**** BEGIN LOGGING AT Tue Aug 16 02:59:57 2011 Aug 16 08:01:37 morning everyone Aug 16 08:42:32 ppisati, hello Aug 16 08:42:41 hope you had a nice vacation :) Aug 16 08:44:02 the kind of vacation where you need some vacation after it to recover :) Aug 16 08:44:10 yep, it was good :) Aug 16 09:07:22 janimo: did you go to the linaro connect meeting? Aug 16 09:09:15 ppisati, no Aug 16 09:10:30 janimo: it's a pity, i would reallt liked to be there but when i booked my vacation i wasn't aware of it Aug 16 09:57:55 there is no ti omap4 ppa for oneiric ? Aug 16 09:58:20 no, TI usually fills up the PPAs around release or shortly after Aug 16 09:58:25 https://launchpad.net/~tiomap-dev/+archive/release/+packages natty/maverick Aug 16 12:01:31 infinity, i'm setting PREINSTALL_POOL_SEEDS="" for ac100 so we dont get build failures all the time (i dont feel like researching why ac100 server fails the pool installation, given the image wont be supported etc) Aug 16 12:33:25 hi, I'm testing youtube videos on pandaboard, ubuntu netbook 11.04 installed, firefox and gnash codec Aug 16 12:34:13 the videos don't play continuously, and stops a lot, even when the download of the data is not the problem Aug 16 12:34:39 does anyone knows any possible solution to this? Aug 16 12:35:23 fix gnash :) Aug 16 12:36:18 I don't know if the problem is the SD card, which one are you using? Aug 16 12:36:30 I'm with Sandisk class 6 Aug 16 12:36:35 honestly thoughm i think gnash is pretty outdated if it comes to compatibility with recent flash, you could try lightspark, which attempts to be compatible with more recent flash but will likely be vers slow due to SW rendering Aug 16 12:40:02 gnash is likely far from being optimized on omap4. what does the CPU look like while doing flash/gnash? Aug 16 12:40:27 it might eat one core :) Aug 16 12:40:45 but it should theoretically work ... kind of at least Aug 16 12:41:19 from what I read, it 'kind of work', no? Aug 16 12:41:28 it works on tegra for exactly one movie here it just doesn never close again after playing that :) Aug 16 12:41:50 so you have the plugin stuck in the bg until you kill your session Aug 16 12:42:00 nice Aug 16 12:42:09 i never tried gnash.. Aug 16 12:42:15 but hey, it works for one film ;) Aug 16 12:42:41 though that was maverick ... i havent tried since i must admit Aug 16 12:43:58 CPU is very busy when using gnash Aug 16 12:44:20 thats expected indeed Aug 16 12:44:31 given it does everything in software Aug 16 12:46:09 do you think a better SD card or disabling swap memory could affect directly in streaming? Aug 16 12:46:54 or just is a SW issue? I'm using Sandisk class 6 Aug 16 12:48:18 i dont think a faster SD will change behavior much Aug 16 12:49:41 janimo, there are un-uploaded changes in livecd-rootfs from you, ok to upload them (or do you need additioanl ones or so) Aug 16 12:49:49 ok, I will try lightspark codec then Aug 16 12:49:57 ogra_, sure, ok to upload Aug 16 12:50:14 hope it work Aug 16 12:50:17 SD card is lilely not involved in this use case. if your CPU is too loaded and if gnash cannot keep up with the rate of the decoder you might expect some freeze (a/v sync) Aug 16 12:50:19 I did not do so as they are not yet used at the moment, and so to go together with possible other changes from others Aug 16 12:51:04 well you could already test a livefs build Aug 16 12:51:15 if your packages are all in the archive at least Aug 16 12:51:40 ogra_, they are in NEW :) Aug 16 12:51:47 still ? Aug 16 12:51:49 bah Aug 16 12:51:59 the mx ones yes, at least they were this morning Aug 16 12:51:59 i thought they were out already Aug 16 12:52:09 binary or source NEW ? Aug 16 12:52:12 NEW queue is quite full, Aug 16 12:52:13 source Aug 16 12:52:17 hrm, k Aug 16 12:52:46 ogra_, https://launchpad.net/ubuntu/oneiric/+queue?queue_state=0&queue_text=mx Aug 16 12:53:16 ogra_, what is the ac100 situation, are those out of NEW? Aug 16 12:53:25 yep Aug 16 12:53:32 nice Aug 16 12:53:54 so you guys make build for MX and ac100 for 11.10? Aug 16 12:53:59 will that be full images? Aug 16 12:54:16 ac100 has to be a 2 part image due to the design of the device Aug 16 12:54:32 and what's the MX? Aug 16 12:54:51 its a tarball installer inside the initrd ... after that bit its like the panda image Aug 16 12:55:04 (oem-config preinstalled image etc) Aug 16 12:55:22 the mx5 will be a full SD image i think Aug 16 12:55:32 both are "community supported" :) Aug 16 12:55:43 cool Aug 16 12:56:39 i am upgrading my 11.10 alpha3 setup, already 377 updates... 219MB to download.. you guys push too many things ;-) Aug 16 12:57:02 heh, yeah Aug 16 12:57:09 upgrade daily ;) Aug 16 12:57:18 weekly gets you such numbers Aug 16 12:57:25 especially around freeze times Aug 16 12:57:36 where the upload frequency doubles or triples Aug 16 12:57:57 that's what i do with my laptop, and it's broken every other day ... so i am a bit more cautious with my panda ;-) Aug 16 12:58:18 lol Aug 16 12:58:26 yeah, one has to set priorities Aug 16 12:58:53 it's just that it takes more time to reinstall on panda than on a laptop! Aug 16 12:59:21 reinstall ? well, that should take 10-20min only from a fresh image Aug 16 12:59:21 do we have image for panda which are not preinstalled? Aug 16 12:59:33 netboot and netinstall are Aug 16 12:59:37 well, I want everything on external USB Aug 16 13:00:00 ok, no real live image? Aug 16 13:00:29 no Aug 16 13:00:43 unless you like 2h installs :) Aug 16 13:09:11 also has a problem with sound through HDMI, not working at all Aug 16 13:10:02 how to redirect music via HDMI cable? Aug 16 13:44:55 janimo, whoops, you forgot the closing ;; for your mx5 case statement Aug 16 13:45:03 * ogra_ uploads a fix Aug 16 14:29:36 ogra_, thanks, sorry Aug 16 15:02:01 ogra_: Why are you doing ac100 server at all? Or is that just for rapid testing turnaround and then you'll switch to a desktop image? Aug 16 15:02:24 i want to have the option while working on the images Aug 16 15:02:37 and i think the server team is actually interested in a server image Aug 16 15:02:53 or at least in a desk topless image Aug 16 15:03:03 hmm, that space wasnt intentional Aug 16 15:03:18 lol Aug 16 15:03:18 *shrug*... Kay... Just seems like a weird target for a "server" image. :) Aug 16 15:03:26 it surely is Aug 16 15:03:47 I should look at WTF is going on with the pool installation sometime this week while I'm multitasking with other stuff. Aug 16 15:03:51 it requires finer grained cdimage changes to supress building it Aug 16 15:04:05 I mean, it works on the only server image we "care" about, but it's effin' bizarre that it doesn't work for you. Aug 16 15:04:12 i'm not that far up the stack yet :) Aug 16 15:05:02 currently i wonder if i should make the initrd building fail for ac100-tarball-installer if the md5 sum is missing Aug 16 15:06:26 * ogra_ just noticed that the hook for installing the actual installer isnt in the package yet so i wonder if i can improve it somehow before adding it Aug 16 15:14:40 argh ! Aug 16 15:15:03 ogra_: argh to you too! Aug 16 15:15:16 df: Warning: cannot read table of mounted file systems: No such file or director Aug 16 15:15:17 stdin: Illegal seek Aug 16 15:15:17 mkinitramfs: for root missing /sys/block/ entry Aug 16 15:15:17 mkinitramfs: workaround is MODULES=most Aug 16 15:15:17 mkinitramfs: Error please report the bug Aug 16 15:15:22 wow Aug 16 15:15:36 heh, hey prpplague Aug 16 15:15:47 ogra_: what's cookin today? Aug 16 15:16:09 infinity, hmm, i supppose using MOYDULES=dep on a buildd isnt so clever Aug 16 15:16:30 * ogra_ wonders why else it could fail Aug 16 15:17:29 Yeah, MODULES=dep isn't going to work on hardware that isn't your target. :/ Aug 16 15:17:47 i'll go with list then Aug 16 15:17:50 that should work Aug 16 15:20:45 well, at least the automatic installing of linux-ac100 seems to work Aug 16 15:27:32 Hi, I was trying to use OTG on panda board as HOST Aug 16 15:27:46 I don't see volatge on the OTG... when I connect a device Aug 16 15:28:01 IS OTG as host supported on ubuntu, Natty Aug 16 15:30:59 check the kernel config in /boot Aug 16 15:31:07 is it enabled ? Aug 16 15:32:59 I have enabled all the options in menuconfig Aug 16 15:33:13 so you recompiled the kernel ? Aug 16 15:33:17 yes Aug 16 15:33:38 as package i suppose so you didnt break the distro mechanics ? Aug 16 15:33:58 I guess not Aug 16 15:34:26 I just enabled all the options under USB and re-built the kernel Aug 16 15:34:44 well, are you sure it is running that kernel ? Aug 16 15:34:58 yes.. Aug 16 15:34:59 (how did you tell it to use the new one= Aug 16 15:35:01 ) Aug 16 15:35:51 I copy the newly built image into SD-card and I boot from SD-card Aug 16 15:35:59 anyway, if you are sure the kernel options are correct, all i can say is that i dont think we tested OTG much Aug 16 15:36:23 well, that leaves you with an out of sync initrd Aug 16 15:37:03 one clarification... have u tested OTG as host Aug 16 15:37:16 you should just put vmlinuz in /boot, unpack a modules tarball and run update-initramfs, that will do all bits you need, including copying the kernel to the vfat Aug 16 15:37:38 ok... I will try that Aug 16 15:38:06 i'm not sure which OTG aspects we tested at all, and unfortunately our QA guy who did that testing is on vacation this week Aug 16 15:38:24 oh... Aug 16 15:38:32 i can tell you on monday ... that wont help you much right now though Aug 16 15:39:02 shall I ping you on monday on this Aug 16 15:39:44 anyway I will try what u have suggested Aug 16 15:40:31 thats just how to do it the clean way for an ubuntu install, i doubt it will affect your problem if you are sure that kernel runs Aug 16 15:41:23 I tested it as device (by connecting it to PC), board powers up Aug 16 16:49:51 infinity, hmm, looking at the live-build output i notice that i have two tarballs Aug 16 16:50:13 livecd.ubuntu-ac100.rootfs.tar-20110816-armel.gz and livecd.ubuntu-ac100.rootfs.tar.gz Aug 16 16:50:44 any idea where the one with the weird timestamp comes from ? Aug 16 16:51:36 (i assume its just a link or so, would be silly to have two actual tarballs) Aug 16 16:52:29 * MrCurious_ wonders about usb speed on pandaboard under ubuntu Aug 16 16:54:56 MrCurious, its slow, known bug Aug 16 16:55:22 MrCurious, if you ping the NIC it will get faster ... alternatively you can run with nosmp Aug 16 16:55:33 oh, i had a different one Aug 16 16:55:38 ah Aug 16 16:55:50 pinging nic didnt speed mine, it was just 1/2 as fast on 11.04 as 10.10 Aug 16 16:56:26 of cource i used wireless, and not ether, so that could explain that bit Aug 16 16:56:35 heh, yeah, definitely Aug 16 16:56:53 it only works if you ping the wired NIC Aug 16 16:56:58 i would happily move back to 10.10 if only the omap addons would work with that version Aug 16 16:57:12 they do ? Aug 16 16:57:21 in fact its the only version that has all bits Aug 16 16:57:21 * MrCurious_ shakes his head not a few weeks back Aug 16 16:57:41 when i tried 3 to 4 weeks back, 10.10 had a dependency issue on omap addons Aug 16 16:57:45 i bug reported that Aug 16 16:57:59 against the TI ppa ? Aug 16 16:58:16 maybe it was 4 - 8 weeks (time has sped up recently) Aug 16 16:58:21 heh Aug 16 16:58:55 i installed 10.10, then tried to add the omap addons meta package, and 2 packages had unresolvable dependencies, as such the omap addons would only artially install Aug 16 16:59:49 you should only need to klick the TI icon actually Aug 16 16:59:58 though it needs universe enabled manually first Aug 16 17:00:10 iirc there was a bug in 10.10 Aug 16 17:01:52 other than the only get 1/2 of ram memory bug? Aug 16 17:02:42 1/2 memory ? Aug 16 17:02:50 its 768M iirc Aug 16 17:03:18 whoops. half of half was missing i stored it in my brain poorly Aug 16 17:04:00 the way the day is going here, its like somebody called do-over on monday Aug 16 17:11:40 ogra_: The ones without timestamps are links to the ones with. Every livefs build looks like that. Aug 16 17:11:50 weird Aug 16 17:12:04 ogra_: Curious that you've only now noticed. :) Aug 16 17:12:19 given you build the one without timestamp from livecd-rootfs Aug 16 17:12:29 is it renamed at some point ? Aug 16 17:12:30 (Though I admit it comes out weird with tarballs, due to the multiple dots in the filename, and the timestamping code being a bit braindead) Aug 16 17:13:42 what is the tarball code in livecd-rootfs for then ? Aug 16 17:14:53 Hrm? Aug 16 17:15:06 (cd "binary/$INITFS/filesystem.dir/" && tar -c *) | \ Aug 16 17:15:06 gzip -9 --rsyncable > "$PREFIX.rootfs.tar.gz" Aug 16 17:15:06 chmod 644 "$PREFIX.rootfs.tar.gz" Aug 16 17:15:20 thats in auto/build Aug 16 17:15:20 That creates the tarball.. Aug 16 17:15:23 I'm having problems with installing ubuntu on beagleboard. I followed the instructions and dded the image to the SD card, but it won't boot. After connecting a serial com link I find that the problem is with booting the kernel. Is there some sort of flag I can use to make it more verbose? Aug 16 17:15:37 And then BuildLiveCD publishes it elsewhere. Aug 16 17:15:56 but it creates it without timestamp Aug 16 17:16:05 you said above the file is the one with timestamp Aug 16 17:16:07 sudo mv ${file} ${PUBDIR}${NOW}/${base}-${NOW}-${ARCH}.${ext} Aug 16 17:16:13 and the link is the one without Aug 16 17:16:17 ln -sf ${base}-${NOW}-${ARCH}.${ext} ${PUBDIR}${NOW}/${base}.${ext} Aug 16 17:16:30 I know what I said. :) Aug 16 17:16:41 ah, so they are shuffled around, k Aug 16 17:17:07 BuildLiveCD takes the non-timestamp one from $chroot/build and publishes it to ~public_html with some mangling. Aug 16 17:17:22 where the heck do you always pull out these code snippets ! Aug 16 17:17:25 Which is what leads to the timestamp being in a weird place. Aug 16 17:17:54 (But the timestamp thing really doesn't matter, since they go away when published again on cdimage...) Aug 16 17:18:02 * ogra_ greps for hours and doesnt find anything Aug 16 17:18:13 ogra_: I pull them out of knowing the code a bit too well? :) Aug 16 17:18:19 damned Aug 16 17:18:39 grr, so why does my update-initramfs fail now Aug 16 17:19:49 so i'm using something like: copy_exec /foo / || echo "foo" Aug 16 17:19:56 in my initramfs hook Aug 16 17:20:19 apparently that makes live-build think the hook exited -1 Aug 16 17:20:32 * ogra_ doesnt get it Aug 16 17:21:42 Why would you want to ignore copy_exec failing anyway? Aug 16 17:21:45 Just out of curiosity... Aug 16 17:22:05 because i need something that installs the md5 file if its there Aug 16 17:22:35 something that doesnt fail if its not there, so i can install the package but dont fail until live-build put it in place Aug 16 17:22:57 And are you sure you don't want copy_exec /foo /bar ? Aug 16 17:23:02 it doesnt fail here locally Aug 16 17:23:14 i want the md5 in / Aug 16 17:23:22 $2 is a target name, not directory Aug 16 17:23:35 ? Aug 16 17:23:44 its usuallsy /sbin or /bin Aug 16 17:24:15 Oh, you're right, the docs are out-of-date compared to the code being changed to allow people to use it like copy. :P Aug 16 17:24:29 s/copy/cp/ Aug 16 17:24:31 yep, since a while Aug 16 17:24:51 anyway, i see no reason why it would exit non zero Aug 16 17:25:27 the actual code is: copy_exec /installer.md5 / || echo "Warning: no tarball md5sum file found !" Aug 16 17:26:21 ... Aug 16 17:26:26 The ! is your problem. Aug 16 17:26:31 huh ? Aug 16 17:26:37 Running that on a local terminal will show you. :P Aug 16 17:26:42 its inside quotes Aug 16 17:26:46 dconrad@cthulhu:~$ copy_exec /installer.md5 / || echo "Warning: no tarball md5sum file found !" Aug 16 17:26:50 bash: !": event not found Aug 16 17:26:57 URGH! Aug 16 17:27:11 * ogra_ shakes head Aug 16 17:27:15 Double quotes get interpreted still. You want single-quotes if you want an unparsed string. Aug 16 17:27:54 i'm really wasting version numvbers today Aug 16 17:29:27 funny though it also works locally with an exclamation mark added Aug 16 17:30:02 ARGH!!!! Aug 16 17:30:44 * ogra_ forhead -> wood x 10 Aug 16 17:31:21 k, it isnt that line failing at all Aug 16 17:32:29 sigh Aug 16 17:37:59 * ogra_ uploads a fix and goes into hibernate until the next publisher is through Aug 16 18:04:18 I'm having problems with installing ubuntu on beagleboard. I followed the instructions and dded the image to the SD card, but it won't boot. After connecting a serial com link I find that the problem is with booting the kernel. Is there some sort of flag I can use to make it more verbose? Aug 16 18:11:12 dabukalam, removing the quiet and spash options from the boot cmdline shoudl make it more verbose Aug 16 18:11:47 how do I do that? beagle sutoboots, if i cut it, where do i go to remove those options? Aug 16 18:11:51 it's not grub :/ Aug 16 18:16:37 janimo, ping Aug 16 18:17:17 dabukalam, right, you need to modify the SDcard image Aug 16 18:17:20 in your computer Aug 16 18:17:48 mount the VFAT partition and alter boot.scr using mkimage Aug 16 18:18:11 there should be resources for editing/updateing boot.scr on omap, I have no link handy though Aug 16 18:18:42 dabukalam, what do you mean by 'sutoboots'? Aug 16 18:18:45 ah, autoboots Aug 16 18:18:48 autoboots ;) Aug 16 18:18:52 yeah, wait Aug 16 18:19:15 so you want me to modify the .img file that I dded to the card? or mount the sd card and modify the files in it? Aug 16 18:19:17 indeed, I would first try a new SDcard/ new image writing Aug 16 18:19:28 tried it 4 times so far Aug 16 18:19:35 those were tested and kernel should not panic on boot unless the hw is broken in some way Aug 16 18:19:42 try a final natty ? Aug 16 18:19:52 always get the same Uncompressing Linux... done, booting the kernel. Aug 16 18:19:58 and then nothing Aug 16 18:20:12 dabukalam, that may be because it takes long (minutes) on first boot Aug 16 18:20:14 it's not a hw problem because I'm doing a reinstall, it was maverick before Aug 16 18:20:16 as it resized the card Aug 16 18:20:20 the partition I mean Aug 16 18:20:21 ah shit Aug 16 18:20:39 so after rebooting 8 times, should I just turn it on now and give it time or flash the image again? Aug 16 18:20:49 should be documented I guess more prominently as it is indeed confusing Aug 16 18:20:52 yes, give it time Aug 16 18:20:59 k i'll brb :D Aug 16 18:21:11 although it can be annoying not knowing exactly how much to wait Aug 16 18:21:14 depends on SD card speed to Aug 16 18:21:16 too Aug 16 18:32:41 janimo: I'm at 10 mins so far, really tempted to pull the plug Aug 16 18:33:28 dabukalam, hmm Aug 16 18:33:42 then try editing the boot.cmd on the SD card Aug 16 18:33:50 and boot without quiet Aug 16 18:34:01 look for how to do it by googling mkimage boot.cmd Aug 16 19:39:57 Hello, I need the usbserial.ko and option.ko modules from ARM Aug 16 19:40:02 someone can help me? Aug 16 21:26:42 rsalveti: is there any particular preseed file you guys use to install when PXE'ing? Aug 16 21:46:18 RoAkSoAx: GrueMaster or mahmoh should know better Aug 16 21:47:08 RoAkSoAx: but were you able to install ubuntu using the netboot files? Aug 16 21:47:56 rsalveti: not really, at least not yet :) I'm trying to test different params for the kernel and see what results are shown cause most of the stuff is added automatically and might be needed to be dropped in cobbler Aug 16 21:48:50 RoAkSoAx: next week, we need to review the shipped cobbler seed Aug 16 21:49:05 Daviey: puufff we ship an ubuntu-server NQA seed Aug 16 21:49:08 pluys the sample seed Aug 16 21:49:16 Daviey: additionally, in orchestra we ship and ensemble seed Aug 16 21:49:17 :) Aug 16 21:49:24 RoAkSoAx: ok, makes sense Aug 16 21:52:42 rsalveti: i just cant get it to display anything on my monitor while it's supposed to be installing the system Aug 16 21:53:35 RoAkSoAx: hm, what is the issue you're seeing? Aug 16 21:53:48 guess I poke me about it, but I'm using another pc atm, so can't check the logs Aug 16 21:53:52 *you Aug 16 21:54:13 could be a kernel issue Aug 16 21:55:42 rsalveti: there's no signal in the monitors. It appears that the monitor obtains signal and then looses it Aug 16 21:55:45 rsalveti: http://pastebin.ubuntu.com/667705/ Aug 16 21:56:12 rsalveti: pxer file: http://pastebin.ubuntu.com/667706/ Aug 16 21:57:10 RoAkSoAx: are you using the dvi or hdmi output at your panda? Aug 16 21:57:27 rsalveti: hdmi to dvi Aug 16 21:57:39 hm, you're missing the vram cmd line settings Aug 16 21:57:47 [ 0.000000] Kernel command line: locale=en_US text auto url=http://192.168.2.118/cblr/svc/op/ks/system/arm9 host2 Aug 16 21:58:06 actually, you're missing all the specific kernel cmd line arguments for panda Aug 16 21:58:38 hm, but it seems your pxe config file is right Aug 16 21:58:55 could be that the kernel is not getting all the arguments or your boot log is not showing the entire line Aug 16 21:59:12 rsalveti: maybe, I';m trying with less arguments Aug 16 22:00:37 jcrigby: do you know if is there any cmdline size limitation when using pxe at u-boot? Aug 16 22:00:59 check RoAkSoAx's logs Aug 16 22:01:53 rsalveti: btw.. i don't know if this is a bug or not Aug 16 22:02:15 RoAkSoAx: Do you have serial console? Aug 16 22:02:23 rsalveti: but when PXE booting, u-boot does not look for ethernet address in the way of 01- Aug 16 22:02:33 rsalveti: it directly searchs for a Aug 16 22:03:29 hm, would be good to check the pxe spec to see what is expected Aug 16 22:03:33 rsalveti, yes it is actually the number of tokens in the line Aug 16 22:03:34 lool should know it Aug 16 22:03:51 jcrigby: because it seems his kernel didn't get the whole cmd line from u-boot Aug 16 22:03:54 it is the same line length issue we have had in other times Aug 16 22:04:12 next u-boot doubles from 16 to 32 Aug 16 22:04:36 you can get around it by spliting into two env vars and combining Aug 16 22:04:48 there is a comment in one of mahmoh 's bugs Aug 16 22:04:51 let me look Aug 16 22:05:51 rsalveti: AFAIK "Ethernet (ARP type 1) with address 88:99:AA:BB:CC:DD it would search for the filename 01-88-99-aa-bb-cc-dd." whicle u-boot is search for 88-99-AA-BB-CC-DD (ALL in uppercase Aug 16 22:05:58 Daviey: yeah that's where I got the log Aug 16 22:07:10 RoAkSoAx: are you tailing the tftpd log to see all the requests it is making? Aug 16 22:07:30 it's unusual if it is only making 1 request Aug 16 22:07:38 it should make about 10. Aug 16 22:08:25 I wonder if it is trying it's mac first, then 01- ? Aug 16 22:09:18 Daviey: i already checked and it is not :) Aug 16 22:09:25 Daviey: it never loks for 01- Aug 16 22:09:30 RoAkSoAx: naturally. :) Aug 16 22:09:50 Daviey: it looks for thenm for IP as in C0ABCETC Aug 16 22:09:55 Daviey: and then it dies :) Aug 16 22:09:59 RoAkSoAx: Just out of interest, can you pastebin the requests it is making? Aug 16 22:10:06 Daviey: sure Aug 16 22:10:10 Daviey: give me a sec Aug 16 22:10:29 I should just reproduce it here.. but i am lazy. Aug 16 22:11:38 rsalveti, see comments #3 and #5 here: https://bugs.launchpad.net/ubuntu/+source/u-boot-linaro/+bug/820116 Aug 16 22:11:39 Ubuntu bug 820116 in u-boot-linaro "usb start in uEnv.txt fails with test - minimal test >>> like /bin/sh" [Medium,Confirmed] Aug 16 22:12:05 RoAkSoAx: There are various non-compliant behaviors of the pxe implementation in u-boot; the path of least resistance if you want to do network boot today is to dump what requests it sends (with tshark or just look at the serial console output) and configure your tftp layout accordingly Aug 16 22:12:31 Daviey: http://pastebin.ubuntu.com/667715/ Aug 16 22:12:35 the whole pxelinux.cfg thing is entirely pxelinux specific for instance, and that doesn't relate to pxe at all Aug 16 22:13:13 lool: path of least reistence = suck. Aug 16 22:13:25 resistance* Aug 16 22:13:28 lool: yes that's what I noticed and ended up writing up a file manually Aug 16 22:14:44 specialcasing cobbler for ARM on this issue just feels rather wrong. Aug 16 22:14:47 Daviey: Yes, I'm not saying we shouldn't fix it, but I can also sympathize with people who want something to work with right now :-) Aug 16 22:15:06 Daviey: Well I've been poking at cobbler exactly to fix this kind of things Aug 16 22:15:16 Daviey: but I don't think it should special case pxelinux.cfg Aug 16 22:15:21 lool: i noticed your recent vital upload :) Aug 16 22:15:22 Daviey: the whole thing: http://pastebin.ubuntu.com/667717/ Aug 16 22:15:31 Daviey: :-) Aug 16 22:15:40 Daviey: I did discover some scary things though Aug 16 22:15:46 lool: oh? Aug 16 22:15:56 Daviey: such as if arch == s309, elif arch == powerpc elif ... kind of things Aug 16 22:16:01 *s390 Aug 16 22:16:27 Daviey: would you happen to know why it doesn't simply use dnsmasq for everything? recent versions provide all of DNS, DHCP and TFTP servers Aug 16 22:16:32 lool: I have the cobbler integration almost done though Aug 16 22:16:37 well that can be handled if we switch to using the tftpd.py rather than hpa-tftpd which is what we are currently doing Aug 16 22:16:46 I couldn't figure the urge for having their own python tftp server in it Aug 16 22:17:14 I have been thinking for a while on the fly TFTP file generation makes more sense Aug 16 22:17:26 Daviey: I would like if we would reuse dnsmasq because it's a popular implementation and is already used by libvirt IIRC Aug 16 22:17:45 lool: it does make losts of sense to have a cgi style generation, rather than flat files IMO. Aug 16 22:18:02 Daviey: Well, I see where you're going to, it's kind of seducing, but it feels a bit gadget-ish given practical needs? Aug 16 22:18:40 two concerns for me with tftpd.py are security and does it scale? Aug 16 22:18:46 the main limitation I would see is if your ruleset becomes really large and you don't want to write a large file and restart daemons Aug 16 22:18:48 security i haven't sniffed at Aug 16 22:19:13 but does it scale? well providing it's multi-threaded it *should* imo.. it's not a heavy task what it is doing Aug 16 22:19:52 well hpa-tftpd or dnsmasq shouldn't require restarts Aug 16 22:20:51 but if fixing uboot is going to be a real pain in the butt, then suggesting use of a patched tftpd.py might make sense Aug 16 22:20:53 Daviey: yup; if usign isc-dhcpd it would though :-/ Aug 16 22:20:54 RoAkSoAx: thoughts? Aug 16 22:21:16 Daviey: patching u-boot is trivial, but implementing something sane isn't :-/ Aug 16 22:21:20 lool: Hah, i didn't know isc-dhcpd supported tftp :) Aug 16 22:21:50 the current implementation is essentially a partial copy of what pxelinux would be doing with a crucial difference: the behavior is hardcoded in your u-boot binary rather than in the first pxelinux bits you load from the network Aug 16 22:22:04 lool: Well s/01-// seems kinda wrong.. so prepending that string might make sense? Aug 16 22:22:40 Daviey: i already though on writing a file for tftp such as rather than 01- for arm cobbler systems Aug 16 22:22:45 Sure, this is trivial; the question is what is this u-boot implementation actually complying to Aug 16 22:22:47 thought* Aug 16 22:22:54 I actually hoped we could use uboot PXE to chainload ipxe.. but ipxe seems to be amd64/i386 specific at this time. Aug 16 22:23:03 (although i did see some work to resolve this) Aug 16 22:23:10 Daviey: haha, I was looking at ipxe/gpxe 10 days ago over Connect Aug 16 22:23:22 lool: success? Aug 16 22:23:23 zero ARM support, and the design will require major code updates to support it Aug 16 22:23:29 lovely. Aug 16 22:23:31 #fml Aug 16 22:23:54 Daviey: But I have good idea on approaching this, in fact you folks just touched what I've been giving thoughts in the last weeks, would you like to get on the phone about it? Aug 16 22:24:01 or mumble Aug 16 22:24:09 RoAkSoAx: well, what is an arm cobbler system? :) Aug 16 22:24:28 lool: are you still on euro time? Aug 16 22:24:39 I am Aug 16 22:24:53 lool / RoAkSoAx: Good for a call tomorrow? Aug 16 22:25:01 Daviey: a cobbler system based of a "arm" import Aug 16 22:25:04 Daviey: sure Aug 16 22:25:08 RoAkSoAx: i see! Aug 16 22:25:22 wfm Aug 16 22:26:09 RoAkSoAx: yep, ultimately we might want to import images, perhaps with a different import backend though Aug 16 22:26:15 but there are so many ARM issues Aug 16 22:26:32 for instance pandaboards currently don't get a good MAC address on boot from an eeprom Aug 16 22:26:54 there's a piece of logic in linux to generate a local address from the dieid (cpuid), but I don't think it's in u-boot yet Aug 16 22:27:01 and even that would be a bit screwy Aug 16 22:27:10 well pandaboards are not /really/ that importiant, it's really to demostrate proof of concept for a real ARM server. Aug 16 22:27:18 So hopefully that gets resolved. Aug 16 22:30:39 lool: Cobbler already "imports" images Aug 16 22:31:24 lool: basically, if everything on http://ports.ubuntu.com/ubuntu-ports/dists/oneiric/main/installer-armel/current/images/omap4/netboot/ is under a path, you can simply import them into cobbler Aug 16 22:31:31