thing i need is two differnet subnets on one vserver box 1109592721 M * bro hiya every1 1109592725 M * maharaja is that possible with 2.4? 1109592734 M * Bertl hiya bro! 1109592743 M * Doener hi bro 1109592766 M * Bertl maharaja: well, there is no significant difference in networking between 1.2 and 1.9 so probably yes 1109592780 M * maharaja ok, thnx 1109592878 M * bro ive encountered some strange stuff with my vserver box on which i use hard cpu scheduling, for some reason(unknown to me) apache hangs and eat's up all tokens, and after that it breaks free from vservers context and eats up cpu on the host machine, so i was wondering if it's my fault somehow or it's a bug 1109592913 M * Bertl hmm .. sounds interesting ... some details like machine, kernel, config? 1109592945 M * bro im using 2.6.9-vs1.9.3, with util-vserver: 0.30.196 1109592974 M * bro machine smp 2x1ghz p3 with 1gb mem 1109593045 M * Bertl hmm, PREEMP enabled? 1109593049 M * Bertl *PREEMPT even 1109593062 M * bro disabled 1109593085 M * Bertl hard cpu and idle task limiting? or just the former 1109593120 M * bro and idle task limiting 1109593154 M * bro CONFIG_VSERVER_HARDCPU_IDLE=y if this is the one u asked 1109593158 M * Bertl any lockups on that machine? 1109593191 M * Bertl or 'just' an escaping apache? 1109593203 M * bro well, not realy, but recently apache began hanging ;/ 1109593247 M * bro just escaping apache, nothing else 1109593260 M * Bertl okay, so what exactly do you observe? 1109593303 M * bro if i somehow manage to log in and enter the *vserver-gone-bad* to kill restart apache, everything continues to work normaly 1109593324 M * Bertl okay, so the apache stays within the vserver, right? 1109593351 M * bro well, yes, but somehow it eat's up host's resources 1109593362 M * bro but it realy shouldnt 1109593370 M * Bertl which resources, cpu? or memory? 1109593377 M * bro cpu && memory 1109593394 M * bro but, memory isn't limited from my side, just cpu 1109593474 M * Bertl could it be that the process starts dancing between cpus? 1109593508 M * bro hmm, how could i check that? 1109593517 M * bro i mean that *dancing* 1109593568 M * Bertl well, not that easy, right ... maybe you see high load on cpu1 but none on cpu0, and a second later the opposite 1109593614 M * maharaja when i have an vserver enabled kernel with no vserver running, i can chroot into one vserver and work as usual, right? 1109593620 M * maharaja except for remouting proc 1109593634 M * Bertl maharaja: yes, that should work 1109593650 M * Bertl (of course you 'might' lose some of the security) 1109593807 M * bro well, once i got lucky and did see vTOP output from host server, but there was only one apache task eating up all the memory and cpu, but, the refresh time wasnt so fast as u'd expect, it took bout 15mins to refresh it, and the picture still was the same 99%apache and all memory gone ;/ 1109593842 M * Bertl ah, okay, now we are getting somewhere ... 1109593853 M * Bertl so your memory basically depleted 1109593871 M * Bertl how much swap space has that machine? 1109593884 M * bro 2.5gb 1109593895 M * Bertl and how much of that is used right now for example? 1109593909 M * bro 120mb 1109593932 M * Bertl do you by any chance remember how much was used when apache 'hang'? 1109593959 M * bro hmm 1109593982 M * bro swap wasnt fully depleted iirc 1109594003 M * Bertl but quite heavily used, I guess 1109594033 M * bro it was 1`n`smtn gig 1109594043 M * Bertl okay, that pretty much explains it 1109594056 M * bro but, at the time it hapened server had almost no load at all 1109594059 M * Bertl let me elaborate on what 'most likely' happened 1109594080 M * bro hmm, it keeps hapening bout 2 weeks in a row, and max 1 once a week 1109594107 M * Bertl after that, I tell you how to verify that this actually happens, when it happens ... 1109594117 M * bro ok 1109594131 M * bro im all ears :> 1109594145 M * Bertl you entire machine has something which is called 'working set' 1109594174 M * Bertl this is the memory (and cpu) used when everything is running normally ... 1109594190 M * Bertl it is also the amount of memory _actively_ used ... 1109594222 M * Bertl and example here: apache might allocate 100MB memory but just use 50MB of it for it's work 1109594260 M * Bertl now the 'real' memory (RAM) consists of basically 4 parts 1109594296 M * Bertl one is used for applications, another is used for buffering, a third for caches (files, inodes) and the 4th is actually unused 1109594314 M * Bertl now the size of those partitions changes over time ... 1109594331 M * Bertl for example application memory has some priority over caches ... 1109594352 M * Bertl caches and buffers OTOH, will use up the free memory as they are created 1109594370 J * _maharaja maharaja@ipax.at 1109594370 M * Bertl (you can see the details of the current usage in /proc/meminfo) 1109594393 J * _are_ ~are@gateway-dsl.lihas.de 1109594410 M * _are_ hi 1109594415 M * Bertl now unless the 'working set' grows larger as the physical memory, almost no swap space will be used ... because swapping is expensive 1109594420 M * Bertl welcome _are_! 1109594450 M * Bertl my 'theory' in your case is the following: 1109594459 J * DaPhreak ~DaPhreak@pc-outside.uni-greifswald.de 1109594464 M * Bertl welcome DaPhreak! 1109594477 M * DaPhreak hello Bertl :) 1109594506 Q * maharaja Quit: 1109594513 M * Bertl either the working set of apache itself, or of more than one vserver grew larger than physical memory (aproxximately almost twice as large) 1109594554 M * DaPhreak yeah, i saw this yesterday on my webserver node :) 1109594565 M * DaPhreak but i think it's due to the mysql server 1109594639 M * Bertl now with the hard cpu scheduler active and a low amount of time available ... the server might end up spending _all_ the time with swapping 1109594642 N * _maharaja maharaja 1109594680 M * DaPhreak heh, thats an effect i saw lately on a vmware gsx server .. haha killed the physical machine :) 1109594682 M * Bertl i.e. apache requests a page ... which takes a lot of time to get swapped in ... at which point the hard scheduler switches to the other vserver 1109594726 M * bro hmm, ok, ive tried to disconnect the server from the internet, so that apache wouldn't get any new requests 1109594730 M * Bertl which needs a completely different page ... and therefore initiates a new swap in 1109594759 M * Bertl and I guess that can pretty easily escalate ... 1109594766 M * bro stayed like that for bout 30 mins, and i didn't have the patience ... 1109594777 M * Bertl so now for how to detect that this happens: 1109594809 M * maharaja re 1109594812 M * Bertl thethe iostat tool shows the pages swapped in and out 1109594816 M * Bertl wb maharaja! 1109594848 Q * DaPhreak Quit: Orks are da best 1109594854 M * Bertl bro: if you see a high swap I/O then 1109594885 M * Bertl it's most likely the issue for the bad responsiveness, which is called 'trashing' 1109594887 M * bro Bertl, yeah, im using it already, and the last time it died, there was only bout 2mb/s read and bout 200kb write, and that's the last line i saw ;/ 1109594925 M * Bertl if that is what heppened, then you can pretty easy stop the misbehaviour in two ways: 1109594930 M * Bertl *happened 1109594950 M * bro vtop at that time was bout loadaverage: 2k, memory all used up, and swap not depleted 1109594965 M * bro fully that is 1109594981 M * Bertl a) you disable the hard scheduler for a few minutes (everything should go to 'normal' but of course, with a large working set you'll see the same when you turn it back on) 1109595021 M * Bertl b) you stop the entire context with sched_pause (which will stop all activity inside that vserver, and thus allow the other servers to page in their memory) 1109595060 M * Bertl of course, none of this is a real solution, the solution would be to add more memory or limit the per vserver memory 1109595336 M * bro hmm, ill try to limit per vserver memory 1109595344 M * bro tnx for advice Bertl :> 1109595351 M * Bertl you're welcome! 1109596797 Q * BWare Remote host closed the connection 1109596865 J * BWare ~bware@212.26.196.154 1109596941 M * Bertl wb BWare! 1109597171 M * BWare thanx 1109598131 J * Val ~val@gj403.loria.fr 1109598134 M * Val Hi :) 1109598139 M * Bertl welcome Val! 1109598150 M * Val hello Bertl :) 1109598239 M * Val lots of mailing-list messages i didn't read yet 1109598244 M * Val ..'m late 1109598273 M * Val are we close to a stable 2.6.X version now ? 1109598282 M * Bertl hmm, no, not really ;) 1109598288 M * Val arg ;) 1109598672 J * ntrs_ ntrs@Dardeene-68.188.50.87.charter-stl.com 1109598672 Q * ntrs Read error: Connection reset by peer 1109598976 M * Bertl okay, back later ... 1109598982 N * Bertl Bertl_oO 1109599774 Q * TheSeer Quit: Client exiting 1109599935 M * _are_ Val. you have bad experiences withvserver and kernel 2.6? 1109600071 M * Val yep :) 1109600103 M * Val with debian : there is no utils-vserver for 2.6.X dev release 1109600117 M * Val so i stay to 2.4.X in production stager 1109600118 M * rs no packaged version 1109600119 M * Val -r 1109600122 M * Val yep 1109600137 M * rs it's not that hard to do its own version actually 1109600140 M * _are_ anything more specific, my vserver boxes run fine on 2.6 so far, one dual-Operon with 64bit-kernel/system, one dual-Xeon 1109600141 M * Val i tried to build one but there's a lot of work to make it "debian compliant" 1109600160 M * _are_ ok, so the main problem is the packaging? 1109600162 M * rs Val: I did one for Lycos, I can send you the debian directory 1109600170 M * Val rs : with make; make install for sure 1109600180 M * rs no, a debianized one 1109600192 M * Val oh yes 1109600218 M * Val can you make it public 1109600222 M * Val ? 1109600236 M * rs sure, it's just a workaround waiting for an official release of the debian package 1109600240 M * rs wait a minute 1109600243 M * Val thx :) 1109600366 M * Val but i saw a long thread about utils-vserver debian packaging for dev. branch 1109600374 M * Val ...is there any news about ? 1109600383 M * rs dunno exactly 1109600386 J * TheSeer ~theseer@border.office.salesemotion.net 1109600390 M * TheSeer re... 1109600404 M * rs http://rs.rhapsodyk.net/vserver/utils/ 1109600413 M * Val rs : thx :) 1109600418 M * rs you just have to remove the -lyceu, but I guess you know that ;) 1109600474 M * Val trhead title on the mailing list was "packaging review for new Debian packages" 1109600542 M * Val rs : yep :) 1109600602 M * Val rs : and i'll read your diff to see how you made it 1109600684 M * rs it's based on the stable package with some changes 1109600694 M * Val okay 1109600712 M * rs hmmm no actually it's not (I'm reading it again :) 1109600718 M * Val :)))) 1109600763 M * Val does it support the new config stuff ? (with sub-directories ...) ? 1109600766 M * rs I made it from scratch, the debian one is really... strange :) 1109600775 M * Val arf, i agree :) 1109600808 M * rs Val: maybe the package won't fit for your needs, my needs are really differents 1109600819 M * Val yes i know :) 1109600825 M * rs the package doesn't have to support the new configuration scheme 1109600831 M * Val i'll test it 1109600836 M * Val arg 1109600840 M * Val :-/ 1109600852 M * rs I didn't do any migration scripts as well 1109600857 M * Val that's what i said : too hard to make it "debian compliant" 1109600871 M * rs it's debian compliant about paths 1109600881 M * Val it's a good start :) 1109600910 M * Val after sarge release i hope this will be done more quickly 1109600974 M * rs do you have some news on this topic ? 1109600980 M * Val yep 1109600994 M * rs I don't really follow the debian devel ml this days 1109601020 M * Val they don't want to build vserver 2.6.X non-stable package 1109601047 M * Val so they're waiting for an upstream stable one 1109601050 M * rs with an explicit name, I don't understand why 1109601083 M * Val for now they have a lot of work for sarge release 1109601085 M * rs yeah, but with a devel package, more people could test it and the stable release would come faster :) 1109601101 M * rs yeah, and do you know when the sarge release should be out ? 1109601113 M * Val so thy don't spend time to build "moving" development branch 1109601125 M * rs quite some time that it planed for... "tomorrow" 1109601133 M * Val yep :-( 1109601150 M * rs it/it's 1109601179 M * Val http://bugs.debian.org/release-critical/ 1109601203 M * Val -> lots of bugs have to be closed before release 1109601223 M * Val i you can see it is not going as fast as we want 1109601227 M * Val -i 1109601317 M * rs Val: with my debian directory, you got our ugly patch to support some kind of bandwith limit :) 1109601369 M * Val great ! :) 1109601378 M * Val ngnet stuff ? 1109601384 M * rs not at all 1109601387 M * rs I said ugly! :) 1109601392 M * Val arf :)) 1109601402 M * rs it's waiting for something better 1109601407 M * Val ok 1109601440 M * rs look at the author of the patch, I think you know him :) 1109601498 M * Val arf, carlos :) 1109601509 M * rs with it, you can put a ratelimit file in the interfaces directory 1109601565 M * Val let's say a big hello to the guys i know in you staff ;() 1109601619 M * Val may be we'll meet us end of this month (i will be in paris for work) 1109601620 M * rs ok :) 1109601634 M * rs yeah would be nice 1109601637 M * Val yep 1109601643 M * Val i'll call tuf too 1109601645 M * Val ;-) 1109601651 M * rs sure 1109601654 M * Val hehe 1109601740 M * rs we continue to talk to him, even if it has betrayed us :) 1109601815 M * Val :) 1109601968 M * rs Doener: are you around ? 1109601974 M * Doener yep 1109601983 M * rs where are you with your ngnet testing ? 1109602003 M * Doener not very far... got some unexpected work to do 1109602010 M * rs ok 1109603665 Q * sladen Remote host closed the connection 1109604253 M * ntrs_ trying to stop a vserver. It cannot completely stop because one process refuses to die. Has anyone ever seen anything like this? How do you really kill a process? 1109604267 M * ntrs_ I tried kill -9 - it doesn't work 1109604289 M * erwan_taf what's its status ? 1109604292 M * ntrs_ postgres 19089 0.0 0.1 10012 1028 ? D 09:11 0:00 postgres: checkpoint subprocess 1109604294 M * erwan_taf if it's "D" 1109604301 M * ntrs_ yes. it is D 1109604305 M * erwan_taf that's D 1109604315 M * erwan_taf kill -9 can't do nothing for that 1109604327 M * ntrs_ ok, so how else do you kill it? 1109604334 M * erwan_taf you just can't 1109604339 M * erwan_taf oh yes 1109604340 M * erwan_taf reboot :) 1109604359 M * rs or fix the reason why it's in D state 1109604360 M * erwan_taf this process is waiting a device 1109604369 M * ntrs_ oops, that is not an option. it is a production server 1109604369 M * rs most of the time, you can't :) 1109604389 M * erwan_taf if the device won't back, you can't kill it 1109604394 M * rs ntrs_: so, you have to move this vserver to another node 1109604396 M * erwan_taf +come 1109604419 M * ntrs_ I wonder what device it is 1109604427 M * rs not necessary a device 1109604437 M * rs can be a network filesystem for instance 1109604488 M * rs which version of the vserver do you use on this node ? 1109604537 M * ntrs_ 1.9.3-rc2 1109605017 J * rs_ ~rs@194.98.28.2 1109605036 M * rs_ re 1109605094 Q * rs Read error: Operation timed out 1109605095 N * rs_ rs 1109605204 Q * maharaja Remote host closed the connection 1109605327 J * maharaja maharaja@ipax.at 1109605682 Q * Val Quit: work off 1109605816 M * prae erwan_taf: :p 1109605849 M * erwan_taf :) 1109606955 J * rusty ~rusty@rev.193.226.233.50.euroweb.hu 1109606957 M * rusty hello folks 1109608114 J * monrad ~monrad@213083190130.sonofon.dk 1109608835 Q * pulsar Remote host closed the connection 1109609286 Q * monrad Quit: Leaving 1109609554 Q * BWare Quit: using sirc version 2.211+KSIRC/1.3.11 1109609612 Q * rusty Remote host closed the connection 1109610979 M * ntrs_ I try to start a vserver and I get this: 1109610982 M * ntrs_ fakerunlevel: open("/var/run/utmp"): No space left on device 1109610990 M * ntrs_ there is plenty of space everywhere 1109611018 M * ntrs_ I use tagxid but the limits are set properly from what I can see 1109611185 M * ntrs_ ok, things are getting worse 1109611198 M * ntrs_ can't even touch a file on the host 1109611203 M * ntrs_ # df 1109611204 M * ntrs_ Filesystem 1K-blocks Used Available Use% Mounted on 1109611204 M * ntrs_ /dev/sda3 4032124 308312 3518984 9% / 1109611204 M * ntrs_ /dev/sda1 101086 14122 81745 15% /boot 1109611204 M * ntrs_ none 1036436 0 1036436 0% /dev/shm 1109611205 M * ntrs_ /dev/sda2 6048352 1017796 4723316 18% /usr 1109611206 M * ntrs_ /dev/sda6 2016016 227052 1686552 12% /var 1109611208 M * ntrs_ /dev/sda7 224131048 23590328 189155460 12% /vservers 1109611212 M * ntrs_ so there is plenty of space 1109611527 M * ntrs_ solved the problem. Please disregard the above. 1109611815 N * Bertl_oO Bertl 1109611825 M * Bertl evening folks! 1109611943 M * erwan_taf evening 1109611943 M * erwan_taf ntrs_: sounds like missing inodes 1109611975 M * Bertl rs: everything fine? 1109612443 Q * _are_ Quit: Disconnecting 1109613106 M * rs Bertl: yeah, hmm the quad is down, dunno why, I will go see what is going on 1109613121 M * rs and I got some trouble with ngnet 1109613206 Q * DuckKing Ping timeout: 480 seconds 1109613257 M * Bertl rs: is the controller still working? 1109613272 M * rs yeah but you can't join it I guess 1109613996 J * rs_ ~rs@194.98.28.2 1109614226 Q * maharaja Remote host closed the connection 1109614478 J * maharaja maharaja@ipax.at 1109614790 Q * Duckx Quit: Leaving 1109614871 Q * rs_ Quit: leaving 1109615276 Q * eyck_ Ping timeout: 480 seconds 1109615292 M * Bertl Doener: you around? 1109615326 M * Bertl mugwump: please contact me when you get back ... 1109615443 M * rs re 1109615461 M * Bertl wb rs! 1109615466 M * rs Bertl: about ngnet, I got bit slowness as soon as I start a vserver 1109615483 M * rs looks like the last test I did with ngnet 1109615488 M * rs maybe because of NFS ? 1109615514 M * Bertl hmm, could be it works in general (even over nfs)? 1109615523 M * rs seems that it's impossible to access the nfs server from inside the context 1109615540 M * rs I didn't test yet with just a context 1109615581 M * Doener Bertl: yep 1109615612 M * Bertl excellent, it looks like there is a real issue with the hard scheduler ... 1109615624 M * rs where ? 1109615626 M * rs with ngnet ? 1109615646 M * Doener i feared to hear that *prepares to dig into the scheduler* 1109615652 M * Bertl hmm, no, in general not ngnet 1109615661 M * rs wow, which kind of issue ? 1109615672 J * eyck eyck@81.219.64.71 1109615680 M * Bertl wb eyck! 1109615752 M * Bertl well, the thing is, ntrs reported that one of his servers did 'hang' every 24 hours 1109615773 M * Bertl (after upgrading to one of the releases) 1109615778 M * rs some lycos users complain about that 1109615780 M * Bertl *recent 1109615780 M * rs one actually 1109615798 M * Bertl now today I did take a look at a server which was running 1109615822 J * ola ~ola@c-adt-5.ataco.se 1109615824 M * Bertl 2.6.9-rc3-bk3-vs1.9.3-rc2 with the hard scheduler enabled 1109615827 M * ola Hello 1109615831 M * Bertl welcome ola! 1109615834 M * ola Thanks. 1109615846 M * ola Working with the util-vserver package for Debian right now. 1109615850 M * rs Bertl: http://www.lycos-vds.com/bugs/bug_view_page.php?bug_id=0000074 1109615851 M * ola Have some questions. :) 1109615855 M * Bertl excellent! 1109615869 M * ola Should all binaries be compiled statically? 1109615885 M * Bertl ola: we fixed a long outstanding bug in the kernel which should make usage of the old and new tools with 2.6 much easier/better 1109615905 M * rs which bug ? 1109615907 M * Bertl ola: basically dietlibc is the best choice ... 1109615935 M * Bertl okay, let's make a short plan what we want to address now and later ;) 1109615947 M * Bertl (otherwise we end up slicing this channel ;) 1109615950 M * rs I never got any issue under debian with shared linked (no dietlibc) binaries 1109615995 M * Bertl Doener, rs, ola: possible topics are: hard scheduler, debian userspace, kernel bugfix (choose one) 1109615996 M * ola My question is just if it should be linked statically ot not. 1109616019 M * ola If it should I'll add overrides to the errors that lintian create. 1109616031 M * rs we can take them in this order :) 1109616037 M * Doener i vote for the last two ;) 1109616047 M * rs Doener: hehe :) 1109616058 M * Doener i guess "kernel bug" is the one we fixed yesterday? 1109616066 M * Bertl yep, precisely 1109616078 M * rs could you shortly describe it ? 1109616084 M * Doener ok, then let's tackle them in reverse order 1109616090 M * rs just for my information 1109616095 M * Bertl is that fine with you ola? 1109616119 M * ola I have not followed the discussion. Rereading now. :) 1109616147 M * ola Let us take one topic at a time. 1109616162 M * Bertl okay, so reverse order it is ... 1109616175 M * ola Ok, starting with kernel bugfix then 1109616190 M * ola What is fixed? 1109616193 M * Bertl the bug was basically wrong return codes which confused the userspace tools 1109616207 M * ola Ok, what trouble did it cause? 1109616211 M * Bertl this way, dynamic contexts could not be entered with 'some' tools 1109616240 M * ola Was this related to 2.4 kernel or 2.6 or both? 1109616242 M * Bertl static contexts were not affected because of a second bug ;) 1109616252 M * ola :) 1109616267 M * rs hehe :) 1109616271 M * Bertl I remmeber (now) that enrico tried to explain the issue to me 1109616291 M * Bertl but I didn't understand it (and he gave up too easily ;) 1109616317 M * Bertl okay, sec emergency! 1109616363 M * Bertl Doener: maybe you can explain the details for me in the meantime ;) 1109616370 M * Doener ok :) 1109616432 M * rs lets hear :) 1109616440 M * Doener the checks for existing contexts were correctly done, i.e. returning -EEXIST if you're trying to create a context with an existing xid 1109616493 M * Doener the 'bug' was, that a previous check in a higher level function already bailed out with -EINVAL because you're not allowed to create a dynamic context using a specified id 1109616549 M * Doener but the tools rely on getting an -EEXIST if a context already exists, to determine if they should try to migrate into a context instead of creating it 1109616587 M * rs ok got it 1109616602 M * Doener with the fix done yesterday you (should?) now correctly get -EEXIST if the context exists, and -EINVAL if the specified value is really invalid 1109616622 M * rs but Bertl said that this bug wasn't triggered with static ids because of another bug? 1109616630 M * Doener it probably got messed up because the legacy api uses a single call for creation and migration 1109616634 M * rs but this bug shouldn't happen with static ids 1109616658 M * rs is that correct? 1109616664 M * Doener hm, let me have a look.. 1109616982 M * Doener ah that's the path i still don't really understand... 1109617008 M * Doener it's basically: suceed if ((not a) and a) 1109617017 M * Doener (at least i read it like that...) 1109617024 M * Doener s/suc/succ/ 1109617038 M * rs hum :) 1109617065 M * rs I've got to go, so you later 1109617069 Q * rs Quit: home 1109617164 M * Bertl okay, now that we discussed? that, let's move on to the debian util-vserver ;) 1109617201 M * ola Not right now... :) 1109617207 M * ola I have to go for an hour. 1109617209 M * Doener phone ;) 1109617222 M * ola But I'll be back later. 1109617241 M * ola Sorry 1109617269 M * Bertl okay, cya, well discuss it later then! 1109617276 M * Bertl *we'll 1109617288 Q * flock Remote host closed the connection 1109618182 J * erwan_ho ~erwan@lns-vlq-39f-81-56-133-136.adsl.proxad.net 1109618437 J * shuri sjnesjd@dsl.speedline209.226.electronicbox.net 1109618671 M * Bertl welcome shuri! erwan_ho! 1109618701 M * Bertl okay Doener, as soon as you are ready we can continue with the scheduler issue 1109618767 M * Doener in about 5 minutes 1109618772 M * Bertl np 1109618921 M * erwan_ho Bertl: \o/ 1109618997 Q * prae Quit: Client exiting 1109619177 J * flock ~restless@l192-117-111-12.broadband.actcom.net.il 1109619264 M * Doener ok, let's go... 1109619423 M * Bertl okay, so today we had a mmachine which 1109619432 M * Bertl showed the following (inside a vserver) 1109619453 M * Bertl apache 32554 0.0 0.3 38588 4084 ? S 07:06 0:04 /usr/sbin/httpd 1109619453 M * Bertl apache 32608 0.0 0.3 38808 4164 ? H 07:06 0:05 /usr/sbin/httpd 1109619454 M * Bertl apache 8734 0.0 0.4 38588 5092 ? H 07:30 0:02 /usr/sbin/httpd 1109619454 M * Bertl apache 8777 0.0 0.3 37492 3756 ? H 07:30 0:03 /usr/sbin/httpd 1109619454 M * Bertl apache 8778 0.0 0.5 38444 5944 ? S 07:30 0:02 /usr/sbin/httpd 1109619456 M * Bertl apache 26051 0.0 0.4 38580 4348 ? H 08:13 0:01 /usr/sbin/httpd 1109619473 M * Bertl basically this goes on and on .. the total is: 1109619491 M * Bertl nr_threads:85 1109619492 M * Bertl nr_running:45 1109619492 M * Bertl nr_onhold:45 1109619501 M * Doener o.O 1109619515 M * Bertl PROC: 85 200 20013933470 1109619517 M * Doener do i fail in math or in linux-knowledge? 1109619563 M * Bertl yeah, well, it's an older kernel (2.6.9-rc3-bk3-vs1.9.3-rc2) 1109619590 M * Bertl and it might actually be that some processes did already terminate, but still are accounted as on hold 1109619600 M * Doener ah ok 1109619653 M * Bertl the funny thing is (and I didn't consider that earlier) 1109619674 M * Bertl there is no practical way to get such processes out of the hold queue/state 1109619787 M * Bertl yeah, that confirms the terminated theory 1109619797 M * Bertl (after raising the scheduler values, I get) 1109619811 M * Bertl apache 8734 0.0 0.3 38588 3204 ? H 07:30 0:02 /usr/sbin/httpd 1109619811 M * Bertl apache 8777 0.0 0.2 37492 2200 ? H 07:30 0:03 /usr/sbin/httpd 1109619811 M * Bertl apache 8778 0.0 0.0 0 0 ? Z 07:30 0:02 [httpd] 1109619828 M * Doener ok 1109619878 M * Bertl so, regardless of _any_ bug, the issue is like this ... 1109619892 M * Bertl - enable hard scheduler, set limits 1109619902 M * Bertl - some processes get 'on-hold' 1109619917 M * Bertl - disable hard scheduler .. 1109619926 M * Bertl - processes will not be rescheduled (I guess) 1109619970 M * Bertl you are good at digging out old irc discussions? 1109619994 M * Doener if i got them i my logs (and preferably remember at least some words) 1109620018 M * Bertl I had a rather lengthy discussion with Sam about the 'future' hard cpu scheduler 1109620033 M * Bertl we considered the advantages of per cpu/per context queues 1109620104 M * Doener October 26 1109620111 M * Bertl if you say so ;) 1109620125 A * Bertl is looking now ... 1109620145 M * Doener about 4am GMT+1 1109620149 J * duckx ~Duck@dyn-83-157-160-90.ppp.tiscali.fr 1109620178 M * Bertl welcome duckx! 1109620264 M * Bertl hmm, seems we have no public logs of that time?! 1109620320 M * Bertl ryan stopped with 17.09.2004 1109620341 M * Bertl pinky started with 15.11.2004 1109620373 M * Bertl rs with 11.11.2004 1109620376 M * Doener http://doener.homeip.net/doener/sched-log.txt 1109620390 M * Bertl tx 1109620677 A * ola is back 1109620683 M * Doener wb ola 1109620689 M * Bertl wb ola! 1109620691 M * ola Thanks 1109620777 M * ola The kernel patch we discussed earlier. The fixed version, was it for 2.6? 1109620786 M * ola And the problem, do it apply for 2.4 as well? 1109620807 M * Bertl it is for 2.6 and it isn't required in 2.4 1109620827 M * Bertl it's just to guarantee easy migration from 2.4 to 2.6 1109620883 M * ola Ahh ok. 1109620888 M * ola Good to know. 1109621024 J * prae ~prae@sherpadown.net 1109621030 M * Bertl welcome prae! 1109621053 M * Bertl ola: so now back to your question regarding static or not static ... 1109621063 M * prae hi Bertl 1109621099 M * Bertl I guess most of the tools do not need to be compiled statically, but Enrico explained why he chose dietlibc over (g)libc 1109621161 M * ola Ok. 1109621177 M * ola Then I assume that they should be statically complied and that it is not a bug. 1109621203 M * Bertl which tools in particular do you look t? 1109621206 M * Bertl *at 1109621231 M * ola In util-vserver? 1109621233 M * Bertl the core utils, I guess? 1109621238 M * ola I 1109621245 M * ola I'll make list. Wait a min. 1109621273 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/capchroot 1109621273 M * ola N: 1109621273 M * ola N: The package installs a statically linked binary or object file. 1109621273 M * ola N: 1109621273 M * ola N: Usually this is a bug. Otherwise, please contact 1109621275 M * ola N: lintian-maint@debian.org about this so that this exception would be 1109621277 M * ola N: recognized in future versions of Lintian, or install an override if 1109621280 M * ola N: your package is an exception. Note that binaries named *-static and 1109621283 M * ola N: *.static are already excluded, just as any binaries in packages named 1109621285 M * ola N: *-static. 1109621288 M * ola N: 1109621290 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/chain-echo 1109621291 M * Bertl ah, I.c 1109621293 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/chcontext-compat 1109621296 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/check-unixfile 1109621298 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/chroot-cat 1109621301 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/chroot-rm 1109621303 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/fakerunlevel 1109621306 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/filetime 1109621308 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/keep-ctx-alive 1109621311 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/lockfile 1109621313 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/readlink 1109621316 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/save_ctxinfo 1109621319 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/secure-mount 1109621321 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/mask2prefix 1109621324 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/exec-ulimit 1109621326 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/rpm-fake-resolver 1109621329 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/vshelper-sync 1109621332 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/sigexec 1109621334 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/vunify 1109621337 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/vcopy 1109621339 M * ola E: util-vserver: statically-linked-binary ./usr/lib/util-vserver/vserver.start.bin 1109621340 M * Doener pastebin is over there ---> ;) 1109621342 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/chbind 1109621345 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/exec-cd 1109621347 M * Bertl yes, that is expected with dietlibc ... 1109621347 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/lsxid 1109621350 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/chxid 1109621353 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vps 1109621355 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/showattr 1109621358 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/setattr 1109621360 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/reducecap 1109621363 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vattribute 1109621365 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vcontext 1109621368 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vlimit 1109621370 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vkill 1109621373 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vnamespace 1109621376 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vrsetup 1109621378 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vsched 1109621381 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vserver-stat 1109621383 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vserver-info 1109621386 M * ola E: util-vserver: statically-linked-binary ./usr/sbin/vuname 1109621388 M * ola They are quite a few. 1109621391 M * ola :) 1109621404 M * Doener hu? did it finish? *g* 1109621417 J * monrad ~monrad@213083190130.sonofon.dk 1109621418 M * Bertl could you paste it again, I didn't get everything? *G* 1109621426 M * Doener ARGH! 1109621438 M * Bertl no nooo! it was a joke! 1109621458 M * Bertl welcome monrad! 1109621463 A * Doener .oO( maybe i should think about that donation again... ;) 1109621483 M * ola :) 1109621511 M * Bertl okay, yes with dietlibc it's static ... 1109621516 M * ola I could have summarized it as /usr/sbin/* and /usr/lib/util-vserver/* :) 1109621537 M * Bertl doesn't mean it's larger than usual (dynamic) 1109621546 M * ola Then it is expected behaviour and I'll handle it. 1109621553 M * Bertl excellent! 1109621570 M * ola I'm not sure if it will be bigger. I do not think it is a big issue though. 1109621602 M * Bertl 11172 Feb 18 20:30 /usr/sbin/vkill 1109621614 M * Bertl 25044 Feb 18 20:30 /usr/sbin/vserver-info 1109621618 J * monski ~monrad@213083190130.sonofon.dk 1109621620 M * Bertl so I guess it's pretty lean 1109621643 M * Bertl 579384 Feb 25 2002 /bin/bash 1109621676 M * Doener the basic idea is to avoid accidental usage of libs from the vservers i guess 1109621694 M * mugwump Morning all! 1109621698 M * Bertl yes, so that preload and such stuff doesn't matter ... 1109621703 M * Bertl morning Sam! 1109621761 M * ola Sounds like a reasonable thing. 1109621816 M * mugwump hey full house today 1109621835 J * ntrs ~ntrs@iits01188.inlink.com 1109621858 Q * monrad Read error: Operation timed out 1109621933 M * Doener hi mugwump 1109621954 M * ola I hope this can give me the opportunity to upload util-vserver (development branch) to exprimental today. 1109622107 M * Bertl that would be great! 1109622249 M * Bertl mugwump: it seems we have troubles with the ahrd cpu scheduler ... 1109622281 M * Bertl so the question is: new design right away, according to our discussion from October the 26th (tx Doener) 1109622311 M * prae Bertl: vserver and i64 arch works ? 1109622366 M * Bertl ia64?, should work but is untested 1109622383 M * Bertl or do you mean x86_64 aka amd64 or emt64? 1109622477 M * prae amd64 1109622538 M * Bertl x86_64 is reported to work with 32/32 and 64/64 but has some issues with 32bit util-vserver tools on 64bit kernel 1109622555 M * prae hmmm, If you want, I can test vserver in amd64 server at mandrakesoft 1109622584 M * Bertl go ahead, what needs to be done is the 32/64 emulation layer 1109622608 Q * duckx Read error: Operation timed out 1109622608 M * Bertl kernel syscall compatibility layer that is 1109622630 M * Bertl as I have no such amchine available, it's hard to test for me ... 1109622738 M * prae if I have time, I will test it on our servers x64 1109622837 M * Bertl btw, is there support for sparc64 (mandrake I mean)? 1109622876 M * prae hmmm, I don't sure 1109622894 M * prae erwan_ho: mandrakesoft is support under sparc64 architecture ? 1109622897 M * prae s/soft// 1109622926 M * prae s/support/supported/ 1109622945 M * mugwump Bertl: is that deschedule oops still giving you troubles? 1109622953 N * monski monrad 1109622954 M * mugwump s/oops/panic/ 1109622976 M * Bertl well, I'm not sure that this is what happens, but I have some other ugly side effects 1109622999 M * Bertl for example we just ended up with a 'hung' context like this: 1109623010 M * Bertl apache 32608 0.0 0.2 38808 2652 ? H 07:06 0:05 /usr/sbin/httpd 1109623010 M * Bertl apache 8734 0.0 0.3 38588 3200 ? H 07:30 0:02 /usr/sbin/httpd 1109623010 M * Bertl apache 8777 0.0 0.2 37492 2196 ? H 07:30 0:03 /usr/sbin/httpd 1109623010 M * Bertl apache 8778 0.0 0.0 0 0 ? Z 07:30 0:02 [httpd] 1109623043 M * Bertl and the 'H' processes are not rescheduled anymore, regardless what you do ... 1109623122 M * mugwump because their bucket never gets recalculated? 1109623147 M * Bertl could be .. setting the bucket by hand didn't help either ... 1109623172 M * Bertl so I guess they just somehow managed to get 'off' the runqueue but still keep the 'H' 1109623180 M * Bertl *holdqueue 1109623284 M * mugwump well, that's no good is it :) 1109623308 M * mugwump there must be something else that can wake a process up, perhaps? 1109623337 M * ola The development libraries for util-vserver. Are they supported in some way? 1109623348 M * mugwump maybe it would be a better idea to never try to deschedule, just move all the processes on the runqueue to a holdqueue 1109623376 M * mugwump (ie, all but the one running) 1109623406 M * mugwump then move the one running to the holdqueue when it is descheduled normally 1109623445 M * mugwump Is there an easy way to reproduce the behaviour? 1109624597 Q * ntrs Quit: Leaving 1109624597 J * ntrs ~ntrs@iits01188.inlink.com 1109624852 M * ola So now an util-vserver package is uploaded to Debian exprimental. Finally! 1109624868 M * ola I hope to upload a kernel patch package soon too. 1109624904 M * ola But it will take some time. And also some extra time as the FTP masters need to approve. 1109624914 M * ola But now it is bedtime. See ya. 1109624954 M * Doener night ola! 1109625082 M * ntrs What happened to Bertl 1109625084 M * ntrs ??? 1109625199 M * ntrs can anyone read this? 1109625276 M * ntrs hello? anyone? 1109625306 M * Doener yep, i can read that :) 1109625325 M * ntrs ok thanks Doener 1109625910 M * Bertl okay, I'm back now ... 1109625922 M * Bertl we had another network outage --- 1109625951 Q * ndim Ping timeout: 480 seconds 1109626172 Q * shuri Quit: 1109626269 J * ndim hun@helena.bawue.de 1109626516 Q * halorgium Ping timeout: 480 seconds 1109626846 J * halorgium tim@halorgium.net 1109627387 Q * Loki|muh Quit: leaving 1109627613 J * ntrs__ ntrs@Dardeene-68.188.50.87.charter-stl.com 1109627613 Q * ntrs_ Read error: Connection reset by peer 1109627843 M * Bertl okay folks, back later ... 1109627848 N * Bertl Bertl_oO 1109628220 J * Loki|muh loki@satanix.de 1109628600 J * duckx ~Duck@dyn-83-157-132-19.ppp.tiscali.fr 1109628711 Q * ntrs Quit: Leaving 1109631123 Q * duckx Read error: Operation timed out 1109631712 N * Bertl_oO Bertl 1109631722 M * Bertl evening folks! 1109631856 M * Doener wb Bertl! 1109631863 M * Bertl ah, thanks! 1109631888 M * Bertl so what are your plans for this evening? 1109631952 M * Doener finally change __loc_vx_info to use ERR_PTR and friends... after that... no idea yet 1109632016 M * Bertl okay, let me know when you approach an idle state ... maybe we can work out a plan for the scheduler or start investigating/hacking on the ngnet 1109632073 M * Bertl btw, I did a nice cleanup of the BME stuff, so I guess either mainline will take it soon, or it should be easily integrated into linux-vserver 1109632127 M * Doener i'd actually prefer ngnet ;) i've started digging into that and the scheduler is something i know this much -><- about 1109632167 M * Bertl no problem with that, I'll dig through the kernel todo list and see what I can mark as done ... 1109632805 M * Bertl hmm, Doener, was that copy_to_user thing resolved by you somehow? is that a pending patch? 1109632836 M * Doener Bertl: hm, according to cscope, __loc_vx_context is used by a legacy call only now. Do you think we should drop __loc_vx_info and modify the legacy vc_new_s_context slightly? 1109632847 M * Doener no, didn't find any time to do that 1109632879 M * Bertl no, I'd say we make that __loc_vx_context() legacy too .. and let it die in piece ;) 1109632891 M * Doener that would mean: use __create_vx_info instead and do a lookup on EEXISt 1109632896 M * Doener ok 1109632916 M * Bertl it's not that large and I guess it will not be changed soon ... 1109632932 M * Bertl I thought about splitting legacy into two parts (net legacy and other) 1109632957 M * Doener guess we can save the time making __loc_vx_info use ERR_PTR then 1109632983 M * Bertl no problem with that either ... 1109633022 M * Doener ok, grepping to make sure, ready for (almost) everything then 1109633043 M * Bertl okay, then let's do the copy_to_user() first and then head for ngnet 1109633130 J * Doener_ ~doener@p5487E9F5.dip.t-dialin.net 1109633149 M * Bertl hmm, wb? 1109633152 M * Doener_ i'll give irssi a spin. xchat uses soooo much space on the desktop 1109633167 M * Bertl hehe, yes I like irssi ... 1109633195 M * daniel_hozac the window is resizable ;) 1109633212 M * Bertl which one? ;) 1109633242 M * prae sleeping time ! 1109633245 M * prae bye all ! 1109633249 M * Bertl night! 1109633264 M * Doener_ daniel_hozac: with 14 channels there's not much to resize, the actual channel space gets too small to be readable 1109633281 Q * prae Quit: Pwet 1109633297 M * Bertl well, that's what the keybindings are for (in irssi) 1109633342 M * mugwump I can even use irssi on my *phone* - only 40x20 1109633345 M * Doener_ i'll say to what i'll stick in the future, let's do something worthwhile ;) 1109633373 M * Bertl sure, proc_vid_info_read() it is ... 1109633404 M * Bertl as you are really good in digging out irc logs, could you have a look who reported that and how he discovered it? 1109633440 M * Doener_ daniel_hozac 1109633441 M * Bertl I'll take a look at existing functions and how they do it now ... 1109633460 M * Bertl ah that nice pal? right over there? 1109633473 M * Doener_ yup yup 1109633478 M * daniel_hozac hehe :) 1109633495 M * Bertl okay, daniel_hozac, how to get the error you have here on my machine? 1109633510 M * Bertl any special compiler/commands whatever? 1109633536 M * daniel_hozac not that i am aware of. 1109633561 M * Bertl so why do you get that message and I do not? 1109633584 P * click [IRSSI] 1109633597 M * daniel_hozac is copy_to_user not defined as __must_check in vanilla sources? 1109633598 M * Bertl btw, what was the message exactly? do you ahve it at hand? 1109633607 M * daniel_hozac i'm rebuilding the tree right now 1109633613 M * Doener_ Feb 12 14:25:17 daniel_hozac kernel/vserver/proc.c: In function `proc_vid_info_read': 1109633616 M * Doener_ Feb 12 14:25:17 daniel_hozac kernel/vserver/proc.c:319: warning: ignoring return value of `copy_to_user', declared with attribute warn_unused_result 1109633617 M * Bertl okay, thanks a lot! 1109633640 M * Doener_ that attribute (warn_unused_result) should be in gcc 3.3 and 3.4 1109633647 M * Doener_ http://www.ohse.de/uwe/articles/gcc-attributes.html#func-warn_unused_result 1109633652 M * Bertl hmm, gcc 3.3.5 here 1109633688 M * Bertl hmm, no 3.3.4 actually on that machine ... 1109633830 M * Doener_ #if __GNUC_MINOR__ >= 4 1109633830 M * Doener_ #define __must_check __attribute__((warn_unused_result)) 1109633831 M * Doener_ #endif 1109633838 M * Doener_ guess that's it ;) 1109633871 M * Bertl so it's 3.4 for that, right? 1109633899 M * Doener_ yep, at least regarding the kernel 1109633944 M * Doener_ maybe that site is just wrong about it being available on gcc3.3 1109633948 M * Bertl hmm, not sure that I want that ... 1109634069 M * Bertl let's see if 3.3. can do that ;) 1109634101 M * Bertl include/linux/fs.h:1621: warning: `warn_unused_result' attribute directive ignored 1109634107 M * Bertl doesn't look like ;) 1109634186 M * Bertl okay, daniel_hozac I guess for now you can test compile on our behalf, right? 1109634193 M * daniel_hozac sure. 1109634203 M * Doener_ seems to be default to return -EFAULT if copy_to_user fails 1109634209 M * Bertl daniel_hozac: excellent! and thanks a lot! 1109634238 M * Bertl yes, but I guess we should change to the buffer thingy 1109634354 M * Bertl btw, does anybody know an apt repository for fc2/fc3 which I could use? 1109634398 M * daniel_hozac fedora.us? 1109634436 M * Bertl okay, is there a way to copy/mirror that? 1109634454 J * johnny ~johnny@ip68-10-185-138.hr.hr.cox.net 1109634457 M * daniel_hozac yeah. mirrors.kernel.org has an rsyncable mirror of it. 1109634476 M * johnny ng9.0 what does that mean? 1109634495 M * Bertl welcome johnny! 1109634503 M * Bertl it means that ngnet is at 9.0 ;) 1109634508 M * johnny oh 1109634514 M * johnny so how is that goin? 1109634531 M * johnny my vserver has been running nice so far :) 1109634541 M * johnny in fact.. using it with gentoo is very nice :)] 1109634565 M * Bertl sounds good, what kernel/patches if I may ask? 1109634571 M * johnny 1.9.4 1109634579 M * johnny on 2.6.11-rc3 1109634583 M * johnny vanilla 1109634610 M * johnny with util-vserver 030.204 1109634619 M * johnny err 0.30.204 1109634643 M * johnny so . how is ngnet going? 1109634753 M * Bertl well, the basic stuff is working ... we are hacking on the details now 1109634787 M * Bertl Doener: proc_pid_attr_read() as example 1109634845 M * Bertl johnny: so you are a happy linux-vserver user now? 1109634930 M * johnny so far 1109634939 M * johnny except about localhost not really existing.. 1109634951 M * johnny err lo 1109634959 M * johnny maybe i should point localhost to a real pi 1109634963 M * johnny err ip 1109634973 M * Bertl yes, that could/would work in some cases 1109634989 M * Bertl Doener: so I guess we remove everything after 1109634990 M * Bertl length = PROC_I(inode)->op.proc_vid_read(vid, (char*)page); 1109634997 M * Bertl and replace it by 1109635002 M * Bertl if (length >= 0) 1109635002 M * Bertl length = simple_read_from_buffer(buf, count, ppos, (char *)page, length); 1109635005 M * Bertl free_page(page); 1109635008 M * Bertl return length; 1109635019 M * johnny is there a roadmap or proposed feature list of ngnet somewhere? 1109635041 M * Bertl not really (no time for that yet) 1109635053 M * johnny what are the real goals then? 1109635068 M * Bertl but for features, it should basically provide this: 1109635083 M * Bertl - separate (virtual/mapping) interfaces 1109635093 M * Bertl - separate accounting of packets 1109635102 M * Bertl - separate routing tables 1109635127 M * Bertl - a virtual switch to let the vservers (and the host) communicate 1109635130 M * Doener_ yep 1109635136 M * johnny so that would allow lo interfaces in vservers? 1109635143 M * Bertl - a virtual arp handler to satisfy arp requests ... 1109635149 M * Bertl yes of course 1109635163 M * Bertl in a later stage we also plan to support 1109635181 M * Bertl - iptables ipv4 and ipv6 1109635198 M * Bertl - bandwidth limiting