**** BEGIN LOGGING AT Thu Jun 09 02:59:57 2011 Jun 09 07:32:43 is there a way to predict when the next version of ubuntu after 11.04 will be released Jun 09 07:32:50 for arm :D Jun 09 07:33:21 i am seeing usb speed issues on 11.04. can anyone replicate? Jun 09 07:36:29 https://wiki.ubuntu.com/OneiricReleaseSchedule+ Jun 09 07:36:32 err Jun 09 07:36:34 https://wiki.ubuntu.com/OneiricReleaseSchedule Jun 09 07:37:16 11.10 daily images are on cdimage.ubuntu.com Jun 09 07:39:21 ty Jun 09 08:13:42 I'm trying to migrate my target system to natty, but I'm having some problems with cross compilation. It seems like libs like libpthread.so.0 and libdl.so.2 are moved to /lib/arm-linux-gnueabi/. OOI why is it done this way? Jun 09 08:23:26 sveinse, I'm not entirely sure, but suspect it's fallout from multiarch. Jun 09 08:23:34 How are you migrating? Jun 09 08:24:27 Well, as discussed earlier, I'm building the rootfs with rootstock (which seems to work) Jun 09 08:24:44 The problems are because my cross compilation of a Qt app fails Jun 09 08:25:07 It can't find libpthreads.so.0 by itself anymore Jun 09 08:25:10 Are you using hrw's cross compilation environment? Jun 09 08:26:06 No, we've made our own. Are you thinking of xdeb or xapt ? Jun 09 08:27:00 No. hrw put together some cross-toolchain packages, which seem to behave sanely. Jun 09 08:27:14 Do you where I can find it? Jun 09 08:27:17 But I don't know much about them (I avoid cross compilation) Jun 09 08:28:19 I think one just installs "gcc-arm-linux-gnueabi" on an i386 or amd64 system. Jun 09 08:28:31 Yes, I know. But still cross compilation is neccessay/useful. Qt for example is built for cross compilation Jun 09 08:28:45 Well, we don't build it that way :) Jun 09 08:28:50 (but, yes, I know) Jun 09 08:29:25 I'm not convinced that the combination of cross-compiled and native-compiled has been extensively tested, nor that there is consistent ABI. I may be mistaken, and I'd be happy if someone did the research to confirm they are the same. Jun 09 08:29:28 Oh yes, of course. It is gcc-arm-linux-gnueabi we're using. The beauty of that cross compiler is that it's configured the same way as the armel native couterpart Jun 09 08:30:33 armel cross compilation is an experienced area in gcc. In fact it has existed even longer as cross compiler than the native version Jun 09 08:31:05 It just recently it has become possible to compile armel code natively Jun 09 08:31:07 Oh, I know. Jun 09 08:31:16 The cross-compiler is likely more mature. Jun 09 08:31:47 MY concern is about mixing (well, also about having everything behave the same so I can ignore which architecture I'm using) Jun 09 08:31:52 So I have no distrust in gcc in that respect. And the issue I have is related to linker search paths, I believe Jun 09 08:31:56 But I'm a bit of a crank about this: feel free to ignore me :) Jun 09 08:32:18 I *thought* hrw's linker was multi-arch aware. Hrm. Jun 09 08:32:24 Maybe catch him when he's around. Jun 09 08:32:47 Surely object, I don't mind (except if it takes me away from my task :D ) Jun 09 08:33:42 In fact, I'd rather merge over to native compilation. But recompiling Qt is a *heavy* task, and we do that a lot (We're not using X11) Jun 09 08:34:21 And such, Qt and our apps are cross build using the armel cross compiler Jun 09 08:34:21 Right. In your very special case, I think you're probably doing the least painful way. Jun 09 08:34:27 In general, I think folk ought use X :) Jun 09 08:34:57 Yes, agree. Unfortunately our target (OMAP3) does not have any accelerated 2D driver for X11 Jun 09 08:34:57 (as I think I argued with you many months ago, only to discover convincing you personally didn't affect the results) Jun 09 08:35:10 hehe Jun 09 08:35:19 Hrm? I thought we had one of those in the archive. Jun 09 08:35:22 pvr-something. Jun 09 08:35:48 powervr or sgx. That a 3D accel, not 2D Jun 09 08:36:22 Thats OMAPs limitation, not Qt nor X11 Jun 09 08:37:10 Ah, yeah. For OMAP3 it appears there are only the EGL and GLES drivers. Jun 09 08:37:54 Anyways since you mentioned our discussion: I am going to test to run stock qt4-x11 on X in one single fullscreen Qt app to bechmark its performance vs. the direct QWS/FB implementation Jun 09 08:39:07 If it works good, we can ditch our own Qt compilation and merge over to pure native compilation Jun 09 08:39:17 persia: we cannot be multiarch with cross compilation yet Jun 09 08:39:24 persia: no support for multiarch headers Jun 09 08:39:46 hrw, Ah. Maybe you know what sveinse needs to do to transition the dev environment to natty? Jun 09 08:40:27 sveinse: you need to update your own cross compiler to handle multiarch paths Jun 09 08:42:46 persia: did you saw http://jeffbastian.blogspot.com/2011/06/storage-speed-on-pandaboard-revisited.html page? Jun 09 08:42:56 I didn't Jun 09 08:43:40 Oh my. That's a bug needs fixing. Jun 09 08:45:26 persia, well, depends which kernel he used, he doesnt give a clue Jun 09 08:45:47 i would suspect its a PM issue Jun 09 08:46:09 if the pinging improves speed that smewhat points to that Jun 09 08:46:44 ogra_, Sure. Something to test against the kernels we have I suppose. Jun 09 08:47:12 well, i would expect that as part of GrueMaster's or janimo's spec work this cycle Jun 09 08:47:22 prpplague wrote on pandaboard ML that issue happens also without smsc on board Jun 09 08:47:40 yes most likely it is. i've modified a pandaboard to remove the Jun 09 08:47:40 LAN9514 and wire a host port connector to the EHCI. even with the Jun 09 08:47:40 LAN9514 removed i am still have similar issues. this leaves the USB Jun 09 08:47:53 i sere something similar on the ac100 .... Jun 09 08:47:55 Jun 09 08:47:58 *see Jun 09 08:48:15 and i actually suspect we have a general issue with USB speed on arm Jun 09 08:48:38 We've had lots of issues before with unexpectedly poor performance from USB storage. Jun 09 08:50:01 welll, i dont think its a storage prob at all Jun 09 08:50:13 i think its on a lower level Jun 09 08:51:23 * persia suspects interrupt handling Jun 09 08:51:42 That said, it's low-level enough that I don't trust myself to guess Jun 09 08:51:50 i also think its related to the USB NIC driver segfaults we see with many differnt drivers Jun 09 08:52:24 And perhaps the reports of horridly distorted audio with some USB audio devices Jun 09 08:53:04 since it seems the majority of USB NICs needs vm.min_free_kbytes set to some high value to not drop packages Jun 09 08:53:39 That's just wrong. Jun 09 08:53:52 thats the only known workaround currently Jun 09 08:54:14 i dont think anyone has tracked that bug further yet Jun 09 08:54:15 How does one debug this sort of thing? Jun 09 08:54:26 no idea, ask the kernel team :) Jun 09 08:54:56 likely by scattering printks all over the code Jun 09 08:55:10 Right. Jun 09 08:55:29 * persia climbs back up the stack into code that doesn't even get compiled Jun 09 08:56:09 on the ac100 is see some funny behavior with PM and wlan ... PM powers down the card faster than a ping returns ... Jun 09 08:56:51 so you get proper throughput when you have a constant stream of data ... but if you ping you get like 3-5sec return times Jun 09 08:57:05 (when pinging without traffic in the card) Jun 09 09:03:27 I'm trying to use arm-linux-gnueabi-gcc without sysroot (since it does not support it). How do you prevent arm-linux-gnueabi-gcc to attempt to link with host libs? Jun 09 09:03:57 It tries to load: "opened script file /usr/lib/x86_64-linux-gnu/libz.so" which is not correct Jun 09 09:07:53 hey, but its 64bit, probably makes your binaries run faster on arm then :P Jun 09 09:08:20 The armel-cross ld shouldn't be searching /usr/lib and /lib, should it ? Jun 09 09:08:28 :P Jun 09 09:08:43 no idea, i dont touch cross stuff Jun 09 09:08:56 hrw ^^ Jun 09 09:22:20 doing arm-linux-gnueabi-gcc -v ... reveals: Jun 09 09:22:32 LIBRARY_PATH=/usr/lib/gcc/arm-linux-gnueabi/4.5.2/:/usr/lib/gcc/arm-linux-gnueabi/4.5.2/../../../../arm-linux-gnueabi/lib/:/usr/lib/x86_64-linux-gnu/ Jun 09 09:22:58 I don't see why the latter is present Jun 09 09:23:25 sveinse: and ' --with-headers=/usr/arm-linux-gnueabi/include --with-libs=/usr/arm-linux-gnueabi/lib' is shown too Jun 09 09:24:13 sveinse: here arm-linux-gnueabi-gcc -v does not show LIBRARY_PATH Jun 09 09:24:48 sveinse: http://pastebin.com/vmHmg8t6 Jun 09 09:24:51 I think you need it in a complete line where it tries to link Jun 09 09:25:41 Try this: arm-linux-gnueabi-gcc -Wl,--no-undefined -Wl,-e,qt_core_boilerplate -Wl,-O1 -shared -Wl,-soname,libQtCore.so.4 -o libQtCore.so.4.7.2 -lpthread -lz -lm -ldl -lrt -v Jun 09 09:25:57 Its does nothing, but it reveals the settings Jun 09 09:27:49 FYI this is arm-linux-gnueabi-gcc 4.5.2 from natty stock Jun 09 09:29:12 hrw, 4.5.2-8ubunutu3 to be more specific Jun 09 09:36:00 thx - it really looks strange Jun 09 10:37:28 hrw, why was --sysroot removed from the toolchain? Isn't it just about not using --sysroot and then it behaves equal to ld without sysroot support? Jun 09 10:41:53 Because when cross building (non debian source), if ld encounters a linker script with say "GROUP ( /lib/arm-linux-gnueabi/libphread.so.0)", using --sysroot would make ld prefix that path with a dir, while it's rather cumbersome without sysroot because I have to alter the linkerscripts. Jun 09 10:42:40 Alternatively, does ubuntu have some chroot tool/environment to do cross building in? Jun 09 10:42:45 sveinse: iirc I already wrote why in your bug report Jun 09 10:43:31 sveinse, qemu-debootstrap Jun 09 10:45:13 hrw, no offence but this means that there is no will to resolve the issue and make sysroot available. With other words: I don't understand why the cross toolchain is broken by a setting which can be omitted. Jun 09 10:46:44 sveinse: Linaro will provide binary toolchain which will be sysrooted Jun 09 10:47:36 Ubuntu cross compilers will still be like they are cause they are more targetted as helping tool for those who want to cross compile for ubuntu then as cross compiler for embedded target developers (like in your use case) Jun 09 10:47:37 do you know if there exists a pre-build PPA for that? Jun 09 10:47:52 no ppa - it will be tarballs not packages Jun 09 10:47:59 ok Jun 09 10:48:19 I can send you link when will produce first working one Jun 09 10:48:24 excellent! Jun 09 10:49:13 but hope that will get sane suggestions/opinions ;D Jun 09 10:50:22 yes, you'll have to excuse my naive questions, I'm just sitting here with a system which does not work. Jun 09 10:50:48 And yes, I hoped the armel-cross were a cross tool similar to codesourcery, but apparently not Jun 09 10:51:10 even though I dont think it's far off being one Jun 09 10:52:22 As a tempoarary workaround (today) I can either create a small tool which will search for all linker-script in my rootfs and replace it with the host path Jun 09 10:52:51 Or, perhaps as ogra_ suggest qemu-debootstrap, but I'm not sure how to use it though Jun 09 10:53:00 like debootstrap Jun 09 10:53:19 behaves the same from a user POV Jun 09 10:54:11 so you can run arbitrary command inside a chroot? Jun 09 10:54:33 you can run a full arm userspace in a chroot Jun 09 10:55:12 Using host executables (and not binfmt/qmeu)? Jun 09 10:55:22 no Jun 09 10:55:28 it indeed uses binfmt Jun 09 10:56:11 So compiling qt will be much slower then Jun 09 10:56:16 sveinse: http://home.haerwu.biz/~hrw/t/ is my experimental one - no warranty that works Jun 09 10:56:34 will be faster than in a qemu VM but yeah, slower than cross Jun 09 10:57:19 hrw, thanks I'll try Jun 09 10:59:55 sveinse: fetch and ping me - I will remove tarballs Jun 09 11:02:00 You know, I actually planned on blogging how we did our setup to make a working cross build environment. But I'm uncertain if there are interest for such a blog because we're using all the tools which is not recommended or "you're on your own"... We use rootstock to create a rootfs-dev/ dir where all -dev packages are installed. And then we used to do gcc --sysroot to that rootfs-dev/ to... Jun 09 11:02:02 ...cross build apps. Jun 09 11:02:16 hrw, eta. 4mins or so Jun 09 11:04:41 I do understand cross building is not the way ubuntu does it or wants to do it, but it's not more dirty to cross build than any other method. Apparently a bit more cumbersome obviously. Jun 09 11:07:04 hrw, done, thanks Jun 09 11:07:20 thx. dropped Jun 09 11:07:26 sveinse: write a post Jun 09 11:08:05 any ideas of a site where I can blog? Jun 09 11:15:43 create own on on wordpress.com? Jun 09 11:20:17 hrw, cc1plus seems to depend on a set of libs. e.g. libgmp.so.10 is missing from my system (and I cant find it on packages.ubuntu.com) Jun 09 11:21:05 sveinse: what is your system? Jun 09 11:21:14 Strange... ldd sais something else... Jun 09 11:21:20 amd64, natty Jun 09 11:21:51 should work under natty. I built under oneiric Jun 09 11:22:47 I can try, and ignore the warning. Jun 09 11:32:50 It seems to fail because of this missing so. At least it does not give any other indication why it fails Jun 09 11:33:38 hrw, http://pastebin.com/1cqLSe85 Jun 09 11:34:25 this is from a configure test within qt in our build setup Jun 09 11:40:16 natty has libgmp.so.3 Jun 09 11:40:29 try in oneiric chroot Jun 09 11:43:43 once I will get it into buildable state I will do build under natty or maybe even lucid Jun 09 12:35:41 ppisati, ping Jun 09 12:36:05 ericm|ubuntu: pong Jun 09 14:28:17 GrueMaster, do you happen to have an up to date netbook/desktop image around on a panda or so ? Jun 09 14:28:59 Not atm, but I can image one and do stuff very quickly. Jun 09 14:29:52 * ogra_ would like to know if Bug 794938 happens on panda too Jun 09 14:29:53 Launchpad bug 794938 in lightdm "lightdm dies with "Failed to fork: Cannot allocate memory" on login" [High,New] https://launchpad.net/bugs/794938 Jun 09 14:30:13 since thats our default login manager since a few days, it would be good if it worked ;) Jun 09 14:30:27 which reminds me ... Jun 09 14:30:42 * ogra_ looks what it takes to switch to the desktop seed, i should do that now Jun 09 14:31:29 let me know what the new name will be. I am also mirroring the kubuntu stuff and want to make sure of no filename conflicts. Jun 09 14:32:08 i hope it will be the same as the main x86 image with just preinstalled added and armel as subarch Jun 09 14:32:26 we'll see what comes out at the rear end i guess :) Jun 09 14:32:30 That's the same name as the kubuntu files. Jun 09 14:32:58 I wish there was a way to differentiate. Would make my life easier. Jun 09 14:33:22 Guess I'll have to modify my mirror script again. Jun 09 14:33:52 * GrueMaster checks to see if the coffee is ready. Jun 09 14:33:53 well, ideally we would have the same name just different subarch Jun 09 14:34:13 but we dont do isos and we're preinstalled Jun 09 14:36:57 oneiric-preinstalled-desktop-armel+omap4.img.gz is the name of the kubuntu image. Jun 09 14:38:59 well, i just fired off a build Jun 09 14:39:11 lets see if we're lucky and the archive is in sync :) Jun 09 14:39:27 if so, we'll see what comes out :) Jun 09 14:39:54 i'm curious where it will be published :) Jun 09 14:44:16 hrw, thanks for the storage speed on panda link, checking it now Jun 09 14:44:23 np Jun 09 14:54:48 hrw, although it indeed looks a board specific issue and not something to take into account in our storage speedup blueprint which si SD centric anyway Jun 09 14:56:10 SD should die Jun 09 14:57:00 0.5$ sd card reader can get better results from simple tests then omap4 controller sometimes Jun 09 15:00:45 hmm, seems my build will survive, its over 20min in Jun 09 15:09:25 Hello. Jun 09 15:09:47 I'm having problem with my Ubuntu on BeagleBoard-xM Jun 09 15:09:59 mmcblk0: error -110 sending read/write command, response 0x0, card status 0x400e00 Jun 09 15:10:18 lots of those messages during boot Jun 09 15:26:00 Garagoth: What image are you using? Jun 09 15:27:00 netinstall Jun 09 15:27:17 it installed fine... I used it for a while... Jun 09 15:27:21 then reboot Jun 09 15:27:33 and it cannot boot as / cannot be mounted Jun 09 15:27:56 netinstall? We don't currently have a supported netinstall. Jun 09 15:28:12 you have... Jun 09 15:28:54 They are enabled and get built with the debian installer, but they are untested and buggy. Jun 09 15:29:10 Howdy folks. Jun 09 15:29:32 http://elinux.org/BeagleBoardUbuntu#Ubuntu_11.04_.28Natty.29 Jun 09 15:29:50 netinstall image Jun 09 15:30:11 s/image/method/ Jun 09 15:32:41 GrueMaster: So netinstall I made is not supported by any means? Jun 09 15:32:57 Not by Ubuntu. We are working on it this cycle. Jun 09 15:34:09 Ah well. 3 days of work lost. Jun 09 15:36:40 I'm following the guide here - https://wiki.ubuntu.com/ARM/RootfsFromScratch Jun 09 15:36:49 And when I try to launch the qemu I am getting kernel panic Jun 09 15:36:51 Any ideas why? Jun 09 16:01:45 Garagoth: Most of the files these instructions use are not even from Ubuntu. I'm looking over the scripts now. There is no way we can support something someone else has cobbled together. Jun 09 16:03:05 We do have a preinstalled image for both headless & netbook for 11.04 though. The headless image uses the serial console to run oem-config, and you can select different tasks from it for server or desktop workloads. Jun 09 16:04:40 GrueMaster: headless install failed for me. Jun 09 16:04:51 Oh? How so? Jun 09 16:05:06 What rev board do you have? Jun 09 16:05:10 I can re-try it now and then I can provide you with details. Jun 09 16:05:32 BB-xM rev C Jun 09 16:06:08 There were some changes apparently. You might need my updated x-loader/u-boot. Jun 09 16:06:17 Let me find the link Jun 09 16:10:00 is x-loader/u-boot update required for headless to install? Jun 09 16:10:25 I have Texas Instruments X-Loader 1.5.0 (Mar 29 2011 - 09:06:55) Jun 09 16:11:18 I'm looking. It is bug 770679 if you want to read up on it. Jun 09 16:11:20 Launchpad bug 770679 in linux "Missing proper support for Beagle XM rev B and C" [Medium,Fix committed] https://launchpad.net/bugs/770679 Jun 09 16:12:55 Looks like a kernel fix. Jun 09 16:13:00 For natty. Jun 09 16:13:57 is it in headless image already, or should I patch it somehow? Jun 09 16:14:31 rsalveti: ^^^ Any hints on this? Jun 09 16:14:47 anyone know of a prebuilt kernel that works with arm qemu? Jun 09 16:15:09 Garagoth: If it requires adding a kernel to the image, I'll bundle it in a tarball with instructions. Jun 09 16:15:55 Thanks. Jun 09 16:17:17 In the meantime I will re-try with headless image Jun 09 16:17:20 GrueMaster: just need to generate an uImage from latest kernel available at proposed Jun 09 16:17:27 it's not going to boot for the first time Jun 09 16:17:39 you need to replace it at the image to make it boot and then install it correctly Jun 09 16:17:49 kind of the same issue we had in the previous cycle Jun 09 16:17:54 ok. I'll pull it in on my XM and create a tarball & instructions. Thanks. Jun 09 16:18:27 GrueMaster: yeah, but best thing is to compile the same kernel that's available with the release, adding the important patches on top Jun 09 16:18:36 as then we can just request the user to download the uImage Jun 09 16:18:36 Hm. I got installer, but installer failed as it had not network. Is it likely the symptom? Jun 09 16:18:43 instead of downloading the deb and installing it Jun 09 16:19:05 GrueMaster: I'll generate the uImage with the same kernel version used at the release Jun 09 16:19:22 GrueMaster: can you just get what kernel version is the one used at the released image? Jun 09 16:20:41 linux-image-2.6.38-8-omap 2.6.38-8.42 Jun 09 16:21:55 thanks Jun 09 16:22:00 When booting current headless image I get: "init: ureadahead-other main process (372) terminated with status 4" Jun 09 16:22:01 give me some minutes Jun 09 16:22:08 that's expected Jun 09 16:22:14 ok. Jun 09 16:23:34 And it says: Beagle unknown 0x02 Jun 09 16:26:19 this should be fixed with the new kernel Jun 09 16:26:28 hold some minutes and I'll have an uImage for you to test Jun 09 16:26:36 Sure. Jun 09 16:37:17 Garagoth, what kernel (uname -a) where you running when you were getting flooded with -110 errors? Jun 09 16:37:31 2.6.39-x1 Jun 09 16:37:51 humm, weird.. which board? Jun 09 16:38:17 BB-xM rev C Jun 09 16:39:39 during install those errors appeared (during boot only I think) Jun 09 16:39:43 very strange.. i haven't seen that yet on my xM's.. i did pushout an 2.6.39.1-x1 update a couple days ago.. if you still have the image, are you able to try it? http://rcn-ee.net/deb/natty/v2.6.39.1-x1/ (install-me.sh script) Jun 09 16:40:01 and after reboot same errors caused / to be unable to be mounted Jun 09 16:40:30 that image does not boot anymore Jun 09 16:40:42 fails with message that / cannot be mounted Jun 09 16:40:46 Sweet, then my 2.6.38 -> to 2.6.39 broke boards.. great.. Jun 09 16:41:37 :-) Jun 09 16:42:04 Your netinstall script was fine Jun 09 16:42:34 everything worked except no slip kernel module Jun 09 16:42:43 and I need it :-) Jun 09 16:43:08 what config do you need? Jun 09 16:43:40 Here's my last xm stress test: http://rcn-ee.homeip.net:81/dl/gcc/sys/dmesg-174678-2.6.39-x1-beagle-xma-512mb.log (why i'm not getting the mmc -110 error..) Jun 09 16:43:45 I have pair of Xbee connected over USB and I need to connect then using slip ... Jun 09 16:44:20 GrueMaster: http://people.canonical.com/~rsalveti/beaglexm/uImage Jun 09 16:44:32 GrueMaster: this is the kernel 2.6.38-8.42 with the rev c patches included Jun 09 16:44:37 Are you using own sd card? I'm using card shipped with beagle... Jun 09 16:45:01 GrueMaster: just update the wiki doc explaining to replace the uImage from the sdcard with this one Jun 09 16:45:10 then the first boot should be fine Jun 09 16:45:33 Garagoth, yeah, i thru away the ones came with the xM's.. to slow.. Jun 09 16:45:42 (thru/burned) Jun 09 16:46:32 rcn-ee: Hm. So maybe this is a reason... I mounted that card on my laptop, fsck'ed it.. copied it fully using dd .. no errors in process. Jun 09 16:46:49 but on BB-xM it fails to work. Jun 09 16:46:51 GrueMaster: the user should also replace the vmlinuz that's inside the second partition, at /boot Jun 09 16:46:59 GrueMaster: replace it with http://people.canonical.com/~rsalveti/beaglexm/zImage Jun 09 16:47:02 Same card works fine with Angstrom Jun 09 16:47:21 GrueMaster: as one step from jasper is to run flash-kernel again Jun 09 16:47:44 replacing both files should be enough to get a working system based on the released images Jun 09 16:47:51 same procedure as we used during maverick Jun 09 16:48:37 GrueMaster: also put a note saying the user should enable natty-proposed in case he's updating his system Jun 09 16:48:39 ok Jun 09 16:48:42 rcn-ee: Also, same sd card 'looks' faster under Angstrom then under Ubuntu Jun 09 16:48:48 as otherwise he'll get a newer kernel but not the one with the fixes Jun 09 16:49:08 Actually, there is no newer kernel excpet in proposed at the moment. Jun 09 16:49:10 and that should do it Jun 09 16:49:18 even better Jun 09 16:49:38 thought we had Ubuntu-2.6.38-9.43 Jun 09 16:49:53 Garagoth, yeah the mmc driver has changed a lot since 2.6.32 in Angstrom... Do you need the other 3 config's enabled http://pastebin.com/5BwQT9aN Jun 09 16:49:57 Haven't seen one. Jun 09 16:50:03 They also boot at 1Ghz.. Jun 09 16:50:50 My xm is currently running natty headless as I am using it for usbboot testing on panda. I have been monitoring updates. Jun 09 16:51:48 rcn-ee: Mm. well, slip was only thing I missed so far; usb serial, fddi and pl drivers were there... does your .39 kernel and build include patches rsalveti included? Jun 09 16:52:02 probably Jun 09 16:52:12 as rcn-ee was the one who created the patches ;-) Jun 09 16:52:36 it was one of those #'s.. https://bugs.launchpad.net/bugs/770679 ;) Jun 09 16:52:38 Ubuntu bug 770679 in linux "Missing proper support for Beagle XM rev B and C" [Medium,Fix committed] Jun 09 16:52:41 rsalveti: rcn-ee; I've been fighting the same battles as Garagoth over the last few days and I'm curious, does u-boot need to be updated to detect bb-Xm rev C also? Jun 09 16:53:03 rsalveti: :-) Jun 09 16:53:13 i thought no... but a couple android people running xM C's with old u-boot these last few days have show otherwise.. Jun 09 16:53:36 mattwaddel: I have bb-xm rev c Jun 09 16:53:39 so yes.. Jun 09 16:54:03 Garagoth: where did you get the updated u-boot? Jun 09 16:54:18 mattwaddel: check bug 770679 Jun 09 16:54:19 Launchpad bug 770679 in linux "Missing proper support for Beagle XM rev B and C" [Medium,Fix committed] https://launchpad.net/bugs/770679 Jun 09 16:54:22 you'll find the patches Jun 09 16:54:30 but I don't think they are really necessary Jun 09 16:54:34 rsalveti: got it, thx Jun 09 16:55:46 Oh rsalveti, heads up a new older beagle "C5" is coming out with new memory, mlo/u-boot patches have been posted to the beagleboard group... Jun 09 16:56:02 rcn-ee: yeah, saw that Jun 09 16:56:17 will be integrating in the following days Jun 09 16:56:27 but thanks for heads up Jun 09 16:56:32 ah okay cool. Been updating my script to include it to.. Jun 09 16:58:44 rsalveti: Do you have a vmlinuz for the patched kernel? Jun 09 16:59:23 GrueMaster: grab the zImage at the same link I sent Jun 09 16:59:48 Ok. I'll just rename it. Make it less confusing for the users. Jun 09 17:03:22 GrueMaster: would be good to rename with the correct version Jun 09 17:03:23 Do we care about fixing maverick? Jun 09 17:03:26 GrueMaster: do you have in hands? Jun 09 17:03:41 like vmlinuz-2.6.38-foobar Jun 09 17:03:54 I can rename at my host area Jun 09 17:03:54 I have it now. Am posting up to my updates directory. Jun 09 17:04:05 ok, cool Jun 09 17:04:26 GrueMaster: for now let's just fix natty Jun 09 17:04:35 ok Jun 09 17:20:38 re Jun 09 17:22:48 GrueMaster: How long till new files for testing will be there? Or should I just grab what rsalveti published? (It is important as I am hungry and I wonder if I have time to go eat something before I get something to test) Jun 09 17:23:23 Garagoth: Done. https://wiki.ubuntu.com/ARM/OmapNetbook Jun 09 17:23:47 Same instructions and files apply to headless (updating that wiki now.) Jun 09 17:27:36 Cool, thanks. Jun 09 17:27:45 Applying... Jun 09 17:29:14 Let me know if there are any issues with the kernel or my instructions. Jun 09 17:29:40 Instructions are fine. Jun 09 17:32:30 one more thing... netbook image does not have /boot/initrd.img-2.6.38-8-omap Jun 09 17:32:39 but there is symlink pointing to it Jun 09 17:34:35 Yes, that is correct. It is made by oem-config. Jun 09 17:38:25 oooh. it detected network. Better. Jun 09 17:39:32 hm, so netbook install now will give me working mouse and keyboard, right? (previously they were dead, just image on screen) Jun 09 17:40:51 They should if you are using the netbook image. If you are installing from the headless image...I don't know. May need to tweak the cmdline. Jun 09 17:41:44 How can I convert headless to give me console? u-boot params? Jun 09 17:41:52 & getty ? Jun 09 17:43:18 After running through oem-config, edit /boot/boot.script and either add console=tty0 or remove the current console==ttyO2,115200n8. Then rerun flash-kernel & reboot. Jun 09 17:43:43 Sounds easy. Thanks. Jun 09 17:43:54 (should be only one "=". I have phat phingers) :P Jun 09 17:44:11 Mm, figured it out. Jun 09 17:45:08 If you remove the "console= " entirely, you should get a nice graphical boot screen. Jun 09 17:45:43 Hm? But I will loose serial console, right? Jun 09 17:45:53 True. Jun 09 17:46:16 Angstrom does have both, serial & X at same time. Jun 09 17:46:30 but their ld is broken somehow. Jun 09 17:46:59 You can add a serial login by copying /etc/init/tty2.conf to /etc/init/ttyO2.conf and editing the file to use ttyO2 115200. Jun 09 17:48:10 You won't see boot messages, but you will have a serial console login. Jun 09 17:48:31 but won't u-boot display there anyway? Jun 09 17:49:04 Yes, that won't change. What you would see is u-boot log output, followed by a login screen. Jun 09 17:49:52 Hmm. Gonna figure out how Angstrom did that... but later, for now headless is fine. Jun 09 17:50:39 By the way, why headless install displays red screen on dvi? Jun 09 17:50:45 bright red Jun 09 17:51:01 because it's written by commies Jun 09 17:52:13 I don't see that. Jun 09 17:52:35 When do you see that? Jun 09 17:52:42 all the time Jun 09 17:52:51 for whole install Jun 09 17:53:11 it is now removing packages, and still bright red Jun 09 17:53:27 Might be due to the last minute color map changes made prior to release. Jun 09 17:53:54 or another rev C difference? Jun 09 17:56:02 Garagoth: that's one of the things I'm seeing with the Linaro images also Jun 09 18:05:46 That's not really surprising: it's the same uboot. Jun 09 18:07:32 It could be that u-boot isn't recognizing the silicon and throwing up a red screen. I noticed on the bug that there is a patch floating around upstream. Shouldn't be critical though. Jun 09 18:10:34 Workaround: either initialise a framebuffer or don't attach a screen :) Jun 09 18:10:59 :-) Jun 09 18:54:36 um... why there is no libserial on ubuntu-arm ? Jun 09 18:54:59 normal natty has it Jun 09 18:56:29 WOHOOOO ! Jun 09 18:56:31 http://cdimage.ubuntu.com/daily-preinstalled/current/ Jun 09 18:56:59 GrueMaster, looks like the name is the same as kde :/ Jun 09 18:57:14 Pffft Figures. Jun 09 18:57:28 * GrueMaster modifies scripts accordingly. Jun 09 18:58:07 * ogra_ is really surprised it built Jun 09 18:58:09 Hey, what about libserial? I kinda need it :-) Jun 09 18:58:24 Garagoth: apt-get install libserial? Jun 09 18:58:38 The headless images are very minimal. Jun 09 18:58:59 ITYM libserial0 Jun 09 18:59:09 E: Unable to locate package libserial Jun 09 18:59:19 ogra_: When will netbook disappear? Jun 09 18:59:33 E: Unable to locate package libserial0 Jun 09 18:59:38 Garagoth: make sure universe is enabled. Jun 09 18:59:38 GrueMaster, after i changed the crontab (which i'll do after the call) Jun 09 18:59:48 libserial-dev is also non-present for arm Jun 09 18:59:52 hmm Jun 09 19:00:25 * ogra_ sees it Jun 09 19:00:27 Garagoth: libserial0 Jun 09 19:00:41 universe you say.... Jun 09 19:00:55 yep. Jun 09 19:01:03 pool/universe/libs/libserial/libserial0_0.6.0~rc1-0ubuntu2_armel.deb Jun 09 19:03:38 My fault for not checking this... I assumed that repo config is same as for desktops Jun 09 19:04:09 It "should" be. Thought we had that fixed. Jun 09 19:04:19 grmbl. Jun 09 19:04:34 we did Jun 09 19:04:47 what image is that ? Jun 09 19:05:00 ubuntu-11.04-preinstalled-headless-armel+omap.img Jun 09 19:05:26 hmm, that should definitely enable universe in your sources.list Jun 09 19:05:49 only restricted was uncommented Jun 09 19:06:20 hmpf Jun 09 19:09:58 oh, wow Jun 09 19:10:05 the first upload of linux 3.0 Jun 09 19:10:11 that was quick Jun 09 19:23:37 netbook image also has only restricted sources enabled. Jun 09 19:27:19 Garagoth: did the image work for you? Jun 09 19:28:08 I just installed it. A second please, I will reboot it now to verify. Jun 09 19:29:02 it still says: "Beagle unknown 0x02" Jun 09 19:29:39 uhm... Jun 09 19:29:50 ALERT! /dev/disk/by-uuid/b1971b59-d07f-4f96-9478-4c96a43e1e56 does not exist. Dropping to a shell! Jun 09 19:30:03 that's weird, same happened with Matt_O Jun 09 19:30:19 eh? Jun 09 19:30:26 first time he tried it showed unknown, then after a reboot it showed rev C Jun 09 19:31:59 well... I still see unknown 0x02... and root device is not found Jun 09 19:32:14 initramfs shell Jun 09 19:32:21 from busybox Jun 09 19:32:27 and that is all I have now... Jun 09 19:33:37 but dmesg shows: mmcblk0: p1 p2 Jun 09 19:35:07 Hm! it has dirrefent uuid ... Jun 09 19:38:58 Hm, no editor in initramfs / ash ? Jun 09 19:39:47 not usually Jun 09 19:39:52 but you can do the usual "dumb" editor Jun 09 19:40:02 echo > file Jun 09 19:40:05 echo >> file Jun 09 19:40:10 ad nauseum Jun 09 19:40:15 I use cat | sed Jun 09 19:40:18 no sed Jun 09 19:40:18 :D Jun 09 19:40:30 yes sed Jun 09 19:40:37 is there sed in that busybox? Jun 09 19:40:42 or initramfs? Jun 09 19:40:44 yes Jun 09 19:40:49 and awk Jun 09 19:40:50 yeah, then cat | sed Jun 09 19:40:52 ash Jun 09 19:41:18 I thought that they still had some nano/pico or vi in there Jun 09 19:41:21 it's gone? Jun 09 19:41:35 cause it was the most useful thing in there, for interactive editing Jun 09 19:41:39 so it seems Jun 09 19:41:50 damn Jun 09 19:41:56 That's unfun Jun 09 19:42:02 * Martyn checks on his Versatile Express Jun 09 19:42:11 I'll be sporked .. yeah .. no vi, no pico Jun 09 19:44:17 hm... where does that path exist? Jun 09 19:44:26 I corrected fstab Jun 09 19:44:29 where else? Jun 09 19:44:51 ahm. initramfs Jun 09 19:47:20 how can I edit this? Jun 09 20:02:39 I have a request... please include nano and flash-kernel into initramfs Jun 09 20:02:55 why ? Jun 09 20:02:59 its on the rootfs Jun 09 20:03:03 just chroot into it Jun 09 20:03:16 chroot... and remount /proc and /dev Jun 09 20:03:27 into chroot Jun 09 20:03:33 mount --bind /dev /root/dev ... Jun 09 20:03:40 before chrooting Jun 09 20:03:45 Mm. Jun 09 20:03:50 then mount sys and proc if you are inside Jun 09 20:04:11 Ok. You are right... Jun 09 20:05:56 rsalveti: Ok. System went up after fixing boot.script & flash-kernel Jun 09 20:06:12 cool Jun 09 20:06:18 for rome reason there was wrong uuid Jun 09 20:06:26 for root device Jun 09 20:07:00 and not whong as a typa.. entirely wrong Jun 09 20:07:04 typo* Jun 09 20:08:57 is mpurate in boot.script a cpu speed? Jun 09 20:09:19 yes Jun 09 20:09:22 it is set to 500... as I have bb-xm, can I change it to 800 or 1000 ? Jun 09 20:09:35 800 is the maximum recommended Jun 09 20:09:37 i think 900 is max with that kernel Jun 09 20:09:47 and 800 the max recommended ;) Jun 09 20:09:56 Stick to 800 Jun 09 20:10:02 you don't want all the heat that you get at 900 Jun 09 20:10:14 (power consumption is a nasty, power function curve) Jun 09 20:10:31 I do not care about power consumption Jun 09 20:10:56 I have attached to beagle 2 microframe motors, each consuming 15 Amps @ 24 V Jun 09 20:11:01 you can also get some weird errors with 900mhz Jun 09 20:11:22 yes, it can cost stability Jun 09 20:11:56 So advertised 1GHz is a myth? Jun 09 20:12:26 No, it's just high sort Jun 09 20:12:36 no, lack proper software support Jun 09 20:12:48 I have an i.mx53 running at 1Ghz .. it's hot, and it's not entirely stable, but with a heatsink it does stabilize and works well Jun 09 20:13:21 Hm, so heatsink would be a solution you say... and what does mpurate=auto do? Jun 09 20:13:33 lets it float Jun 09 20:13:43 from to where? Jun 09 20:13:54 dunno .. I'd have to dig around the kernel code :) Jun 09 20:14:02 Garagoth: bug 771537 Jun 09 20:14:03 Launchpad bug 771537 in linux "mpurate=1000 fails on beagleXM" [Medium,In progress] https://launchpad.net/bugs/771537 Jun 09 20:16:06 between 500 and 700 Jun 09 20:16:07 Smart reflex? Jun 09 20:16:21 Martyn: thanks. I will set 800 then... Jun 09 20:16:57 Garagoth: dynamic voltage and frequency support Jun 09 20:17:19 http://focus.ti.com/general/docs/wtbu/wtbugencontent.tsp?templateId=6123&navigationId=12032&contentId=4609 Jun 09 20:18:33 Yeah, found the same link... had no idea that linux has this... Jun 09 20:19:08 I assumed it is a hardware solution, not software... Jun 09 20:19:35 you also need support at the software side Jun 09 20:35:56 mpurate is 800... what about core clock? (332) Jun 09 20:39:42 Martyn, the PXE code ... Jun 09 20:40:28 Martyn, does that send any identifier to the DHCP server during the communication so we can make out what device sent the request ? Jun 09 20:40:40 yes Jun 09 20:40:53 although I don't have the code in front of me Jun 09 20:41:04 best person to ask is Jason -- jason.hobbs@calxeda.com Jun 09 20:41:05 i.e.a vendor-class identifier or some such Jun 09 20:41:43 yea Jun 09 20:42:06 how does it set that ? Jun 09 20:42:13 based on the board name ? Jun 09 20:42:17 yes Jun 09 20:42:18 CONFIG_BOOTP_VCI_STRING Jun 09 20:42:24 awesome Jun 09 20:42:39 for versatile express: U-boot.ca9x4_ct_vxp Jun 09 20:42:46 thanks :) I knew you were online :) Jun 09 20:42:55 I just wasn't sure if you were watching IRC :) Jun 09 20:43:50 there's also a client architecture field that can be used, it's stuff like "x86_32" or "x86_64" usually, but doesn't have official definitions for ARM parts Jun 09 20:44:49 yeah, that doesnt help our usecase ... but a vendor-class-identifier will Jun 09 20:45:26 we need to know the subarch, not the arch :) Jun 09 21:58:42 gruemaster: if i install a headless image, there is no reason that i cant have it upgrade the packages to be an image with display, but after the initial install right? Jun 09 21:59:08 as in, there is no special bit of magic outside of the package manager installed bits differentiating head and headless installs Jun 09 21:59:14 Just run tasksel to select what environment you want. Jun 09 22:00:36 The only other tidbit is you will need to edit /boot/boot.script and either delete the console= section or add console=/dev/tty0 Jun 09 22:03:33 cool, so not involved at all Jun 10 00:31:46 Are there any known usb-storage issues on natty/pandaboard ? I have all sorts of interesting issues w/ various chipsets I use in linux/windows on a regular basis, the JM20336 always drops fairly soon after plugging it in, the CY7C68300A normally works but drops sometimes as well ( have rootfs on it so thats troublesome ) ... another gives an error and disables all URBs and never works Jun 10 00:38:03 the jmicron chipset gets a reset then tells me the device offlined, not ready after error recovery Jun 10 00:44:31 Seems like it might just not like the SSD connected.. Jun 10 01:12:55 * jburkholder42 feels left out with no hardware Jun 10 01:33:50 Anyone know a solid guide for getting an ARM system running in qemu? Jun 10 01:46:27 Jef91: https://wiki.ubuntu.com/ARM/RootfsFromScratch Jun 10 01:47:03 I've got that, when I try to launch the qemu I'm only getting a black screen Jun 10 01:48:40 did you try the chroot? Jun 10 01:49:31 IIRC both the system emulator and userspace emulator worked last time I tried Jun 10 01:50:32 What are you trying to do with the emulator? Jun 10 01:53:44 just launch the system Jun 10 01:54:25 I'm launching "qemu-system-arm -M versatilepb -cpu cortex-a8 -kernel ./vmlinuz-lucid -hda arm.img -m 256 -append "root=/dev/sda mem=256M devtmpfs.mount=0 rw" Jun 10 01:54:25 " Jun 10 01:54:55 Oh, I meant at a much higher level. There's been a few people by who wanted to do stuff and it turned out the stuff they wanted to do didn't require the emulator they thought it might initially. Jun 10 01:55:14 That ought work. Jun 10 01:56:04 I want to create a full root FS Jun 10 01:56:15 want to build some packages in the ARM arch. as well for a repo Jun 10 01:57:44 If you have hardware, I'd recommend doing the building there. Jun 10 01:58:35 If you don't have hardware, I'd suggest using sbuild or pbuilder with a cross-chroot, and not running full qemu. Jun 10 01:58:58 I'd like to do it in a full qemu as I'm going to be building a rootfs from it eventually Jun 10 02:00:43 You plan to build a rootfs based on the deritus of the environment in which you've been running builds? Jun 10 02:01:41 No I plan to have two environments Jun 10 02:02:41 A build environment and a target environment, for deployment? Jun 10 02:02:48 Correct. Jun 10 02:04:04 Thats a good idea right? Jun 10 02:04:43 OK. So, for the build environment, I'll recommend *not* using qemu's system emulation, and just use the syscall emulation with pbuilder or sbuild's cross-chroot support. Jun 10 02:05:13 That's more likely to have been tested (yes, it's noisy about some things, but still). Jun 10 02:05:13 Alrighty - noted. Jun 10 02:05:27 I still need a qemu though Jun 10 02:05:34 for setting up a depolyment system Jun 10 02:05:49 For testing, or for chroot creation? Jun 10 02:06:13 Both I would think Jun 10 02:07:36 In my experience, running on hardware and running in qemu have surprisingly different performance characteristics. Jun 10 02:08:12 One of the key bits is that most Ubuntu environments end up not being able to run with the 256M of memory available to qemu's versatilepb environment. Jun 10 02:09:25 I don't care about performance Jun 10 02:09:35 I just want to physically setup the system Jun 10 02:09:44 in a qemu Jun 10 02:10:08 And there was recently a debate about how to generate rootfs images for production in this channel, in which the consensus seemed to be "get some hardware". If you're playing, rootstrap or qemu-debootstrap or multistrap or debootstrap in an emulator can all work, but nobody seemed confident about standing behind any of them. Jun 10 02:10:50 Ah. If you really want to use qemu, then you have to use qemu to do that :) In which case, the instructions on the page already mentioned are supposed to work. Jun 10 02:11:08 I have to run for now - and I am aware they are "suppose" to work. Jun 10 02:11:17 If they did in fact work for me though, I wouldn't be here ;) Jun 10 02:11:28 I guess I'll be back later when hopefully someone can help :-/ Jun 10 02:11:37 Thanks for your input persia Jun 10 02:12:06 Sorry I'm not able to help you reach your goal. Jun 10 02:12:52 blank screen might be because the console is not going to the right place Jun 10 02:15:04 I thought console went to qemu console by default. Jun 10 02:16:12 (which may not be the default shown in the SDL window) Jun 10 02:18:54 the kernel might be using the serial port though Jun 10 02:19:21 I seem to remember having to fool with redirecting the serial port to stdout to see the console messages Jun 10 02:24:52 either with -nographic or -serial stdio qemu options Jun 10 02:29:02 building a toolchain in the system emulator took forever so I switched to the chroot environment **** ENDING LOGGING AT Fri Jun 10 02:59:57 2011