[00:07] Nick change: arekm_ -> arekm [00:08] hi arekm! [00:13] hi arekm [00:19] ah... finally that download completed :) [00:19] the binutils? [00:20] yepp [00:20] hmm...shorewall is only accepting subnets to masquerade ... wish me luck that it accepts a single ip too ;) [00:20] well, bad news, next step is the compiler :( [00:20] hmm... --target=%{CROSS_ARCH}-linux ... that doesn't make too much sense if i just use it 'as is' [00:20] yep, you replace the %{CROSS_ARCH} by the arch name, what arch would you prefer? [00:21] currently mips,mips64,hppa,hppa64,ppc,ppc64 and m68k is not working [00:21] then i'll try with ppc [00:22] and which one is working besides ix86 ? [00:22] okay, just let me check that I'm not talking nonsense ... [00:22] ia64 ? [00:24] for ia64 we didn't manage to build a crosscompiler yet ;) [00:24] what kernel by the way would you prefer? 2.4 or 2.6? [00:25] me or him or both ? [00:25] ahh, well yes both, all? [00:25] 2.6 :) [00:25] to many tasks at the moment, I'm trashing ... [00:26] okay, I started a test run, well know for 2.6.2-rc3 in a few minutes ... [00:26] i will try it first with 2.4 ... i don't quite trust 2.6 till now ;) [00:27] Doener: they had some arch fixes in the changelog, so maybe some archs work now, just verifying ... [00:27] you will have to get the gcc sources too, so you can start a download for gcc-3.3.2 (if you do not have the sources around) [00:29] when my servers are running i will set up a gentoo install for the testing [00:29] great! [00:31] ah... kickin the other machines from the router did the trick ;) now downloading @~90kb/s ... guess someone was doing filesharing or such... [00:31] Doener: you should implement the wondershaper ;) [00:33] btw ... the arch tests .... for the stable or the dev branch of vserver ? [00:33] once we got the basic setup working, it will be done for _all_ verser releases automagically ... [00:34] (at least the cross compile tests) [00:34] ok but for which one first ? [00:34] hum, don't know, why? [00:35] doesn't matter ? [00:36] okay, maybe I missed something, please elaborate? [00:36] ok, got gcc [00:36] you need the binutils compiled first ... [00:37] i was just curious because you didn't mention which version should be tested ... or did I miss that ? [00:38] ahh, okay, no the first step for the cross compiling is to get the vanilla kernel working ;) [00:38] [ARCH alpha/alpha] building kernel ... succeeded. [00:38] [ARCH hppa/parisc] building kernel ... failed. [00:38] [ARCH hppa64/parisc] building kernel ... failed. [00:38] [ARCH i386/i386] building kernel ... succeeded. [00:38] [ARCH m68k/m68k] building kernel ... failed. [00:38] [ARCH mips/mips] building kernel ... failed. [00:38] [ARCH mips64/mips] building kernel ... failed. [00:38] [ARCH ppc/ppc] building kernel ... succeeded. [00:38] [ARCH ppc64/ppc64] building kernel ... failed. [00:38] current status for 2.6.2-rc3 ;) [00:39] :( [00:39] so ppc64 is an option ;) [00:39] ok, doing next configure run [00:40] you can install more than one cross compiler, they do not interfere ... [00:40] (same for the binutils) [00:41] Cyrix: does this make sense to you? [00:41] or did I confuse you completely? [00:42] ok, time for coffee... forgot to make clean... next run... [00:43] _shuri (~shushushu@vserver.electronicbox.net) joined #vserver. [00:43] hi _shuri! [00:45] you mean a vanilla kernel without any vserver-patch ? [00:46] yes, it doesn't compile with the default config on most architectures (default config for the specific arch) [00:46] I guess in most cases only small adaptations are required to make it compile ... [00:47] (something like selecting or deselecting a driver ..) [00:49] <_shuri> hi [00:49] <_shuri> my ipv6 tunnel was down [00:49] <_shuri> was not able to connect to irc [00:49] how unfortunate! [00:52] http://vserver.13thfloor.at/Stuff/Cross/compile.info [00:52] here is the current status ... [00:52] <_shuri> ok [00:53] <_shuri> where is the defaut comfig? [00:53] you get the default config with: [00:53] yes '' | make oldconfig [00:53] (at least that is what I do ;) [00:53] the 2.4 vanilla kernels do not compile on all architectures with their defaults ?! then why they even release them ? ;) [00:54] Cyrix: do not ask me .. maybe I'm doing something wrong ... maybe it doesn't work with cross conmpiling config, I don't know ... [00:54] or you could do a make menuconfig and immediately save without making any changes ;) [00:55] yes, you have to make sure to specify ARCH= and CROSS_COMPILE=-linux- [00:55] so the line actually is: [00:56] yes '' | make oldconfig ARCH=ppc64 CROSS_COMPILE=ppc64-linux- [00:56] for example ... [00:56] and kernel compile is the done with [00:57] make ARCH=ppc64 CROSS_COMPILE=ppc64-linux- [00:57] this remindes me of make menuconfig ARCH=um *crouch* ;) [00:57] well, yes it's the same kind of cross compiling ... [00:57] ok bertl, finally binutils make suceeded [00:58] okay, you have to install it, and compile the gcc (cross) [00:58] distclean seems to be broken (or got broken when i pressed ctrl+c :) [01:03] Bertl: should i apply the powerpc fix patch? or is that just ppc and not ppc64? [01:03] apply it, doesn't hurt ... [01:08] for hppa and hppa64 (parisc) there are specific kernel patches, which are required to make it work ... [01:10] MASQ and SNAT are now running side by side without interfering :) [01:11] wonderful ... [01:11] now I just need to implement something which restores my SNAT-rules when I restart my firewall due to configuration changes ... [01:19] Cyrix (~master@hmbg-d9ba829b.pool.mediaWays.net) left irc: Remote host closed the connection [01:21] suhcoolbro (~Suh@216-161-89-245.ptld.qwest.net) joined #vserver. [01:22] hi suhcoolbro! [01:23] Cyrix (~master@hmbg-d9ba829b.pool.mediaWays.net) joined #vserver. [01:24] damn my w2k-client got fucked by a blue screen of death ... i should reinstall linux on my working partition as quickly as possible [01:24] BSOD 8-) [01:26] yepp ... BSOD ... i wasn't aware that this acronym is known in this linux-only chat 8-) [01:26] well, there is a screensaver for X ;) [01:27] aha...i saw only one for win itself ;) [01:27] and one bash script providing the same *fg* [01:28] well, some time ago, a co-worker rebooted my linux workstation because of that screen saver ;) [01:28] .oO(or better say a parody of it) [01:29] well, actually he pulled the plug, because the reset and power switch was disabled :( [01:29] omg [01:30] okay, updated http://vserver.13thfloor.at/Stuff/Cross/compile.info [01:30] was ext2 running ? [01:30] it was ext2/ext3 mixed setup, no issues with that ... [01:31] maybe not....after an 2 hour reboot (and repair) time ;) [01:32] 2.4 has more fails than 2.6?! O_O [01:32] The following requested languages were not found: con [01:32] The following languages were available: c ada c++ f77 java objc treelang [01:32] Configure in /home/doener/vserver/gcc-3.3.2/gcc failed, exiting. [01:32] no, that is quite fast, usually I have /, /usr and /var on ext2, the rest is ext3 [01:33] Doener: hmm, how did you configure? [01:33] i have every partiton >128M at ext3, rest ext2 [01:33] i took the line from http://vserver.13thfloor.at/Stuff/Cross/gcc-cross.spec [01:34] are the binutils installed? [01:34] does 2.6 not have any modules anymore or why they are only mentioned for 2.4 ? [01:35] 2.6 has a new build system, it doesn distinquish (at least not for make) between kernel and modules ... [01:35] +t [01:37] binutil are installed [01:37] s/util/utils/ [01:39] okay, configure should work now ... [01:39] so for 2.6 make moduels got into make and make modules_install in make install ? [01:40] hmm, at least that is what I assume ... but let me check, maybe I'm wrong, never built 2.6 modules ;) [01:40] forget it... copy'n'paste messed up the configure... three times in a row... [01:40] Cyrix: yes, modules are now in the default target [01:41] doing make of gcc now... [01:45] @echo 'Other generic targets:' [01:45] @echo ' all - Build all targets marked with [*]' [01:45] @echo '* vmlinux - Build the bare kernel' [01:45] @echo '* modules - Build all modules' [01:45] @echo ' modules_install - Install all modules' [01:46] so it seems modules_install is still a separate target ;) [01:46] ah... should read more careful... [01:47] so you've built a pure monolithic kernel ? [01:49] for my vserver tests on 2.6 yes ... for this compile test, probably not, don't know haven't checked yet ... [01:49] I uploaded the logs and configs for both kernels and all archs [01:49] http://vserver.13thfloor.at/Stuff/Cross/ [01:51] it seems that the default config doesn't use modules ... [01:54] hmm... whole bunch of errors on gcc make... [01:54] errors or warnings? [01:54] errors [01:54] did you use the make line from the spec? [01:55] make -j 5 TARGET_LIBGCC2_CFLAGS='-Dinhibit_libc -D__gthr_posix_h' [01:55] let me check... [01:55] ok, now i don't trust in copy'n'paste anymore... [01:58] .oO(?) [01:59] hmm ... .oO(?) [02:04] gthr-posix.h:37:21: pthread.h: No such file or directory [02:04] gthr-posix.h:38:20: unistd.h: No such file or directory [02:04] hmm... [02:04] do i need anything besides the big source package for gcc? [02:04] hmm, well working glibc headers ... [02:05] seems that your distro doesn't define what gthr-posix.h requires ... [02:05] let me check where the are on mandrake ... [02:06] glibc devel is installed, right? [02:08] libc6 devel is, there is no glibc devel package... [02:08] okay, /usr/include/pthread.h does exist? [02:09] yepp [02:09] hmm, interesting ... maybe some separate pthread package? [02:09] at least nothing with pthread in its name... [02:10] maybe you should try to convert the rpm with alien to a tar? [02:10] the cross rpm's are binaries right? [02:10] yes [02:10] Action: Doener just had the same idea ;) [02:10] then i'll try to convert them to deb [02:10] ;) [02:11] I always thought that debian and source based systems are better in compiling from source, but recently I'm not so sure ... [02:12] util-vserver didn't compile on several distros or had some issues with broken headers etc ... [02:12] now plain gcc-3.2.2 doesn't compile on debian ;) [02:12] 3.3.2 I mean ;) [02:12] i wouldn't say so for debian ... not for woody/stable trying to compile the newest sources [02:13] hehe, a world without problems would be... boring i guess [02:13] sure, where would be the challenge ... [02:13] i'm on sid ;) woody for a desktop box is not what i consider fun [02:13] but i don't expect problems with gentoo compiling such sources [02:14] umm...which one was testing and which one unstable regarding sid and sarge ? [02:15] sid is always unstable [02:15] current testing is sarge [02:15] never used it...although i should mention that i use woody on a server system without X ;) [02:17] my only linux desktop system for now is Knoppix ... once i had SuSE installed and another time gentoo ... but compiling X and KDE isn't that much of fun [02:29] hmm... last tar package was messed up... i hope the new one is fixed as the error keeps me from installing any debs... [02:29] hum, which tar package? [02:30] the deb package for tar itself [02:30] it claimed a directory that is needed for the install process of any deb... so dpkg tries to protect that directory and keeps me from installing... [02:31] cool! [02:32] not really... as tar is needed for the install process this could be mean big f&§$ up... [02:32] well I assume you can fix that with vi if it is installed ;) [02:34] ;) [02:35] shorewall is a jungle of shell scripts but i bashed my way through and now the SNAT rule is applied also every startup [02:37] huh .. sorry, but if that would take so long, I would have written my own fw script, hmm, now I remember I did! ;) [02:38] i also did once, but shorewall is far easier to maintain and normally provides all configuration options you need ;) [02:39] but i'll speak to the maintainer that he forgot something ;) [02:40] ok, current tar package is fixed [02:48] ok guys, it's been a long day at which end i'm pleased now so I think I'll retire... there's still a lot of work to do but on a solid foundation this should be no problem, many thanks especially to Bertl ;) I'll see you hopefully next week ready to do the m68k-tests, bye & gn8. [02:48] good night, and have fun! [02:49] we'll see ... tomorrow ... ;) [02:49] Cyrix (~master@hmbg-d9ba829b.pool.mediaWays.net) left irc: Quit: Leaving [02:51] xsbyme (xsbyme@D5761031.kabel.telenet.be) joined #vserver. [02:52] hi xsbyme! [02:52] hello :-) [02:52] ok, bertl, everything seems to be working, finally... [02:52] Bertl [02:52] u know how to work with [02:52] its4 [02:52] ... [02:53] ts4 - scans C(++) source for potential security vulnerabilities. [02:53] its4 - scans C(++) source for potential security vulnerabilities. [02:53] ... [02:53] isn't that some security tool? [02:53] :-) [02:53] yes [02:54] its4 - scans C(++) source for potential security vulnerabilities. [02:54] and did you scan the kernel? [02:54] no [02:54] hmm, would be interesting ;) [02:54] erm [02:55] okay, what is your point?! [02:55] if you scan the kernel that sort tools get a huge rapport than [02:55] k [02:55] you know other tools to that [02:55] you know other than that one [02:56] you mean searhing for bad lengths and uninitialized stuff, buffer overrungs etc? [02:56] yep [02:56] yes [02:57] hmm, I know there are some folks doing such code audits, with a 'secret' tool on the kernel ... [02:57] but I don't know any 'available' alternative, atm ... [02:57] k [02:57] if you know [02:57] tell me [02:57] or mail me pleas [02:58] well, I would google for that ... [02:58] k [02:58] Bertl [02:58] u know the name of that tool [02:58] nope sorry, I meant google for such tools in general ... [02:59] k [02:59] http://www.linuxjournal.com/article.php?sid=5673 [02:59] well, I'm good at googling ;) [03:00] hmm [03:00] thnx [03:00] ps [03:00] Bertl [03:01] vserver is running smooth [03:01] on traffic4all [03:01] the vservers * [03:01] Bertl: what comes after gcc compilation? make menuconfig still only gives me x86 targets... [03:01] 22:57 < Bertl> yes '' | make oldconfig ARCH=ppc64 CROSS_COMPILE=ppc64-linux- [03:02] 22:58 < Bertl> make ARCH=ppc64 CROSS_COMPILE=ppc64-linux- [03:03] ouch... [03:04] darn copy'n'paste... [03:04] guess i just dumped a servers config... [03:04] we should get as close as possible to a vserver config anyway ... [03:05] with all useful stuff, like filesystems enabled and such .. [03:05] later I'm planning to do a monolithic and a module version ... [03:05] if we get that right for some archs, the test can be done automated ... [03:07] by the way, I appreciate your help ... [03:07] np ;) [03:08] btw, today i got permission to give the code of our vserver-interface out to the guys of pbvcs [03:08] hey great, hopefully they will make good use of it ;) [03:09] hmm... i guess shade won't care if i give you access to his test account, so if you're interested... [03:09] sure, a sneak peek ... [03:13] Bertl u know sutch programmas but also for windows [03:14] sutch [03:14] no sorry, don't do windows ... [03:14] k np [03:15] but I guess with cygwin you would be able to compile most tools ... [03:15] k [03:15] Doener [03:15] yeah? [03:15] u maked vserver-interface ? [03:15] webbased or ??? [03:16] yepp, php-based [03:16] hmm [03:16] screenshots ? [03:16] hmm... let me see if i got some... [03:16] but it's only in german language atm [03:16] np [03:17] traffic4all is seeking vserver-interface that are webbased ... [03:18] I made a vserver interface in php about 2 years ago ... [03:18] hmm [03:18] was targeted at the vserver reseller ... [03:18] show ... [03:18] pleas [03:19] unfortunately it was done under closed/proprietary license [03:19] k [03:19] and the company I wrote it for, still refuses to use or publish it :( [03:19] <_shuri> :( [03:20] <_shuri> this is bad [03:20] that was the point where I decided to stop writing proprietary stuff where possible ... [03:20] Bertl but you are the main coder right ? [03:20] well, I'm the auther, I own the copyright, but the right on selling that product are with the company ... [03:21] yes [03:21] that you have the right on your code ... [03:21] xsbyme: nope, got no screenshots... the code got published to the makers of pbvcs, an upcoming open-source interface, just today, so you may want to keep an eye on their project [03:21] deadguy [03:21] Doener [03:21] site [03:22] http://vserver.wireless-winds.de/PBVSC [03:22] Bertl [03:22] than you have the rights on your code to be gnu or closed sources etc ... [03:23] well, yes basically I could re-release it as GPL, but if I make it available, they would sue me for lost profit (which is illy, because they do not use/sell it) [03:24] +s [03:24] but anyway, I guess the next generation interface will be much better and smarter than my interface ever was ;) [03:24] :-) [03:25] how far are you with the new one [03:26] well, I'm not writing on it .. [03:26] ah k [03:26] the pbvsc (i hate to type this ;) maintainer is Sh[a]de [03:27] but I will have a look at it from time to time and add my suggestions/ideas .. [03:28] <_shuri> pbvsc look good [03:30] yepp ;) if not, i wouldn't have bugged the one who paid me for my interface to allow me to provide the source to them [03:30] hrhr, ppc64 had agp in its default config [03:32] I guess we might be able to feed back those changes to the kernel mainline ... [03:32] nn [03:32] aal [03:32] all [03:32] g'night xsbyme [03:32] xsbyme (xsbyme@D5761031.kabel.telenet.be) left irc: [03:32] don't drink and irc ;) [03:44] Bertl: the build logs for 2.6.2-rc3 you provide are made with default config? [03:44] hmm, yes they should be ... [03:44] the config is there ... [03:46] hmm... differs from what i got from oldconfig... [03:46] did you clean the .config first? [03:46] otherwise it will use your x86 config ;) [03:47] i used a fresh 2.6.0 source and applied patches up to 2.6.2-rc3 and made mrproper and clean... but i'll try again... [03:47] might easily be that I made a mistake somewhere in my scripts ... [03:48] in your config proc support is turned off, whilst in mine it is turned on [03:49] interesting, let me check taht again ... [03:52] hmm funny did that on the command line, and yes, it differs ... [03:54] I even get a bunch of modules, I never saw before ... [03:57] ahh, okay, I'm stupid ... [03:57] I did some hack for the mips, which do not recognize mips/mips64 correctly, and seem to require something else ... [03:58] unfortunately a missing .conf isn't treated the same way then an empty one ... [03:58] the empty one seems to disable almost everything ... [03:58] I'll fix that and let the monster have another chew ... let's see what archs work now ;) [04:16] http://doener.homeip.net/ppc64/errors [04:17] that makes no sense to me ;) [04:17] hmm, sounds weird ... [04:18] hmm, why is zImage called anyway? [04:18] does ppc support zImages? [04:22] ah ... [04:22] CROSS32_COMPILE = [04:23] # NOTE: this code is built for 32 bit in ELF32 format even though [04:23] # it packages a 64 bit kernel. We do this to simplify the [04:23] # bootloader and increase compatibility with OpenFirmware. [04:23] so you need to build the ppc binutils too ... [04:23] and specify CROSS32_COMPILE=ppc-linux- [04:26] seems to be a ppc64 oddity ... [04:31] so ppc64 and ppc binutils? [04:31] guess so ... mdon't think they need the ppc compiler too ... [04:32] but ppc64 is the only arch which does define CROSS32_* [04:33] that's because userland on ppc64 is usually 32 bit [04:34] hmm, didn't know that, thanks for the info Rik! [04:34] 32 bit apps run way faster than 64 bit ones [04:34] on ppc [04:34] ppc-gcc is also needed... [04:35] did you alienate the ppc64 compiler? [04:35] different for x86 vs x86-64 since the 64 bit x86 version has twice as many registers [04:35] hmm, yes registers are important for performance ... [04:36] rik, by the way, do you know what needs to be changed to build a ia64 cross compiler? [04:36] Action: riel runs away screaming [04:36] the 'usual' fix I did to most archs didn't work for ia64 ;) [04:36] yepp, i got alienated gcc and binutils for ppc and ppc64 now [04:37] of course, ia64 is "different" [04:37] unsinkable, iceberg resistant, etc [04:37] yeah, binutils worked out of the box, gcc no success so far :( [04:37] ok, got a build now, will go back to clean and try again to be sure [04:37] why are you even trying to build ia64 ? ;) [04:38] well, we are doing many archs, I set up an automated cross build test space ... [04:39] http://vserver.13thfloor.at/Stuff/Cross/ [04:39] http://vserver.13thfloor.at/Stuff/Cross/compile.info [04:39] (that is currently without vserver :( ) [04:39] 2.6 kernel I hope ? [04:39] (2.4 kernel might not build for ia64 without patches) [04:40] well, I'm expecting troubles on 2.6 and 2.4 ... but as I said, I didn't succeed to compile gcc-cross for ia64 ... [04:41] but it's funny that for most archs 2.6.2-rc3 does not build with the default config ... [04:41] maybe the cross compiling is to blame but I doubt that ... [04:46] ok, config change number one confirmed, expected that as there is no agp header file in asm-ppc64... waiting for next build... [04:46] okay ... [04:47] second one also confirmed... [04:47] kernel built, no more errors [04:47] great! [04:48] please send me the config or a short overview regarding the changes, or maybe we should put that on the wiki? [04:48] http://doener.homeip.net/ppc64/ppc64-config-changes [04:49] plus the cross32_compile of course [04:49] ah okay ... [04:49] great ... [04:50] many thanks, hmm, wanna do another one? [04:51] i'll see if i can manage one more tomorrow, my girl-friend just came home from a party and i'm 'supposed' to go to sleep now :) [04:51] ah, okay, great, have a good night then ... [04:52] g'night [04:52] I'll update the infos on the url, so you can see what has been done ... [04:52] the compile.info one? [04:52] yes ... [04:52] ok [04:52] it's currently running ... seems better now ... [04:52] Nick change: Doener -> doener_zZz [04:52] cya tomorrow [04:55] riel: do you know how the .config and make oldconfig logic is supposed to work? [04:56] (in the kernel 2.6 and 2.4) [09:08] noel- (~noel@pD952C364.dip.t-dialin.net) joined #vserver. [09:16] noel (~noel@pD9FFA52B.dip.t-dialin.net) left irc: Ping timeout: 504 seconds [09:16] kestrel (athomas@home.swapoff.org) joined #vserver. [09:58] talon (talon@host-63-149-223-100.irwinresearch.com) joined #vserver. [10:02] i have pre context quotas working with linux 2.4.24-vs1.24 and the context quota patch (patch-2.4.23-vs1.20-q0.12.diff) plus the fix for the quotaoff hang that was mentioned on the mailing list archives (patch-q0.12-hang-fix.diff) the block quotas seem to work quite well. but the inode quotas dont. [10:08] Action: talon shrugs [10:08] well im going ot keep playing with it. i figured id ask here to see if this has come up before. [10:09] i plan on making a context quota howto eventually. [10:11] hi talon, was afk for a moment ... [10:11] hmm, let me catch up ... [10:12] oh yeah i forgot ot mention i was using the bind mount extensions patch too. [10:12] and UID24/GID24 maping. [10:12] http://vserver.13thfloor.at/Experimental/delta-2.4.25-pre7-vs1.24-q0.12-t1.diff [10:12] that would be a good quota patch, I hope ... [10:13] yeah, that contains the fix, and is suited for 2.4.25-pre7++ which also contains some quota fixes ... [10:13] Bertl: repatch a clean kernel with that patch instead of the original quota path ? [10:13] the ro bind stuff (bme) should work with that too .. [10:13] yup [10:14] cool i will give it a shot now and let you know. [10:14] guess 2.4.25-pre8 should work with that one too ... [10:14] hmm, wait let me check that ... [10:18] http://vserver.13thfloor.at/Experimental/patch-2.4.25-pre7-vs1.24-q0.12.diff [10:18] http://vserver.13thfloor.at/Stuff/patch-q0.12-hang-fix.diff [10:19] http://www.13thfloor.at/vserver/s_release/v1.24/patch-2.4.25-pre6-vs1.24.diff [10:19] okay that should be all you need [10:19] starting with the vs1.24 patch, then the q0.12 patch and the fix ... [10:20] talon: sounds good? [10:22] sorry i was away for a bit. [10:23] yeah sounds good. [10:23] in that order. [10:23] vs1.24, q0.12, fix ... [10:23] then optional bme ... [10:24] ok thats how i was doing it before. [10:24] hmm, I guess if that bme stuff is actually used, I have to update the patches ;) [10:24] yes its quite useful. [10:24] for me too ... [10:24] id rather have read only mounts than doing that unification stuff. [10:24] its also useful outside of vserver. [10:25] it was developed outside ;) [10:25] im taking a template distro, mounting bin sbin usr lib and then giving users an /top partition and bind -o rw mounting /opt/local to /usr/local [10:26] /opt i mean. [10:26] with their own copies of etc and var. [10:26] yeah, that is the idea among others ... [10:27] okay, I'm off to bed for now ... cu later? [10:27] sure, when are you on normally? [10:27] i will let you know how it worked out. [10:28] it's about 8:30 am CET here [10:28] I usually sleep the whole day, and will be back on around 19:00 CET [10:28] okay, have a nice one, cu! [10:29] thanks. [10:29] Nick change: Bertl -> Bertl_zZz [13:38] kestrel (athomas@home.swapoff.org) left irc: Quit: kernel upgrade [13:44] kestrel (athomas@home.swapoff.org) joined #vserver. [14:17] miller7 (none@213.239.180.106) joined #vserver. [14:17] hello people! :-) [14:18] hi there [14:18] hey kestrel [14:19] hi there, brb, upgrading my radeon drivers [14:19] kestrel (athomas@home.swapoff.org) left irc: Quit: brb [14:29] kestrel (athomas@home.swapoff.org) joined #vserver. [14:48] kestrel_ (athomas@home.swapoff.org) joined #vserver. [14:48] kestrel (athomas@home.swapoff.org) left irc: Read error: Connection reset by peer [15:28] miller7 (none@213.239.180.106) left irc: Ping timeout: 492 seconds [15:40] arekm_ (misiek@ikar.t17.ds.pwr.wroc.pl) joined #vserver. [15:40] arekm (misiek@ikar.t17.ds.pwr.wroc.pl) left irc: Read error: Connection reset by peer [16:09] Nick change: doener_zZz -> Doener [16:09] hi [16:10] anyone else doing cross compiling stuff or can i still trust the sucess/failure results given by bertl? [16:38] mcp (~hightower@wolk-project.de) left irc: Ping timeout: 492 seconds [18:17] serving (~serving@213.186.191.180) left irc: Read error: Connection reset by peer [18:47] mcp (~hightower@wolk-project.de) joined #vserver. [19:19] miller7 (none@213.239.180.106) joined #vserver. [19:20] hi [20:13] serving (~serving@213.186.191.180) joined #vserver. [20:19] Nick change: Bertl_zZz -> Bertl [20:19] hi everyone! [20:22] hey bert [20:23] how's it going? [20:23] well, I just woke up, so it's going ... slowly ;) [20:23] :) [20:23] it's 7:25pm!!! [20:24] no [20:24] 6:25 pm [20:24] now you woke up?! [20:24] :D [20:24] yeah, better ... early in the morning, right? [20:24] yeah [20:25] well, I wen around 8:40 am to bed .. so ... 9-10 hours sleep ... [20:25] not bad! [20:25] during he week I don't get that much ... usually ... [20:25] +t [20:28] it's good to keep off the computer for a while :) [20:29] is it? ;) [20:29] don't ask me that question cause my answer will be NOOOOOOOOO [20:29] :P [20:37] okay, anybody interested in testing new 2.6 stuff? [20:38] anybody interested in helping with the cross compiling configs? [20:39] what's that? [20:39] well about a week ago, we started a cross compiling attempt, to allow kernel builds on x86 for all archs vserver supports ;) [20:40] http://vserver.13thfloor.at/Stuff/Cross/ [20:40] ic [20:40] here are packages and spec files for the cross compiling tools ... [20:40] http://vserver.13thfloor.at/Stuff/Cross/compile.info [20:40] shows how the vanilla kernels compile with the 'default' config [20:41] yes '' | make oldconfig ARCH= [20:41] I'll take a look at it later and see what's it about [20:41] unfortunately this fails for some archs ... [20:41] and probably needs some tweaking in the config files ... [20:41] okay, np, I'm going to get some breakfast ... [20:41] Nick change: Bertl -> Bertl_oO [21:03] Nick change: Bertl_oO -> Bertl [21:03] ahh, much better ;) [21:04] :-) [21:04] hey Bertl [21:04] i guess we are to blame for the cross compilation failures ;) [21:05] hmm, how so? [21:05] sec, got phone call... [21:14] ok, it's all about the make oldconfig [21:14] hmm, please elaborate (hi, by the way!) [21:14] hi ;) [21:15] grml... same one on the phone again... [21:18] must be microsoft user-specific that no one actually reads what's on the screen... [21:19] back to topic... the kernel's make files handle config/menuconfig/oldconfig in a common way, that tries to use as much as possible from the current kernel config [21:20] well, usually nobody is able to interpret what is written on a microsoft screen ... [21:20] what oldconfig does is that it's only asks questions which have been added since the kernel version that is currently installed [21:20] so it's just a shorter version of make config [21:21] Doener: but in the current test, there is no .config, when make oldconfig is called ... [21:21] basically for us that means that the cross compiled kernels had the configuration that our currently running kernels have [21:21] the running one ;) [21:21] from /boot or whereever yours is [21:21] hum ... [21:21] what we need to do is make defconfig [21:23] i realized this when i did make config (or whatever) and a message about an undefined config symbol PRISM54 hushed over the screen [21:23] since my box uses a prism54 card but the drivers are not part of vanilla [21:24] hmm, okay, maybe we should investigate this a little more ... [21:24] make help was my source of knowledge ;) [21:24] hmm, damn, never tried that one ;) [21:24] oldconfig - Update current config utilising a line-oriented program [21:24] defconfig - New config with default answer to all options [21:25] okay that _is_ an option for 2.6 then ... [21:25] allnoconfig - New minimal config [21:25] but it won't work for 2.4 ... (I already knew the allyes/noconfig) [21:25] okay, I adjust my script to do make defconfig and give it another run ... [21:26] oh, as you mention it, now i can decipher allnoconfig, the target name made no sense to me ;) [21:27] hehe [21:29] okay, started a 2.6.2-rc3 testrun with defonfig ;) [21:29] +c [21:34] Doener: but I can't believe that it's looking at /boot/ and such ... [21:35] great [21:35] erm... wrong channel [21:37] scripts/kconfig/conf -o arch/ppc64/Kconfig [21:37] # [21:37] # using defaults found in /boot/config-2.6.1 [21:37] # [21:37] /boot/config-2.6.1:49: trying to assign nonexistent symbol X86_PC [21:37] there goes a whole bunch of errors [21:37] hmm, funny ... okay that convinces me ... [21:38] what I don't understand is, why doesn't an empty .config work, and produce a default config [21:38] it's also there on oldconfig, but we didn't see it cause of the lot of messages it produces with yes '' | ... [21:39] I always thought that the # blabla is not set had a function ... [21:40] probably the options aren't set in our kernel configs either [21:40] well, it doesn't make much sense to have [21:40] a) CONFIG_BLABLA=y [21:41] b) # CONFIG BLABLA is not set [21:41] c) - empty - [21:41] if b == c [21:42] it would make sense, if b) would be used as do not set it, and c) would mean, new option, use the default ... [21:44] heh, some orkut photos are creative ... [21:44] somehow that is... when using oldconfig, a) will stay a), b) will stay b) and c) will be questioned [21:44] well, but not with default values ... [21:45] otherwise I have no explanation why the first try has such a small config ... [21:45] (because I accidentailly wrote an empty .config file instead of 'no' config file ...) [21:46] and most option seem to default to 'no' if there is an empty config file, but to yes, if there is no config file at all ;) [21:47] but maybe that is because then the build stuff sniffed around in /boot ;) [21:48] the whole build system (especially the default system) tries to be too smart IMHO ... [21:48] well hppa,hppa64,m68k,mips,mips64 also fails with defconfig :( [21:48] s/fails/fail/ [21:49] hmm... then i'll go for hppa(64) now [21:49] I would suggest mips, because the parisc page states, that the kernel needs some patching to make it work ;) [21:50] according to that page, the current kernel might not compile without that patches ... [21:51] basically I do not understand why those updates are not fed back into the mainline, as a non working kernel arch only is bloat IMHO, Rik any ideas about that? [21:52] ok, hppa was a really short build ;) [21:54] Bertl: because the ia64 "maintainers" hacked up the architecture independant code, instead of changing the ia64 code to not need those changes [21:54] so their changes are effectively unmergeable upstream, because they would break the other architectures [21:54] erhm, hppa=parisc not ia64 ?! [21:57] Doener: wat for a moment, I guess I screwed up the script again ;) [21:58] k [22:08] okay, started new run, will take some time ... [22:09] infowolfe (~infowolfe@pcp04891550pcs.frnkmd01.md.comcast.net) left irc: Read error: Connection reset by peer [22:29] Doener: okay, m68k, mips, mips64 failing with default config ... [22:30] for mips it seems to be caused by a special cpu=xxx statement which isn't recognized, maybe the cross compiler needs some extra options?! [22:38] ok, i'll see what i can find out [22:39] I assume that you are located in germany, right? [22:43] Guten Tag Deutschland. Wie Gehts es Ihnen? [22:44] hi mids! [22:46] Gruss Gott Bertl [22:46] :) [22:47] please refresh my memory, what did we work on last time? [22:47] ext2/ext3 quota deadlock [22:47] ahh, yes, thanks ... [22:48] Bertl: me? [22:48] yeah, but I fugured that already ... [22:48] s/fu/fi/ [22:48] Doener: added you to the Hall of Fame ... [22:50] ah i see :) [22:51] maybe we should add the IRC nick names on that page too ;) [22:54] mids: are you interested in 2.6 vserver stuff? [22:54] btw: my name actually is written with o-umlaut as yours is ;) [22:55] hmm, okay so Björn, right? [22:55] yepp [22:55] I'll correct that on the next run ... [22:56] thos funny umlauts .. they always make troubles ... [22:56] I didnt try 2.6 yet. I was planning to stick to 2.4 for a little while [22:56] i avoid to type ö whereever possible since as i always have the wrong character coding ;) [22:56] s/since// [22:57] yeah, know that, I changed the email back to 'oe' some time ago, as the lkml (and other list) people do not know about umlaut o ... ;) [23:11] Doener: yeah, it seems that the make defconfig on requires the CROSS32_COMPILE to make ppc64 work ;) [23:12] yeah, i included that on my test run this morning [23:13] as that was mentioned *wherever you found the hint last night* i didn't try without CROSS32_COMPILE [23:14] well, had a look at the Makefile ;) [23:21] Doener_zZz (~doener@pD9588172.dip.t-dialin.net) joined #vserver. [23:21] Nick change: Doener_zZz -> _Doener_ [23:28] Doener (~doener@p5082D3E6.dip.t-dialin.net) left irc: Ping timeout: 480 seconds [23:28] serving (~serving@213.186.191.180) left irc: Ping timeout: 501 seconds [23:34] okay, updated the overview ... (not many changes though ...) [23:39] Bertl: those new patches didnt help. in fact the new quota patch actually didnt compile.though it applies cleanly to 2.4.25-pre8 and pre7 once i fixed that i was right back where i started with inode quotas not working. [23:40] hum, okay, should not happen, we will try to fix that now, you have time? [23:41] yeah. [23:41] im patching a clean kenrle tree with those patches now so i can show you the line i had to change to make the kernel compile after applying the quota patch. [23:42] it was referring to a dquot sturcture that no longer exists. [23:42] okay, let me see, I'll update all patches to the latest version, and you try to describe/reproduce the issue with the core tools only, is that possible? [23:43] should be. [23:45] basicly inside a context it doesnt keep track of inodes unless you are removing files. if you ad new ones the quota system doesnt notice. until the next time quotacheck is run. [23:45] you can copy tons of files to your home directory and the inode count never shows up in quota. [23:46] how about outside a context, when you add a quota hash for the host (xid=0)? [23:46] does this also show that effect? [23:46] Bertl: im going to check that out now. [23:48] although im not sure how well that will work since im not sure i can use quotas outside of a context like tehy are used inside with the vroot and using fstype ufs. id have to try it with ext3 mounting. [23:48] well, it doesn't matter, you can change the type in the host /etc/mtab similar to the vservwer [23:48] you could even use the vroot device, although that has no effect on that ... [23:49] by the way, which quota format version do you use? [23:49] UID24/GID24 for now. i tried 32/32 before and had problems with it. [23:50] hmm, I was referring to the quota file format, so v2 or v0 (or even v1 ?) [23:51] Bertl: how do i figure that out? [23:52] hmm, a) how are your quota files named? (the ones created by quotacheck) [23:52] b) do we talk about quota or disk limits? ;) [23:52] aquota.user and group [23:53] okay, that looks like the newer format ... [23:53] what did you specify in the kernel config? v0 quota? [23:53] i enabled teh v0 support. [23:53] but i dont remember seeing an option to select a specific format. [23:53] okay, so you are most likely using v0 format now, which is okay ... [23:54] you could make sure that it actually is v0 if you specify -f vfsv0 on every quota* invocation ... [23:56] i hope i dont turn out to waset your time on this. [23:56] hey, no problem, the quota stuff needs to be updated anyway ;) [23:57] and documented better i think :) which id be glad to do once i figure it all out myself. i at least got block quotas working which is better than nothing. [23:57] well, others got all quota working ... but they do not use the latest version ... [23:57] mids: is your inode quota working or not? [00:00] --- Mon Feb 2 2004