[00:00] oh, but you've seen the segfaults? [00:00] that doesn't look like you're even getting into qemu [00:01] yeap.. it use to segfault after apt-geting everything in the second stage... [00:01] have you used up all your loop devices? [00:01] well, I seem to have a workaround for the segfaults [00:02] something with the lucid userspace confuses qemu; like a libc mismatch or something [00:02] nope.. none's showing up with mount... (the script auto umounts any it finds..) [00:02] if you exec into a new script after debootstrap finishes, you can proceed normally [00:03] rcn-ee, add a -x to the set -e in rootstrap, and send me the next log you get from it [00:04] I'm using a fork rootstrap, but I doubt they've changed a whole lot; it's still just a glorified wrapper around deboobstrap :) [00:04] forked, rather [00:05] yeah mine is too, the useful stuff i send up to oliver... [00:05] oh, I see what's going on there now [00:05] dying at gnupg [00:05] or was that the old segfault? [00:06] i don't remember.. ;) i'll fire up my script, i got a full mirror here so it's prety quick.. [00:07] set -xe in rootstock as well as the installer script that iirc is created from within rootstock [00:08] also, somebody should make an smp versatile kernel :p [00:08] i think lool's trying to get 512Mb to work first.. [00:08] meh, just make a big swap device on a tmpfs [00:08] swappiness=100, and you're laughing [00:09] I've gotten a nice speed bump from this.. ;) http://bazaar.launchpad.net/~beagleboard-kernel/+junk/image-builder/annotate/head:/patches/05-use-real-hardware.diff (i give it nice fast harddrive..) [00:11] I'd expect that to be similar to enabling writeback caching on a normal device file? [00:11] performance-gain-wise [00:11] -drive file=${IMAGENAME},aio=native,cache=writeback [00:11] is what I use [00:12] i think so... it would also help if i used that drive as my filesystem, it's about 2x faster... [00:16] I spoke too soon as usual [00:16] I got significantly further along, but it still ended in a segfault [00:18] mine just reached apt-get... should crash any minute... [00:19] there isn't any qemu special sauce compiled into those qemu kernels is there? [00:19] I guess when I put it like that.. [00:20] i think the cortex-a8 emulation is relatively new in qemu... [00:20] but a karmic debootstrap is rock solid [00:21] yeap.. and all those packages where armv6.... [00:21] ahhhhhhhh [00:21] well that makes some sense of it [00:21] so one question that comes to my mind... is qemu emulating all the cortex-a8 errata properlly.; ) [00:22] * cwillu_at_work giggles and groans [00:22] and we have the thumb2 bits on too.. which lucid is the first distro to ever try that... Angstrom is just arm mode... [00:22] I see now why we build our packages on actual arms [00:23] now that 2.6.34's almost in shape, i'm thinking of taking some time to make a 'arm' rootstock... [00:24] wouldn't be hard; you could basically search and replace the qemu calls with chroot [00:25] exactly... chroot's are easy.. and we aren't emulating anything.. [00:25] I just... wish I could saturate my server doing builds rather than using the real hardware [00:26] that's one of the big advantages of rootstock... [00:27] stupid segfaults [00:28] * cwillu_at_work tries cortex-a9 for kicks [00:28] have you tried the beagle qemu target? [00:28] didn't see such a thing [00:29] i see lool running "qemu-system-arm -M beagle" every once in a while, not sure if it's mainline.. [00:29] doesn't show up in qemu -M ? [00:35] yeap.. just segfauted... [00:35] what's weird, it gets so close to being done... [00:36] feels racy [00:37] didn't get the whole log.. here's right after apt-get was done in the internal script... http://pastebin.com/V3LTkeEZ [00:38] ya, not much interesting going on there :/ [00:38] of couse, when i do the next run, it'll segfault earlier... for some strange reason on this pc... the first run will get that far... [00:39] when they get back from uds, fingers crossed for a duh moment by a qemu guy.. [00:40] * cwillu_at_work sighs [00:40] thanks for reminding me that I'm a programmer, and really should be able to solve this sort of thing myself :p [00:40] * cwillu_at_work apt-get sources qemu [00:41] nah.. 'arm' roostock... then once qemu is fixed, you can run it in emulation.. ;) [00:42] I can hurt you... :p [00:44] one thing i haven't looked direcly into.. is the qemu-kvm a branch/fork of qemu [00:46] -kvm is the mainline of it now, isn't it? [00:47] well kvm is in the mainline kernel... so qemu-kvm would just also include the extensions needed to utilze that.. [00:48] the only really big thing, in lucid rootstock gets alot farther then qemu in squeeze.. [00:50] what release date was squeeze approximately? [00:50] or it's not out yet? [00:50] mid summer... still over 1K bugs.. half patched in last report.. [00:51] sorry, it's due this summer? [00:51] yeah, they were saying summer 2010... i'm thinking winter 2010/11.. ;) [00:52] ls [00:52] other then being slow compared to karmic/lucid it's been reliable on my beagles.. [00:53] nothing much in debian/patches [00:55] there's been some activity... http://git.kernel.org/?p=virt/kvm/qemu-kvm.git [00:55] to rebuild or not rebuild.. ;) [00:55] ya, just not in the patches dir :p [00:56] 13 days ago [00:58] qemu in lucid is 0.12.3 whereas git is 0.12.4 + 2 weeks of patches.. [00:59] or the head is taged, not out.. [03:37] mmm, butter. [03:37] Good with toast and garlic. [03:37] Or more precisely, a French baguette, I think. [08:13] how _do_ you update our qemu package? [09:26] Guest83715, so about your suid problem, doesnt genex2fs just take the permissions that are created when creating the chroot ? [09:27] (i.e. does it reset anypermissions that already exist ?) [09:35] ogra: sorry, internet here is **** [09:35] * NCommand2r is on a 3G card now [09:36] NCommand1r, i answered the mail, that should make it better ;) [09:36] ogra: we can't extract the raw files preserving permissoins without unsquashfs or tar running as root. Its unclear if genext2fs saves permissions as it builds the filesystems [09:36] why would you extract anything ? [09:37] ogra: you have to extract the squashfs with the preinstalled system on it, no? [09:37] livecd-rootfs should just provide us a generated ext3 image [09:37] i think you are fiddling with the wrong bit here :) [09:37] ogra: OH, that will work [09:37] ogra: I think your right. [09:37] * NCommand2r goes to hide in shame [09:37] nah [09:38] was a valid issue ... just the wrong design assumption [09:38] ogra: yeah, fiddling with the wrong bits, although if we're doing this on the image build system, we can just loopmount [09:38] which solves a lot of headache IMHO, and lets us use EXT4 [09:38] i'll write up the spec tomorrow if i'm back from my fathers place, that should make everything clearer [09:39] ogra: yeah, I find writing code during UDS can be counterproductive when one's mind is exhausted [09:39] for code sanity we should just use genext2fs instead of mksquashfs in livecd-rootfs [09:39] ogra: but if we generate the extX image on the live image builder, then we should just use a loop mount and ext4 [09:39] that can then keep a similar command structure [09:39] */2 cents* [09:40] loop mounting will add a new function, using genext2fs will just be a case check and one additional commandline [09:40] ogra: well, sorta, we still need a new code path to only download that live image, and then excute the boot scripts. Its just not a drop in replacement (sadly) [09:40] it is [09:40] ogra: in d-cd? [09:40] hrm [09:40] let me wait for your spec [09:40] i'll show you some code3 next week :) [09:40] then it will be much clearer [09:40] -3 [09:41] ogra: yeah, lets save this for next week, I think we're both exhausted from UDS :-/ [09:41] (where are you anyway?) [09:41] #right, i'm massively sick ... [09:41] at home [09:41] carried a bad cold with me ... but dont want to miss my fathers 80est so i'll vanish now [09:41] ogra: ah, you have the ubuflu [09:42] * NCommand2r has it too [09:42] i'll write up the spec tomorrow and do some PoC code for livecd-rootfs [09:42] ogra: I'll assist with the d-cd hacking once I understand your design and help test it all out [09:43] its identical to the other livecd builds, just that the image name doesnt end with the squashfs suffix [09:43] use the existing omap code [09:44] one issue we have here is that MLO needs to sit in the first block of the vfat [09:44] and that u-boot.bin needs to be in the vfat as well ... beyond that it should be pretty much identical [09:44] oh, wait, no ... [09:44] we need code for a second partition to dd the ext2 into ... [09:45] so it should be a merge of the existing omap and imx51 code [09:45] ogra: the issue is that we don't want the code to add a pool or anything to the image [09:45] ogra: that needs to be removed [09:45] right, i'm talking aboout live [09:46] we need a two partition image ... code is in imx51 [09:46] but here the first partition isnt raw but vfat [09:46] ogra: but the live image still has a pool on it and other files beside the squashfs [09:47] ah, right, that will need special casing, we dont want to use the ship seed [09:47] or ship-live or however it is called atm [09:47] there should be code in d-cd wheer you can just disable it [09:48] ogra: yeah, so this isn't quite as trivial as we wish, but we'll figure it out. lets save it for Monday [09:48] well, we have all code already, its really more a copy paste job and one extra special case [09:49] anyway, i'm off [12:43] ogra, you on perchance? [12:43] I still have your serial cable... [13:01] I managed to get qemu-arm running with more than 256mb internal memory [13:03] at least kernel reports the additional memory ;-) [13:06] anyone knows a quick check to see whether or not the 256+ mb memory can actually be used? [13:08] this is qemu-0.10.6 using versaltilepb on arm with 850mb memory by the way [13:44] Hi all. The arm ubuntu 10.04 info is hard to find. Any change that the link on the main pages can be updated to not point at the Lucid info and not the Karmic info? [13:44] umm _to_ point to the Lucid info [13:47] So, I have been playing with the Lucid netbook install image for beagle [13:48] I have the netbook install on one SD and am installing to another in a USB reader. [13:49] From what I can tell the installer does nothing to actually set up u-boot to actually boot the new image, correct? [13:50] If so I would like to update the wiki page to describe how to do that. [13:51] I just don't want to do that if I missed something. (I always walk away while the files are coping. I have never seen it finish. I alway come back to just the purple wallpaper. [15:09] > ogra, does netbook install adjust u-boot parameters at all? [16:11] rcn-ee, checked latest qemu git tag, still segfaults [16:17] well it's consistent.. ;) [16:17] :) [16:19] * cwillu resigns himself to a future of beaglestock [18:01] cwillu: what machine are you emulating on qemu? [18:07] Omegamoon, rootstock context, so versatile.. [18:34] rcn-ee: I just tried to cross the 256mb memory boundary on versatilepb in qemu [18:35] how'd it go? it didn't fail like before? [18:35] kernel is reporting 850mb now, and I'm building 'apt' just to see if it works [18:36] cool [18:36] I'm hitting a different boundary now, it seems... [18:36] cpu is 100% all the time ;-) [18:38] the qemu-system-arm process is using 306mb of memory already, so I guess it works [20:35] ogra : Are you on? [20:36] I've got your USB -> RS232 adapter. If I can get your address before leaving Belgium, I can pop it in the mail [22:09] * Viduka Discounts!! Our Special Limited Time Offers Up To May,22!!!New BranD!! Notebooks,Plasma and LCD TV's.Buy your electronic needs at our unique prices. Laptop Sony VAIO® VGN-FW590FFD-575,57$!!!Apple MacBook® Air MC234LL/A-695,27$!!! http://www.elplace.com/