1120435446 N * Bertl_oO Bertl 1120435464 M * Bertl back now ... did take some while ... 1120435496 M * Bertl looks like most folks are already asleep ... 1120436800 Q * complexho Ping timeout: 480 seconds 1120437019 J * complexho ~mark@funk.gotadsl.co.uk 1120437530 Q * click Ping timeout: 480 seconds 1120437606 Q * rs Quit: rs 1120438847 J * eXplasm2 explasm@p549F799F.dip.t-dialin.net 1120439286 Q * eXplasm Ping timeout: 480 seconds 1120443520 M * kestrel is it possible to do bind mounts into a vserver while stilling using namespaces? 1120443549 M * kestrel ie. bind mount a directory from the ctx 0 into ctx 2, for example 1120443668 M * kestrel s/ie/eg/ 1120443792 M * Bertl yes 1120443803 M * Bertl but you have to do them in the 'correct' namespace 1120443932 M * kestrel okay 1120443989 M * kestrel the namespace is the same as the ctx? 1120444013 M * Bertl yes, it is stored within the context, so you can enter it with the xid 1120444018 M * kestrel cool 1120444037 M * Bertl you need to use vnamespace for that 1120444040 M * kestrel so, i've managed to mount lvm volumes using vnamespace, but bind mounts are eluding me.. 1120444063 M * kestrel i am doing: vnamespace -e 3 mount -o rbind /srv/arch/cache/ /vservers/retina/var/cache/pacman 1120444099 M * kestrel hmm 1120444108 M * kestrel wait a minute, i might just be insane! 1120444209 M * kestrel that is exceedingly cool herbert 1120444324 M * Bertl well, thanks! 1120444330 M * kestrel heh 1120444344 M * kestrel ah, actually, i have one problem 1120444371 M * kestrel does the source path have to exist in the new namespace? 1120444376 M * kestrel [root@sclera:/srv/arch/cache]vnamespace -e 3 mount -o rbind /srv/arch/cache /vservers/retina/var/cache/pacman 1120444377 M * kestrel mount: special device /srv/arch/cache does not exist 1120444394 M * kestrel there is no /srv/arch/cache in ctx 3 1120444419 M * Bertl well, you can mount it there, no? 1120444446 M * Bertl or you could mount such stuff _before_ creating the namespace :) 1120444492 M * kestrel but why is it failing? 1120444516 M * kestrel i can do both of those things, but the first one will mean i'll have to symlink it to the correct place which is a bit messy 1120444544 M * Bertl huh? 1120444576 M * kestrel well, it has to be in /vservers/retina/var/cache/pacman, that's where the package manager requires it 1120444594 M * kestrel does that make sense? 1120444626 M * Bertl fine, a bind mount is not an option? 1120444711 M * matti Bertl: :)* 1120444712 M * matti :> 1120444762 M * Bertl hey matti :] 1120444770 M * matti :] 1120444826 M * kestrel a bind mount is what i'm trying to do... 1120444873 M * kestrel [root@sclera:/srv/arch/cache]vnamespace -e 3 mount -o bind /srv/arch/cache /vservers/retina/var/cache/pacman 1120444874 M * kestrel mount: special device /srv/arch/cache does not exist 1120444891 M * kestrel is that the correct way to do a bind mount in another namespace? 1120444909 M * Bertl if /srv/arch/cache does exist in that namespace, yes 1120444927 M * Bertl ah, yeah, and of course /vservers/retina/var/cache/pacman 1120444933 M * kestrel heh, yeah 1120444958 M * kestrel however, this works: 1120444959 M * kestrel [root@sclera:/srv/arch/cache]vnamespace -e 3 mount -o bind /srv/arch /vservers/retina/bar/ 1120444959 M * kestrel [root@sclera:/srv/arch/cache]ls /vservers/retina/srv/arch 1120444959 M * kestrel ls: /vservers/retina/srv/arch: No such file or directory 1120444966 M * Bertl not sure why you are using this obsolete syntax instead of the mount --bind :) 1120444989 M * kestrel just habit i guess :) 1120445008 M * Bertl why not use bash and look around in that namespace? 1120445063 M * kestrel i have been...anything specific? 1120445112 M * kestrel i don't understand why i can mount /srv/arch but not /srv/arch/cache 1120445125 M * kestrel even though /srv/arch doesn't exist at all in namespace 3 1120445212 M * Bertl well, probably because of the way 'mount' does lookup dentries 1120445250 M * kestrel hmm, okay 1120445302 M * kestrel so there's no work around except for symlinkink or mounting prior to namespace creation...is that a fair summary? 1120445393 M * Bertl symlinking will buy you nothing 1120445411 M * Bertl but I still do not see what your problem is ... 1120445451 M * kestrel my problem is, i can't bind mount a particular directory from ctx 0 into another directory in ctx 3 1120445469 M * Bertl no, but why would you want to do that? 1120445492 M * kestrel to share a directory amongst multiple vservers and the root context 1120445530 M * Bertl that just requires that you either mount it before you spawn off the namespaces, or mount it in every namespace 1120445566 M * kestrel yes, well, the former is one of the options i just mentioned 1120445582 M * kestrel but the latter doesn't work unless i mount the physical volume 1120445603 M * kestrel ie. vnamespace -e 3 mount /dev/vg00/cache /vservers/retina/var/abs/cache 1120445622 M * kestrel ^^ that works, but the cache is not at the root of the volume 1120445641 M * kestrel so i'd have to mount it somewhere else in the vserver, then symlink it to /var/abs/cache in retina 1120445681 M * Bertl no, you can mount it outside the guest, just as on the host, no? 1120445693 M * Bertl then you 'just' --bind mount it into place 1120445727 M * Bertl you could also (as a third alternative) mount it inside the guest, and then --rbind the subdirectory over the mountpoint :) 1120445790 M * kestrel what do you mean by "mount it outside the guest"? 1120445808 M * kestrel i have it mounted on /srv/arch in the host at the moment... 1120445809 M * kestrel ? 1120445833 M * Bertl yeah, change into the namespace 1120445839 M * Bertl (with bash) 1120445842 M * kestrel okay 1120445845 M * Bertl then look at /srv/arch 1120445849 M * Bertl is it there? 1120445865 M * kestrel yep 1120445874 M * kestrel well, the pth is there, but it's empty 1120445881 M * kestrel [root@sclera:/srv/arch/cache]vnamespace -e 3 /bin/bash 1120445881 M * kestrel [root@sclera:/vservers/retina]cd /srv/arch 1120445881 M * kestrel [root@sclera:/srv/arch]ls -l 1120445881 M * kestrel total 0 1120445881 M * kestrel [root@sclera:/srv/arch] 1120445885 M * Bertl so mount it there ... 1120445897 M * kestrel ah 1120445909 M * Bertl then bind mount it into the guest 1120445949 M * kestrel AHA! 1120445955 M * kestrel success! 1120445972 M * Bertl congrats! you mastered namespaces lvl 1 :) 1120446012 M * kestrel hmm yes 1120446042 M * Bertl nah, actually it's lvl 2 ... lvl 1 was the rbind 1120446295 M * matti ;p 1120446318 M * matti namespace is a really black magic stuff. 1120446319 M * matti ;] 1120446356 M * Bertl indeed 1120446409 M * Bertl and a bunch of different issues can arise from small mistakes 1120446797 M * kestrel yeah 1120446859 M * kestrel i was under the misapprehension that namespace inheritance from ctx 0 was live, but i assume not 1120447120 M * kestrel instresting 1120447502 M * Bertl indeed! :) 1120450827 M * Bertl okay, off to bed now ... have a nice one ... 1120450833 N * Bertl Bertl_zZ 1120451464 J * shuri_ shuri@64.235.209.226 1120451466 Q * shuri_ Quit: 1120452682 Q * shuri Remote host closed the connection 1120454881 Q * SNy uranium.oftc.net keid.oftc.net 1120454881 Q * maharaja uranium.oftc.net keid.oftc.net 1120454881 Q * sladen uranium.oftc.net keid.oftc.net 1120454881 Q * neofutur uranium.oftc.net keid.oftc.net 1120454881 Q * albeiro uranium.oftc.net keid.oftc.net 1120454948 J * albeiro ~albeiro@graffias.estrefa.pl 1120454948 J * sladen paul@starsky.19inch.net 1120454948 J * neofutur ~neofutur@neofutur.net 1120454948 J * maharaja maharaja@ipax.at 1120454948 J * SNy ~mfr@bmx-chemnitz.de 1120456093 M * Hollow morning 1120456433 M * are|lunch morning 1120456440 N * are|lunch _are_ 1120456441 M * DaPhreak lo Hollow, are 1120457016 J * maple ~maple@strace-nat0.ua-online.net 1120459671 Q * _are_ Quit: bbl 1120460441 J * DaPhreak_ ~phreak@styx.xnull.de 1120460442 Q * DaPhreak Read error: Connection reset by peer 1120460665 J * sukria ~sukria@sargon.lncsa.com 1120461444 J * Doener_ ~doener@p54875C79.dip.t-dialin.net 1120461554 J * cryo ~say@212.86.243.154 1120461885 Q * Doener` Ping timeout: 480 seconds 1120463180 J * prae ~prae@ezoffice.mandriva.com 1120463370 J * gregster ~gregor@greart.de 1120463376 M * gregster good morning 1120463403 J * maple_ ~maple@strace-nat0.ua-online.net 1120463403 Q * maple Read error: Connection reset by peer 1120463419 J * [maple]_ ~maple@strace-nat0.ua-online.net 1120463419 Q * [maple] Read error: Connection reset by peer 1120464300 Q * Aiken Ping timeout: 480 seconds 1120465596 M * prae morning :) 1120466019 J * kenzzo ~x@82.78.136.86 1120466045 P * kenzzo 1120466917 N * maple_ maple 1120467579 M * DaPhreak_ lo prae, gregster 1120467677 N * DaPhreak_ DaPhreak 1120467953 J * comdata ~mertins@Daf24.d.pppool.de 1120468055 M * comdata moin 1120468096 M * comdata Doener_: are you awake? 1120468314 M * prae DaPhreak: morning :p 1120468431 M * comdata anybody knowing what to do when I get: chbind: vc_set_ipv4root(): Function not implemented 1120469450 M * DaPhreak comdata: could you do a zcat /proc/config.gz | grep "CONFIG_VSERVER_NGNET" and tell me/us what it returns (well depends if you enabled the /proc/config stuff in your kernel config) ? 1120469860 M * comdata DaPhreak: # CONFIG_VSERVER_NGNET is not set 1120469906 M * DaPhreak which is good .. 1120469944 M * comdata curr I can't build the util-vserver tools 1120469965 M * comdata In file included from /usr/include/linux/spinlock.h:16, 1120469965 M * comdata from ./kernel/dlimit.h:5, 1120469966 M * comdata from ./linuxvirtual.h:24, 1120469967 M * comdata from lib/syscall.c:28: 1120469970 M * comdata /usr/include/asm/processor.h: In function `__monitor': 1120469970 M * comdata /usr/include/asm/processor.h:288: error: `asm' undeclared (first use in this function) 1120469971 M * comdata /usr/include/asm/processor.h:288: error: (Each undeclared identifier is reported only once 1120469971 M * comdata /usr/include/asm/processor.h:288: error: for each function it appears in.) 1120469971 M * comdata /usr/include/asm/processor.h:288: error: parse error before "volatile" 1120469972 M * comdata /usr/include/asm/processor.h: In function `__mwait': 1120469972 M * comdata /usr/include/asm/processor.h:296: error: `asm' undeclared (first use in this function) 1120469973 M * comdata /usr/include/asm/processor.h:296: error: parse error before "volatile" 1120469974 M * comdata /usr/include/asm/processor.h: In function `prefetch': 1120469975 M * comdata /usr/include/asm/processor.h:665: error: `asm' undeclared (first use in this function) 1120469975 M * comdata /usr/include/asm/processor.h:665: error: parse error before "volatile" 1120469977 M * comdata /usr/include/asm/processor.h: In function `prefetchw': 1120469977 M * comdata /usr/include/asm/processor.h:679: error: `asm' undeclared (first use in this function) 1120469979 M * comdata /usr/include/asm/processor.h:679: error: parse error before "volatile" 1120469979 M * comdata In file included from /usr/include/linux/spinlock.h:43, 1120469981 M * comdata from ./kernel/dlimit.h:5, 1120469981 M * comdata from ./linuxvirtual.h:24, 1120469983 M * comdata from lib/syscall.c:28: 1120469983 M * comdata /usr/include/asm/spinlock.h: In function `_raw_read_lock': 1120469985 M * comdata /usr/include/asm/spinlock.h:217: error: `asm' undeclared (first use in this function) 1120469985 M * comdata /usr/include/asm/spinlock.h:217: error: parse error before "volatile" 1120469987 M * comdata /usr/include/asm/spinlock.h: In function `_raw_write_lock': 1120469987 M * comdata /usr/include/asm/spinlock.h:225: error: parse error before "volatile" 1120469989 M * comdata In file included from ./kernel/dlimit.h:5, 1120469989 M * comdata from ./linuxvirtual.h:24, 1120469991 M * comdata from lib/syscall.c:28: 1120469991 M * comdata /usr/include/linux/spinlock.h: In function `bit_spin_lock': 1120469993 M * comdata /usr/include/linux/spinlock.h:544: error: `TIF_NEED_RESCHED' undeclared (first use in this function) 1120469993 M * comdata /usr/include/linux/spinlock.h: In function `bit_spin_trylock': 1120469995 M * comdata /usr/include/linux/spinlock.h:561: error: `TIF_NEED_RESCHED' undeclared (first use in this function) 1120469995 M * comdata /usr/include/linux/spinlock.h: In function `bit_spin_unlock': 1120469997 M * comdata /usr/include/linux/spinlock.h:579: error: `TIF_NEED_RESCHED' undeclared (first use in this function) 1120469997 M * comdata In file included from lib/syscall-wrap.h:35, 1120469999 M * comdata from lib/vserver-internal.h:33, 1120469999 M * comdata from lib/syscall.c:33: 1120470001 M * comdata lib/syscall-alternative.h:1041:1: warning: "__syscall_return" redefined 1120470001 M * comdata In file included from lib/vserver-internal.h:25, 1120470003 M * comdata from lib/syscall.c:33: 1120470003 M * comdata /usr/include/asm/unistd.h:304:1: warning: this is the location of the previous definition 1120470005 M * comdata In file included from lib/syscall.c:33: 1120470005 M * comdata lib/vserver-internal.h: At top level: 1120470007 M * comdata lib/vserver-internal.h:39: warning: invalid storage class for function `vc_noop0' 1120470007 M * comdata lib/vserver-internal.h:239: warning: invalid storage class for function `vserver' 1120470009 M * comdata In file included from lib/syscall-compat.hc:24, 1120470009 M * comdata from lib/syscall.c:36: 1120470011 M * comdata lib/safechroot-internal.hc:44: warning: invalid storage class for function `vc_tell_unsafe_chroot' 1120470011 M * comdata In file included from lib/syscall.c:36: 1120470013 M * comdata lib/syscall-compat.hc:31: warning: invalid storage class for function `vc_new_s_context_compat' 1120470013 M * comdata lib/syscall-compat.hc:41: warning: invalid storage class for function `vc_set_ipv4root_compat' 1120470015 M * comdata make[2]: *** [lib/lib_libvserver_a-syscall.o] Error 1 1120470015 M * comdata make[2]: Leaving directory `/root/compile/util-vserver-0.30.207' 1120470017 M * comdata make[1]: *** [all-recursive] Error 1 1120470017 M * comdata make[1]: Leaving directory `/root/compile/util-vserver-0.30.207' 1120470019 M * comdata make: *** [all] Error 2 1120470036 M * comdata I did replace all kernel-files in /usr/include with the patched vserver ones 1120470037 M * TheSeer next time, use a pastebin website 1120470055 M * comdata TheSeer: sorry 1120470259 M * DaPhreak hmm .. well the 207, are fine here and i've only heard from some persons, while using older (which means 2.4) headers, that the utils don't work/build 1120470363 M * DaPhreak comdata: which kernel are you running ? 1120470732 Q * comdata Ping timeout: 480 seconds 1120471965 J * rs ~rs@ATuileries-153-1-30-20.w82-123.abo.wanadoo.fr 1120472273 Q * sukria Quit: see you 1120474526 J * _are_ ~are@gateway-dsl.lihas.de 1120474564 M * _are_ hi 1120474708 M * DaPhreak lo _are_ 1120474772 Q * rs Ping timeout: 480 seconds 1120475760 J * eXplasm explasm@p549F799F.dip.t-dialin.net 1120475760 Q * eXplasm2 Read error: Connection reset by peer 1120475883 J * nox_ ~nox@noxlux.de 1120475895 Q * nox_ Quit: 1120476643 Q * Vudumen Ping timeout: 480 seconds 1120477505 J * Vudumen vudumen@perverz.hu 1120479696 J * sukria ~sukria@213.223.184.194 1120480155 J * rs ~rs@ATuileries-153-1-30-20.w82-123.abo.wanadoo.fr 1120480956 M * Doener_ hey folks 1120481325 M * Doener_ Hollow: hm, including the vserver kernel headers with #include <...> isn't a good idea... 1120481347 M * Doener_ that way you need to have the headers installed prior to building libvserver (which itself installs the headers...) 1120481351 M * Doener_ chicken, egg ;) 1120481481 M * Doener_ the static int rc will probably cause problems if we ever decide to have some convenience functions that call over functions in the same file 1120481680 M * DaPhreak lo Doener_ ;) 1120481698 M * Doener_ hey DaPhreak 1120481825 M * DaPhreak nah, bash is getting me nuts 1120481955 M * DaCa-fi DaPhreak: its designed for it :p 1120481970 M * DaPhreak yeah .. that's what i already thought ;) 1120482087 M * DaPhreak NAME="foo/"; echo ${NAME##*/} should give me an result of "foo" (without the trailing /) .. but it doesn't ... 1120482334 J * baggins baggins@kenny.mimuw.edu.pl 1120482398 M * kestrel you want ${NAME%/*} 1120482432 M * baggins hi, does anyone use "Persistent Inode Context Tagging" on amd64? 1120482520 M * Doener_ DaPhreak: NAME="foo/"; echo ${NAME%/} 1120482534 M * Doener_ ## is prefix, % is suffix 1120482539 M * Doener_ s/%/%/ 1120483204 M * DaPhreak ah .. thanks ;) well it worked until i touched it again ;) 1120483329 M * Doener_ *sigh* why does i have to be 30+° C outside today? I'll get a mud pack in 20 minutes :( 1120483335 M * DaPhreak whoooa ... sf-cvs is down *grml* 1120483421 M * DaPhreak thanks kestrel, Doener_ ;) 1120483460 M * DaPhreak so i can get rid of NAME="`echo ${opt_val} | sed 's:/$::'`" 1120483523 M * Doener_ hm, i didn't notice that kestrel already answered... oops :) 1120483528 M * DaPhreak heh ;) 1120483718 M * kestrel :) 1120485703 Q * eXplasm Remote host closed the connection 1120486082 Q * Vudumen Ping timeout: 480 seconds 1120486296 N * _mountie mountie 1120486505 J * Vudumen vudumen@perverz.hu 1120488012 Q * [maple]_ Ping timeout: 480 seconds 1120488012 Q * maple Ping timeout: 480 seconds 1120488393 J * [maple] ~maple@strace-nat0.ua-online.net 1120488697 J * maple ~maple@strace-nat0.ua-online.net 1120489058 N * Doener_ Doener 1120489645 J * eXplasm explasm@p549F799F.dip.t-dialin.net 1120489659 M * Doener DaPhreak: here's one for you ;) 1120489661 M * Doener One thing I have to say is that I've learnt a lot more shell tricks. 1120489662 M * Doener Now I'll just have to unlearn them, so that I won't have nightmares. 1120489662 M * Doener --Linus 1120489677 M * DaPhreak heh ;) 1120489681 M * DaPhreak yeah, thanks Doener *g* 1120489894 M * DaPhreak hopefully I won't get nightmares from those shelltricks ;) 1120489951 M * DaPhreak well maybe some /. ambushing me ;P 1120490347 Q * maple Read error: Connection reset by peer 1120491106 N * Bertl_zZ Bertl_oO 1120491127 Q * eXplasm Remote host closed the connection 1120492562 J * tbenita ~tbenita@atreal.net2.nerim.net 1120492615 M * tbenita Hi, I didn't find how to have a vserver that doesn't start automatically on boot, with new config. Any idea ? 1120492935 M * daniel_hozac echo "default" > /etc/vservers//apps/init/mark 1120492944 M * daniel_hozac and make sure the vservers-default init script is ran on boot. 1120493032 M * tbenita ok thanks daniel_hozac ! 1120493123 M * daniel_hozac oh sorry, misread. that'll make it start on boot. 1120493148 M * tbenita np daniel_hozac I did a translation and will remove default ;) 1120493202 M * tbenita on a host that has 2 network cards, one is the default gw to the internet. I can't get the vservers that have a private IP on the other NIC access to the internet. Any idea ? 1120493247 M * tbenita On another host, it seems to work fine, but both kernel (all 2.6) and firewalling are different. 1120493282 M * tbenita But I can't find what does this difference.. 1120493731 M * tbenita in order to benefit the default route, _must_ I add a network address mapped to the NIC used for the default route ? 1120493830 M * Zoiah tbenita: are you using SNAT or MASQUERADING on the box where it doesn't work? 1120493839 M * tbenita no 1120493858 M * Zoiah tbenita: then it will send packets to the internet with the internal IP of the vservers. 1120493930 N * Bertl_oO Bertl 1120493944 M * Bertl evening folks! 1120493960 M * tbenita sorry Zoiah I added the masquerading on the output card, and also ip_forward to 1, but the packet isn't sent to the output card... 1120493965 M * tbenita hi Bertl 1120493965 M * Vudumen hi Bertl :) how are you? :) 1120493982 M * Bertl fine thanks! 1120493993 M * Zoiah tbenita: do you see the packets in tcpdump? 1120494017 M * Doener evening Bertl! 1120494026 M * Bertl tbenita: nothing is forwarded in that case ... 1120494040 M * Bertl tbenita: add a POSTROUTING SNAT rule to your public ip 1120494055 M * Bertl evening Doener! 1120494187 M * Doener Bertl: http://pastebin.ca/16851 -- any idea? i don't understand what's going on... 1120494202 M * tbenita thx Bertl I try 1120494282 M * Bertl Doener: uint32_t is defined in stdint.h, no? 1120494322 M * Bertl (as it is a C99 type) 1120494346 M * Doener hm, it works for all the lib files, but not for the tools with ... but yes, that works, thanks :) 1120494359 M * Doener (i.e. i have both in there now...) 1120494444 M * Bertl tbenita, Doener: yw :) 1120494710 M * tbenita it works Bertl thx a lot ! 1120494783 M * baggins is there someone using vservers on amd64? 1120494797 Q * rs Ping timeout: 480 seconds 1120494824 M * Bertl baggins: yes a bunch of folks do, actually ... 1120494852 M * baggins Bertl: I seem to be having UID/GID problems 1120494870 M * Bertl care to give some details? 1120494876 M * baggins [root@duch root]# ls -ld /vservers/test/media/ 1120494876 M * baggins drwxrwxr-x 4 root 4278206271 31 Jul 4 18:23 /vservers/test/media/ 1120494903 M * baggins it should be: 1120494905 M * baggins drwxrwxr-x 2 root disk 0 Jan 8 15:14 /media 1120494911 M * Bertl kernel version, tools version, filesystem and xid tagging? 1120494945 M * baggins 2.6.11.12smp, util-vserver-0.30.207, XFS, UGID24 1120494965 M * baggins the same effect is with xid tagging disabled :o 1120495009 M * Bertl which context id does your guest use? 1120495044 M * baggins that was with system-assigned, with context=200 this gid is 16777214 1120495047 Q * _are_ Ping timeout: 480 seconds 1120495079 M * Bertl interesting ... 1120495093 M * Bertl Vudumen: is the sun shining again? 1120495111 M * Vudumen Bertl: not yet :( sorry :( 1120495139 J * Pazzo ~Pazzo@host130-250.pool8172.interbusiness.it 1120495148 M * Pazzo hi guys! 1120495155 M * Doener welcome Pazzo 1120495158 M * Pazzo hey, rc5 :-) 1120495161 M * Pazzo hi Doener! 1120495164 M * Bertl baggins: okay, could you give the testme.sh a spin, and upload the output for me (pastebin.com or so)? 1120495167 M * Pazzo everything fine? 1120495170 M * Bertl hey Pazzo! 1120495171 M * Pazzo Bertl! 1120495212 M * Doener sure, just that the sun should've been shining on my birthday and not today when i had a mud pack :/ 1120495226 A * Doener was sweating like hell... 1120495227 M * Pazzo :-) 1120495230 M * Pazzo *gg* 1120495231 M * Bertl Doener: ah, details :) 1120495242 M * _ag_ Bertl: The ng9.5 code needs to be ported to 2.0-rc5, because netfilter stuff has incredibly change over the transition 2.6.11 -> 2.6.12. Am I wrong? 1120495255 A * Pazzo opens a bottle of wine 1120495265 M * Bertl _ag_: yes, you are wrong, it was broken before, it needs to be rewritten :) 1120495268 M * Pazzo a little bit late but nonetheless: happy birthday Doener! 1120495274 M * Doener heh, thanks :) 1120495293 M * Pazzo any important changes from rc4 to rc5? 1120495300 M * Doener you still beat my older sister ;) 1120495313 M * Pazzo :) 1120495355 M * Pazzo hey, has someone of you guys some experience with SPEC benchmarks? 1120495419 M * Bertl yes, they suck, why? 1120495420 M * baggins Bertl: any specific options? becouse testme.sh says succeeded on all run tests 1120495433 M * Bertl no, just upload the output ... 1120495436 M * _ag_ Bertl: i suppose the code forked from shaper.c in vnet.c is ok, though :) 1120495451 M * Bertl _ag_: a lot of the existing code is probably fine ... 1120495464 M * Bertl I'll attack it right after the 2.0 release 1120495620 M * Bertl another OS X data integrity secret uncovered: ... UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY, next command, fsck -f -y :) 1120495680 M * tbenita Bertl, Zoiah , I found that a MASQUERADE rule was previous to the SNAT rule ; that was the issue ! thanks for you help ! 1120495687 M * baggins Bertl: uploaded to pastebin.com 1120495706 M * Bertl tx 1120495762 M * Bertl baggins: hmm, interesting, where is that PLD kernel from? 1120495822 M * Pazzo re 1120495890 M * baggins Bertl: PLD distro, www.pld-linux.org 1120495932 M * baggins Bertl: vserver patch 2.0rc4 1120495976 M * Bertl yeah, just wondered why they 'hide'/remove the version 1120495980 M * Pazzo Bertl: I have to reach a SPECint_rate_base2000 value of 28 on a dual xeon 3.0 GHz (2MB Cache, Supermicro board) using windows server 2003 standard with all available patches 1120495998 M * Bertl Pazzo: congrats! 1120496016 M * Pazzo Bertl: I have to reach it, didn't do so 1120496028 M * Bertl yeah, that's why I said, congrats! 1120496046 M * Pazzo problem is that without commercial libraries (smartheap...) you aren't able to do so 1120496058 M * Doener for sure you'll have a lot of fun :) 1120496125 M * Bertl baggins: but you did enable xid tagging on your filesystem, yes? 1120496175 M * micah new debian vserver kernel patch uploaded (updates patch against kernel-source-2.6.8 debian version -16 and fixes extended fs. attribute problem) 1120496179 M * Pazzo fortunately it's not my job to do so, I just have had to test some linux distro on this servers - but this poor windows boys are somewhat clueless now as they realized what they have to do and as they are my friends (now that they're using at least firefox :) I would like to help them... 1120496238 M * Pazzo I have no idea where to get this stupid libraries as MicroQuill doesn't even have an online shop and phoning to USA makes no sense today :( 1120496238 M * Bertl micah: did you read my comment about the FOR-2.0 patches last time? 1120496274 M * baggins Bertl: ah, not. but: XFS: unknown mount option [tagxid]. 1120496288 M * Bertl Pazzo: well, last time I did install windows libraries it was for win3.0 IIRC ... 1120496296 M * Pazzo *gg* 1120496367 M * baggins Bertl: XFS needs patching to understand it? 1120496367 M * Pazzo I also offered them to run the benchmark on a 64bit linux of their choice but they're forced to prove this benchmark results on windows 1120496367 M * micah Bertl: you mean about porting vs2.0 to the debian kernel version by applying each individual diff one at a time? 1120496415 M * Bertl or at least including some of them ... improving the kernel? 1120496423 M * Bertl (yes that's what I meant) 1120496430 Q * prae Quit: Execute Order 69 ! 1120496440 M * micah Bertl: oh! i didn't catch the part about including some of them 1120496448 M * micah Bertl: tell me more 1120496455 M * micah (sorry I missed it before!) 1120496481 M * Bertl well, I'm not going to 'maintain' a distro kernel, but it seems you are 'trying' to do so ... so you should try to keep up, no? 1120496491 M * micah indeed 1120496510 M * micah well, i am not maintaining a whole kernel, but want to keep the patch available 1120496568 M * Bertl well, that 'just' leads to me telling folks: remove that crappy debian kernel, install the vanilla 2.0 release, no? 1120496687 M * Bertl I mean, if I have a bunch of patches, which fix issues, or extend features, and I _know_ they are missing, then the choice is easy for me ... 1120496737 M * Bertl OTOH, if you want to provide a high quality 'debian' kernel with vserver support, you have to maintain it somehow ... 1120496753 Q * sukria Quit: see you 1120496773 M * micah yes, well there are some people who do not wish to have the latest kernel, but instead use the vendor provided source (and I know at least one person who is required to at his job!) 1120496779 J * rs ~rs@ATuileries-153-1-30-20.w82-123.abo.wanadoo.fr 1120496823 M * micah so, is the purpose of the FOR patches to fix problems? 1120496843 M * Doener micah: then those folks should bug their maintainer, not Bertl ;) 1120496874 M * micah Doener: i am the maintainer of the kernel patch at this moment, that is why i want to make it better 1120496915 M * Bertl evening rs! 1120496931 M * Doener ok, somehow i got the impression that you don't intend to apply the FOR-2.0 patches... 1120496946 M * Bertl micah: the FOR-2.0 patches are a mixture of fixes, cleanups, enhancements,e tc 1120496975 M * Bertl (basically they are what was added after 1.9.x to make a stable 2.0 release) 1120497001 M * Bertl some of them will apply without any issues, others might cause you grief ... 1120497023 M * micah Doener: not at all! I definately want to keep the patch up to date and would like to work on adding the FOR pieces 1120497033 M * micah what does FOR mean btw? 1120497056 M * Bertl for means 'for' as in 'waiting for' 1120497064 M * micah so the FOR-2.0 patches are all the enhancements that make up vs2.0? Or are they changes to prepare for 2.0? 1120497101 M * Doener basically 1.9.5 + all FOR-2.0 should give you the latest 2.0 rc atm 1120497129 M * Doener (maybe it's starting from 1.9.5.x or sth. like that, dunno...) 1120497145 M * Bertl micah: it might be that I included some of them in the debian version, so you have to expect that a subset _is_ already applied 1120497363 M * micah ok 1120497386 M * micah so the FOR patches make vs2.0 rc 1120497418 M * Bertl yes, they are broken down deltas 1120497439 M * Bertl everytime we change something, we make a patch and put it there 1120497459 M * Bertl (of course this doesn't include changes in the baseline kernels) 1120497494 M * micah ok, so I will take the debian 2.6.11 kernel and apply the vs1.9.5 patch and then begin to apply the FOR patches 1120497602 M * Bertl yeah, take the latest patch/fix I did, apply it to the sources, then check for every FOR-2.0 patch, if it is already there, and if you can incorporate it without a) breaking things, and b) rewriting everything ... 1120497629 M * Bertl if you have troubles incorporating a specific delta, you can always ask me ... 1120497768 M * micah ok, I will begin working on this 1120497966 Q * tbenita Quit: Leaving 1120498333 M * Bertl hmm, seems enrico is now doing it the debian way ... bug #12920 1120498406 M * micah Bertl: the patch? 1120498416 M * micah or do you mean util-vserver things? 1120498431 M * micah mmm, i did not know there was a bug tracker for the vserver project? 1120498450 M * Bertl it's util-vserver related and it was a bad joke :) 1120498460 M * micah heheh 1120498513 M * Bertl but at least he is addressing the bugs, so woho! 1120498551 M * Doener https://savannah.nongnu.org/bugs/index.php?func=detailitem&item_id=12920 1120499539 M * Greek0 Bertl: do you have time for some kernel-patch questions? 1120499570 M * Bertl sure, online from MQ or did you skip that? 1120499625 M * Greek0 no, I'm at the MQ currently? 1120499634 M * Greek0 do you intend to come too? 1120499663 M * Bertl not today, but maybe next week ... 1120499687 M * Greek0 ic 1120499717 M * Bertl so what's up, kernel wise? 1120499750 M * Greek0 well, I thought to better understand vserver and libvserver it might be a good idea to read the kernel patch 1120499778 M * Greek0 and I got some more or less intelligent and important questions 1120499793 M * Greek0 arch/i386/kernel/cpu/proc.c <-- what are you doing there? 1120499851 M * Bertl you should use the broken down version for this purpose ... 1120499923 M * Greek0 well, I'm about 10% through, and it works quite well so far.. 1120499967 M * Bertl k, so you are referring tothe varhz part, yes? 1120499980 M * Bertl - c->loops_per_jiffy/(500000/HZ), 1120499983 M * Bertl + HZ*(c->loops_per_jiffy >> 3)/62500, 1120499985 M * Bertl ? 1120499989 M * Greek0 yes 1120500004 M * Bertl it is a rewrite to allow higher HZ rates 1120500022 M * Bertl it doesn't change the results for existing HZ values (1000) 1120500052 M * Greek0 just to prevent overflow or ZeroDivision? 1120500058 M * Bertl yep 1120500102 M * Greek0 ok 1120500106 J * AprilDL___ ~chatzilla@ip68-9-200-247.ri.ri.cox.net 1120500121 M * Greek0 vroot.c:220 (( memset(vr, 0, sizeof(*vr)); )) 1120500171 M * Greek0 you already do memset(vroot_dev, 0, ..) above, so vr should already be zeroed here anyway, or did I miss something? 1120500214 M * Bertl ad varhz: http://www.13thfloor.at/vserver/d_rel26/v1.9.5/split-2.6.11.5-vs1.9.5/26_2.6.11.5_varhz.diff.hl 1120500348 M * Greek0 where can I find the split up patches? 1120500355 Q * AprilDL Ping timeout: 480 seconds 1120500365 M * Greek0 or how to use your online diff viewer? 1120500378 M * Bertl ad vroot: yes that seems to be superfluous initialization, which results from vroot_dev originally being a pointer array 1120500404 M * Greek0 k 1120500411 M * Bertl but it is completely uncritical, as it happens only on kernel startup (once) 1120500419 M * Greek0 mm 1120500449 M * Bertl http://vserver.13thfloor.at/Experimental/del-vs2.0-rc3/ 1120500456 M * Bertl (last public splitup) 1120500485 M * Greek0 __vx_info_map_pid, the if (pid == 0) check is also unnecessary afaics, since the case is covered by the later return pid; anyway. 1120500503 M * Bertl baggins: it should be supported in vs2.0-rc4/5 1120500528 M * Bertl Greek0: sec, checking ... 1120500551 M * Greek0 hmm 1120500566 M * Greek0 oh, except vxi->vx_initpid == 0, my fault 1120500571 M * Greek0 donno if that case is possible 1120500583 M * Bertl that is the default :) 1120500616 M * Bertl i.e. if no init process was declared, then it is 0 1120500620 M * baggins Bertl: I found a patch in FOR-2.0, but It's only half of what XFS needs 1120500635 M * Greek0 mm, sorry, missed that at the first pass 1120500658 M * Bertl baggins: would it be possible to try with a vanilla kernel + vs2.0-rc5 patch, and with xfs as well as ext3? 1120500686 M * Bertl (just to verify that it is an xfs issue) 1120500716 M * baggins Bertl: not now unfortunatelly :( maybe when $VENDOR sends us new server 1120500742 M * baggins Bertl: the funny thing is I don't have this problem on i686 1120500755 M * Greek0 I was also wondering a bit about the 20:12 < baggins> Bertl: the funny thing is I don't have this problem on i686 1120500760 M * Bertl well, there could be 64bit issues ... 1120500762 M * Greek0 arrr 1120500763 M * Greek0 sorry 1120500782 M * Greek0 I was also wondering a bit about the (pid == vxi->vx_initpid) case in __vx_info_rmap_pid 1120500860 M * Bertl hmm? 1120500892 M * Greek0 you return -1 then, if I parsed that correctly (~0U) 1120500948 M * Bertl yes, basically 1120500965 M * Greek0 can that case ever happen actually, unless you have a kernel bug? 1120501021 M * Bertl yes, you could for example specify the 'real' pid of the init 1120501067 M * Bertl host: init(1), guest_init(10), guest: guest_init(1), pid=10:evil 1120501203 M * Bertl but an interesting detail is, that vx_rmap_tgid() seems now unused :) 1120501274 M * Bertl (means a patch to remove it might be welcome :) 1120501365 M * Greek0 ok, off for a while 1120501722 M * Bertl k, cya 1120501878 M * Doener Bertl: do all vc's have the xid == -1 means current semantics? 1120501941 M * Bertl hmm ... most should, I guess 1120501950 M * Bertl (but it's not standartized atm) 1120501953 M * Doener ok 1120501966 Q * rs Quit: rs 1120501973 M * Doener Hollow: made quite a few functions in libvserver fail if xid <= 1 1120501976 M * Doener thus i ask ;) 1120502039 M * Bertl hmm, for wrappers the lib should not introduce 'new' policy IMHO 1120502057 A * Doener agrees 1120502072 M * Bertl i.e it should make the kernel syscall, and evaluate the return code 1120502083 M * Hollow Doener: no time atm, i'll read the backlog 1120502088 M * Doener np 1120502109 M * Doener I'm fixing libvserver atm to work with the changed prototypes/functions 1120502117 M * Bertl baggins: so any things we could test? 1120502121 M * Hollow ok 1120502464 M * DaPhreak lo ;) 1120502502 M * Bertl evening DaPhreak! 1120502535 M * DaPhreak lo Bertl :) 1120502553 M * DaPhreak yay finally back home .. (damn back treatment) ;) 1120502791 M * Bertl you survived again! great! 1120502823 M * DaPhreak yeah ;) 1120502927 M * DaPhreak and how thinks are going ? 1120503052 M * Bertl quite fine ... 1120503080 M * DaPhreak any sign of live of enrico ? 1120503267 M * Doener 19:32:18 Bertl hmm, seems enrico is now doing it the debian way ... bug #12920 1120503272 M * Doener 19:35:17 Bertl but at least he is addressing the bugs, so woho! 1120503272 M * Doener 19:35:56 Doener https://savannah.nongnu.org/bugs/index.php?func=detailitem&item_id=12920 1120503547 M * DaPhreak thanks Doener :) 1120503557 M * daniel_hozac he is pretty active on the fedora lists as well. 1120503578 J * _are_ ~are@dsl-084-056-149-128.arcor-ip.net 1120503624 M * Bertl evening _are_! 1120503694 J * godmode0 debian-tor@seppia.noreply.org 1120503713 M * Bertl welcome godmode0! 1120503927 P * godmode0 1120504092 M * _are_ hi 1120504500 M * Doener well, godmode0's hostmask said everything ;) 1120504768 M * _are_ ;) 1120505719 M * Bertl okay, off for now ... back later ... 1120505725 N * Bertl Bertl_oO 1120506527 Q * Zoiah Ping timeout: 480 seconds 1120508410 Q * Hollow Ping timeout: 480 seconds 1120509636 J * prae ~prae@sherpadown.net 1120510124 J * Aiken ~james@tooax6-121.dialup.optusnet.com.au 1120510715 Q * monrad Quit: Leaving 1120510804 J * monrad ~monrad@213083190130.sonofon.dk 1120512231 J * are|lunch ~are@dsl-084-056-129-055.arcor-ip.net 1120512610 Q * _are_ Ping timeout: 480 seconds 1120512812 J * Zoiah Zoiah@matryoshka.zoiah.net 1120513119 N * Bertl_oO Bertl 1120513125 M * Bertl back now ... 1120513357 M * Doener wb Bertl 1120513362 M * Aiken Bertl where are most of the vserver people? 1120513371 M * Doener geographically? 1120513378 M * Aiken by the way you go to bed as our day is starting I am wondering Europe somewhere 1120513404 M * Bertl hmm, hard to tell ... 1120513413 M * Bertl 'our' means? 1120513419 M * Doener hm, actually most of the time, Bertl's personal timezone isn't aligned to his home town's timezone ;) 1120513433 M * Aiken east coast australia 1120513436 M * Aiken ah 1120513450 Q * AprilDL___ Read error: Connection reset by peer 1120513453 J * AprilDL____ ~chatzilla@ip68-9-200-247.ri.ri.cox.net 1120513454 N * AprilDL____ AprilDL 1120513499 M * Bertl wb AprilDL! 1120513512 M * Bertl Aiken: and with 'vserver people' you refer to the developers? 1120513541 M * Aiken you, Doener mainly 1120513548 M * Aiken just curious 1120513561 A * Doener is from Germany 1120513577 M * Bertl well, Enrico too, and I'm from Austria 1120513590 M * Doener http://linux-vserver.org/Hall+of+Fame 1120513594 M * AprilDL thanks BertL 1120513673 M * Aiken that is the general area I had guessed 1120513709 M * Bertl but not from the time when I go to bed and return to the channel I hope? 1120513748 M * Bertl (because if so, it was a really lucky guess :) 1120513759 M * Aiken your times did not help :) 1120513834 J * rs ~rs@imhotep.rhapsodyk.net 1120513874 M * baggins Bertl: tell me what needs testing and I do as much as I can 1120513925 M * Bertl welcome rs! 1120513940 M * Doener evening rs 1120513961 M * Bertl baggins: well, I already suggested two things, what could you do in that direction? 1120514009 M * Aiken Bertl you had said there could be a work around for alpha + fakeinit, any news on that? 1120514034 M * Bertl well, I know what needs to be done, but it wasn't done yet 1120514073 M * baggins Bertl: I can try vanilla+rc5 on xfs, and maybe ext3 via loop device 1120514074 M * Aiken ah, I am happy to do any testing with it 1120514108 M * Bertl k, I can do a quick hack to verify my theory if you like to test it ... 1120514123 M * Aiken ok 1120514132 M * baggins Bertl: fortunately it's vacation time at univ so less people to anger with reboots ;) 1120514158 M * Bertl baggins: okay, sounds like what I suggested ... so go ahead :) 1120514216 M * baggins Bertl: when I get to work :) I'll get back to you tomorrow, now time to get some sleep 1120514233 M * Bertl good night then! 1120514310 M * baggins good night 1120515125 M * Bertl Aiken: do you know details about the sys_getxpid syscall? 1120515142 M * Bertl (i.e. is it documented?) 1120515155 M * Aiken don't know 1120515193 M * Bertl the thing is, it is written in assembler, and I suspect it does exactly the same as the generic sys_getppid() 1120515509 M * Bertl Aiken: okay, you got a recent kernel source tree, best in already compiled state? 1120515572 M * Doener Bertl: hm, http://lkml.org/lkml/2005/5/1/23 -- there sys_getxpid is 'compared' to sys_getpid... so maybe the 'x' is a joker, i.e. it does all the sys_get?pid things? 1120515628 M * Aiken the alpha has a 2.6.11.11-vs2.0-rc4 kernel and source tree 1120515664 M * Aiken only kernel source tree on it at the moment 1120515666 M * Bertl okay, that should be sufficient for a test 1120515691 M * Bertl Doener: http://vserver.13thfloor.at/Stuff/sys_getxpid.txt 1120515708 M * Bertl (check/compare it with sys_getppid in kernel/timer.c) 1120515738 M * Doener hm, that's the vanilla source, right? 1120515751 M * Bertl ld* are load operations, st* store 1120515780 M * Bertl (notation is right to left) 1120515790 M * Doener what's cmpeq? i'm confused by it taking 3 arguments... 1120515845 M * Bertl compare $3 with $4, result goes to $4 1120515856 M * Doener ah, separate result... 1120515891 M * Bertl there are no 'condition' flags on the alpha IIRC 1120515989 M * Bertl Aiken: okay, you know how to wield vi/emacs/YEOF ? 1120515999 M * Bertl s/YEOF/YEOC/ 1120516046 M * Aiken happy with vi, have not used the other 2 1120516082 M * Doener YEOC is probably Your Editor Of Choice ;) 1120516090 M * Bertl hehe, k, let's disable the existing sys_getxpid() 1120516132 M * Bertl arch/alpha/kernel/entry.S 1120516160 M * Bertl by just adding an additional underscore right after the xpid 1120516160 M * Aiken I have to disappear for a bit, looks like I have either a 5.4 or 10.8km walk ahead of me shortly 1120516180 M * Bertl hmm .. okay ... 1120516451 M * Aiken the sys_getxpid: line? 1120516487 M * daniel_hozac and the .end sys_getxpid line ;) 1120516499 M * Bertl .globl sys_getxpid 1120516499 M * Bertl .ent sys_getxpid 1120516499 M * Bertl sys_getxpid: 1120516507 M * Bertl .end sys_getxpid 1120516510 M * Aiken all of them 1120516517 M * Bertl all of them, just add an underscore 1120516541 M * Bertl then in kernel/timer.c 1120516558 M * Bertl around line 966 1120516576 M * Bertl #ifndef __alpha__ 1120516580 M * Bertl change that to 1120516582 M * Bertl #ifdef __alpha__ 1120516604 M * Bertl and a few lines down, the 1120516605 M * Bertl sys_getppid 1120516624 M * Bertl change that to 1120516628 M * Bertl sys_getxpid 1120516716 M * Bertl this will give some 'additional' and superfluous syscall procedures but they should not hurt 1120516726 M * Aiken done 1120516738 M * Bertl compiling should be quite fast too 1120516752 M * Bertl (only two files to recompile, plus linkage) 1120516899 M * Aiken the new kernel is smaller 1120516928 M * Bertl hmm, unusual ... 1120517180 M * Aiken http://pastebin.com/307513 1120517211 M * Bertl not too shabby for a Q&D hack :) 1120517284 M * Bertl Aiken: okay, would be great if you could test some (p)pid related tools like top/ps/pstree if they work as expected 1120517310 M * Bertl maybe compare them between original and modified kernel 1120517367 M * Bertl the only thing I could not clarify yet, is why the memory barrier (used in the smp case) does not show up in the generated assembler code 1120517435 M * Aiken I'll do that when i get back, my wife is waiting for me so I think I really should go 1120517448 M * Bertl (somehow I suspect some kind of alpha bug there) 1120517468 M * Bertl yeah, okay, is perfectly fine, have fun and thanks! 1120517550 M * Bertl Doener: any idea why the rmb() doesn't result in an 'mb' assembler statement? 1120517552 M * Aiken just started a job I wanted to run and top on the host seems to be working as it should 1120517675 M * Bertl Doener: hmm, give me a minute, it seems SMP was disabled ... 1120517736 M * Doener and that's the only thing i see... ;) 1120517769 M * Aiken something broke http://pastebin.com/307516 :( 1120517790 M * Aiken back in an hour or so 1120517962 M * Bertl hmm, maybe it really combines getpid and getppid 1120518010 M * Bertl but it looks more like it takes an argument and stores something there 1120518017 M * Bertl stq $1, 80($sp) 1120518032 M * Bertl actually that should be the tgid 1120518053 M * Bertl guess we need a sys_getxpid() manpage or so ... 1120518170 M * Bertl somehow I have the feeling the alpha folks have forgotten why they did that :) 1120518302 Q * rs Quit: rs 1120518308 M * Doener http://66.102.9.104/search?q=cache:xnsTEcYxk_wJ:www.iglu.org.il/lxr/source/arch/alpha/kernel/osf_sys.c%3Fa%3Dsparc64+osf+sys_getxpid&hl=de 1120518326 M * Doener the sys_getxpid there might tell something... 1120518348 M * Doener it stores the parent's tgid in some reg and returns the process' tgid 1120518627 M * Bertl yeah, well, doesn't tell us much about the linux case ... 1120518823 M * Doener the guy on lkml said it's for osf compatibility, so i guess it's pretty much the same 1120518915 M * Bertl well, there is no 'compatibility' with osf in this syscall range 1120518945 M * Bertl what he probably meant is, that they 'just' wanted to use a similar syscall than osf did (at the initial port) 1120518964 M * Bertl (so that the libc didn't need too much modifications) 1120519082 M * Bertl I find it very intersting that this syscall seems to be ignored by recent glibc ... 1120519108 M * Doener well, then let's take alpha/get[p]pid.S from dietlibc :) 1120519113 M * Bertl grep -r getxpid * 1120519114 M * Bertl ChangeLog.6:these syscalls in terms of getxpid/getxuid/getxgid. 1120519114 M * Bertl sysdeps/unix/sysv/linux/alpha/sysdep.h:#define __NR_getpid__NR_getxpid 1120519114 M * Bertl sysdeps/unix/sysv/linux/alpha/sysdep.h:#define SYS_getpid__NR_getxpid 1120519124 M * Bertl (this is in glibc 2.2.5 for example :) 1120519151 M * Bertl so they make it a synonym to getpid() 1120519166 M * Doener because it simply returns the process' pid 1120519179 M * Bertl yes, but how to get the ppid then? 1120519188 M * Doener and getppid calls the same syscall, but gets its result from register $26 1120519205 M * Doener ehrm, $20 1120519217 M * Bertl ahem, it's never stored there on alpha 1120519222 M * Doener .global getppid 1120519222 M * Doener getppid: 1120519222 M * Doener lda $sp,-8($sp) 1120519222 M * Doener stq $26, 0($sp) 1120519223 M * Doener jsr getpid 1120519224 M * Doener ldq $26, 0($sp) 1120519226 M * Doener lda $sp, 8($sp) 1120519228 M * Doener cmovge $0, $20, $0 1120519230 M * Doener ret 1120519247 M * Bertl look at the assembler code I uploaded 1120519265 M * Doener got the entry.S opened in vim 1120519276 M * Bertl do you see any $20 there? 1120519277 M * Doener maybe i misread the cmovge? 1120519306 M * Bertl we are obviously looking at different code ... 1120519315 M * Doener the above is from the dietlibc 1120519319 M * Bertl http://vserver.13thfloor.at/Stuff/sys_getxpid.txt 1120519325 M * Bertl this is the kernel side ... 1120519325 M * Doener there's no cmovge in the kernel source 1120519376 M * Doener i'm trying to get an idea by looking at the way dietlibc deals with it 1120519382 M * Doener sorry for the confusion 1120519383 M * Bertl I agree that the kernel is storing something on the stack 1120519402 M * Doener ah, $sp is stack pointer, right? 1120519406 M * Bertl and IMHO it is the TGID 1120519411 M * Bertl yes 1120519423 M * Doener ok, yes, it stores the TGID on the stack then 1120519427 M * Bertl the TGID of the parent to be precise 1120519442 M * Bertl and it returns the processes TGID 1120519452 M * Doener agreed 1120519465 M * Bertl now that requires that the syscall has at least one argument 1120519476 M * Doener why? 1120519479 M * Bertl (the place to store the stuff to) 1120519493 Q * prae Quit: Pwet 1120519510 M * Doener the above code from dietlibc reserves some space on the stack AFAICT 1120519551 M * Greek0 re 1120519558 M * Bertl wb Greek0! 1120519645 M * Bertl or maybe that's the reason why a) it is coded in assembler and b) nobody (especially not the arch maintainer(s)) wants to talk about it :) 1120519647 M * Greek0 Bertl: following up on my last sentence, can there ever be a pid 10 in the guest? IMHO not, or am I wrong there? 1120519665 M * Greek0 host: 10 -> 1 in guest. 1120519673 M * Bertl Greek0: no, there can not be one, but it might be 'requested' no? 1120519681 M * Greek0 this means that pid 10 is globally taken, and it's aliased .. 1120519682 M * Greek0 ah 1120519712 M * Bertl (for example by a simple ls /proc/10) 1120519753 M * Doener dietlibc: it decreases the $sp by 8, stores $26 there (whyever...), calls getpid (==getxpid), load the value on the stack into $26, restore the $sp and then that strange cmovge with $20 which i don't understand... 1120519788 M * Bertl IMHO that is just protecting against the clobbering ... 1120519810 M * Bertl make space on the stack, save the clobbered register, call, restore register 1120519863 M * Bertl the cmovge might be the syscall return code check ... 1120519912 M * Doener hm, maybe 80($sp) in the syscall matches $20 in the saved userspace registers? 1120519933 M * Doener that would then match the osf getxpid implementation google came up with 1120519937 M * Bertl possible, even likey ... 1120519942 M * Bertl *likely 1120519957 M * Doener which means that the result of getpid is returned in $0 and the one from getppid in $20 1120519960 M * Bertl but that would only feed my theroy ... 1120520009 M * Doener no argument required... 1120520027 M * Bertl severely broken interface :) 1120520056 M * Doener i agree on that one... it's not high level language friendly at all 1120520180 M * Doener stq $20, 80($sp); \ 1120520183 M * Doener yep, they match 1120520559 M * Bertl okay, as I do not intend to code the initpid lookup in assembler too, we will make a wrapper for this strange calling/return scheme ... 1120520940 M * Bertl okay, is delayed until I find some time to dig into alpha assembler 1120521114 M * Greek0 Bertl: kudos for your kernel patch, it's surprisingly clean and easy to read.. 1120521222 M * Bertl thanks! was quite some work to make it that clean ... 1120521286 M * Bertl (and we are still cleaning up here and there :) 1120521370 M * Greek0 do you have time for another round of questions? 1120521378 M * Bertl sure 1120521450 M * Greek0 what are you doing in fs/devpts/inode.c? 1120521457 M * Bertl I always appreciate folks reviewing my patches and asking some questions ... 1120521501 M * Bertl first, I'm moving the magics where they belong 1120521518 M * Bertl then I add an additional 'permission' check (for devpts isolation) 1120521547 M * Bertl then we have to reintroduce the devpts_readdir (because the generic one isn't generic enough) 1120521558 M * Greek0 I'm about 20% through, and the devpts stuff (well, the readdir stuff there) is the only major code block I don't get