**** BEGIN LOGGING AT Wed Aug 08 02:59:58 2012 Aug 08 05:35:25 hi,all. Does anybody know to know the version of armcc ? Aug 08 05:35:48 how I can know the version of armcc ? Aug 08 05:36:27 benn, you mean gcc for arm? Aug 08 05:37:13 scientes: I want to know the version of armcc in realview Aug 08 05:37:54 scientes: but I cannot find any option like '--version' or '-V' Aug 08 05:38:19 i don't know what armcc is Aug 08 05:38:21 and somebody ask me that whit is armcc v2.0 Aug 08 05:38:45 the compiler provided by arm Aug 08 05:39:01 just use the linaro toolchain in ubuntu Aug 08 05:39:11 gcc or gcc-arm-linux-gnueabihf Aug 08 05:40:20 We are using linaro gnu toolchain. I asked this question just because somebody ask me that what is armcc v2.0 ? Aug 08 05:40:33 i have no idea Aug 08 05:40:56 scientes: okay. thanks Aug 08 05:42:08 clang is version 3.1, and gcc is version 4.6/4.7 Aug 08 06:36:27 Hello everyone, i have installed Pre-Installed OMAP/OMAP4 ubuntu 12.04 on begalboard its working fine. Now i have a sensor driver which compiles fine with normal gcc but when i insert it says - invalid kernel module. Aug 08 06:37:36 i know, i have to cross compile it, can anyone tell me how to cross compile it... Aug 08 06:41:48 please do suggest me how to install that driver. Aug 08 06:42:55 is there any other channel to discuss this kind of topic ? Aug 08 06:52:45 aum_ goes so fast Aug 08 07:03:29 aum__: why crosscompile if you have ubuntu in place.. just compile on your beagle. Aug 08 07:08:37 aum__, oh hes back.... you don't need to cross compile, and you have to use the same version of the kernel to compile against, otherwise the special string in the module wont match Aug 08 07:16:25 LetoThe2nd, in that pricompiled image the is no arm specific compiler. normal gcc compiler compiles it though kernel rejects it to insert. Aug 08 07:17:04 aum__: apt-get build-essential. done. Aug 08 07:18:00 scientes, can you please elaborate the process... should i compile it on begalboard itself or on another x86 desktop.? Aug 08 07:18:43 LetoThe2nd, i do have normal x86 gcc compiler which is not arm specific compiler Aug 08 07:18:58 aum__: i doubt that *on* your beagleboard. Aug 08 07:19:23 aum__: just log into the beagle, do apt-get install build-essential, compile like you would on your desktop. Aug 08 07:20:05 LetoThe2nd, thats what i did but the kernel rejects it to insert. Aug 08 07:20:43 then the module does not fit the kernel and you have to go hunting why it doesn't, instead of talking about crosscompile magic. Aug 08 09:18:00 aum__ even i have same problem , i m trying to compile a kernel module for arm but compilation terminates with error "scripts/recordmcount: not fount" Aug 08 09:21:38 i have tried to generate recordmcount using "make ARCH=arm CROSSCOMPILE=arm-linux-gnueabi- prepare " but it still not working Aug 08 11:30:42 ogra_: ping Aug 08 11:31:05 do you think there is an available hardware acceleration (3D, gfx) for embedded boards like the raspberry pi in case Linux? Aug 08 13:06:22 djszapi, no idea, we dont support RPi, ask in an RPi channel Aug 08 13:43:21 ogra_, are you having nightmares involving RPi already ? Aug 08 13:43:31 lol Aug 08 13:43:49 no, my current nightmares revolve more around flash-kernel currently Aug 08 13:43:59 at one point you'll realize it's easier to bootstrap an RPi image than telling people it's unsupported :) Aug 08 13:44:09 haha Aug 08 13:44:13 what's flash-kernel doing that's nasty? Aug 08 13:44:32 I finally went with modifying the precise version after talking to you last time. Works fine on the transformer Aug 08 13:44:45 I may look at how the quantal version needs changing Aug 08 13:45:44 ogra_, where do you think default bootcfg files for abootimg (and maybe similar ones for other bootloader tools) should be provided? Would flash-kernel itself not be a good container for them to have all board specifics in one package? Aug 08 13:46:24 f-k 3.0 doesnt use any input config files anymore Aug 08 13:46:49 you have to use abootimg directly on the boot partition to change anything Aug 08 13:47:14 it doesnt currently support that for any bootloader btw Aug 08 13:48:11 fo u-boot where we just switched to uEnv.txt and preEnv.txt i actually plan to store the cmdline in /etc/default/flash-kernel ... if you think we need something like that for abootimg too, we can indeed add it Aug 08 13:55:10 ogra_, we need something like that for abootimg right? In case the user would like to change a kernel parameter (console, rootfs) they should be able to do it just like with grub Aug 08 13:55:47 well, we could also just document how to do it with abootimg Aug 08 13:55:57 ogra_, also something more generic than 'copy to this device' may be needed. On the transformer you actually copy the bootimg to a certain offset on the mmc Aug 08 13:56:00 i think thats the debian approach Aug 08 13:56:02 as the boot partition is hidden Aug 08 13:56:30 but yeah, i would pretty much like to have a generic place Aug 08 13:56:37 ogra_, so a kernel update would extract the bootimg and update the initrd and zimage in it and reuse the bootimg? Aug 08 13:56:52 thats what it does on ac100 currently Aug 08 13:56:56 the bootimg.conf I mean Aug 08 13:57:04 the cmdline is set at install time and f-k doesnt touch it Aug 08 13:57:10 I thought I saw it use /boot/bootimg.conf for the ac100 Aug 08 13:57:12 on precise Aug 08 13:57:30 with the pre-3.0 f-k Aug 08 13:57:44 3.0 completely changed Aug 08 13:57:55 it doesnt generate boot.scr either anymore Aug 08 13:58:06 only updates uImage uInitrd Aug 08 13:58:09 any rationale for not supporting this,is it by design or just new codebase and noone hit this yet? Aug 08 13:58:21 i think that was by design ... Aug 08 13:58:32 which doenst mean we cant improve it ;) Aug 08 13:58:56 as i said, for u-boot i plan to go with /etc/default/flash-kernel Aug 08 13:59:01 well I wonder what the rationale was, it may be something I did not think of Aug 08 13:59:03 working on this over here atm Aug 08 13:59:10 janimo, ask lool Aug 08 13:59:15 he designed it that way Aug 08 13:59:36 if we can freely diverge from debian it may not be much benefit for them rearchitecting it carefully :) Aug 08 14:00:22 with the new uEnv/preEnv.txt stup in u-boot its just a cat $cmdline >$preEnv.txt ... i would like to have a similar simplicity in abootimg Aug 08 14:00:48 well, we already diverted 17 versions from debian Aug 08 14:01:14 and beyond things like highbank/armada support, i dont expect them to take many of tehse changes Aug 08 14:04:25 ogra_, janimo: Essentially the idea was to live with static boot scripts as much as possible Aug 08 14:05:05 generally flash-kernel lacks a range of functions for setting the cmdline, multiple kernels, and allowing the end-user to change these at boot time Aug 08 14:05:06 lool, to avoid bricking and support issues? Aug 08 14:05:19 janimo: generally keeping it simple stupid, yes Aug 08 14:05:52 if defaults are simple that should work no? While allowing those who need it override the actual boot files Aug 08 14:06:10 janimo: It's to avoid multiple code pathes Aug 08 14:06:24 many combinations of options Aug 08 14:06:37 also, handling of the cmdline is quite complex to do in a generic way Aug 08 14:07:15 In Ubuntu, we only face U-Boot, but then some platforms might be using other bootloaders; then do you express the fact that some config items don't work on certain platforms? Aug 08 14:07:30 I am not aware of the wider set of scenarios flash-kernel is used in. On the ac100/abootimg case I do not see multiple codepaths, just a way of telling what goes into the image - every image. With a sane default and the option of editing the file thus overriding it Aug 08 14:07:32 how do you deal with cmdline that might be set in U-Boot environment Aug 08 14:07:32 well, our discussion revolves around abootimg :) Aug 08 14:07:34 really like grub.cfg Aug 08 14:07:51 then there's also what do you offer in terms of cmdline: override, append, prepend, everything? Aug 08 14:07:58 some bootloaders hardcode root= in their environment Aug 08 14:08:02 same thing as grub does Aug 08 14:08:21 override would be the most commonly used I'd say Aug 08 14:08:26 janimo: Yup abootimg is fairly clean there Aug 08 14:08:31 /etc/default/flash-kernel carries a single var that has the cmdline Aug 08 14:08:41 it is much simpler to use than uboot Aug 08 14:08:47 there's also a default kernel cmdline in the kernel itself Aug 08 14:08:48 lool, ogra_: in case you weren't aware u-boot supports syslinux menus now Aug 08 14:09:01 robher, nice ! Aug 08 14:09:09 robher: Awesome, I've followed the email thread where missing features were discussed as well Aug 08 14:09:12 robher: and the bug Aug 08 14:09:14 but not a level of complexity i want to play with right now Aug 08 14:09:43 I lack a bit of time to discuss this here Aug 08 14:09:49 but I'd like to discuss a new project with you folks Aug 08 14:09:53 it's a bit early to discuss it though Aug 08 14:09:59 lool, np. thanks for your input anyway Aug 08 14:10:03 Linaro is working on a GRUB port to U-Boot Aug 08 14:10:17 looks like for abootimg we could do what ogra suggests without complicating the code Aug 08 14:10:21 This would not only change the game for the flash-kernel flexibility issues I mentioned earlier Aug 08 14:10:44 (basically flash-kernel would just be used to write a grub payload, then the usual grub configuration interface would apply) Aug 08 14:11:04 it would also change the game around the syslinux/pxelinux compatibility problems Aug 08 14:11:14 bugs and lacking features can be fixed... Aug 08 14:11:45 Instead of doing pxelinux "emulation", we'd be using grub's netboot support and making sure that e.g. cobbler generates grub config giles Aug 08 14:11:58 robher: Yes, but consider that we'll be using grub on top of uefi Aug 08 14:12:24 and that uefi's PXE support (or basically any plain PXE implementation) isn't enough for booting a kernel with an initrd and cmdline etc. Aug 08 14:13:04 Ok; now that I've launched this grenade, I'll go prepare a meeting :-) Aug 08 14:13:19 robher: I still think that it's great that we have this pxelinux "emulation" right now Aug 08 14:13:32 robher: it's definitely a great way of delivering the feature for U-Boot platforms Aug 08 14:14:37 so what's the pxelinux solution in an ARM UEFI world? Aug 08 14:15:30 robher: above, I'm proposing not to use pxelinux Aug 08 14:16:06 robher: but rather grub's script/menu config format, and implement support for that in e.g. cobbler or other places generating pxelinux ocnfigs Aug 08 14:16:57 your DHCP server would server a pre-built grub for u-boot or grub for uefi ARM binary that would optionally bundle a config (or the config can be read from the network), the config would show a menu allowing to chose between multiple kenrel + initrd images loaded from the network Aug 08 14:17:50 (I've also looked at the syslinux source code to check how much effort it would be to port the actual code to ARM and it would be crazy -- 12 kLOC of x86 assembly) Aug 08 14:18:19 (reimplementing the support for the format was definitely the right thing for a pragmatic U-Boot implementation) Aug 08 14:18:39 does grub do tftp/http loading? Aug 08 14:18:46 robher: not http Aug 08 14:19:27 Another option is to write another pxelinux/syslinux config parser as an UEFI application and/or as a GRUB command Aug 08 14:20:11 but since GRUB EFI is likely to be what distros use on installed systems anyway and since GRUB has a config / script / menu language already, I personally it is a bit more natural to just generate GRUB configs Aug 08 14:20:37 need to run, bbl Aug 08 14:44:33 speaking of grub-efi is that working with uefi on arm yet? Aug 08 14:59:55 ogra_, so regarding abootimg. If you add seek and bs args to the dd invocation it will be reusalbe on the tf101 too with non-zero seek offset :) Aug 08 15:03:03 janimo, dd invocation ? Aug 08 15:03:08 there is no dd anymore Aug 08 15:03:14 just a single abootimg call Aug 08 15:03:36 ogra_, hmm. Not good for the tf then. Aug 08 15:03:48 Oh yes, I remember you directly pass the block to abootimg Aug 08 15:03:55 instead of using a tmpfile as in precise Aug 08 15:03:58 i dont :) Aug 08 15:04:11 my code used dd and created an image first ;) Aug 08 15:04:20 yes Aug 08 15:04:22 janimo, actually there is still a way of using it Aug 08 15:04:27 thats all debians reworking that dropped this Aug 08 15:04:30 that woudl have worked for the tf Aug 08 15:04:35 lilstevie, which one? Aug 08 15:04:44 janimo, there is a hacked up driver which exposes boot/recovery Aug 08 15:04:57 lilstevie, ah indeed you mentioned that, and I forgot Aug 08 15:05:09 but with something like syslinux or grub on u-boot I could see that being irrelevant anyway Aug 08 15:05:21 that would be good. Although maybe risky to expose those, especially recovery Aug 08 15:05:21 just repartition with nvflash for initial install Aug 08 15:05:46 lilstevie, syslinux or grub on uboot are way in the future I guess Aug 08 15:05:49 janimo, we currently expose them in android on the tf201 in cm Aug 08 15:05:54 certainly not on existing hardware Aug 08 15:05:56 waaaay in the future Aug 08 15:06:08 hm Aug 08 15:06:42 didn't I just read that someone cobbled syslinux menu parsing into u-boot Aug 08 15:07:04 but you would need syslinux first :) Aug 08 15:07:07 ah Aug 08 15:07:31 ok I missed that, I thought it handled that part Aug 08 15:07:34 that could be nasty Aug 08 15:08:10 how is grub looking with uefi on arm? same deal? Aug 08 15:08:37 janimo, anyway, my current task is /etc/default, if you want changes to abootimg handling, i'll happily take patches (or you just upload them yourself :) ) Aug 08 15:08:44 lilstevie, having the code that exposes those extra partitions would be nice. Although it may confuse existing tools that expect certain partition naming order Aug 08 15:08:57 as long as you dont break ac100 handling, feel free to change it over Aug 08 15:09:23 janimo, it tacks them to the end Aug 08 15:09:27 so no, it wouldn't Aug 08 15:09:36 ogra_, sure. I was saying this so you can take it into account now. If ac100 still used dd, we would not need an extra snippet for transformer, as it would be just different parameters to the same method Aug 08 15:09:44 lilstevie, good Aug 08 15:10:04 janimo, but more importantly tf101 gets nvflashed, which it is easy to just move the partitions to the end, rather than using hacked up drivers Aug 08 15:10:05 maybe the kernel solution is better Aug 08 15:10:05 janimo, right, i dont mind either way Aug 08 15:10:28 something like tegrapart would surely be betterm yeah Aug 08 15:10:29 lilstevie, not all tf can be nvflashed. Mine which is a tf101g cannot Aug 08 15:10:37 janimo, well it can Aug 08 15:10:45 just nobody has put the effort in Aug 08 15:11:22 well, cannot at the moment, which is what I care about :) Aug 08 15:11:26 * lilstevie is still waiting for someone to try and figure out how the nvflash stuff on the tf201 works and port it to other devices Aug 08 15:18:25 lilstevie, so that patch is like ac100's tegrapart? Aug 08 15:23:18 janimo, not exactly Aug 08 15:23:36 why Aug 08 15:23:51 it hooks into the gpt driver Aug 08 18:49:47 Hello, while trying to boot ubuntu-server on my pandaboard for the first time i get a blank screen. I can switch to tty2 and get a password prompt, but i don't know the password as it has not been set yet. I would really appreciate some pointers as to what might be wrong. Aug 08 19:03:32 Limer: The server images need to run through the install setup first, which defaults to the serial console. Aug 08 19:03:56 That will set your locale, network, system name, and user info. Aug 08 19:07:19 Is there a way to reach the setup without the need of a serial cable? Aug 08 19:22:04 Limer: If you remove the console= bits from the kernel command line, it'll stop trying to do it on serial. Aug 08 19:52:56 But in order to remove those i need to make a new image? It is unfortunate that the normal image isn't installable without a serial cable. Aug 08 19:55:27 infinity: Noob question, bought the damn serial cable, can't connect to the iMX though. First some message with "Broken stream" flashes and then I get "Could not find PTY". I used sudo screen /dev/ttyUSB0 115200 Aug 08 19:56:14 Quintasan: And then screen exits? Aug 08 19:56:18 Yes Aug 08 19:56:32 Quintasan: If so, that's a problem with the cable/driver/host machine, not with the board. Aug 08 19:56:59 Quintasan: (plugging in a USB/RS-232 dongle and firing up screen on the device will "work" regardless of if you attach it to something on the other end) Aug 08 19:57:11 hurr Aug 08 19:57:13 infinity: http://wklej.org/id/806837 Aug 08 19:57:17 Or permissions. Aug 08 19:57:22 I plugged it in and it got discovered Aug 08 19:57:33 First it didn't want to eunumerate for some reason Aug 08 19:58:19 Permissions will/can also be an issue. We don't add the default user to the "dialout" group, which is required to access that device. Aug 08 19:58:45 "sudo adduser $me dialout" and log out/in, should that seem to be an issue for you. Aug 08 19:59:16 Yeah Aug 08 19:59:20 Not in dialout Aug 08 20:06:58 uhh Aug 08 20:06:59 infinity: http://wklej.org/id/806855 Aug 08 20:07:07 Any idea why do I get that? Aug 08 20:07:23 It happens everytime I copy big stuff to sdcards Aug 08 20:07:25 like images Aug 08 20:07:39 * Quintasan should probably ask in #ubuntu-kernel Aug 08 20:08:45 Your SD cards suck. Or your reader. Or it's a kernel bug. Aug 08 20:08:57 Pick one or more of the above, season to taste. Aug 08 20:09:16 Time to give up on this and buy a serial-cable tomorrow ;) Aug 08 20:09:54 infinity: Oh, ok. It's class 4 SD card Aug 08 20:10:01 i've had really bad luck with kingston brand sd Aug 08 20:10:25 Quintasan, yeah you need class 10 Aug 08 20:10:37 If you're going to rewritre frequently, I swear by Lexar (or, in fact, anything targetted at photography professionals). Aug 08 20:10:49 If you're writing it once, pretty much any junk will do. Aug 08 20:11:25 Will sudo screen /dev/ttyUSB0 115200 suffice to connect or I need to invoke some other black magic? Aug 08 20:11:41 bah Aug 08 20:11:43 Quintasan: That'll do. No need for the sudo if you're in dialout, mind you. Aug 08 20:11:57 Stil the same error Aug 08 20:12:09 So my cable is crap Aug 08 20:12:47 Ha ha, Aug 08 20:12:51 Now it won't eunumerate **** ENDING LOGGING AT Wed Aug 08 20:13:49 2012 **** BEGIN LOGGING AT Wed Aug 08 20:14:16 2012 Aug 08 20:17:34 infinity: THE WHOLE ARM BUSINESS IS IN CONSPIRACY AGAINST ME Aug 08 20:17:36 * Quintasan panics Aug 08 20:17:55 I feel the same way right now ;) Aug 08 20:18:30 Limer: What, another proud owner of iMX53? Aug 08 20:18:35 Or something worse? Aug 08 20:19:04 I am using a pandaboard, so things should work. However they don't. Aug 08 20:19:11 Logic. Aug 08 20:19:32 Limer: Your issue is just that the server image defaults to serial out, as we said. Aug 08 20:19:35 ARM = Automatically Rebooting Machine. Aug 08 20:19:42 Limer: You can change the boot.scr and fix that. Aug 08 20:20:14 if i change anything in boot.scr it refuses to boot at all Aug 08 20:20:28 Limer: Well, yes. boot.scr has a magic binary header on it. Aug 08 20:21:25 What is this black magic? How can i make a boot.scr that works? Aug 08 20:21:45 Limer: Strip off the binary header, edit the text to be what you need, save that as, say, boot.script, then: Aug 08 20:21:53 mkimage -A arm -T script -C none -n "Boot Image" -d boot.script boot.scr Aug 08 20:22:28 mkimage is in the u-boot-tools package. Aug 08 20:31:55 wow, thanks a lot! **** ENDING LOGGING AT Wed Aug 08 20:41:31 2012 **** BEGIN LOGGING AT Wed Aug 08 20:42:22 2012 Aug 08 21:47:38 Why does there seem to be a limited number of video modes availible in the linux images for my MK802? **** ENDING LOGGING AT Thu Aug 09 02:59:59 2012