[02:34] ErKa (keryell@keryell.pck.nerim.net) joined #fai. [02:37] ok, my fai cd creation script now burns a cd with a customized initrd on it (built with mkinitramfs-kpkg) [02:37] so, write cdrom drive detection routines, then make a mirror FAI will be happy with, then make FAI happy. [02:37] this isn't easy :) [02:55] ErKa (keryell@keryell.pck.nerim.net) left irc: Ping timeout: 480 seconds [03:02] hmm. extract the initrd, add a 32 byte file, re-compress it, and its suddenly too large. [04:07] ok, looks like i'm rebuilding the cpio archive using the wrong format... [04:31] and now in the wrong file order.. [05:23] aha. but that worked. [05:25] now i can start inserting code into the executable path. :) [08:56] ErKa (keryell@keryell.pck.nerim.net) joined #fai. [09:00] MT (~MT@dove.informatik.tu-muenchen.de) joined #fai. [09:23] Mrfai (~lange@129.215.255.137) joined #fai. [09:23] Mrfai (~lange@129.215.255.137) left irc: [09:24] Mrfai (~lange@129.215.255.137) joined #fai. [09:38] MT_ (~MT@80.109.167.79) joined #fai. [09:46] MT (~MT@dove.informatik.tu-muenchen.de) left irc: Ping timeout: 480 seconds [10:05] ErKa (keryell@keryell.pck.nerim.net) left irc: Ping timeout: 480 seconds [10:16] MT_ (~MT@80.109.167.79) left irc: Ping timeout: 480 seconds [10:34] ErKa (keryell@keryell.pck.nerim.net) joined #fai. [11:00] ErKa (keryell@keryell.pck.nerim.net) left irc: Ping timeout: 480 seconds [12:50] horst (irc@shining.mistrust.net) joined #fai. [13:11] MT (~MT@M343P031.adsl.highway.telekom.at) joined #fai. [13:12] Nick change: MT -> Guest1502 [13:13] Nick change: Guest1502 -> MT [13:13] Mrfai? [13:14] ja [13:15] ich denke, die partitioning-info kannst Du zu #380629, #330915, #277045, #356862, #416633 [13:15] hinzufügen [13:15] #330915 und #356862 sind wohl davon schon erledigt [13:15] und wenn Du schon beim Aufräumen bist [13:16] #364763 ist IMHO kein Bug [13:16] einfach eine falsche Config [13:16] und wenn wir schon dabei sind [13:17] ich denke darüber nach, die diversen Skript-Teile für shdd2 in ein einziges Skript zusammenzuführen [13:17] dann ist das zwar recht lang, aber das war das alte ja auch [13:17] (und die Version von Sam ist nicht besser in dieser Hinsicht) [13:17] und um das ganze gleich noch weiterzuspinnen [13:18] was hältst Du davon, das neue Tool in FAI aufzunehmen [13:18] aber (vorerst) nicht zu verwenden [13:18] MT du kannst die partitioning-info auch hinszufuegen. #364763 schau ich mir jetzt an. [13:18] ok, mache ich [13:18] so, jetzt eins nach dem anderen. Mein Fifo ist sonst ueberfuellt. [13:19] sorry :-) [13:19] also um die Infos kümmere ich mich [13:19] hast Du also auch noch nicht gemacht, oder? [13:19] weil die BTS-E-Mails manchmal ja etwas länger dauern... [13:20] #364763: ich werde mal mailen das es unreproducible ist oder moreinfo verlangen. Wenn er sich dann nicht meldet schliessen wird den Bug. [13:24] so. next itme. [13:24] Ich werde mir auf der debconf (bzw. im debcamp) shdd2 ansehen. Ich moechte es auch bald in FAI einbauen ohne es per default zu nutzen. [13:25] Ich werde aber erst mal sehen ob der Codingstyle so OK ist und ob ein grosses oder mehrere kleine files besser sind. Also aendere noch nichts daran bevor ich mir das genauer angesehen habe. [13:25] ok! [13:28] FRage ist auch ob nicht ein besserer/schoener name fuer das script nett waere. So was wie TDT "THE disk tool" oder movotst wie in "michael's own version of the setup_harddisk tool" :-) [13:29] da würde ich aber tdt bevorzugen :-) [13:30] oder ddo: disk data organizer, oder pt "The partition tool" wobei es ja mehr als nur partitioning macht. [13:30] wenn Du code-reviews machst, schreib einfach in den Code rein, was Du schlecht findest oder nicht verstehst oder so [13:30] werde ich machen [13:30] Ich werde in rot reinschreiben ;-) [13:30] wenn DU TODO oder FIXME reinschreibst genügt's schon, das markiert mein vim gleich fett gelb [13:31] das versuche ich mir zu merken [15:16] MT (~MT@M343P031.adsl.highway.telekom.at) left irc: Quit: MT [15:43] meandtheshell (~markus@85-124-207-255.dynamic.xdsl-line.inode.at) left irc: Quit: Leaving. [15:48] meandtheshell (~markus@85-125-192-9.dynamic.xdsl-line.inode.at) joined #fai. [15:56] Mrfai: wie heit denn dieses script, das der toni da bearbeitet hat? [15:59] Ich meine es war dies hier: http://faiwiki.informatik.uni-koeln.de/index.php/Source:SW_RAID_LVM [15:59] ah, habs schon... [16:03] oder das andere... davon sind jetzt auch schon zwei Versionen im Wiki... [16:05] Mrfai: hoer ma auf mit dem bugfixen - die ganzen mails nerven :) [16:40] nausicaa_ (~popo@AAmiens-153-1-51-40.w90-18.abo.wanadoo.fr) joined #fai. [16:40] hi [16:40] Nick change: nausicaa_ -> nyal [16:40] better :) [16:41] i don't understand [16:41] The programm eject is not needed inside the nfsroot, but inside the [16:41] install eject onto your install client and have it in a package_config [16:41] file. [16:41] eject is used in subroutine script [16:41] not in the chroot [16:44] so i should have eject package in the nfsroot [16:47] Mrfai ? [16:51] nyal: The only use in the nfsroot I can imagine is when doing fai-cd install - to eject the cd after the install... [16:52] nyal: but maybe I don#t get the whole picture you are talking about... [16:52] yes exactly lazy [16:53] i use fai-cd [16:53] so i need eject because i use bootcd [16:53] nyal: Hi [16:54] nyal: eject is used in subroutines, but there a small shell script is created which will be executed using chroot [16:55] i don't really want eject in my system [16:57] nyal: you could not use eject from inside the nfsroot, because then you will execute a command that will remove itself from the directory tree. That does not work. You have to copy eject, add needed libs and the /dev and maybe the /proc into a small chroot environment and then execute eject from this chroot. [16:57] It's easier to use the eject that we install onto your new system. [17:02] i don't have it [17:02] ok i understand [17:03] i can't cut my legs :) [17:06] thanks Mrfai [17:16] nyal (~popo@AAmiens-153-1-51-40.w90-18.abo.wanadoo.fr) left irc: [17:36] ok, i have a makeinitramfs-kpkg generated initrd on my CD.. now to get it to scan for a cdrom. [17:53] juri_: makeinitramfs-kpkg? Why not just using mkinitramfs? Oh I see. There are pretty much the same [17:59] juri_: I've put another "scan-for-cdrom-devices" example to http://www.informatik.uni-koeln.de/fai/download/linuxrc/. It's called live, and it has some nice code to create the list of all removable devices. [17:59] have a look at find_livefs() { [17:59] excelent. [18:00] I hope to talk to the maintainer tomorrow. He's also here at debconf [18:00] i just wrote a stub to drop me to a prompt when it goes to mount the root FS. hopefully, i can mount it manually this time, and it should try to start FAI. [18:01] it was easy enough to stub. just stop passing a root=, and pass a "boot=cdrom", and it looks like it will execute a script in scripts/cdrom. [18:01] i expect to be installing from CD in about 7 hours. :) [18:03] :-) [18:06] i'm testing that last reported behavior as we speak. [18:07] mind you, i'm making a metric ton of assumptions, which happen to be correct for my environment (cdrom is ide, etc etc), so the result of what i'm doing right now will b basically "heres one way to do it, in an ideal world". :P [18:08] which is progress, of a sorts. :) [18:11] juri. IMO initramfs is very good and modular. Loading of kernel modules of all sorts of CDROM is already done my initrafs-tools. So if you just iterate over the list of removable devices (also usb sticks) this would be fine. Do not iterate over '/dev/hd* /dev/sd*' like some linuxrc script are doing this [18:12] sounds reasonable to me. i was thinking of iterating over all removable devices without partition tables. [18:15] this script you passed me does everything, being called as init.. i'm assuming you'd like me to pull the relevent sections out, and add them to my "modify initrd" procedure? [18:15] right now, my make-fai-cd.sh script is really 40 or 50 lines of no error checking banging through. :P [18:16] juri_: yes pull the relevant sections out. [18:17] My (long term) goal is to push this lines of code into the official initramfs-tools. Then fai-cd only needs to call mkinitramfs [18:18] actually, the approach i'm using is just dropping a single file (scripts/cdrom) into the initrd. seems easy enough to get merged upstream. [18:19] despooling and rebuilding the ramdisk properly was a headache. :) [18:20] I think mkinitramfs will do everything for you if you just drop your script into the correct directory. Something like .../initramfs-tools/scripts/local-premount or a different directory [18:21] sounds likely to me. [18:22] no-one ever said i was doing it right. :) [18:29] well, good news is that kinda worked. it sourced my cdrom script. which has an error in it.. ;P [18:34] ah. } out of place. [18:34] I'm extending faimond. I can now call fai-chboot to change the pxelinux.cfg file if an install client has finished. Question, should be that default behaviour, or only enabled by an option? I can implement both. [18:35] sorr, i'm a cd user. i have no input here. :) [19:20] ErKa (keryell@keryell.pck.nerim.net) joined #fai. [19:25] yay. that approach i outlined earlier works. :) [19:39] hmm. i see no pivot_root magic here. [19:46] here? Where? [19:55] Mrfai: i guess an option would be the best. since at staging time (where you test all kinds of stuff, including faimond) it would be a horrible PITA to do fai-chboot -I ... again :) [19:55] we will only call fai-chboot -d [19:56] this can be reversed with fai-chboot -e [19:57] but it would have to be done manually. i'm happy enough to only change configuration space and just reboot the client [19:59] juri_: I really like to look into your code ASAP. Please email it to me even if it's not finished or perfect. I just like to see how it works. [20:03] mrfai: as soon as i've got fai itsself running, sounds great. :) [20:04] right, the initrd refuses to run /etc/init.d/rcS, claiming its not executable (its a symlink), but when i send it directly to /usr/sbin/fai, fai reports "Please give more parameters if not run from the nfsroot". [20:09] juri_: the initrd should not run rcS. The initrd hook you write should only find the file system on the removable device, mount it. Then the default initramfs scripts will chroot/pivot_root into it and run the rest. You do not need to do this. [20:09] juri_: Please show me what you've done so far. Otherwise you may waste time with things, you do not need to do. [20:10] If you do not like to send me an email, use paste.debian.net [20:10] i'mm blast a mail to the list, so i can humiliate myself publicly. :) [20:11] i'm not doing that part. i passed the default scripts etc/init.d/rcS to run, and it paniced and died, claiming it was un-executable. [20:11] now, i don't know if its chrooting/pivot_rooting, but i do know /usr/sbin/fai, when passed as an init, is claiming not to have enough arguments. [20:12] my next step is to see what arguments its getting passed, and what it expects. [20:13] fai does this: if [ ! -f /.THIS_IS_THE_FAI_NFSROOT ....] [20:14] oh. i changed that to ".THIS_IS_THE_FAI_CD. ;P [20:14] juri_: if you show what you've done, I may help. [20:15] Action: juri_ fixes. :) [20:15] that should be the "last bug" then. that, and i havent started cutting and pasting your detection routines. right now it invokes an sh, i mount the cd, and it continues. :) [20:24] Action: juri_ rebuilds [20:51] juri_: please show it to me! [20:54] yay! that was it. fai is now firing up. [20:54] er. [20:55] it still looks like the initrd isnt pivoting root, and instead is just calling /usr/sbin/fai. [20:55] Action: juri_ prepares a big mail [20:56] or not. lets just paste.debian.net [20:59] juri_: I just ask the live-helper maintainer. chroot and pivot_root is dead. Now they use mount -move [21:00] well, it looks like thats not getting done at the bottom of init (the sh script in / on the initrd) [21:01] MT (~MT@dove.informatik.tu-muenchen.de) joined #fai. [21:02] uhm, so how do you use this damned thing? :P [21:02] this mount -move is not in the default initramfs script, but its in the live script [21:03] i pasted my code, and it just returned me to the page to enter. i would have assumed it would have given me a url. [21:03] But I'm also wondering how this works. [21:03] I will ask him tomorrow [21:04] juri_: where did you paste? [21:04] paste.debian.net [21:04] oh, ok. [21:04] can;t see anything? Which name? [21:05] http://paste.debian.net/30134 [21:05] thats the script i use to build a CD. [21:06] my scripts/cdrom is 30135. [21:08] my menu.lst is 30136. everything else is pretty normal. [21:08] i'm using the last entry in menu.lst for testing. [21:08] i told you this is pretty ugly. :) [21:09] where is the code for detecting the cdrom? cdrom-top, cdrom-bottom [21:09] dosent exist. [21:09] i was just copying what the nfs stuff does. [21:09] it calls a top and bottom that don't exist, as well. [21:09] i just type "mount /dev/hdc /root" when that sh prompt comes up, and exit. that mounts the drive. [21:10] arent you glad i don't commit often, with that coding style? :) [21:11] no comment :-) [21:12] but I like to save you time, but pointing you to the right direction for a solution. [21:12] i'd prefer if fai-cd was a usable thing again. :P [21:13] i don't know why /usr/sbin/fai is now freaking out, i'll know more in about 15 minutes, once my next CD builds. [21:13] i'm going to try firing up a bash, instead of fai, to see what the bash says the environment looks like. [21:14] juri_: it is. Didn't you read http://www.uni-koeln.de/bin2/maillist/linux-fai/20070523.114038/175201 ? [21:14] sure. if you want to install sarge packages. :P [21:15] that pegs out my "unacceptable" meter. :) [21:16] yes, i'd rather spend all weekend rebuilding everything from scratch than use packages that are no longer in stable. i'm strange. :) [21:16] ok. No problem for me. I just want to point to a fast solution. I also prefer a clean solution. [21:18] well, this at least shows the "easy" place to hack the normal debian initrd solution. [21:31] now i need to fix the cdrom script to do scanning.. [21:43] Mrfai (~lange@129.215.255.137) left irc: Quit: leaving [22:23] MT (~MT@dove.informatik.tu-muenchen.de) left irc: Ping timeout: 600 seconds [22:55] hmm. ok, its running fai properly, but the filesystem is completely read-only. [23:11] touch /boot/RUNNING_FROM_FAICD [23:32] ErKa (keryell@keryell.pck.nerim.net) left irc: Ping timeout: 600 seconds [23:47] so, right now, my FAI CD dosent run properly, cause its read only. [00:00] --- Mon Jun 11 2007