**** BEGIN LOGGING AT Mon Feb 06 02:59:58 2012 Feb 06 04:39:02 I am trying to build a new kerne Feb 06 04:39:08 kernel* Feb 06 04:40:21 I am surprised that when I run debian/rules editconfigs it picksup armel as well when I am running armhf Feb 06 04:50:15 All I know is I get the upstream kernel and run "make deb-pkg" Feb 06 05:03:28 twb: I am not sure if upstream has all the omap changes Feb 06 05:03:32 that ubuntu has Feb 06 05:03:49 Shrug Feb 06 05:59:58 Is anyone using any method to cool the pandaboard Feb 06 06:01:46 I am building the kernel natively and dmesg is full of thermal messages Feb 06 06:01:48 http://pastebin.com/JxrEkATz Feb 06 06:07:00 krosswindz: have you got yours inside a case at all? Feb 06 06:07:14 Does the panda's power brick have active cooling? Feb 06 06:07:20 (I don't have one) Feb 06 06:29:53 twb: i dont have the panda in a case Feb 06 06:30:04 twb: the power brick doesnt have active cooling Feb 06 06:30:44 What's the ambient temperature? Feb 06 06:31:05 If it's like 40\degC, that wouldn't be so suprising Feb 06 06:32:38 room temp is much lower Feb 06 06:32:53 I guess I might have to move the board from its current location so that it gets more free air Feb 06 06:33:10 room temp is probably in the 20s Feb 06 06:33:56 Also worth checking if throttling governor &c are on Feb 06 06:35:32 i think throttling governor is on which is what is switching the frequency Feb 06 06:35:52 the omap_thermal_step_freq_down Feb 06 06:39:23 /sys/devices/system/cpu/**/cpufreq/scaling_governor Feb 06 06:40:06 ondemand Feb 06 06:40:15 That's OK then Feb 06 06:40:19 I dunno what else to check Feb 06 06:46:05 twb: not a problem Feb 06 06:46:08 I was just wondering Feb 06 06:47:40 FWIW native compiles on TF101 don't overheat that Feb 06 06:56:08 hmm Feb 06 06:56:10 ok Feb 06 08:28:40 mmm, nice. Feb 06 08:28:56 i'm trying to find ARM benchmarks... Feb 06 08:29:09 a passmark would work pretty well Feb 06 08:29:27 or some other general all-around benchmark that has a baseline on a VAX Feb 06 08:29:43 or, barring that, some way to get onto an ARM box to run a benchmark Feb 06 08:29:58 hosting service that'll take cash, or ARM enthusiast that'll take beer Feb 06 08:31:15 doug: dmips is easy enough Feb 06 08:31:51 http://homepages.cwi.nl/~steven/dry.c Feb 06 08:32:18 Other (non-synthetic) benchmarks usually involve compiling GCC &c, but they cost money Feb 06 08:33:46 The third run in "sh dry.c" output for me, on a dual-core A9, is Microseconds for one run through Dhrystone: 0.4; Dhrystones per Second: 2602811 Feb 06 08:36:05 With -O3 I get down to 0.3us / 3003003 DMIPS Feb 06 08:36:28 Er, 3003 DMIPS, since M is million Feb 06 08:37:11 infinity, can you post that "magic chroot" code again? Feb 06 08:37:35 nvm Feb 06 08:37:48 scientes: As in, copy /usr/bin/qemu-arm-static to chroot/usr/bin? Feb 06 08:37:57 yes, i remember Feb 06 08:38:01 that is so slick Feb 06 08:38:18 Note that prior to 1.x qemu-arm-static had some bugs Feb 06 08:38:36 Most obviously triggered when using hard float Feb 06 08:38:40 thing with --no-install-recommends, is that emdebian grip actually changes the apt default to that, so i got lazy Feb 06 08:38:48 im using 11.10 Feb 06 08:39:08 fortunately it's static so you can simply build qemu-arm-static and copy the binary wherever you want Feb 06 08:39:17 twb, indeed Feb 06 08:39:57 or copy the binary/install the .deb anywhere Feb 06 08:41:19 The .deb is marked as belonging the the host arch, though, so you can't sensibly install it inside the chroot Feb 06 08:41:37 you only need it inside the chroot, you understand, the only thing the host part cares about is the binfmt-support hooks Feb 06 08:44:18 well, ar x and tar works file too Feb 06 08:44:31 how do you extract only one file from a tar on the command line? Feb 06 08:45:06 (however many probably dont know you can do that with .deb files) Feb 06 08:45:27 scientes: if you're not going to dpkg -i it there is no point carrying around more than the single qemu-static-arm binary Feb 06 08:45:37 scientes: dpkg -x, rather. Feb 06 08:45:50 *That is, dpkg -x rather than ar+tar. Feb 06 08:45:57 of course, but im talking about the something you generally are doing one-off Feb 06 08:46:25 twb, ahh that would probably be cleaner than using ar+tar, thx Feb 06 08:46:38 You can't extract a single file unless using ar+tar AFAIK, in which case you just do it the tar way Feb 06 09:00:13 hi Feb 06 09:00:26 i have a beagleboard and I want to install the sgx drivers Feb 06 09:00:32 the software renderer is slow Feb 06 09:03:11 i'd like to provide a SD image with preinstalled omapfb drivers, is that possible? is there a toolchain to build custom ubuntu installation images? Feb 06 09:03:39 carli2: what do you mean exactly? Feb 06 09:05:29 carli2: is this what you are looking after http://omappedia.org/wiki/Add_Packages_To_Ubuntu_Preinstalled_Images Feb 06 09:07:27 ndec: ah thanks :) Feb 06 09:07:41 but isn't chroot architecture specific? Feb 06 09:07:52 i should run chroot on a arm system Feb 06 09:07:59 nope. Feb 06 09:08:04 you can cross chroot Feb 06 09:08:05 or does it include a qemu-call? Feb 06 09:08:08 with qemu Feb 06 09:08:45 ok... it seems that the wiki only explains the native chroot. Feb 06 09:08:56 but it would work with cross chroot too. Feb 06 09:10:05 how can I delete all users created at installation time and turn the system back into OEM status? Feb 06 09:10:28 you basically need to install qemu-user-static, and then sudo cp /usr/bin/qemu-arm-static //usr/bin/, then you can cross chroot from x86 into the ubuntu arm rootfs Feb 06 09:11:17 i don't think you can. or at least i don't know. perhaps infinity would know that... but I still think it's best to start with the preinstalled image before doing the installation. Feb 06 09:11:22 as per the wiki i gave Feb 06 09:15:10 configure a preinstalled ext2/3 system image "jasper" Feb 06 09:17:35 how much faster would the hardfloat image be? Feb 06 09:23:30 and no libreoffice Feb 06 09:23:57 but it's installable :) Feb 06 11:01:11 after installing the sgx drivers, the monitor stays black Feb 06 18:31:01 hi Feb 06 18:31:27 I can't login to Ubuntu on my Pandoboard Feb 06 18:31:31 Login timed out after 60 seconds. Feb 06 18:31:31 Feb 06 18:31:31 Ubuntu 11.10 localhost.localdomain ttyO2 Feb 06 18:31:31 Feb 06 18:31:31 localhost.localdomain login: root Feb 06 18:31:32 Password: Feb 06 18:31:38 kindly help!! Feb 06 18:33:11 micadeyeye_: Did you run through the oem-config installer? It sets up a default account and hostname. Feb 06 18:33:22 Which image are you using? Feb 06 18:35:44 GrueMaster, ubuntu 11.04 from here http://www.omappedia.org/wiki/OMAP_Ubuntu_Core Feb 06 18:36:09 I didn't run the oem-config. Feb 06 18:37:16 You also didn't read the instructions there that point out that you need to either remove or set root's password. Feb 06 18:37:53 i didn't see that. Feb 06 18:38:40 The page recommends just making root passwordless. I'd recommend setting a password or adding a user, but whatever works better for your use-case. Feb 06 18:38:59 This is not an "Ubuntu supported" installation method. My best suggestion is to either use one of the preinstalled images from cdimage.ubuntu.com or follow those instructions carefully and ask the author for support. Feb 06 18:39:40 what's the quickest(/best?) way for me to get an account on an ARM box? Feb 06 18:40:08 Buy an arm box and install Ubuntu on it? Feb 06 18:40:33 I thought it wasn't using a password. I followed the instructions and did this "edit the file /etc/shadow and remove the '*' character in between the semi-colons. You" Feb 06 18:40:44 I am running ubuntu on pandaboard Feb 06 18:41:18 hm, what's the cheapest arm box that i can buy (today)? Feb 06 18:42:05 I got the rootfs from - http://cdimage.ubuntu.com/ubuntu-core/releases/11.10/release/ubuntu-core-11.10-core-armel.tar.gz Feb 06 18:42:42 doug: It really depends on your needs. I think the beaglebone (http://beagleboard.org) is probably the cheapest that we can support in Ubuntu, but there are better depending on hw needs. Feb 06 18:42:43 micadeyeye_: GrueMaster is right, we don't support core as a bootable/installable OS. It's meant for people to build on top of. Feb 06 18:43:09 micadeyeye_: If that's not really working out for you, you might want to try a more full-featured installer, like ubuntu-desktop or ubuntu-server. Feb 06 18:43:46 infinity, am not sure they would run on my omap4 device. Feb 06 18:44:04 my need is certainly to find something dirt cheap Feb 06 18:44:10 doug: Some board from Arduino family Feb 06 18:44:19 ubuntu on arduino? Feb 06 18:44:21 micadeyeye_: You're using a non-standard kernel? Feb 06 18:44:29 doug: no ubuntu, but dirst cheap Feb 06 18:44:34 I think so. Feb 06 18:45:01 yeah, need something that'll run a linux, something with an MMU Feb 06 18:45:06 wget http://ports.ubuntu.com/ubuntu-ports/dists/precise/main/installer-armel/current/images/omap4/netboot/MLO Feb 06 18:45:06 wget http://ports.ubuntu.com/ubuntu-ports/dists/precise/main/installer-armel/current/images/omap4/netboot/u-boot.bin Feb 06 18:45:07 wget http://ports.ubuntu.com/ubuntu-ports/dists/precise/main/installer-armel/current Feb 06 18:45:08 Well, the "cheapest" is certainly less than that (raspberry PI comes to mind), but we only support hardware that has ARMv7 technology. Feb 06 18:45:18 pi isn't armv7? Feb 06 18:45:20 doug, try Pandaboard Feb 06 18:45:26 micadeyeye_: Right, well, if you need to build something from scratch, just chroot into your ubuntu-core system on another box and "passwrd root" before you boot it. Feb 06 18:45:33 they now have pandaboard es Feb 06 18:45:42 micadeyeye_: Err, that's the standard Ubuntu omap4 kernel. Feb 06 18:45:44 doug: If you go with something less, you can always run debian. And no, Raspberry Pi is Armv6. Feb 06 18:45:56 I don't know that would be the cheapest, but beaglebone at 89$ is probably closest Feb 06 18:45:58 micadeyeye_: If you use one of our full-featured images, it'll be exactly the same kernel. Feb 06 18:45:58 but it's requesting a pwd. Feb 06 18:46:04 cheapest to run ubuntu, that is Feb 06 18:46:08 micadeyeye_: What platform are you running on? Feb 06 18:46:17 yeah, beagleboard is what, $89? 3 times the advertised price of the (low-end) pi Feb 06 18:46:33 pandaboard (omap4) Feb 06 18:46:42 micadeyeye_: Please, just use an image with an installer. Feb 06 18:46:44 Ubuntu 11.04 Feb 06 18:46:47 well lets see what the pi price will turn out tu be in reality Feb 06 18:46:52 micadeyeye_: Then use one of our omap4 preinstalled images. Feb 06 18:47:12 micadeyeye_: http://cdimage.ubuntu.com/releases/11.10/release/ Feb 06 18:47:32 micadeyeye_: Either ubuntu-desktop omap4 or ubuntu-server omap4, depending on if you want pretty things or not. :P Feb 06 18:47:51 okay Feb 06 18:51:52 micadeyeye_: If you want greater speed and performance, you can also try our latest alpha 2 of 12.04. http://cdimage.ubuntu.com/releases/precise/alpha-2/ Feb 06 18:52:52 According to Phoronix, the performance improvements are pretty good. http://www.phoronix.com/scan.php?page=article&item=ubuntu_1204_omap4460&num=1 (note that we did not run or commission these tests - they just look awesome). Feb 06 19:11:08 any progress on ALSA lib main.c:260:(execute_sequence) unable to open ctl device 'hw:Panda' Feb 06 19:11:12 ? Feb 06 19:24:27 pbuckley: Not yet. The kernel team just got the bug late last week, and I am sure they are looking at it. Feb 06 19:25:07 k.. :) thanks for the follow up.. what was the bug id again? I forgot to bookmark Feb 06 19:25:30 bug 925069 Feb 06 19:25:31 Launchpad bug 925069 in linux-ti-omap4 "No analog audio on omap4 panda" [Undecided,Confirmed] https://launchpad.net/bugs/925069 Feb 06 19:25:50 thank you Feb 06 19:26:50 ooh Feb 06 19:26:55 looks like they might already have a patch Feb 06 19:27:17 any gotcha's about compiling ubuntu kernels? Feb 06 19:29:44 Yea, I don't do them. :P Seriously, I don't know other than they take a while on panda. Feb 06 19:39:41 well Feb 06 19:39:52 at least they compile the kernels with /proc/config.gz support Feb 06 19:39:56 that makes it alot easier in theory Feb 06 19:51:02 lag: greetings Feb 06 19:51:28 prpplague: Howdy Feb 06 19:51:47 prpplague: Are you well? Feb 06 19:52:02 as well as can be expected, hehe Feb 06 19:52:11 Struggling on by Feb 06 19:52:13 :) Feb 06 19:52:26 You're not at Connect are you? Feb 06 19:53:18 lag: not this time, too many irons in the fire Feb 06 19:53:30 prpplague: Right, no probs Feb 06 19:53:45 prpplague: I wanted you to bring a Flyswatter so we could test it on Snowball Feb 06 19:53:54 prpplague: I have some other toys to play with too Feb 06 19:53:59 lag: ahh, i'll be at ELC on thursday of next week Feb 06 19:54:00 kernel source tree has changed pretty drastically since last time i looked Feb 06 19:54:04 where is my make oldconfig :( Feb 06 19:54:19 prpplague: Unfortunately I'm giving this one a miss Feb 06 19:54:36 lag: bummer Feb 06 19:54:53 prpplague: I know dude - I usually like to attend Feb 06 19:55:07 lag: i've already allocated my budget of freebie flyswatter2's for this month, let me see what i can do to get one to you for testing Feb 06 19:55:11 hm, what's the cheapest i can get an ubuntu-support ARM box for? Feb 06 19:57:08 35 bucks? Feb 06 19:58:09 assuming small quality Feb 06 19:58:16 bulk orders im sure you could find cheaper even Feb 06 19:58:25 actually there is a 25 dollar version too Feb 06 19:58:27 ;) Feb 06 19:59:08 (raspberry pi) which i assume will have ubuntu on it at somepoint Feb 06 19:59:14 though i doubt it will ever run x Feb 06 19:59:20 pbuckley: The Pi can't run Ubuntu. Feb 06 19:59:33 pbuckley: It's ARMv6, we only support v7. Feb 06 19:59:34 prpplague: I didn't mean to keep - would just be good to get some shots up of it working great with snowball :) Feb 06 19:59:39 really? Feb 06 19:59:44 damn so much for that idea Feb 06 19:59:44 pbuckley: Really. Feb 06 19:59:51 pandaboards then Feb 06 19:59:54 but those are like 180 Feb 06 19:59:59 doug: The cheapest Ubuntu-supported board you can get is probably the beaglebone. Feb 06 20:00:00 lag: no worries, we want to make sure it works with as many items as possible Feb 06 20:00:09 oh right Feb 06 20:00:13 forgot about the beagle's Feb 06 20:00:13 doug: But the Panda or mx53 quickstart are better options, IMO. Feb 06 20:00:38 oh and there is the pandaboard and pandaboard es Feb 06 20:00:48 so many flavors Feb 06 20:00:54 hm. Feb 06 20:01:00 beaglebone is $89, right? Feb 06 20:01:08 i personally use the pandaboard es as a desktop Feb 06 20:01:13 is there any particular reason precise kernel is missing the pandaboard OTG port kernel module Feb 06 20:01:51 i'm particularly interested in running stuff as a server Feb 06 20:01:52 krosswindz: File a bug against linux-omap4 please. Feb 06 20:01:56 headless, etc. Feb 06 20:02:00 GrueMaster: sure Feb 06 20:02:29 doug: Depending on the type of server, you would probably be better off with the mx53 as it has native SATA. Feb 06 20:02:47 its a freescale chip no? Feb 06 20:03:02 Having said that, I can also say that the Panda does really well on server loads too. Feb 06 20:03:09 yes, Frescale. Feb 06 20:03:37 yeh.. the pandaboard sucks at disk io :( Feb 06 20:03:41 Here is a list of tests I ran last cycle for arm server. https://wiki.ubuntu.com/ARM/QA/Server Feb 06 20:03:42 even usb drives are pretty bad Feb 06 20:03:53 though 12.04 seems to have improved sd performance at least Feb 06 20:04:22 Note that I also ran Raid, iSCSI (client & host), and CEPH ClusterFS, all on pandas. Feb 06 20:04:38 oh nice Feb 06 20:04:44 i was thinking about iscsi Feb 06 20:04:46 how did that go? Feb 06 20:05:32 Well, it initially had issues (same as on x86 when I ran it), but those have been fixed. I haven't tested it again this cycle, but it is on the todo list. Feb 06 20:05:35 Requires a bit of manual configuration. iSCSI-root fails to boot. See LP:838809 Feb 06 20:05:41 ah ok Feb 06 20:06:32 It will still need either an SD for u-boot or optionally it can boot from a host pc through the OTG port. Feb 06 20:06:44 (documentation on that feature coming soon). Feb 06 20:07:35 hm, where's a good place to buy mx53 quickstarts from? Feb 06 20:07:48 Digikey. Feb 06 20:07:48 does digikey sell them? Feb 06 20:07:51 if so i would go there Feb 06 20:08:16 Digikey sells everything anyone could ever want. If you want it and they don't have it, you're wrong about wanting it. Feb 06 20:08:23 * infinity nods. Feb 06 20:08:23 hahaha Feb 06 20:08:28 Yes, I bought two through digikey (and one direct from Freescale). Feb 06 20:08:52 how does the mx53 compare to the panda as far as x performance goes? Feb 06 20:09:04 that sata port is tempting Feb 06 20:09:07 GrueMaster: against linu-omap4 or linux-ti-omap4? Feb 06 20:09:17 Note that the newer START-R model doesn't run Oneiric, and until we get a new kernel in the pool, it won't run Precise. Feb 06 20:09:26 oh nm Feb 06 20:09:45 krosswindz: linux-omap4. Send me the bug number and I can make sure it gets triaged. Feb 06 20:09:52 ok Feb 06 20:09:54 you can take precise from my cold dead fingers Feb 06 20:10:49 heh. Feb 06 20:20:53 pbuckley: X performance on the quickstart is good. Its only real failing is the slower CPU. Feb 06 20:21:26 GrueMaster: Does the new 3.1.x kernel we just shoved in still not boot the START-R? Feb 06 20:21:47 When did that get uploaded? Feb 06 20:22:02 I processed it through NEW yesterday. Feb 06 20:22:08 And updated the meta yesterday. Feb 06 20:22:19 Oh, well I will check it out then. Feb 06 20:22:32 (I wasn't working this weekend). Feb 06 20:22:53 Yeahp, fair enough. Feb 06 20:23:16 Oh, speaking of not working on weekends, can you give me a reminder of those flash-kernel/preseed bug numbers? Feb 06 20:23:18 GrueMaster: bug#: 927860 Feb 06 20:23:26 I'll get to them when I'm bored here at Connect. Feb 06 20:23:59 bug 927860 Feb 06 20:24:00 Launchpad bug 927860 in linux-meta-ti-omap4 "Missing musb-hdrc module required by Pandaboard OTG port" [Undecided,New] https://launchpad.net/bugs/927860 Feb 06 20:24:34 GrueMaster: ubuntu-bug files it under linux-meta-ti-omap4 Feb 06 20:24:38 GrueMaster: not sure why Feb 06 20:24:53 Don't worry, I can fix it. Feb 06 20:25:01 GrueMaster: thanks Feb 06 20:26:22 GrueMaster: it only affects Precise, oneiric 3.0 kernel has it compiled into the kernel Feb 06 20:26:39 Interesting. Feb 06 20:34:21 Ok, I tweaked the bug report and assigned it to our kernel engineer. We should see an update later this week. Can you enable it and test it in the mean time? If it works, post your results to the bug. Feb 06 20:34:55 GrueMaster: Yes I am trying to build the kernel by myself Feb 06 20:35:14 infinity: I'll try to get to the f-k-i fixes this afternoon (after lunch). Feb 06 20:35:19 GrueMaster: The first time I built it and was doing testing I didnt realize it was missing. Feb 06 20:35:40 Ah. Feb 06 20:35:46 GrueMaster: I will keep you updated on the progress, I am at work now :( the board is at home. Feb 06 20:35:50 brb - need to feed. Feb 06 20:36:13 oh lord, who let jkridner in here? Feb 06 20:36:23 :) Feb 06 20:36:35 * prpplague shakes his head at the low requirements for entry into #ubuntu-arm Feb 06 20:36:49 lol Feb 06 20:37:24 GrueMaster: I had it mostly coded in my head anyway, just forgot the bug numbers and lost them in scrollback. Feb 06 20:41:50 hm, i found one ms53 board retailing for $149... Feb 06 20:42:12 from newark.com Feb 06 20:43:09 never used em Feb 06 20:43:21 digikey has always done right by me Feb 06 20:45:13 hm, digikey's MCIMX53-START-R-ND is also showing up as $149 Feb 06 20:45:33 there you have it Feb 06 20:46:55 now if i could just find a good source for pricing on the cpu itself Feb 06 20:47:11 digikey says "call" Feb 06 20:49:22 it depends on the quanity Feb 06 20:49:43 (usually) Feb 06 20:51:58 well sure Feb 06 20:56:24 if i apt-get install linux-source-3.2.0 will that give me the correct branch? or should i pull it out of a git repo somewhere? Feb 06 21:02:35 pbuckley: apt-get source linux-ti-omap4 Feb 06 21:03:07 ah thanks Feb 06 21:03:19 Which will get the meta. ;) Feb 06 21:03:26 apt-get --only-source source linux-ti-omap4 Feb 06 21:03:42 should i also do a apt-get build-dep linux-ti-omap4? Feb 06 21:04:54 NOTICE: 'linux-ti-omap4' packaging is maintained in the 'Git' version control system at: Feb 06 21:04:58 neat Feb 06 21:05:00 since when did apt support git? Feb 06 21:05:19 or am i not reading that right? Feb 06 21:06:05 (because i basically wrote my own package format to get around deb's limitation of only one installed version of a package at a time) Feb 06 21:06:18 and i assume if git is in apt they have worked around that Feb 06 21:06:59 No, it is just indicating that our kernel trees are in git as opposed to bzr. Feb 06 21:07:13 ah ok Feb 06 21:07:28 got all excited Feb 06 21:07:29 ;) Feb 06 21:09:07 also dont know if this is a packaging legacy defect Feb 06 21:09:14 but it points me to http://kernel.ubuntu.com/git-repos/ubuntu/ubuntu-maverick.git Feb 06 21:09:25 is that the active git repo or is it ubuntu-precise.git? Feb 06 21:10:02 ubuntu-precise.git Feb 06 21:10:06 k Feb 06 21:15:10 thank you again :) Hopefully those two patches on that bug report help with the sound issue im having.. going a bit mad without audio Feb 06 21:18:48 So I've been trying usbboot. It works, but seems to find the boot script on the sd card and uses that. Is there any way to override the boot script? Feb 06 21:19:27 yea, remove the sd. Feb 06 21:19:35 Without removing the SD :) Feb 06 21:19:49 picky picky. :P Feb 06 21:20:00 The reason is that I want to boot off the SD except when automation wants to netboot Feb 06 21:20:18 Right now I have to assume that the machine isn't bricked and can be shelled into to change the SD to netboot Feb 06 21:20:45 I understand. I want to use it to recover from a botched netinstall. Feb 06 21:20:55 Yeah, exactly :) Feb 06 21:21:08 Right now the only way I can think of is to hack u-boot to never read a script Feb 06 21:21:27 (and supply the hacked u-boot over usb) Feb 06 21:21:32 I am hoping that there's a better way. Feb 06 21:21:37 I don't know how the new usbboot works, but the omap4boot utility used abootimg to build an image. This included a boot script or kernel cmdline. Feb 06 21:25:27 This seems to use mkimage to convert u-boot.bin into a u-boot.img (with a load address of 0x80E80000) Feb 06 21:25:42 I take it this is a different image format to abootimg? Feb 06 21:26:26 Yes. mkimage only creates a checksum wrapper. Feb 06 21:42:19 so if i wanted to submit patches to http://kernel.ubuntu.com/git-repos/ubuntu/ubuntu-precise.git Feb 06 21:42:26 is there a doc i can read that explains the process? Feb 06 21:46:16 pbuckley: https://wiki.ubuntu.com/Kernel/Dev/KernelPatches Feb 06 21:46:46 That's probably the easiest. Feb 06 21:49:03 brilliant thank you :) Feb 06 22:12:17 GrueMaster: I edited the kernel config and started a fresh build using the current sources. I will let you know how it goes Feb 06 22:12:44 GrueMaster: I will update the bug once I reinstall kernel Feb 06 22:12:47 Ok. Feb 06 22:24:35 so compiling on the panda.. make -j2,-j3,-j4? Feb 06 22:26:07 Depends on your storage medium. If SD, I would not multi-thread. Otherwise -j2 or -j3. I think the rule of thumb (thumb2?) is # cores +1. Feb 06 22:26:26 k Feb 06 22:26:29 ty Feb 06 22:28:13 also make oldconfig generates warning: (USB_WUSB) selects UWB which has unmet direct dependencies (EXPERIMENTAL && PCI) Feb 06 22:28:57 That is a question for #ubuntu-kernel. Feb 06 22:29:01 k Feb 06 22:33:44 sorry, if ask: but is thumb(2) something like mmx/sse/...? Feb 06 22:33:50 not at all Feb 06 22:33:51 *if i ask Feb 06 22:34:03 on ARM, instructions have fixed size Feb 06 22:34:08 ie 32 bits Feb 06 22:34:32 they found that sometimes you don't need full-featured instructions, and you could greatly reduce size with 16 bits instructions Feb 06 22:34:36 that's thumb Feb 06 22:35:05 then, they decided they could make the best of both worlds with thumb2, when using a 32bits instruction can be more efficient, they use a 32bit, else a 16bits one Feb 06 22:35:16 shortly said: you get smaller binaries that load faster into ram Feb 06 22:35:44 (and sometimes also run faster) Feb 06 22:35:59 (because it's faster to load from ram to cpu) Feb 06 22:37:18 infinity: New mx53 kernel still appears not to enable usb on the START-R rev of the Quickstart. Feb 06 22:37:33 hm... thanks for the explanation =) Feb 06 22:38:18 GrueMaster: Irksome. I'll hunt down the landing team this week and see if we can get to the bottom of this. Feb 06 22:38:51 k. Feb 06 22:39:08 I'll see if I can generate some kernel output. Feb 06 22:39:16 That could be helpful. Feb 06 22:41:57 infinity: http://paste.ubuntu.com/832005/ Feb 06 22:44:47 GrueMaster: Hrm. I don't see an obvious kernel failure there at all. Feb 06 22:47:05 infinity: Starting at line 305: usb 2-1: device descriptor read/64, error -71 Feb 06 22:48:41 Does it boot from SD, then? Feb 06 22:48:55 Might make it easier to debug the USB issues if one can hotplug a non-root USB device. Feb 06 22:49:02 I have a few more tests to run, but the only difference between my quickstart (qs1) and quickstart-r (qs2) setup is that qs1 is on port 2 of my kvm, and qs2 is on port 4. Feb 06 22:49:36 It has always booted from sd. The problem is that it doesn't see keyboard/mouse. Feb 06 22:50:07 Oh. Key. I use mine headless, would probably not even notice. ;) Feb 06 22:50:33 Well, considering we only build desktop images for the board... Feb 06 22:51:30 Yeah, I don't test images on my mx53, it's my local mirror. Feb 06 22:51:40 It was also my first armhf system, before I had images. Feb 06 22:51:51 (and I have a slight shortage of sata drives atm). Feb 06 22:51:54 Speaking of, we have armhf+mx5 images now. Feb 06 22:52:11 cool. I'll test tomorrow. Feb 06 22:58:52 hi, someone up? Feb 06 22:59:07 i would like to know if someone had a i.mx515 image working Feb 06 22:59:13 infinity: Just to salt the wound a little, the dev image that came with it works ootb. Kernel 2.6.35.3-1129-g691c08a Feb 06 22:59:35 methril: mx51? Like babbage3? Feb 06 22:59:47 GrueMaster: I've adjusted u-boot-linaro-omap4-panda-splusb's default boot script to do pxe only. Is there any situation where this shouldn't be the case, ie. should I propose this as a permanent fix? Feb 06 23:00:13 GrueMaster, yes, and Sharp PC-Z1 Feb 06 23:00:34 i would like to update it Feb 06 23:00:45 it's a powerful processor Feb 06 23:00:52 methril: Ah, yes. That device. persia used to bring his to UDS all the time. Feb 06 23:01:45 GrueMaster, so nobody is working on it, isn't it? Feb 06 23:01:55 Image wise we don't support that platform. Package wise, it should handle everything current. Feb 06 23:02:45 any guid to start porting it "image wise"? Feb 06 23:03:05 s/guid/guide Feb 06 23:04:38 Not really. iirc, that had Jaunty, with was armv5. I don't know if you can get a kernel for it or not. Feb 06 23:05:31 GrueMaster, there is a 2.6.31 port, and a 2.6.35 WIP Feb 06 23:07:24 If the 2.6.31 port works and is at least compiled for armv7, try it. We can go from there. Feb 06 23:08:02 ok, thank you GrueMaster Feb 06 23:11:09 i will try to work on it, and come back when something is working Feb 06 23:11:55 If you can boot with an armv7 kernel, the next step would be to see if you can chroot into an ubuntu-core image and do stuff. Feb 06 23:12:46 ok, i need to buy a bigger sd card Feb 06 23:13:52 The core images are fairly small. ~35M compressed. Feb 06 23:14:08 Note that these are NOT bootable images. Feb 06 23:14:09 uhm... Feb 06 23:14:22 i've a bootable sd card Feb 06 23:14:32 i only need to boot it with a newer kernel image Feb 06 23:14:44 Ok. Feb 06 23:15:00 Older binaries "should" still work. Feb 06 23:15:22 the older is what i'm going to try Feb 06 23:15:31 to get a new version Feb 06 23:15:40 i need it for traveling Feb 06 23:16:19 Understand. From what I've seen of it, it seems fairly capable. Feb 06 23:16:49 i think so Feb 06 23:17:00 let's see what we could get from this old device :) Feb 06 23:26:17 GrueMaster: I filed bug 927956 Feb 06 23:26:18 Launchpad bug 927956 in u-boot-linaro "USB SPL boot should be able to override local SD card on panda" [Undecided,New] https://launchpad.net/bugs/927956 Feb 06 23:26:21 (with patch) Feb 06 23:26:33 ok Feb 06 23:34:34 you werent joking about kernel compiles taking ahwile on the panda Feb 06 23:36:02 i might have to break down and spin up an ec2 instance and start doing my compiles there.. i assume there is a cross compile toolchain somewhere? Feb 06 23:42:44 pbuckley: There is, but I couldn't tell you how to do a cross compile. Try doing a google search on the ubuntu wiki. Feb 07 00:42:12 pbuckley: export $(dpkg-architecture -aarmel); export CROSS_COMPILE=arm-linux-gnueabi-; fakeroot debian/rules clean binary-omap4 Feb 07 00:42:27 pbuckley: you'll need gcc-arm-linux-gnueabihf installed Feb 07 00:42:57 err, gcc-arm-linux-gnueabi Feb 07 00:43:08 for armhf, s/armel/armhf/ and it's gnueabihf Feb 07 00:44:52 thank you Feb 07 00:45:18 im stil waiting for my first compile i kicked off about 4 hours ago Feb 07 00:45:32 so cross compiling is really the only way Feb 07 00:45:39 i can do this in any sort of real time frame Feb 07 01:29:55 i wonder if distcc is still around Feb 07 01:31:03 sweet it is.. anyone have success with it on arm? Feb 07 01:31:51 it works Feb 07 01:32:18 ty Feb 07 01:32:19 it's not very useful though, if you have 40 arm boxes to hook up with it you're probably also in posession of a rather powerful PC too :) Feb 07 01:32:26 well Feb 07 01:32:46 true.. but even in ec2 the kernel is taking forever.. ive gotten spoiled with near instant builds Feb 07 01:32:58 going to spin up a couple more instances Feb 07 01:33:02 and distcc this shiznit Feb 07 01:33:49 Neko: some stuff can't be cross-compiled, though Feb 07 01:34:10 Although the same stuff probably also has trouble with distcc... Feb 07 01:34:18 and that;s where you break out qemu to emulate arm enough to get the bits that can't be cross-compiled done Feb 07 01:34:41 I guess. Feb 07 01:34:49 Not all arches get so much love in qemu :-) Feb 07 01:34:58 suse manage it in their build service just fine, ubuntu are getting this stuff up and running. compiling natively on arm boxes is just looking for a long wait no matter how many you have. Feb 07 01:35:40 when you hit quad or eight core 1.8GHz monsters with 4GB RAM and real hard disks, yeah, it'll make sense again, but they don't exist yet. Not quite yet :) Feb 07 01:35:43 AFAIK Debian still relies heavily on native builds Feb 07 01:36:03 and it'd still be faster on a PC. Nobody does a "native" Android build Feb 07 01:36:10 * pbuckley drools about the thought of an eightcore 4gb ram monster Feb 07 01:36:14 yeah we own the armhf build farm.. Feb 07 01:36:18 Well, android are crack-smoking java monkeys, what do you expect Feb 07 01:36:35 and it works.... but there's a better way Feb 07 01:36:49 whats the better way? Feb 07 01:36:55 cross-compile :] Feb 07 01:37:01 heh Feb 07 01:37:25 use your 8-core hyperthreaded xeon with 32gb ram and a 4TB raid array and build all of android in 25 minutes. Feb 07 01:37:34 I expect it'll become trendier in Debian once multilib stuff is done Feb 07 01:38:12 Neko: that still takes 25 minutes? You can build openwrt on an *atom* in that time Feb 07 01:38:37 openwrt is a little smaller though Feb 07 01:39:03 building a 6MB ROM for a router is a bit different to something that comes out about 300MB worth of apps and java crap Feb 07 01:39:37 spitting a kernel out takes a couple minutes for a full distclean build.. you'd never do that on ARM even if you had 50 boxes doing it Feb 07 01:40:16 im at about 5 hours for building 3.2.0 on my pandaboard Feb 07 01:40:48 it takes 80 minutes on an Efika MX :) Feb 07 01:40:55 Those extra 294MB are just bloat ;-) Feb 07 01:41:11 twb: thats the java ;) Feb 07 01:41:11 have 50 of them and it might take half an hour, maybe 10 minutes if you're lucky Feb 07 01:41:23 twb, I did say java crap :D Feb 07 01:41:27 efika mx you say Feb 07 01:41:29 * pbuckley googles Feb 07 01:41:53 Now now, just because dalvik walks like a duck and quacks like a duck, doesn't mean Oracle can sue Google ;-) Feb 07 01:42:06 pbuckley: The Efika's much slower than your Panda, I wouldn't bother with the googling. Feb 07 01:42:17 pbuckley: If it has taken 5 hours on your panda, you're doing something wrong. https://launchpad.net/ubuntu/+source/linux-ti-omap4/3.2.0-1405.7/+build/3121438 Feb 07 01:42:56 And that system isn't even optimized. I think it is running natty. Feb 07 01:43:12 GrueMaster: apt-get --only-source source linux-ti-omap4 (something); cd /usr/src/linux; make oldconfig; make Feb 07 01:43:18 is basically all i did Feb 07 01:43:38 Are you running on SD or USB Sata? Feb 07 01:43:44 SD Feb 07 01:44:09 GrueMaster: he is probably building with debug on Feb 07 01:44:12 Ah. That explains it. You'll see almost 10x performance increase on USB. Feb 07 01:44:14 That ALWAYS gets me Feb 07 01:45:01 I'm regretting buying a pandaboard, not because it's slow (usb disk makes it fun to use again) but because everything just seems not to be as good as you'd think Feb 07 01:45:07 awesome CPU, awful TI peripherals Feb 07 01:45:15 except the wireless, that works really well Feb 07 01:45:17 Looking at the buildlog from the link I posted, the buildd is running Natty (Kernel version: 2.6.38-1209-omap4 #20-Ubuntu). Feb 07 01:46:03 The panda is essentially a cellphone dev platform. When thought of along those terms, it is actually quite good. Feb 07 01:46:16 it's a shame nobody's making a real omap4 desktoppy kind of system that's been optimized for usage and not just developing phones Feb 07 01:46:49 the other TI dev board is much better than the panda for that though too Feb 07 01:47:03 like.. if you had the choice you'd buy the MX53 EVK and not bother with the Quickstart Feb 07 01:47:18 if you were developing phones that is Feb 07 01:47:18 Actually, there are a lot of non-phone systems out there (kindle fire, nook tablet). And I heard of one system coming soon that will just plug into a spare hdmi port on your tv. Feb 07 01:47:36 I thought omap4 was a better platform than the dragonball Feb 07 01:47:38 not a big enough market for it Feb 07 01:47:42 tablets are just big phones at the end of the day, same development model Feb 07 01:48:03 omap4 needs a trimslice, efika mx kind of thing Feb 07 01:48:10 True. Can't wait for omap5. Could be interesting. Feb 07 01:48:41 well I'm duty bound to wait for MX6, but I can't help thinking most companies will run to nVidia before TI at the moment Feb 07 01:48:54 all the new tablets in the pipeline seem to be tegra3, if that is at all realistic Feb 07 01:49:06 I remember when they were all going to be MX51, hahahaha.. Feb 07 01:49:19 what happens at CES, stays at CES Feb 07 01:49:52 Yea, well.... Feb 07 01:50:00 a tegra3 pandaboard like board would be fun Feb 07 01:50:11 *I* remember in like 2008 when asus demoed an 9" eeepc running arm Feb 07 01:50:28 And then I had to wait four years for the tf101 Feb 07 01:50:39 so whats the trick to turn debug mode off? Feb 07 01:51:14 pbuckley: make menuconfig, find it in the kernel hacking section Feb 07 01:51:32 It basically means whether it builds with -g Feb 07 01:53:21 twb: That is on by default in all of our kernels, from what I can tell. Feb 07 01:53:47 GrueMaster: it is on standard ubuntu and debian kernels, yes Feb 07 01:54:07 but it makes compile take an order of magnitude longer, ESPECIALLY if you have slow I/O Feb 07 01:54:13 ah k Feb 07 01:54:13 thanks Feb 07 01:54:30 CONFIG_DEBUG_KERNEL=y I guess, I don't have menuconfig in front of me right now Feb 07 01:54:42 But IME you can't just edit .config you need to use the UI Feb 07 01:55:11 E: Package 'ncurses' has no installation candidate Feb 07 01:55:15 oh wait Feb 07 01:55:16 that changed to Feb 07 01:55:31 I think you can if you run make oldconfig afterwards. Feb 07 01:55:49 ncurses-base Feb 07 01:55:54 wow that package name changed again Feb 07 01:56:24 libncurses5-dev or so Feb 07 01:56:33 though im a bit surprised it doesnt get installed with apt-get build-dep Feb 07 01:56:39 ncurses-base is just terminfo files Feb 07 01:56:50 build-dep of a kernel *source* package? Feb 07 01:57:02 well know.. but of the linux-kernel package Feb 07 01:57:05 s/know/no Feb 07 01:57:16 linux-kernel is a metapackage, no build deps Feb 07 01:57:21 ah Feb 07 01:58:11 libncurses5-dev isn't a build-dep of the kernel because the kernel build doesn't call menuconfig. Feb 07 01:58:19 fair enough Feb 07 02:00:19 infinity: right, thanks Feb 07 02:00:36 flag changed.. thank you for the tip Feb 07 02:00:36 :) Feb 07 02:01:05 and on that note its time to go home Feb 07 02:12:11 If building natively, you can also do "make localmodconfig" and "make localyesconfig" to compile all used modules in, and disable all unused modules Feb 07 02:31:06 GrueMaster: the rebuilt kernel works Feb 07 02:31:11 GrueMaster: the device is detected Feb 07 02:31:20 Excellent! **** ENDING LOGGING AT Tue Feb 07 02:59:58 2012