**** BEGIN LOGGING AT Thu May 20 03:00:03 2010 May 20 03:40:46 playya_: what are you doing with ubuntu on pre/milestone? **** BEGIN LOGGING AT Thu May 20 03:51:34 2010 May 20 06:51:23 tmzt, nothing yet. we're using OE to build our distributions May 20 07:19:17 ogra_cmpc: do you have time? May 20 07:19:51 ogra_cmpc: i saw this error "Uninitialized omap_chip, please fix!" in the dmesg you posted May 20 07:27:14 Hey, trying to install Lucid on a BeagleBoard (i.e., an OMAP device), it gets as far as the partitioner without issue, but then cannot find the SD card, depite having booted off it (though my understanding is that only the boot loader touches it then, so that's not conclusive about the kernel finding it) May 20 07:28:22 The more annoying part for me is that an identically formatted SD card worked on another BeagleBoard here with no problem. May 20 07:32:42 gsnedders, which BB revision do you have? May 20 07:33:12 gsnedders, and which lucid image did you try: netbook or server? May 20 07:35:24 zyga: C4, I use using the image from https://wiki.ubuntu.com/ARM/BeagleNetInstall May 20 07:36:26 gsnedders, strange, anything seemingly relevant in the system log? May 20 07:36:47 Not that I saw May 20 07:37:22 did you try _exactly_ same sd card in other devices? May 20 07:37:32 perhaps something is wrong with that one card? May 20 07:38:14 Oh, now I see something in dmesg, looking again, from a long time after the initial boot May 20 07:39:04 it picks up the SD card, and it's size, then some line of output saying just "mmcblk0: p1" and then "retrying using single block read: May 20 07:40:02 But no, I haven't tried the same SD card in other devices. It's expected I can't boot from one already with Ubuntu installed on it like that, right, as it boots from NAND? May 20 07:40:30 * gsnedders should get a serial cable and try and boot another BB with that SD card May 20 07:40:48 gsnedders, you can always boot from an SD card I think May 20 07:41:07 gsnedders, just push the user button/reset button together May 20 07:41:16 gsnedders, and try another SD card (the one that worked in other devices) May 20 07:47:24 * gsnedders was unaware of that May 20 07:48:32 pressing both buttons it still beats fron NAND May 20 07:48:58 morning May 20 07:49:31 er, hold user, press reset, release reset, wait a bit. May 20 07:49:43 then release user. May 20 07:49:57 ogra: ping May 20 07:51:02 beats from NAND? boots. May 20 07:52:10 zyga: Another SD card with Ubuntu already installed on it? May 20 07:52:37 gsnedders, just put the netbook/server/netboot install image on some sd card that worked before May 20 07:52:50 gsnedders, and follow DanaG's instructions on how to boot from SD May 20 07:52:55 hrw, morning, how are you? May 20 07:52:59 hrw, flood is getting worse May 20 07:53:15 hrw, did you manage to boot your BB May 20 07:54:24 Hmm, the only SD card I have that I know worked before has a full Ubuntu install on it, so that's not really an option May 20 07:54:43 zyga: I even installed netbook on it. but did not rebooted to it yet May 20 07:55:08 gsnedders, get some more SD cards, I know it sucks but they _are_ cheap May 20 07:55:22 zyga: took me some time to make my ubuntu less free&open May 20 07:55:29 gsnedders, and after you install some image you can boot from NAND and use USB storage for main filesystem May 20 07:55:44 gsnedders, (so you don't have to use SD cards all the time) May 20 07:56:13 hrw, are you drafting any specs? May 20 07:56:39 zyga: There's plenty of SD cards around here, and using them all the time isn't actually a big problem May 20 07:56:44 (even if they are slow) May 20 07:56:53 zyga: cross compiler packages one now and was asked for arm toolchain one May 20 07:58:25 * gsnedders wonders if we have any sort of USB storage around in the office he can use May 20 08:00:24 gsnedders, thumb drive ;-) ? May 20 08:00:39 zyga: thumb drive? May 20 08:00:49 hrw, yeah May 20 08:00:53 why so limiting... May 20 08:01:15 hrw, 1) no power reqs, 2) silent, 3) could be fast enough if you have good device May 20 08:01:18 grab usb-storage raid5 ;d May 20 08:01:47 Ubuntu can't fit on nand. :( May 20 08:01:59 * gsnedders notes that the other BB he has has just power, an SD card, and a USB network interface plugged in, and he kinda likes that minimalisim May 20 08:02:09 well, initramfs can... but I don't know how to make it use the initramfs. May 20 08:02:49 zyga: I think I'd still have to go out and buy one anyway :P May 20 08:03:14 DanaG, you don't have to fit the whole thing in nand, having boot/kernel parts there is enough May 20 08:05:34 * gsnedders concludes we must have other SD cards in the office… May 20 08:09:41 Hmm, now a bunch of I/O errors trying to get at the SD card May 20 08:11:30 Then trying to install onto a USB thumb drive I found it gets stuck at computing partitions May 20 08:11:35 Something really weird is going on May 20 08:13:33 Stuck in such a way that I can't change terminal and see what's going on May 20 08:14:20 So after 3s or so of having the USB thumb drive plugged in it seems to freeze May 20 08:15:50 lag, hey May 20 08:17:30 ogra: I've been told to speak with you May 20 08:17:48 ogra: I've just successfully compiled Ubuntu for ARM May 20 08:18:09 ogra: How would I get that on the a board? May 20 08:18:42 TFTP? uBoot? May 20 08:18:52 compiled ubuntu for arm ? May 20 08:19:27 Yeah May 20 08:19:44 what do you mean by that ? May 20 08:19:51 you rebuilt the archive ? May 20 08:20:03 The ti-omap head May 20 08:23:51 lag: you build an omap kernel, not ubuntu-on-arm :-p May 20 08:23:56 *built May 20 08:24:21 amitk, did he ? i'm still trying to decypher :) May 20 08:24:56 amitk: Yes, with all the extra Debian stuff in May 20 08:25:05 My apologies May 20 08:25:52 lag: ok, so you built an omap kernel .deb (ubuntu-way of kernel building). ogra he now wants to put it on his (presumably) beagle board. May 20 08:25:59 ok, so you have a rootfs, a kernel and an initrd ? May 20 08:26:23 Nope, if I had those I'd know what to do with them May 20 08:26:29 if its a beagle, just use the lucdi image and then replace the kernel with your package May 20 08:26:36 *lucid May 20 08:26:40 I have 34 .udeb files and 3 .deb files? May 20 08:26:46 lag: no apologies required, it's still early in the morning :) May 20 08:27:11 amitk: :) May 20 08:27:22 lag, install one of the images from here https://wiki.ubuntu.com/ARM/Beagle May 20 08:27:36 after that just install the linux-image deb in the new install May 20 08:31:49 The Beagle board was just an example - I don't actually have one of those here May 20 08:32:02 Can I test my new kernel in qemu? May 20 08:36:32 there is a qemu-maemo somewhere but i dont know how well it works May 20 08:36:42 (its not in ubuntu though) May 20 09:05:02 ogra: hi could help to me test again? May 20 09:05:28 indeed May 20 09:07:06 ogra: http://people.canonical.com/~roc/kernel/omap4/ May 20 09:07:30 ogra: i saw this error "Uninitialized omap_chip, please fix!" in the dmesg May 20 09:07:31 you posted May 20 09:07:46 oh, i missed that May 20 09:08:20 it means the kernel does not think the chip is an OMAP chip May 20 09:09:08 so cpu1 was not up May 20 09:09:10 yeah May 20 09:13:34 cooloney, hmm, looks slightly better but i get a ton of garbage on the serial console and no login prompt May 20 09:13:46 i see two penguins on the LCD May 20 09:14:04 which apparetnly indicates both cores were initialized May 20 09:14:22 butu it seems to hang May 20 09:14:37 ogra: too bad May 20 09:14:48 any chance to get the dmesg? May 20 09:15:39 well, its cut off due to the garbage May 20 09:16:41 ok, let me check my serial port config May 20 09:17:58 well, the output is fine up to a certain point May 20 09:18:11 something gets enabled that trashes it May 20 09:18:28 let me try to write to a log and see if i get more May 20 09:20:00 ogra: thanks a lot May 20 09:35:13 I have a question, but you have to promise not to laugh! =:-) May 20 09:35:56 Is the omap-5912 supported by us? I'm guessing it's too LP. May 20 09:37:24 thats ARM926EJ-S ? May 20 09:37:46 that would be supported with 9.04 only May 20 09:38:19 lag: that is ARM9, so no. We only care about ARMv7-based processors May 20 09:38:49 amitk: Thanks :) May 20 09:38:53 omap5912 is omap1 - arm926 May 20 09:39:16 one of oldest omap devboards May 20 09:39:17 A relic then May 20 09:39:19 arm926 was supported in jaunty May 20 09:39:34 It's the only omap board I own :( May 20 09:39:42 lag: you can run Debian on it or go for Ångström May 20 09:39:45 get a beagle May 20 09:39:56 right, or run debian/amgstrom May 20 09:40:58 I'm sure Ubuntu will give me some hardware to play with sooner or later May 20 09:41:01 I would take angstrom but thats because I used it on too many devboards already May 20 09:41:32 I need to run Ubuntu May 20 09:41:43 The priority is the OS, not the board May 20 09:41:44 :) May 20 09:42:15 +1 May 20 10:10:14 arm gets me all confuzzled... armv7 arm9 arm11.... sigh... May 20 10:11:19 jussi: http://en.wikipedia.org/wiki/ARM_architecture May 20 10:12:53 arm9 == v4 & v5 May 20 10:13:07 so basically we care about cortex...? May 20 10:13:17 cortex-a8/9 May 20 10:13:21 ARMv7 May 20 10:13:37 Bingo May 20 10:13:38 :) May 20 10:15:41 lag: v4, v4t, v5 I would even extend May 20 10:15:43 what about those with armv6 :p May 20 10:15:47 and what is i.mx51? May 20 10:15:53 lilstevie: 9.10 only May 20 10:15:56 jussi: armv7a May 20 10:16:00 hrw: really? May 20 10:16:09 jussi: i.mx5xx are cortex-a8 May 20 10:16:15 hrw: Agreed May 20 10:16:27 lilstevie: 10.04 is for armv7a, 9.10 was armv6 May 20 10:16:35 ah :) May 20 10:16:42 jussi: there are i.mx51, i.mx53, i.mx508 May 20 10:16:45 hrw: Well read ;) May 20 10:16:53 jussi: i.mx61/63 will be A9 May 20 10:17:07 so lucid runs on imx51? May 20 10:17:12 yes May 20 10:17:13 yes May 20 10:17:23 only on the babbage board though May 20 10:17:57 ok May 20 10:18:01 there are companies like pegatron that build stuff based on the imx51 design but its not quite the same which ednts up to break without images May 20 10:18:12 *ends May 20 10:18:24 s/without/with our/ May 20 10:18:30 * ogra goes for more coffee May 20 10:18:40 during uds-m guy from Freescale said that there can be i.mx51 based beagleboard-like device May 20 10:19:02 the sharp is quite close to a real babbage May 20 10:19:46 jussi: you could take a custom kernel and use it with lucid & maverick rootfs going forward May 20 10:19:49 but uses special patches as well May 20 10:20:28 hrw, probably ... but likely not in the same price segment May 20 10:22:22 amitk: ahh, that sounds interesting May 20 10:22:58 ogra: yep May 20 10:23:34 jussi: today's hardware does not usually require specific packages to boot (other then kernel) May 20 10:23:35 the babbage is actually not far away from what the XM offers in functionallity, on board sockets etc May 20 10:23:41 but its $750 May 20 10:27:48 hrw: what exactly do you mean by that? May 20 10:28:05 a bootloader ? May 20 10:28:37 are normal packages apart from the kerneal portable across arm arches? May 20 10:29:05 jussi: yes, they are as long as they are not machine specific May 20 10:29:10 everything we provide in lucid will work on all v7 CPUS May 20 10:29:31 jussi: think: opengl support or HW accelerated audio/video codecs May 20 10:29:43 hrw: how do we tell what is machine specific? May 20 10:29:46 jussi: or gpio/i2c/regs related debug tools May 20 10:30:11 jussi, everything that needs HW accel May 20 10:30:21 What about HW FP? May 20 10:30:25 jussi: s3c_gpio tool does not work on non-samsung devices. but thats just example rather not present in ubuntu May 20 10:30:25 ogra: ahh, that makes sense. May 20 10:30:27 (or HW debug tools) May 20 10:30:55 gsnedders: you mean VFP? May 20 10:31:02 hrw: yeah May 20 10:31:08 thats handled by the kernel May 20 10:31:24 libc hooks into a common API May 20 10:32:25 gsnedders: so far I did not heard about armv7a without vfp. marvell has a line with different vfp then other cortex-a8 but I do not know how much it affects userspace May 20 10:32:54 a lot :) we had massive probs in the past with the dove ports May 20 10:33:12 but they usually pointed at HW issues that were resolved in a HW update May 20 10:33:36 its what you get if you dont stick 100% to the specs :) May 20 10:34:10 ah.. right Dove was marvell May 20 10:34:55 dove/armada is an awesome platform but has its specialities :) May 20 13:00:29 ehlo! May 20 13:00:45 Does Ubuntu 10.04 run on a C4 beagle board? May 20 13:00:54 i.e. TI OMAP3 May 20 13:01:09 jeremiah: yes May 20 13:01:15 w00t May 20 13:01:33 Can you point me to an image or ISO I can burn to a flash card? May 20 13:01:48 since day one: pick your poison http://elinux.org/BeagleBoardUbuntu or https://wiki.ubuntu.com/ARM/Beagle both have images.. May 20 13:02:36 fantastic May 20 13:02:42 thanks! May 20 16:03:26 have a nice rest of day May 20 16:10:32 have a nice day! **** BEGIN LOGGING AT Thu May 20 16:18:43 2010 May 20 16:23:09 grmbl May 20 16:23:29 why does x-loader always use u-boot headers in a hardcoded mannaer May 20 16:23:47 io.h -> ../../../u-boot/include/asm/io.h May 20 16:23:51 hmpf May 20 16:24:06 thaat will be a big quilt patch May 20 21:29:42 rcn-ee, you said rootstock on beagle was faster? May 20 21:30:11 5 hours later I'm still watching "setting up " May 20 21:40:48 cwillu_at_work, just wait for https://blueprints.edge.launchpad.net/ubuntu/+spec/preinstalled-sd-card-images-for-omap to be implemented and you wont have to worry anymore ;) May 20 21:41:08 ogra, that won't help me in the slightest May 20 21:41:15 why ? May 20 21:41:18 I'm actually making my own images May 20 21:41:36 so having a bare minimal ubuntu image wouldnt help you ? May 20 21:41:45 nope May 20 21:42:00 how do you build your ubuntu images atm ? May 20 21:42:01 note how rootstock works :) May 20 21:42:13 i know how it works, i wrote it :) May 20 21:42:35 okay, so how does having the bare minimal install help me when I need pieces removed from it and lots of other pieces added to it? :p May 20 21:42:59 you remove pieces from a bare debootstrap ? May 20 21:43:10 wow May 20 21:43:22 it makes me feel special :p May 20 21:43:31 you are special then :) May 20 21:43:46 I'm also using a btrfs root May 20 21:44:03 hmm May 20 21:44:15 i was always pondering to add a minbase option to rootstock that makes it use the minbase install of debootstrap May 20 21:44:25 would that help you ? May 20 21:44:29 I just hack up rootstock to serve me May 20 21:44:52 I should compare mine and your current to see if there's anything useful to push back May 20 21:44:52 its a lot smaller but really misses functionallity due to that omission May 20 21:44:58 yeah May 20 21:45:05 I've got a bit of a module system, which is really just a lightweight dpkg May 20 21:45:08 maverick is open, let in the crack :) May 20 21:45:44 note that there are chances that ubuntu might default to btrfs with maverick May 20 21:46:00 (sadly depending on grub support) May 20 21:46:11 well, grub is one requirement May 20 21:46:13 it's not the only one May 20 21:46:20 right May 20 21:46:23 there's _lots_ of oops left May 20 21:46:26 but the biggest one i heard May 20 21:46:47 not really; the required headers are already available under an appropriate licence May 20 21:46:51 from the ubuntu foundation team side at least May 20 21:47:00 for grub2 ? May 20 21:47:02 some of it would have to be reworked, but that's not a huge deal May 20 21:47:03 yes May 20 21:47:10 i think there is some integration work left May 20 21:47:40 in any case it was heavily discussed at UDS May 20 21:47:43 I haven't heard cmason weigh in yet, but there's a few people who have a negative opinion of ubuntu in the btrfs dev community May 20 21:47:50 and the intrest is very high May 20 21:48:06 perhaps some devs should start hanging out in #btrfs, and announce themselves as such :) May 20 21:48:55 yeah, not my headdache though ... the foundation team will do the most of the work as well as the decision of the default ... i'll just follow that decision in my little arm world :) May 20 21:49:05 :) May 20 21:49:17 though i'D love to see us switch May 20 21:49:35 me too, but I think 10.10 is premature May 20 21:49:49 might be ... we'll see what they decide May 20 21:49:51 although there's lots of patches flying around right now, if they land and things get much better, then maybe May 20 21:50:02 I'm talking from the upstream perspective May 20 21:50:10 remember that that was also one of the requirements :) May 20 21:50:16 well, we have some QA force out there at least :) May 20 21:50:17 ("upstream needs to be happy about this" May 20 21:50:31 lots and lots of testers May 20 21:51:01 I'm going to tar up a copy of my rootstock infrastructure if you want to take a look May 20 21:51:22 indeed upstream needs to be happy but what better can you get than having millions of users that help improving your code :) May 20 21:51:30 that would be great May 20 21:51:36 well... May 20 21:51:52 i wont get to much rootstock work before end of next week though ... May 20 21:51:53 millions of users is worse than useless if you already know you're not ready for mainstream May 20 21:52:02 indeed May 20 21:52:12 if you are though ... May 20 21:52:22 I'm just thinking I can give you a tour of what I have, just to see if there's anything worth grabbing May 20 21:52:38 module system sounds definately intresting May 20 21:55:29 basically it's a separate dpkg; moving files in before vm/native, kicking off scripts and installing debs once in vm/native, and then running some post jobs after, when actually writing the tarball to an sd card, so that you can have a single root tarball, and finish up setting the hostname, installing ssh keys, etc when you're actually writing it out May 20 21:56:18 separate dpkg ? May 20 21:56:34 you mean you copy in .debs and install them ? May 20 21:56:38 part of my use is so that when I write an image to go to a customer's site, I can set up the keys for it to log into our monitoring system, and at the same time notify the monitoring system of the new target, print out some paperwork, and so forth May 20 21:56:49 no, well yes, well no. May 20 21:56:53 heh May 20 21:57:12 a module can include a deb which will be installed, but most don't May 20 21:57:24 so just scripts May 20 21:57:37 that sounds overall definately intresting May 20 21:57:39 I use it to install a hacked up pixman deb to help with some performance issues, for instance May 20 21:58:16 set up specific users + environment May 20 21:58:38 really, it's nothing that dpkg shouldn't be able to do, but yet I find it convenient to be able to perform jobs outside of the image May 20 21:58:38 you could do that with the --script option already May 20 21:58:50 well, kinda May 20 21:59:00 i guess you're not alone with that requirement May 20 21:59:21 the idea is that you keep related stuff together, even though the pieces for a given module need to be performed at different times May 20 21:59:25 not that i would need it but i see how it is useful, so a tarball would indeed make sense May 20 22:00:38 right ... i was thinking about stealing the plugin system i wrote for ltsp, it does something similar ... each plugin has several stages (pre-install, install, after-install) where you can add different script snippets May 20 22:00:45 yep May 20 22:00:50 but its very complex May 20 22:01:11 and i didnt want to make rootstock to hard to understand for novice users May 20 22:01:25 do you have a standard tool to write out an image? May 20 22:01:37 its its strenght currently that everyone can easily understand it May 20 22:01:45 i.e., my mkcard ties into this May 20 22:02:11 no, because i dont want to clash with the actual image building tools we have in ubuntu May 20 22:02:39 rootstock was never meant to be an image builder, something writing images should just wrap around it May 20 22:03:08 sadly someone noticed that i create qemu images on the go and sent a patch :P May 20 22:03:55 the initial purpose was never to write to SD or to create disk images though ... but i'm bad at saying no if someone sends good code ;) May 20 22:05:22 so if you have any good code that ties in with the std ubuntu tools (parted for example instead of fdisk/sfdisk etc) i'D happily merge it May 20 22:06:06 but effectively i still think writing to disk should be done by a wrapper May 20 22:07:00 in any case a tarball would be good if it saves you work to modify it everytime the upstream code changes May 20 22:10:39 grrr May 20 22:10:41 what was the last thing I said? May 20 22:10:48 netsplit fun :) May 20 22:11:00 i.e., my mkcard ties into this May 20 22:11:08 and then i sadi a lot :) May 20 22:11:11 *said May 20 22:11:25 I missed everything you said after mkcard, and you missed everything I said :) May 20 22:11:48 pasted you a PW May 20 22:11:50 err May 20 22:11:52 PM May 20 22:11:53 I went onto a tangent about update-initramfs being interesting, but being structured inverted from how I would do it May 20 22:12:24 likewise :) May 20 22:12:59 heh May 20 22:13:21 why not actually tie in with initramfs tools ? May 20 22:13:35 for the reasons I mentioned :) May 20 22:13:44 seems logical to me if you want to use any kind of raid May 20 22:13:55 I hate how I have to "find | grep btrfs" if I want to tell what jobs actually exist and where they are May 20 22:14:30 well, in the case of initramfs you only have two scripts usually May 20 22:14:37 the hook and the actual script May 20 22:14:53 the hook puts binaries you wil use in place, the script executes the, May 20 22:14:55 *them May 20 22:15:10 yes, but there's 11 subfolders in /scripts/ where your file could be hiding May 20 22:15:16 i find that very simple once you understand how it works May 20 22:15:23 absolutely, but it's not convenient May 20 22:15:29 indeed because there are several stages in the boot May 20 22:15:40 you don't understand my point :) May 20 22:15:50 and several ways of booting (local/network etc) May 20 22:15:59 I'm not saying that you don't need those hooks May 20 22:16:08 I'm saying that the directory structure is inverted May 20 22:16:28 I want to have a modules.d/btrfs/ folder with script names that match the hook I want May 20 22:17:14 it's ungodly annoying to never be able to do a simple "ls" to see what's what :) May 20 22:17:20 well, you could rework intiramfs tools, heve it that way and work with links May 20 22:17:30 pretty much May 20 22:17:43 make a proposal :) we might change it May 20 22:17:45 did you catch the part where this is basically how my module system works already :) May 20 22:18:00 its a valid concern that might simplify life for everyone May 20 22:18:25 yes May 20 22:19:15 well, roll a tarball, that definately sounds intresting :) May 20 22:19:51 okay; I'll send you something over the weekend; I have a site visit tomorrow that's probably going to be a long night tonight and a long day tomorrow :p May 20 22:20:00 no hurry May 20 22:20:23 i wont get to rootstock work before end of next week anyway May 20 22:20:59 incidentally, a nice thing about knowing how to write the images is that you can punt alot of the work out to the first boot transparently, if for example qemu decides to not work with lucid's userspace :p May 20 22:21:18 indeed May 20 22:21:35 i wish we would find that darn bug May 20 22:22:18 i also wish i could do without a VM at all May 20 22:22:44 it wouls already be possible if ubuntus desktop and netbook metapackages wouldnt pull in mono :/ May 20 22:22:59 hard to, while any package could contain a binary which it calls in a post-install script May 20 22:23:14 hmm? May 20 22:23:14 no prob at all in a chrooted env May 20 22:23:29 the first stage of rootstock operates completely in a chroot May 20 22:23:36 ten times faster than a VM May 20 22:23:45 yes, but that's what debootstrap does already, no? May 20 22:23:50 i only use the VM to actually install packages May 20 22:24:14 its qemu-user in a binfmt handler that enables that May 20 22:24:28 okay, but that's still a vm, no? May 20 22:24:41 nope May 20 22:24:46 syscall translation May 20 22:24:52 as i said 10x faster May 20 22:25:19 install qemu-arm-static and try out qemu-debootstrap --arch armel May 20 22:25:21 I'm missing something May 20 22:25:46 and that doesn't suffer the same fate under lucid? May 20 22:25:52 right May 20 22:25:58 it has other probs though May 20 22:26:05 hmm May 20 22:26:08 such as? May 20 22:26:27 your /proc in a chroot comes from the host May 20 22:26:38 which is x86 ... May 20 22:26:58 (or whatever) May 20 22:27:05 packages like mono that have a builting garbage collector that operates on /proc/self explode for example May 20 22:27:14 *builtin May 20 22:28:13 i'm planning to play a bit with LXC if i find the time during maverick development ... to see if that could help May 20 22:28:52 if you can fake certain parts of /proc we might not need VMs anymore May 20 22:29:09 it's not a complete solution, unfortunately, based on the discussion at UDS. We'd still need to fake /proc May 20 22:29:14 back... cwillu_at_work i hope you did run it on a beagle with an sd card... using fast sata drives with usb converters... May 20 22:29:15 and operatre nearly at host speed for 80% of the stuff you do May 20 22:29:45 the gzip routine also kills it when it's working on a 1gig folder.. May 20 22:29:56 persia, so you should finally sit down with lifeless and finish that fuse thingie :) May 20 22:30:08 sorry May 20 22:30:13 ice was falling from the sky May 20 22:30:22 ice into rum and coke. .;) May 20 22:30:23 lifeless isn't going to do it. We almost got StevenK to do it, but there was an interruption. May 20 22:30:48 bah and then he wandered off to new challenges May 20 22:31:00 which surely occupy him May 20 22:31:30 i need a fatter pipe.. 305Mb xfce demo image... May 20 22:31:47 local mirrors or package proxies help :) May 20 22:31:51 so, what you're saying is... :) May 20 22:32:22 yeah... my outside server can handle it.. it's the upload at 25kb's.. ;) May 20 22:32:28 ah May 20 22:32:40 * ogra pats his SDSL May 20 22:32:59 costs a fortune but is symmetric :) May 20 22:33:18 i wish we had something like that... but i'm in the middle of farm land.... ;) May 20 22:33:43 yeah,i guess you can be lucky you dont need to morse your packages :) May 20 22:33:52 so, here's a thought: rootstock which can selectively work via qemuvm, chroot, or by writing out to a card to finish up properly on native hardware? May 20 22:34:28 the latter might be handled by the jasper tool i'm just writing May 20 22:34:42 (described in the spec i posted initially) May 20 22:35:06 its not its purpose but i could add hooks :) May 20 22:35:37 the former two already happen based on what you have installed on the host May 20 22:36:01 i'd like more chroot in that scenario but that breaks on mono May 20 22:36:07 * cwillu_at_work checks May 20 22:36:24 oh, and that's why you still drop into qemu May 20 22:36:32 right May 20 22:36:35 only for mono May 20 22:36:46 else i'D rip out the whole VM May 20 22:37:00 and overcome all speed issues May 20 22:37:10 well, that makes me feel a little bit silly May 20 22:37:29 being as it is that I'm writing the software that I'm installing at that point May 20 22:37:40 gcc works under qemu-static? May 20 22:37:46 sure May 20 22:38:03 but compiling works rather at VM speed, the gain isnt big there May 20 22:38:26 even so, it works at vm speed with the image on 8gb of ram May 20 22:38:38 as opposed to an mmc reader, or even fast sata drives on usb May 20 22:38:43 * cwillu_at_work glares at rcn-ee briefly :p May 20 22:38:51 *grin* May 20 22:39:21 if I do this, and it finishes before the beagle finishes its install, I'm either going to smack you or kiss you :p May 20 22:39:51 there is some tool lool worked on that might enter maverick that inserts a cross into that picture May 20 22:39:55 *cross gcc May 20 22:40:17 heh May 20 22:40:31 * ogra cant remember the exact name ... was it cacao ? May 20 22:40:36 persia, ^^^ ? May 20 22:40:45 you know, all you need to do is copy in the cross compiler into the chroot with a divert, or into /usr/local/bin :D May 20 22:40:51 croco May 20 22:40:57 croco, right May 20 22:41:02 But it's not cross, really. May 20 22:41:05 cwillu_at_work, sadly thats not enough May 20 22:41:14 you need libs and stuff May 20 22:41:32 What it does is that it intercepts calls to a whitelist of programs and uses native binaries (at the cost of a larger chroot) May 20 22:41:45 So you don't end up running emulated bash, for instance. May 20 22:42:20 Or you use a cross-compiler instead of a regular compiler. May 20 22:42:24 * ogra gets called by GF ... its nearing 1am over here May 20 22:42:53 i guess i have to call it a day, else she gets angry May 20 22:42:56 But really, these tools are in an early stage: there's no guarantee that the results of such builds are going to be ABI-compatible with the regular Ubuntu builds. May 20 22:42:59 heh May 20 22:43:25 ya, I generally stick with native builds (or qemu builds at least) when I can May 20 22:43:27 They are mostly only useful to do test-builds to see if something would work in the absence of hardware. May 20 22:43:42 right May 20 22:44:11 okay, so tell me if I have this right: May 20 22:44:15 anyway, /me waves and vanihes May 20 22:44:28 with qemu-kvm-static, I can just chroot into the arm image? :) May 20 22:44:29 or not ... let me wait for your listing :) May 20 22:44:35 right May 20 22:44:42 okay, you may leave :) May 20 22:44:45 :D May 20 22:44:48 * ogra does that all the day with his SD cards May 20 22:44:48 8D May 20 22:44:53 heh May 20 22:44:55 bye May 20 22:45:09 >8D May 20 22:45:25 * cwillu_at_work hacks up run-vm May 20 22:46:13 it's funny, I misnamed my newer fork of rootstock "rootstock.chroot"... and now I'm making it use a chroot :) May 20 22:46:14 cwillu_at_work: the actual package name you need is qemu-kvm-extras-static May 20 22:46:24 persia, I know, it's already installed, I just forgot the exact name May 20 22:46:37 OK. Just wanted to make sure :) May 20 22:47:31 Also, you will need to copy the static qemu interpreter into the chroot: take a look at the qemu-debootstrap implementation for which binary needs copying. May 20 22:47:47 k May 20 22:47:56 If you use qemu-debootstrap, it's safe to expect this to be done for you, but if you're doing something more complex... May 20 22:48:29 I don't mind knowing how to duplicate the magic May 20 22:51:02 * cwillu_at_work kicks off a build May 20 22:55:11 * cwillu_at_work adds a chroot call into run-vm and kicks off another build :p May 20 22:55:46 I: Installing core packages May 20 22:55:48 \o/ May 20 22:56:02 god I hope I'm not overwriting my server with arm binaries :p May 20 22:56:04 makes you wonder... give me a co-arm processor.. ;) May 20 22:56:31 btrfs snapshot? May 20 22:56:37 god May 20 22:56:44 I wish I thought of that before I started the build May 20 22:57:03 unsurprisingly, this is the server that is also using a btrfs root :) May 20 22:57:10 surprisingly, not _all_ of my machines are May 20 22:57:20 specifically, my main desktop at work and at home don't May 20 22:57:43 which bootloader then? grub2 doesn't have support yet i heard... i have new harddrive so it's tempting me... May 20 22:59:19 I just use a boot partition May 20 22:59:41 ah duh.. May 20 22:59:51 you have to fight with update-grub a little to get it to acknowledge the existence of grub partitions May 20 23:00:19 Selecting previously deselected package linux-image-2.6.33.4-l3. May 20 23:00:19 (Reading database ... Unsupported ioctl: cmd=0xffffffffc020660b May 20 23:00:22 repeat x100 May 20 23:00:31 seems to be proceeding though May 20 23:00:35 yeap, that's normal inside the machine.. it's weird.. May 20 23:01:21 amusingly, my server has already caught up with the beagle May 20 23:01:26 the beagle's been running 6 hours May 20 23:01:52 i'm tempted to ask for your package list for comparisson... I'm slowly untarring 1Gig onto a sd card... May 20 23:02:54 I'm not using sata drives :p May 20 23:04:03 re: btrfs, the annoying thing is that there's a tonne of features which are going to be easy to implement, but very few _have_ been implemented yet May 20 23:04:15 basically the answer to every cool question is "not yet" May 20 23:04:30 separate raid levels per folder/file? not yet May 20 23:04:46 even so, it looks interesting.. May 20 23:04:47 online scrubbing? not yet (although you can fake it from userspace) May 20 23:04:50 oh, absolutely May 20 23:05:06 it's only very recently that you could actually delete snapshots :) May 20 23:05:33 hence why I didn't think about it until I already started: been using it 9 months on this box, and it's only been possible to delete in the last month :) May 20 23:06:37 honestly, you should start hanging out in #btrfs; the visibility will help, and you'll pumped hearing about it :) May 20 23:07:59 "cwillu: hey, if I saw ubuntu/canonical devs in here I'd be all for it." May 20 23:08:06 "but from what I see, it'd be a bunch of uninform May 20 23:08:07 ed n00bs in here complaining about how btrfs broke their systems and us telling May 20 23:08:07 people to update and no updates put into any ubuntu kernels May 20 23:08:07 " May 20 23:08:23 this is an image problem that could be solved very easily :) May 20 23:09:32 those groups of people are always fun to deal with.. ;) May 20 23:09:55 the people in question are quite nice actually :p May 20 23:10:00 oh, _that_ group May 20 23:10:51 the guy who said that spent a couple hours with me updating the btrfs wiki with debian/ubuntu instructions to update btrfs via dkms (which is currently the only recommended way of using btrfs) May 20 23:11:05 update btrfs from git via dkms, rather May 20 23:12:15 dkms makes sense if your using a normal supported ubuntu kernel.... quick updates... May 20 23:12:40 Now we just need dkms for powervr. =þ May 20 23:13:07 we need mimotrace for arm... ;) then reverse engineer the suckker...; ) May 20 23:13:16 From what i heard at UDS, we ought be able to do something with DKMS+jockey for powerVR for OMAP. May 20 23:13:42 Hallelujah. May 20 23:13:56 Also, DKMS should work for *any* kernel, as long as one packages the kernel and installs the headers, etc. May 20 23:13:57 Since the TI build script is just a miserable failure. May 20 23:14:07 well the TI SGX bits are gpl.. (i've just recently add those modules into my dev tree 2.6.34's..) it's the binary blob libGL and loader that's the problem.. May 20 23:14:16 (kernel modules bits) May 20 23:14:48 I'm also curious how GL ES works... and if I could use it on Radeon KMS. May 20 23:15:09 Right. The loader would be the bit that would be DKMS, and the binary blob would be handled by jockey. The analogy that I heard at the OMAP session at UDS was to how nVidia works on x86 (assuming no nouveau) May 20 23:15:20 i think we can... but i don't know if r300g is hocked up to the egles bits.. (both are in mesa0 May 20 23:15:26 DanaG: GL ES is just a subset of GL. May 20 23:15:43 actually the loader doesn't need DKMS.. just a binary deb ... May 20 23:15:44 So anything that supports GL automatically supports GL ES May 20 23:15:45 My hardware is RV635, to be specific. May 20 23:15:56 hmm, I wonder if fglrx would support it on bare console. May 20 23:15:58 persia, did they say anything about how to acquire the binaries in the first place though? May 20 23:16:09 rcn-ee: Well, except it's hard to distribute raw binaries :) May 20 23:16:27 persia, the typical process involves sending a request to ti with a corporate email address purporting to be doing oem design work May 20 23:16:48 cwillu_at_work: No. If we can find some that are licensed appropriately, we can toss them in multiverse, which would seem the least painful solution to me. May 20 23:17:07 that's a pretty big if May 20 23:17:21 yeah.. specially with TI's export license... the key thing, if the gpl modules are built with the kernel, then all you need is the loader/libGL which don't need gcc to build.. May 20 23:17:25 persia: something that would work: a script like fwcutter. May 20 23:17:33 Well, there were folks from TI at the session, and they seemed interested in making it work. I'm not sure how far anyone will get. May 20 23:17:34 Feed it the x86 .run. May 20 23:17:49 DanaG: How do you mean? May 20 23:17:57 DanaG, fwcutter is an option because you have the right to have the binary you're cutting from May 20 23:18:03 that isn't the case here May 20 23:18:06 I mean a bit more manual, though. May 20 23:18:14 Have people download from TI, then feed it to a script. May 20 23:18:15 or rather, finding a binary where it _is_ the case is the tricky part May 20 23:18:39 If TI can post it for anyone to download, TI can probably allow it to be in Multiverse. May 20 23:18:39 or we got to talk ti into makeing a arm based *.bin loader.. ;) May 20 23:18:49 instead of the pure x86.. May 20 23:19:03 I think the issue is that TI is constrained (it's not actually a TI part, just bolted into a TI part) May 20 23:19:04 (the binary they distrubt that contains all the junk) May 20 23:19:28 yes, that's exactly the problem May 20 23:19:35 yeap, and that's the same problem intel (pousble) and freescale have... May 20 23:19:38 they licenced the powervr core, just like they licenced the arm core May 20 23:19:56 and the powervr core's terms are fairly onerous May 20 23:20:03 i've also heard from rumours, mali isn't much better.. so there really isn't a good one yet.. May 20 23:21:20 Well, the solution for other architectures has often been to expose a bus and make it someone else's problem :) May 20 23:22:22 I wonder how similar gma500 and this SGX are.... May 20 23:22:37 it'd be amusing if the GL code were exactly the same, except for target architecture. May 20 23:22:38 * rcn-ee chears come on oem-config.... you can load! May 20 23:23:06 the arch between the cores according to wikipedia looks similar... May 20 23:23:43 perhaps everyone's favourite space tourist should buy a redistributable licence :p May 20 23:23:45 From rumours I've heard, PSB is the reference SGX implementation. Dunno how true this is. May 20 23:24:06 I'd rather have a radeon. May 20 23:24:17 ARM + Radeon would be fun. May 20 23:24:30 cwillu_at_work: May not be possible, really. Some of this stuff is so many levels deep in restrictions that it would take years for the legal work, even if everyone wanted it to happen yesterday. May 20 23:24:37 * cwillu_at_work inquires about the possibility of adding a pci-e port to the beagle-xl :p May 20 23:24:44 i'm tempted to get one of those mini-pci express to pci express adapter do that that once my tegra 2 comes... May 20 23:25:00 DanaG: You can do that with the MVL Orions (but not with Ubuntu, as that's ARMv5). They have PCIe. May 20 23:25:17 hmm, is there yet an armv7l marvell? May 20 23:25:31 more importantly, is there a multithreaded dpkg yet? May 20 23:25:36 the armanda's i haven't seen any reference designs yet wit pci x.. but they will be coming... May 20 23:25:47 image building is fast enough that I can get impatient with it :) May 20 23:25:50 DanaG: Not that I've heard about in retail. There's a "dove" kernel in the archive. May 20 23:26:04 PCIe, rather. May 20 23:26:12 "X" is that old long server PCI thingy. May 20 23:26:17 "extended". May 20 23:26:23 128 bit, wasn't it? May 20 23:26:42 Or, rather, isn't it (still available on some boards) May 20 23:26:52 yeah and 3.3volt signaling... yuck.. pain in the rear on my server... May 20 23:28:13 Well, pre multi-lane PCIe, it was better than what else was available :) May 20 23:29:02 i'm just annoyed, as it the only slot aviable on my 1U, and all i can get is network/pata/sata boards.... May 20 23:30:59 * rcn-ee calling it good, starts a slow 305Mb xfce demo image upload to rcn-ee.net... ;) May 20 23:31:28 * cwillu_at_work starts downloading kernels from rcn-ee.net in a loop May 20 23:31:54 * rcn-ee hopes cwillu_at_work doesn't find the upload point and kill his cable modem... May 20 23:32:56 * rcn-ee 3hours.... i'll be using wget -c for this.. May 20 23:35:44 cwillu_at_work, say with 2.6.34 have you run into this? http://pastebin.com/iqAHMVD4 May 20 23:36:16 on reboot? May 20 23:36:16 only occurs say after 10minues uptime, and then issuing a reboot... May 20 23:36:41 I haven't, but I can't say I've been exercising that codepath May 20 23:36:44 i know it happens on beagles... does the overo suffer the same fate? May 20 23:37:09 it's actually happening on all mine... including ones in a headless setup.. May 20 23:37:34 i've pinged linux-omap, but not that may people have jumped to 2.6.34 yet either.. ;) May 20 23:38:22 haven't noticed it, but I haven't been doing a whole lot with the overo's in the last two weeks either May 20 23:38:33 the zippy's came in, and I've been doing those sd burn tests May 20 23:41:35 noob question, how to get the network-manager thing to show up in xfce on the bottom. ;) May 20 23:41:47 nm-applet May 20 23:42:38 geez, I've run this 3 times with the chroot in the time it took the original beagle to get to the point that the first chroot died (at which point it also died) May 20 23:43:01 taking into account that the beagle had a 6 hour head start May 20 23:43:05 i think the beagle lost.. kill it.. May 20 23:43:15 quite May 20 23:43:34 Emulation isn't always this much faster. It really depends on the RAM requirements for the build. May 20 23:43:42 yay, nm-applet runs, it doesn't like to run by default.. May 20 23:43:45 I know May 20 23:43:53 If you compare 1000 iterations of a simple hello.c compile, it's about the same. May 20 23:43:56 I'm very very io bound May 20 23:44:06 Heh. May 20 23:44:13 and doing it in ram is very very much faster than doing on an sd card May 20 23:44:42 it's funny though May 20 23:44:43 rcn-ee: if you've not tried it before, #xubuntu is also a good place to ask XFCE questions May 20 23:45:03 I've been tripping over that code for a while now, never bothered to actually try it :) May 20 23:45:18 granted that in my case, it still would've been doing lots from a normal vm anyway May 20 23:45:26 but I might have wised up May 20 23:46:10 yeah, i should, i'm starting to use xfce alot on the beagle..... ;) i've gone from providing just console image to full blown xfce images... give the people what they want.. ;) May 20 23:48:47 I found xfce not worth the bother... by the time I got it the way I liked it, I might as well have used gnome. May 20 23:49:35 If you're going that deep, #xubuntu-devel may be of interest, but from what I've heard, Xubuntu isn't quite as optimised for low-spec machines as it was in the past. May 20 23:50:00 nope May 20 23:50:01 (where Xubuntu is XFCE+everything to make it the way folks like it) May 20 23:50:15 gnome... i did that once on a beagle Bx... Debian Lenny kinda like watching paint dry... but i got firefox up and took a pic.. May 20 23:51:09 rcn-ee, my ui is actually built entirely on firefox :p May 20 23:52:10 nice.. it's actually not that bad on Karmic/Lucid.. but on lenny with armv4 unoptimzation... ;) May 20 23:52:15 heh May 20 23:52:42 i do really like the speed of chrome on arm thou... May 20 23:52:54 it wouldn't have had the neon accelerated routines in libpixman either May 20 23:52:59 that makes a huge difference May 20 23:53:26 hmm May 20 23:53:32 Setting up openssh-client (1:5.3p1-3ubuntu3) ... May 20 23:53:32 qemu: uncaught target signal 11 (Segmentation fault) - core dumped May 20 23:54:02 come on beagle.. ;) May 20 23:54:12 [31576.626128] kernel tried to execute NX-protected page - exploit attempt? (uid: 0) May 20 23:54:12 [31576.626140] BUG: unable to handle kernel paging request at ffffea0006704898 May 20 23:54:12 [31576.626147] IP: [] 0xffffea0006704898 May 20 23:54:12 [31576.626159] PGD 28402067 PUD 28403067 PMD 800000002de001e3 May 20 23:54:12 [31576.626168] Oops: 0011 [#1] SMP May 20 23:54:13 [31576.626174] last sysfs file: /sys/devices/virtual/block/loop1/queue/hw_sector_size May 20 23:54:15 [31576.626178] CPU 2 May 20 23:54:58 not sure if that happened at the same time, or in response to the umount May 20 23:55:44 * cwillu_at_work tries it again without the tmpfs May 21 00:01:39 or maybe my server is dying of btrfs-related conditions May 21 00:02:27 it looks like qemu... the loop mount... May 21 00:05:01 I got a bunch of paging request failures below though May 21 00:05:11 ah May 21 00:05:13 and all my terminal windows locked up May 21 00:05:22 I'm actually suspecting the hardware May 21 00:05:31 but why fix it if I can just reboot every night? :p May 21 00:06:21 * cwillu_at_work orders pizza May 21 00:07:09 exactly.. it's what those kitchen timers are for... (specially on cable modems that lock up every 7 hours) May 21 00:07:38 I actually bought one yesterday :) May 21 00:07:46 it's rebooting a beagle for me May 21 00:08:45 i picked up one of those ipower devices, can turnon/off 4 nodes over ip.. it's rebooted many beagles while i've been away... May 21 00:09:22 how much are those? May 21 00:09:46 I want to reboot the thing every ten minutes, and bring it back up right away, but I also don't want to pay more than 7.95 to do it May 21 00:11:06 laughs, more $60-70 than that.. but geeks has them on sale every once in a while.. http://www.geeks.com/search.asp?QUERY=ipower then a nice shell script to access them.. http://azug.minpet.unibas.ch/~lukas/myfree/ May 21 00:11:21 hmm :/ May 21 00:11:33 I think I'll just wire up a relay to a gpio May 21 00:12:06 that was my first thought... old arm board with ethernet, gpio to relays/etc... May 21 00:12:43 no ones really makes it, might be a nice hobby kit to sell... May 21 00:22:35 * cwillu_at_work nums on pizza May 21 00:25:52 pizza sounds good right about now May 21 00:26:29 I've started on the long, tedious process of finding duplicated symbols in the arm tree May 21 00:26:53 It's going to be like zapping flies .. killing them one at a time until all the ARM architectures can be compiled in at the same time May 21 00:27:04 (making a stupidly large, but VALID kernel) May 21 00:29:03 sounds like fun! Martyn is your current git tree public? May 21 00:31:00 rcn-ee : No, I need to github the work as it goes along May 21 00:31:22 i'll do that this weekend. This is _literally_ the first night I've started to work on what I promised to do @ UDS May 21 00:31:44 I just cloned the linux-arm kernel tree at 2.6.34 current May 21 00:32:42 well you've been busy, sounded like uds was crazy for arm.. ;) i'm mostly asking, cause my tegra 2 is to arrive tomorrow or saturday.. May 21 00:33:08 ( I have an even higher priority task to do . I need to get a basic patchset to add the arch/arm/mach-sstone subarch before the 2.6.35 window closes ) May 21 00:33:09 rcn-ee : AWESOME May 21 00:33:13 We are all going to have those little things May 21 00:33:25 nVidia have been DRAGGING their fucking feet on getting the linux SDK out May 21 00:33:42 frankly, I think we might be able to button up lucid lynx on it faster than they can May 21 00:33:54 I already have arm-server working correctly on it May 21 00:34:18 yeap, they have... some of thought they'd deliver... May 21 02:20:44 root@lucid-zippy:/etc/init# xsplash --display=:0 May 21 02:20:44 No protocol specified May 21 02:20:46 odd May 21 02:23:13 hmm May 21 02:23:14 xterm opens May 21 02:23:19 firefox doesn't, with the same error May 21 02:23:36 did we add more xauth sillyness in lucid?\ **** ENDING LOGGING AT Fri May 21 02:59:57 2010