1114214526 M * Doener Bertl: "vx_state is too early", i think i didn't get that one... did you mean vs_context_state()? if not, how can a variable be early? 1114215097 M * Bertl ah, okay, yes, I meant vs_context_state 1114215194 M * Doener ok... we can't use the vxi in the wait function, as it would keep the context alive... and once the context is freed, we don't have the vx_state any longer nor do we have the vx_exit wait queue... 1114215254 M * Doener assuming that only few tasks will ever wait for a context to exit, do you think it is fine to use a shared global wait_queue? or will we be ran over by a thundering herd? 1114215347 M * Bertl hmm ... sec 1114215561 M * Bertl I'd say we do not care about the real disposal ... which could take much longer, we are interested in the point where the context is unhashed 1114215580 M * Bertl (i.e. where it disappears from global scope) 1114215590 M * Bertl so looking at unhash_vx_info() 1114215623 M * Bertl I'd say it should happen right after the __unhash_vx_info() 1114215676 M * Bertl and as the vc_wait_exit() will keep a reference to the context 1114215689 M * Bertl it should not be released before that function returns ... 1114215722 M * Bertl so the context waitqueue should be woken after __unhash_vx_info() outside the hash lock of course 1114215726 M * Doener ok, that sounds easier... 1114215752 M * Bertl add a __wakeup_vx_info() there 1114215765 M * Bertl (or whatever name you consider appropriate ;) 1114215781 M * Bertl and put that __wake* into kernel/vserver/signal.c 1114215827 M * Doener btw, my .12-rc2 port has an uptime of 12 days now :) 1114215845 M * Doener 2.6.12-rc2-vs1.9.5.12 that is 1114215859 M * Bertl excellent! 1114215875 M * Bertl we will need it soon .. (once .12 is released) 1114215894 J * sim0n ~simon@212.76.255.43 1114215901 M * sim0n hi all 1114215903 M * sim0n hi bertl 1114215906 M * Doener welcome sim0n 1114215934 M * Bertl hey sim0n! 1114215998 M * sim0n i got an OT question... 1114216026 M * sim0n i have an old binary... OS: Linux, ABI: 2.0.0... and can't get it to work on a recent system 1114216063 M * sim0n when i set "LD_ASSUME_KERNEL=2.4.5" before calling it, it seems to work, but it doesn't use the command line parameters passed to it 1114216089 M * sim0n if i don't pass that ld_assume... i get a relocation error: symbol errno, version GLIBC_2.0 not defined in file libc.so.6 with link time reference 1114216111 M * sim0n does anybody have an idea how i can get that binary running correctly ? is it even possible ? 1114216141 M * sim0n (if i set ld_assume, it looks like it just ignored the params passed to it... unfortunately it's a closed-source software :-( ) 1114216426 M * sim0n anybody ? 1114216454 M * Bertl hmm, you should probably put in old libc5 or earlier libraries 1114216488 M * sim0n i tried that, but it tried to use libc.so.6 1114216508 M * Bertl so it has 2.0.0 abi but uses libc6? 1114216510 M * sim0n the prog uses libc6 1114216515 M * sim0n yes...weird 1114216525 M * Bertl where do you get such a beast? 1114216526 M * sim0n ldd request_ 1114216526 M * sim0n linux-gate.so.1 => (0xffffe000) 1114216526 M * sim0n libc.so.6 => /lib/tls/libc.so.6 (0xb7ecb000) 1114216526 M * sim0n /lib/ld-linux.so.2 (0xb7feb000) 1114216554 M * Bertl linux-gate even more interesting ... 1114216570 M * sim0n it's an online payment cgi ... a customer has to use it... the bank hasn't released a newer version so far 1114216605 M * Bertl and for what system is it written for? 1114216612 M * Bertl (i.e. where did it work?) 1114216644 M * sim0n dunno... i have no specs of it... it works on older linuxes... dunno redhat 6 or so :-) 1114216669 M * Bertl do you have such a distro? 1114216675 M * sim0n no 1114216688 M * sim0n is there no way to force it to use a compatibility libc5 ? 1114216701 M * sim0n i tried ld_preload .. didn't work.. 1114216711 M * Bertl well, of course, you could write your own preload wrapper to do so 1114216730 M * Bertl but I guess it would be a lot easier to get the 'right' libs and chroot the entire beast 1114216766 M * sim0n chrooting is far from simple in this case :-)... 1114216774 M * sim0n how would a wrapper look like ? 1114216860 M * sim0n waht i'm wondering too is, why ldd tells me the binary wantd libc6, if it's such an old program.. hmm 1114217038 M * Bertl well, ldd shows you what the binary get's not what it wants ;) 1114217135 M * sim0n k... but then if i set ld_preload=/path/libc.so.5 ... it should use libc5 ? 1114217221 M * Bertl no, it will load the libc5 before any other ... 1114217236 M * Bertl so terminals there will be used ... 1114217369 Q * nayco Quit: Bonne nuit ! 1114217463 M * sim0n ld_preload works... it complains about "Symbol `sys_errlist' has different size in shared object".... but anyway... it still ignores command line arguments passed to it :-( 1114217512 M * Bertl well, probably it needs some more stuff, check with strace 1114217913 M * sim0n strace doesnt give any usable info 1114217938 M * Bertl well, then gdb ;) 1114218031 M * sim0n wait... it does work now... 1114218033 M * sim0n hmmm 1114218042 M * sim0n it doesn't ignore the params anymore 1114218097 M * Bertl excellent ... 1114218122 M * sim0n i currently have a wrapper script which looks like this: 1114218129 M * sim0n #!/bin/sh 1114218129 M * sim0n cd /home/sites/tuning-land.com/www/web/cgi-bin/e-transactions/payment/bin 1114218129 M * sim0n env LD_ASSUME_KERNEL=2.4.5 1114218148 M * sim0n but it doesn't pass the params... 1114218157 M * sim0n do you know how i can pass the params ? 1114218180 M * daniel_hozac $* 1114218217 M * sim0n great! thnx 1114218226 M * sim0n it really seems to work now ;-) 1114218296 M * sim0n argf...so it was my script not passing the args... ;-) 1114218324 M * Bertl hehe 1114218373 M * Bertl Doener: we should also look for stuff like this: http://vserver.13thfloor.at/Experimental/FOR-2.0/delta-vroot-fix01.diff 1114218391 M * Bertl (i.e. if compiled as module, it requires __iget) 1114218644 M * Doener k 1114219040 Q * sim0n Quit: Leaving 1114219724 M * Doener http://vserver.13thfloor.at/Stuff/SYSCALL/vserver.h 1114219734 M * Doener --> _syscall3(int, vserver, uint32_t, cmd, uint32_t, id, void *, data); 1114219756 M * Doener are those too many commas, or is there something i don't know about C function headers? 1114219767 M * Doener s/those/there/ 1114219789 M * Doener (the latter is easily possibly, thus i ask ;) 1114219821 M * Bertl that's something you don't know about CPP macros ;) 1114219864 M * Bertl but it's easily demystified ... 1114219874 M * Bertl look at the syscall.h file 1114219887 M * Bertl and any definition of the _syscall3 macros in there 1114220009 M * Doener ah i.c. 1114221103 M * Doener hm, yeah, first test works... 1114221451 Q * ndim Ping timeout: 480 seconds 1114221512 M * Bertl Doener: cool! 1114221660 J * ndim hun@helena.bawue.de 1114221668 M * Snow-Man Bertl: You read Enrico's note about Debian and glibc? :) 1114221758 M * Doener http://www.13thfloor.at/~doener/vserver/tools/vx_wait_test.tar.bz2 1114221769 M * Doener http://www.13thfloor.at/~doener/vserver/patches/diff-2.6.11.7-vs2.0-pre2-wait.diff 1114222151 M * Doener multiple waiting processes for one context also seem to work fine, as well as multiple processes waiting for different contexts, and a mixture as well... 1114222214 M * Doener s/(waiting) (processes)/$2 $1/ 1114222712 M * Bertl Snow-Man: yeah, but I do not consider it a good reply :/ 1114222760 M * Bertl or do you really want to restrict debian folks to debian vservers? (an thus cultivate debian images for redhat, suse, mandrake ...) 1114222806 M * Snow-Man Bertl: Well, short answer is- it doesn't affect me, so... :D 1114222826 M * Snow-Man Bertl: The other relatively simple solution is, don't use vrpm & co. 1114222851 M * Snow-Man It still seems to me like it'd be easy enough to just exec rpm instead of doing it internally... 1114222861 M * Snow-Man Which would alliviate the problem I believe... 1114222869 M * Snow-Man At least, if I understood it correctly. 1114222890 M * Bertl Doener: good work, will look at it soon ... 1114222917 M * Doener np, take your time... 1114227905 J * logger ~rs@vds.pas-mal.com 1114228990 M * Bertl Doener: 1114228991 M * Bertl +extern void __wake_vx_exit(struct vx_info *vxi); 1114229006 M * Bertl +__wakeup_vx_exit(vxi); 1114229032 M * Doener woops :) 1114229201 M * Doener http://www.13thfloor.at/~doener/vserver/patches/diff-2.6.11.7-vs2.0-pre2-wait_fixed.diff 1114229204 M * Doener there we go... 1114229330 M * Bertl okay, now it doesn't really matter as I have to cleanup the signal part anyways, but the _internal_ __wakeup* does not directly belong between the syscall commands ;) 1114229428 M * Doener true... i never managed to get such stuff in the right place... 1114229464 M * Doener and the "fixed" patch still misses the linux/vserver/signal.h include in context.c 1114229479 M * Doener so, if you're at it anyways.... ;) 1114229512 M * Bertl yup yup ... I'll clean it up ... 1114229535 M * Bertl btw, for the userspace tools, did you have a look at the vdlimit-0.03 package? 1114229560 M * Doener ah, i knew there was something i forgot... 1114229599 M * Bertl no problem, it just represents the easy way to write arch independant hack tools with almost no efford ... 1114229641 M * Bertl i.e. you 'copy' the include/vserver dir to the tools dir, add the syscall.h + vserver.h (as you did) and be done ... 1114229688 M * Doener ok. i used the kernel/ dir from util-vserver and cleaned it up as much as possible 1114229721 M * Doener plus the syscall.h+vserver.h of course... 1114229726 M * Bertl yeah, I hope that util-vserver will soon switch to using the include dir and the new syscall.h ... 1114229751 M * Bertl (should make life easier on various archs) 1114229794 M * Bertl okay, I guess I'm off to bed for tonight ... 1114229805 M * Doener sweet dreams Bertl! 1114229846 M * Bertl if you find some time (and some fun in doing so) it would be interesting to switch the reboot cycle to a wait started from the reboot-vshelper call, which restarts the vserver once it is gone ... 1114229860 M * Bertl that said ... have a good night! 1114229866 N * Bertl Bertl_zZ 1114229961 Q * maharaja Ping timeout: 480 seconds 1114230229 M * Doener i'm off to bed too... good night! 1114230231 N * Doener Doener_zZz 1114231026 J * albeiro_ albeiro@linux.gentoo.pl 1114231031 Q * albeiro Remote host closed the connection 1114231042 N * albeiro_ albeiro 1114231142 J * duckx ~Duck@dyn-83-157-153-195.ppp.tiscali.fr 1114231406 Q * DuckKing Ping timeout: 480 seconds 1114240102 J * monrad ~monrad@213083190130.sonofon.dk 1114243366 Q * monrad Quit: Leaving 1114249832 J * kjo ~krischan@p5484CF8B.dip.t-dialin.net 1114251241 J * erwan_ho ~erwan@lns-vlq-39f-81-56-133-136.adsl.proxad.net 1114252113 Q * kjo Quit: Verlassend 1114253104 J * atsab ~as@lotes.vtu.lt 1114259021 J * nayco ~nayco@82.249.35.252 1114259028 Q * nayco Quit: 1114260119 J * hws123 ~sebastian@p508CEDBD.dip.t-dialin.net 1114260121 M * hws123 hi 1114260177 M * hws123 I am trying to start my vserver but he complains the file/directory /etc/vservers/.defaults/run.rev/2 - what must this thing point to, how do I create this file? I think I deleted it accidently?! 1114260383 M * hws123 Sorry, I mean the directory/file /etc/vservers/.defaults/run.rev/2 is missing, how do I fix this? 1114260804 M * hws123 Do I need /etc/vservers/template/fstab? What must this file contain? 1114260919 M * hws123 Anyone? 1114260970 M * kalou_ Hi hws123 ! 1114260995 M * hws123 hi! 1114261011 M * hws123 Are you able to help me? 1114261051 M * kalou_ about the fstab, It depends on your startup scripts 1114261058 N * kalou_ kalou 1114261109 M * kalou With no fstab, a vserver can work :) 1114261120 M * hws123 For some bizarre reason when I start the vserver I get "save_ctxinfo: symlink("/etc/vservers/template","/etc/vservers/.defaults/run.rev/2"): No such file or directory" and the directory "/var/run/vservers.rev" does not exist.... 1114261171 M * kalou which version of the vserver-utils are you using ? 1114261211 M * hws123 util-vserver-0.30.204 on gentoo 1114261247 M * hws123 the file /etc/vservers/.defaults/run.rev is pointing to /var/run/vservers.rev which doesn't exist 1114261271 A * kalou downloading tools 0.30.204 .. :) 1114261272 M * daniel_hozac so create that directory. 1114261438 M * hws123 Ok, that works 1114261444 M * hws123 but what about the fstab file 1114261468 M * hws123 if I delete it I get an error message "context: execvp("/sbin/rc"): No such file or directory" 1114261535 M * kalou did you try with an empty fstab ? 1114261587 M * hws123 yes, that's what I did as well 1114261639 M * kalou probably your init scripts need to be fixed 1114261717 M * kalou I suppose that /sbin/rc is launched from etc/inittab inside your vserver 1114261865 Q * hws123 Quit: Leaving 1114262209 Q * erwan_ho Remote host closed the connection 1114262235 J * erwan_ho ~erwan@lns-vlq-39f-81-56-133-136.adsl.proxad.net 1114265146 N * Doener_zZz Doener 1114268271 M * DaPhreak morning Doener ;) 1114268314 M * Hollow morning 1114268333 M * Hollow DaPhreak: http://bugs.gentoo.org/show_bug.cgi?id=90122 1114268370 M * DaPhreak ah, your vserver handbook :) 1114268447 M * Hollow :) 1114268621 J * kalou_ ~kalou@ALamentin-101-1-5-114.w81-48.abo.wanadoo.fr 1114268736 Q * kalou Ping timeout: 480 seconds 1114269435 M * Doener Hollow: in section 7.e. you could add a note about the vprocunhide settings being quite reasonable 1114269518 M * Hollow Doener: Part 1 is just c&p from bertls vserver paper 1114269727 M * DaPhreak http://gentoo.home.xnull.de/doc/en/vshandbook/index.xml?part=3&chap=3 <-- is much more readable as the flower page ;) 1114269771 M * Hollow hehe 1114271288 J * knoppix_ ~knoppix@dsl-213-023-144-127.arcor-ip.net 1114272807 Q * grecea Remote host closed the connection 1114273483 J * grecea ~grecea@h-195-22-237-74.mdl.net 1114273788 N * Bertl_zZ Bertl 1114273799 M * Bertl morning folks! 1114274206 Q * erwan_ho Remote host closed the connection 1114274537 M * Doener hey Bertl! 1114274583 M * Bertl morning Doener! 1114274774 N * ntrs__ ntrs 1114275009 M * Doener ok, port to .12-rc3 should be pretty easy... let's tackle that now... 1114275031 M * Bertl excellent, I'll focus on the signla cleanup 1114275076 M * Doener after that is done, I'll see how much trouble it is to get from 1.9.5.12 to 2.0pre2 1114275427 M * Hollow morning Bertl 1114275441 M * Bertl hey Hollow! everything fine? 1114275489 M * Hollow yeah, chillin' a bit, you? 1114275545 M * Bertl chillin'? 1114275552 M * Hollow yup ;) 1114275954 M * Bertl how is 2.0-pre2 doing on gentoo? 1114276502 M * Bertl Doener: hmm, looking at the unhash_vx_info() I think we need to do something for the case where nobody is waiting for the context 1114276533 M * Doener hm... why? 1114276556 M * Bertl because if nobody is waiting, then the __unhash_vx_info() will free the context, no? 1114276591 Q * eyck Remote host closed the connection 1114276600 M * Doener hm.. lemme check that... 1114276673 M * Doener no... 1114276689 M * Doener __unhash_vx_info is called via the __release_vx_info path 1114276721 M * Doener in __release_vx_info only the task count is decremented, but the task still has the vx_info assigned and thus usecnt > 0 1114276766 M * Doener clr_vx_info is always called after release_vx_info 1114276804 M * Bertl hmm ... 1114276875 J * eyck eyck@81.219.64.71 1114276927 M * Hollow Bertl: havn't tried pre2 yet 1114276978 M * Bertl Doener: you're right ... 1114276984 M * Doener :) 1114277036 M * Bertl nevertheless, I thought about the following simplification: 1114277072 M * Bertl what if we remove the get/put from the __hash/__unhash 1114277093 M * Bertl and move it up higher to the callers ... 1114277105 M * Bertl (probably outside the hash lock too) 1114277209 M * Bertl should reduce the time the hash lock is hel without any unwanted side effects ... 1114277439 M * Doener do we need the get/put there at all? 1114277469 J * sOCiETY ~society@p549AD799.dip.t-dialin.net 1114277475 M * Bertl just thought the same ... 1114277478 M * Doener welcome sOCiETY 1114277479 M * sOCiETY hi 1114277490 M * Bertl welcome sOCiETY! 1114277551 M * Doener while we decide if the context is going to stay, we already get'ed it and have to unhash anyway if we decide to drop it after we hashed it 1114277558 M * Doener (in the creation part that is) 1114277586 M * Doener and if we unhash, it is still get'ed by one task and thus won't go away accidently 1114277668 M * Doener i'd say we can be sure that a context that is hashed won't be invalided (i.e. freed) anyway... (unless we have a bug elsewhere of course) 1114277851 M * Doener AFAICT we basically do this atm: ...->get->(lock->get->unlock)->... and ...->(lock->put->unlock)->put->... (the brackets symbolize [un]hash_vx_info) 1114278374 M * Bertl sounds good 1114278492 M * Doener so let's drop the get/put in [un]hash_vx_info, right? 1114278512 M * Bertl yup, and add a short comment comments 1114278709 M * Bertl hmm ... too many comments ;) 1114278771 M * Doener http://www.13thfloor.at/~doener/vserver/patches/patch-2.6.12-rc3-vs1.9.5.12.diff 1114278776 M * Doener currently test compiling... 1114279178 M * Doener getting to 2.0pre2 from there seems doable as well... 1114279506 M * Doener compile suceeded 1114281023 Q * sOCiETY Quit: Verlassend 1114281601 M * Doener http://www.13thfloor.at/~doener/vserver/patches/patch-2.6.12-rc3-vs2.0pre2.diff 1114281607 M * Doener test compile is running 1114282121 M * Doener suceeded as well 1114282247 M * Doener ok, testboot here we come :) 1114282250 N * Doener Doener|gone 1114282507 N * Doener|gone Doener 1114282664 M * Doener Bertl: testme.sh works on .12-rc3-vs2.0pre2... gonna do a qemu build now to test -L 1114282677 M * Bertl good! 1114283307 M * Doener ok, -L also suceeded without errors 1114283446 M * Bertl Doener: hmm, if we remove the get from the hash ... the __loc_vx_info() might return a vxi with refcnt=0 ... 1114283478 M * Doener hm... did i miss-double-check that?... 1114283498 M * Bertl well, maybe I missed something ;) 1114283515 M * Doener it calls: __hash_vx_info(get_vx_info(new)); 1114283528 M * Bertl oops, right ... 1114283543 M * Doener yeah, also had to double check ;) 1114283553 M * Bertl maybe I should do some ikebana today, not coding ;) 1114283575 M * Doener heh :) 1114283596 M * Doener time to finally eat something ;) back in a few 1114283599 N * Doener Doener|gone 1114285029 N * Doener|gone Doener 1114285036 M * Doener re 1114285062 M * Bertl wb Doener! 1114285541 A * Doener .oO( hmm... what do i do next... ) 1114285697 M * Doener {025} vc_wait_context() sync command -- this can be marked as done, right? 1114285766 M * Bertl yup, please remove it .. although I adjusted the implementation a little 1114285798 M * Doener hm, no... the one we did was vc_wait_exit... is that the same? 1114285829 M * Bertl it's older, the name was not fixed at the time I entered that 1114285835 M * Doener ah ok 1114286071 M * Doener hmm... i'll need tcl knowledge soon (university), so i'll try hacking a small vserver configuration utitity in tcl... 1114286209 M * Bertl OMG ;) 1114286217 M * Doener hu? 1114286224 M * Doener such a bad idea? 1114286233 M * aba Doener: don't you need something reasonable like python :P 1114286274 M * Doener aba: no... no yet... maybe i'll start hacking, for example, trac someday... then i'll need python... 1114286280 A * Bertl is removing tcl wherever he can ... 1114286291 M * Doener but for the robotics stuff at university i'll need tcl 1114286347 A * Doener still waits for the day when university tells him that they're going to teach brainfuck :) 1114287629 M * albeiro Doener: here we are doing robotics in pascal so you know... 1114287636 M * albeiro and asembler 1114287658 M * Doener eeek! 1114287665 M * Doener ;) 1114287700 M * albeiro and applications calculating automatic control problems (i am studing it) we are not aloved to write in C. 1114287707 M * albeiro alowed 1114287758 M * albeiro because C is good for computer science students and here we care about algoritms and no about programing theory. 1114287762 M * albeiro ;p 1114287879 J * Nik ~Nik@cable-153-130.online.bg 1114287883 M * Nik hi all 1114287886 M * Bertl welcome Nik! 1114287903 M * Nik Bertl :-) 1114288476 M * Bertl Doener: as you followed the acc_integral changes ... where did it go? was it replaced by a similar accounting? 1114288508 M * Doener i don't remember, would need to look it up again... i just did my best to verify that i don't have to adjust anything ;) 1114288551 M * Doener but i can do that in a few minutes... 1114288639 M * Bertl okay, would be great ... because I ehard some rumours that there is some kind of accounting done for the vmkilelr 1114288653 Q * Nik Ping timeout: 480 seconds 1114289131 J * Nik ~Nik@cable-153-130.online.bg 1114289388 M * Doener Bertl: the function is not gone, but the calls are relocated.. guess i forgot to rebuild my cscope database when i did the .12-rc2 port 1114289445 M * Bertl so where do they happen now? 1114289458 M * Doener acct_update_integrals is only useful to call if stime changes otherwise 1114289459 M * Doener it will simply return. It is therefore best to relocate the function call 1114289459 M * Doener to acct_update_integral into the function that updates stime which is 1114289459 M * Doener account_system_time and remove it from the vm code paths. 1114289486 M * Doener calls are in account_system_time, do_execve and do_exit 1114289533 M * Doener (the above is from the .12-rc1 changelog) 1114289710 M * Bertl ah, okay, makes sense ... 1114289717 M * Bertl thanks! 1114289761 M * Doener you're welcome 1114290192 M * Doener hm, i'll cancel the tcl stuff... ;) 1114290368 Q * Nik Ping timeout: 480 seconds 1114291272 J * Nik ~Nik@cable-153-130.online.bg 1114291763 Q * Nik Ping timeout: 480 seconds 1114291892 M * Bertl Doener: do you still ahve the test setup for the wait_exit tests? 1114291937 M * Doener kernel is there and i did the tests with my plain qemu setup... so yes ;) 1114291965 M * Bertl okay, just because I modified some things, and I would be happy if you could look at them/test them ... 1114291970 M * Doener the tests were all done by hand ;) 1114291977 M * Doener sure 1114291989 M * Bertl will be ready in a few minutes (maybe 10 or so) 1114292798 M * albeiro just got a phone from friend. lol. firebridage just came back from one of students houses. it wouldn't be strange, we have smoke alarms in every room directly connected to their headquaters and they do activate few times a week (when someone is smoking in room that's it) but now three brigades came ! 1114292859 M * albeiro no doubt they will prosecute person making this alarm. by now everybody knows which room it was (424 if you would like to know ;p) 1114292886 M * Bertl your room? ;) 1114292904 M * albeiro nah, i am above one kilometer from this place. 1114292917 A * albeiro is not smoking :] 1114292933 M * Bertl well, but amybe playing with the smoke detectors :] 1114293019 M * albeiro oh, mayby once i almost smoked microwave ;p but i know the rules and was bright enough to make aeration big enough all the smoke came through widely opened doors ;p 1114293039 M * albeiro in fact i opened everything what would be opened ;p 1114293045 M * albeiro would/could/ 1114293147 M * albeiro last thing - you know how they test smoke detectors ? there is a man visiting each room from time to time with candle in kind of can and putting it near detector to see if it works ;p 1114293684 M * Bertl Doener: okay, against 2.0-pre2 1114293707 M * Bertl delta-signal-clean01.diff 1114293711 M * Bertl delta-state-clean01.diff 1114293719 M * Bertl delta-wakeup-fix01.diff 1114293728 M * Bertl delta-hash-clean01.diff 1114293742 M * Bertl (in this order) 1114293843 M * Doener hm, one of the killers did a lot of context creations/destructions + proc access, right? maybe that would be a good test for the hash cleanup 1114293856 M * Bertl yeah, excellent idea ... 1114293871 M * Bertl I will prepare such a test run ... 1114293880 M * Doener great 1114293983 M * Bertl should strike on the quad pretty soon ... 1114294002 J * Nik ~Nik@cable-153-130.online.bg 1114294007 Q * Nik Quit: 1114294011 M * Bertl (if we forgot something, of course ;) 1114294541 M * albeiro could i have a kernel releated question and not necesary vserver ? 1114294562 M * Bertl yeah, let me get one for you ;) 1114294564 M * albeiro what lock_kernel() do ? 1114294586 M * Bertl that does lock the kernel against concurrency 1114294643 M * albeiro that's like BKL ? 1114294650 M * albeiro or it is BKL ? 1114294656 M * Bertl yup, that's the BKL 1114294664 M * Bertl (which was removed in 2.6 ;) 1114294697 M * albeiro right, thx :] 1114294716 M * albeiro much of it in 2.4 btw 1114294716 M * Doener Bertl: hm? it's not completely removed, is it? 1114294732 M * albeiro Doener: it can be preempted afaik 1114294741 M * Bertl no, that was why added the (by myself overused) smiley ;) 1114294741 M * Doener at least the new option to make it preemptible would make no sense then ;) 1114294757 M * Doener ah ok :) 1114294855 A * Bertl .o( we need more native english speakers around, gramma already degrades :/ ) 1114294997 M * Vudumen hi allz :) 1114295028 M * Bertl welcome Vudumen! 1114295029 M * Doener Bertl: hm, is that a proof for your theory? ;) 1114295110 M * Vudumen Bertl: is it normal if Vserver kills my process instead of swapping out if i set up an RSS limit and it reaches it? 1114295157 M * Bertl yup, that's normal 1114295192 M * albeiro Bertl: but englishmans. us people are speaking in some strange way and it is said that many folks learning english as additional language are doing better. so far i have heard the best english from japan people :] 1114295199 M * Vudumen if i remember well it's not the same behaviour as in the simple kernel without any virtualization 1114295302 M * Bertl Vudumen: you do not ahve an RSS limit in a simple kernel? 1114295375 M * Vudumen well i check my files. a moment. it's possible that i said something wrong ;) we saw this on the beginning of the week :) 1114295421 M * Vudumen Bertl: yes you are right 1114295473 M * Vudumen mayeb it was AS? dunno :/ monday i ask my colleague :) 1114295489 M * Vudumen would it be hard to make an option to swap it out instead of killing it? 1114295575 M * Bertl yup 1114295692 M * SiD3WiNDR :) 1114296175 Q * knoppix_ Quit: Verlassend 1114297194 J * clcont ~xxx@nts-100.7-185-64.nts-online.net 1114297213 N * clcont clanon 1114297217 M * clanon hi guys 1114297227 M * Bertl welcome clanon! 1114297279 M * SiD3WiNDR hey, I have a printer just l.. hmm, nevermind 1114297282 A * SiD3WiNDR HUMOR 1114297284 M * SiD3WiNDR *cough* 1114297461 M * clanon so this thing seems interesting 1114297466 M * Doener Bertl: + if (vx_info_state(vxi, VXS_SHUTDOWN|VXS_HASHED) == VXS_SHUTDOWN) 1114297480 M * Doener what's the idea behind having VXS_HASHED there? 1114297503 M * Bertl it is not supposed to return unless the vxi is unhashed 1114297529 M * Bertl (the __wakeup* was generalized and might fire more often in the future) 1114297586 M * Doener ah, i.c. == not & ... 1114297664 M * clanon hmm 1114297693 M * Bertl clanon: you bet it is! 1114297850 M * clanon so this would be like good to have a lot of users on a shared box and nobody could mess with other peoples servers and get them defaced? 1114297985 M * Bertl yup .. soemthing along this road 1114298076 M * clanon thats good. :o 1114298113 M * Doener ok, finally building the qemu kernel... 1114298771 M * clanon god i hate when i forget passwords 1114298863 Q * wurd Ping timeout: 480 seconds 1114298868 M * Doener use public keys ;) 1114298879 M * clanon i been trying 1114298883 M * clanon but i keep forgetting the password 1114298999 M * Doener Bertl: tests were fine 1114299045 M * Bertl excellent, so I didn#t mess something up then ;) 1114299162 M * clanon its possible to run multiple SSL sites on one IP, right? 1114299180 M * Vudumen clanon: it isn't 1114299221 M * clanon hmm 1114299221 M * Vudumen clanon: maybe you can play with certificates signed for *.domain.name common names but i don't think even a CA who signs a cert like this. 1114299247 M * Vudumen of course you can define more vhosts for port 443 and enable SSL for all of them but browsers will complain about the invalid certificate 1114299251 M * clanon well i read that it wasnt possible to do that 1114299265 M * clanon (have SSL sites using a shared IP) 1114299271 M * clanon but plesk does it 1114299285 M * Doener Vudumen: hm, are certificates bound to ip addresses now? 1114299302 M * Doener i always thought they use the domain as the common name... 1114299320 M * Doener so as long as you have different domains there shouldn't be any problems 1114299322 M * Bertl yeah, the certificate is not a problem, the ssl connection is ;) 1114299388 M * Doener Bertl: hm? please elaborate... 1114299411 M * Bertl if you have 3 different sites, on one IP 1114299431 M * Bertl then the https (per definition) first builds the ssl socket 1114299443 M * Bertl (ssl connection from client to host) 1114299458 M * Bertl and then transmits the http request for the domain 1114299462 M * Vudumen exactly as Bertl sais 1114299480 M * Bertl so it's pretty ahrd to figure the 'correct' certificate ;) 1114299493 M * Vudumen and the hostname for "Virtual Domains" are located in the http request (http 1.1 at least) 1114299506 M * Doener ah, right 1114299516 M * clanon sorry i was talking about named virtual hosting with SSL 1114299521 M * Bertl btw, that's a design flag in https ;) 1114299527 M * Bertl s/flag/flaw/ 1114299542 A * Bertl .o( what am I typing today ... ) 1114299543 M * clanon i was wondering if it was now possible 1114299550 M * Doener clanon: we're talking of that, too ;) 1114299574 M * Bertl clanon: no it is not possible, unless you figure a smart way to tell in advance which domain will be selected 1114299607 M * Bertl i.e. you 'know' at the time the ssl connection is done, which request will be issued in the future 1114299609 M * Vudumen Bertl: i don't think it's a flaw. it makes sure that _everything_ in the http communcation will be encrypted. including the full requiest URL (including variables passed with GET method) 1114299637 M * Bertl Vudumen: well, the design flaw is that the ssl doesn't allow to 'select' the domain 1114299676 M * clanon i dunno 1114299686 M * Vudumen Bertl: ssl is a general solution. and it's applied to http. the flaw is in the application i think :) but it was the easier for the developers 1114299703 M * Doener Vudumen: you could also use a separate server certificate that is used to encrypt the first step (i.e. figuring out the domain) 1114299704 M * Vudumen the application == the method of application 1114299857 M * Vudumen Doener: for me it seems to be the best solution to "redesign" https to be able to get datas before the SSL negotiation. But in this case it's an interesting thing if the hostname is a sensitive information and it's transferred without encryption and befor the authentication of the server :) i don't think there are any scenarios like this but maybe for someone... 1114299869 M * clanon hmm 1114299920 M * clanon whatever plesk is doing, its doing it with 1 IP 1114300007 M * Bertl well, then it must be magic ;) 1114300016 M * clanon probably 1114300082 M * clanon i dont see anything magical 1114300167 M * Bertl well, then let's hear how it does it ... 1114300194 M * clanon lol i asked in the #apache room 1114300199 M * Bertl because if that works, it should work quite fine on any linux host 1114300229 M * Vudumen or maybe any other not just linux 1114300260 M * clanon i dunno 1114300322 M * clanon 1114300323 M * clanon DocumentRoot /home/httpd/vhosts/webmail 1114300323 M * clanon ServerName webmail 1114300350 M * clanon ServerAlias webmail.* 1114300388 M * clanon it has that for lists too 1114300394 M * clanon oooh 1114300408 M * clanon lol that ip doesnt work :-D 1114300480 M * clanon *hides* 1114300481 M * clanon so 1114300487 M * clanon what else is up fellas? how are yall this saturday morning? 1114300529 M * Bertl hmm, it's already sunday here ... 1114300545 M * Vudumen here too :) 1114300547 M * clanon i have 3 virtual SSL hosts on that IP too 1114300552 M * clanon thats cool so where are you Bertl? 1114300574 M * Bertl I'm at home right now .. *G* 1114300586 M * Vudumen :)) 1114300682 M * clanon i dont get it 1114300787 M * clanon oh ok 1114300794 J * bRuTaL`script ~12weq@mail.ibs-bg.net