**** BEGIN LOGGING AT Mon Apr 16 02:59:59 2012 Apr 16 06:28:55 <_av500_> mranostay: post was saved by professional commenting skills Apr 16 08:11:35 hmm, looks like that thread about stable will never end ... Apr 16 08:11:51 so, it's a stable thread Apr 16 08:12:08 ;) Apr 16 08:12:08 does it get packports from mainline threads? Apr 16 08:12:12 back* Apr 16 08:43:39 * koen flips a coin Apr 16 08:44:01 another 15 minutes to read lkml :) Apr 16 08:44:13 only stupid stuff Apr 16 08:45:11 git for runaways ;) Apr 16 08:46:33 aholler: hmm, the thread is at its end now Apr 16 08:46:41 hopefully Apr 16 08:47:32 He now has something that he can show his grandchildren... Apr 16 08:53:25 like "see I was called a f* m* by Linus" ;) Apr 16 08:54:17 I wonder what he has done before to get called such ;) Apr 16 08:58:55 aholler: I guess you haven't dealt with felipec :) Apr 16 09:51:20 Is someone has a good link that explains how to compile or activate omap_mux module into a BeagleBoard xM? Apr 16 09:56:15 make menuconfig Apr 16 09:57:03 so you need to learn how to compile a kernel Apr 16 09:57:20 sounds exiting Apr 16 10:16:48 Hi, anyone here knows how to configure the McASP in CCS? Apr 16 10:19:03 aholler: ok I will search what i can do Apr 16 10:19:38 is anyone know why omap_mux isn't integrated in the default kernel? Apr 16 10:24:25 to force people to learn how to compile a kernel ;) Apr 16 10:25:01 lol ok... any clues how to start? googling out too mucj result :( Apr 16 11:22:34 Crofton|work: I owe you a beer Apr 16 11:30:08 rofl Apr 16 11:30:26 I see Apr 16 11:36:15 Where is the latest kernel and uboot hosted for beagleBone? Is the latest kernel and uboot hosted at https://github.com/beagleboard? Apr 16 11:37:43 prasant: http://www.angstrom-distribution.org/building-angstrom Apr 16 11:41:55 koen: If I want only the kernel and uboot and not the entire distribution. Is it possiblt to get/extract the updated kernel and uboot from the same setup? Apr 16 11:51:55 prasant, read the recipes for the kernel and u-boot Apr 16 11:54:23 koen + Crofton|work : thanks, I got the tar files. Yes, I should understand bitbake, openembedded [never worked with either :-( ]. Apr 16 11:56:11 prasant: grep the tarball and patches from the recipes Apr 16 11:56:31 and pretend its 1990 again Apr 16 12:02:00 I forget if koen has a repo somewhere with all the patches applied Apr 16 12:03:14 something github something Apr 16 12:03:27 yes, but that's only for dumb and lazy people Apr 16 12:03:37 and not tested Apr 16 12:06:08 applying a billion patches by hand is annoying Apr 16 12:07:27 very Apr 16 12:09:25 so don't do it by hand Apr 16 12:09:52 there's a while buildsystem around those 'recipes' to do it for you Apr 16 12:10:10 white? Apr 16 12:10:11 what if I don't want to use that? Apr 16 12:10:18 and attentive readers will notice a script that applies the patches for you Apr 16 12:10:25 mru: then don't use it Apr 16 12:10:55 which takes us back to the nightmare of finding and applying the patches Apr 16 12:11:05 why can't you just keep a git branch with the patches applied? Apr 16 12:12:39 I do Apr 16 12:12:44 since when? Apr 16 12:12:51 since a few months Apr 16 12:13:09 where is this git repo? Apr 16 12:13:16 koen, if you just want to mess with the kernel, OE is overkill Apr 16 12:13:18 really Apr 16 12:13:18 on github/koenkooi Apr 16 12:13:42 mru: it would be in github/beagleboard if jkridner would give me access Apr 16 12:14:21 Crofton|work: 90% of the people who popup here to ask for the kernel are better of using OE Apr 16 12:14:53 that is scary Apr 16 12:15:12 it is Apr 16 12:15:49 I've also found a correlation between the nicks used here and support requests at circuitco a day later Apr 16 12:15:52 Scary Patches IV. Apr 16 12:16:41 the requests are usually like "I have this thing from 'git' now, plz send me exact steps to compile it" Apr 16 12:16:59 koen: in this case, both github kernel and OE are wrong Apr 16 12:18:25 repo? Apr 16 12:18:34 rather branch Apr 16 12:18:48 it's hidden by design :) Apr 16 12:19:02 Crofton|work: of course not repo, buildroot! Apr 16 12:19:09 https://github.com/koenkooi/linux/branches Apr 16 12:19:56 compiling the kernel is running gag Apr 16 12:20:42 every since I'm here Apr 16 12:20:49 -y Apr 16 12:21:35 especially now that the bone kernel is compiled in thumb2 mode to save ~700KiB Apr 16 12:21:54 pretty much every prebuilt toolchain chokes on that Apr 16 12:21:54 that works with every compiler? Apr 16 12:21:56 so it has hollow bones? Apr 16 12:22:11 and if it doesn't, it has a binutils that chokes on the broken TI code Apr 16 12:22:46 hm, 700, thats quite a lot Apr 16 12:22:57 and how much faster it is? :) Apr 16 12:23:06 I never got anything to run with either thumb or thumb2 ;) Apr 16 12:23:07 faster? Apr 16 12:23:34 ynezz: ext2load was slow in uboot, so I needed the kernel as small as possible Apr 16 12:23:50 ext4load is a lot faster now, but 700KiB on a total of 3MiB is still a lot Apr 16 12:23:59 yes, it is Apr 16 12:24:11 3.7 -> 3.0, not 3.0 -> 2.3 Apr 16 12:26:45 koen, with the smaller kernel, how much of an improved ext2load time did you see? before and after numbers? Apr 16 12:28:12 bradfa: http://dominion.thruhere.net/koen/cms/kernel-compression-experiments Apr 16 12:28:26 that's not the thumb2 numbers, but it does include load times Apr 16 12:28:34 Oh my! that is quite a long time Apr 16 12:28:50 bradfa: the upcoming uboot release makes ext2load 20 times faster by not loading it block-by-block Apr 16 12:29:14 koen, which "upcoming" release is that? 2012.0x? Apr 16 12:29:20 yes Apr 16 12:29:50 bradfa: http://patchwork.ozlabs.org/patch/149244/ Apr 16 12:30:05 koen, I will do a quick test, if I can, I'm doing ext2load with uboot 2011.09 from the arago repo, I'm not of the impression my 2MB kernel takes very long, but I don't have absolute numbers to back that up Apr 16 12:30:25 bradfa: and http://patchwork.ozlabs.org/patch/149244/ Apr 16 12:30:37 ehm Apr 16 12:30:55 https://plus.google.com/100242854243155306943/posts/hWysPLZY1f4 Apr 16 12:31:25 koen, that's a huge difference... I'm going to try grabserial to see what my performance looks like Apr 16 12:31:28 thanks! Apr 16 12:31:29 with the patched uboot on beaglebone ext2load is as fast as fatload and there's an ext4load to load from ext4 Apr 16 12:31:39 I'd love an ext4 load! Apr 16 12:31:55 Currently running 3 partitions (fat, ext3, ext4) and would love to only have 2 Apr 16 12:32:08 the ext4 patchset will change and fold into the regular ext2 system Apr 16 12:32:15 so 'ext2load' will work on 2,3 and 4 Apr 16 12:32:36 koen, cool! Thanks for the info! Apr 16 12:37:57 koen, I'm seeing about 650 ms to ext2load a 2.1MB kernel on my BeagleBone off an ext3 partition. That's not screaming fast, but fast enough that I'm not going to try to optimize it (yet) Apr 16 12:39:55 GrabSerial pastebin: http://pastebin.com/QbpXBbBS Apr 16 12:40:27 if that's a useful data point or not, I'm not sure Apr 16 12:51:49 When it comes to DMA on the beaglebone, am I supposed to use the stuff in arch/arm/plat-omap or the generic linux stuff? Apr 16 12:52:04 define DMA Apr 16 12:52:51 bradfa: FWIW, gcc 4.4 is the worst you can get for ARM Apr 16 12:53:40 transfering data from the gpmc to memory at a very fast rate without hogging the cpu Apr 16 12:54:20 koen, Yeah, but I don't have to build it, Debian does it for me. I'm also compiling for armv5t and running on Cortex-A8. Currently optimizing for development speed with long-ish term support rather than execution speed or bleeding edge software. Apr 16 12:54:38 gcc 4.4 is still horrible Apr 16 12:54:43 get 4.5 or 4.6 Apr 16 12:54:53 linaro has nice builds you can use Apr 16 12:55:00 mru, fair enough :) I'll get a newer one with Ubuntu 12.04 or Debian Wheezy Apr 16 12:55:12 gcc 4.4 isn't even supported any more Apr 16 12:55:41 mru, My only qualm with Linaro is that in 2 years, am I still going to get software security updates? With Debian / Ubuntu LTS, there's a good chance I will. Apr 16 12:55:42 so if you're looking for long-term support, you're better off picking a fairly recent version now Apr 16 12:55:56 I'm happy with the 4.5 I'm using, but I'm going to jump to 4.7 next month Apr 16 12:55:59 there will be no more updates for gcc 4.4 Apr 16 12:56:02 it is end of life Apr 16 12:56:12 No, but Debian will put out security updates for it Apr 16 12:56:18 lol Apr 16 12:56:20 till Squeeze ends life Apr 16 12:56:52 I'm moving to Ubuntu 12.04 or Wheezy armhf soon, don't worry :) Apr 16 12:57:04 no matter how you spin it, gcc 4.7 will receive support longer than 4.4 Apr 16 12:57:52 mru, good to know, I'll keep that in mind. For now, I have a working toolchain that I'm comfortable with and that's more important at this point to me. Apr 16 13:13:08 jsabeaudry: which generic linux stuff? Apr 16 13:15:11 jsabeaudry, for now, you use arch/arm/common/edma.c in conjunction with the DMA API Apr 16 13:15:33 at some TBD date in the future you will use the generic DMA Engine API in conjunction with the DMA API Apr 16 13:15:37 mdp, I was under the impression that there was some form of abstraction layer over different dma but now I understand this is highly device dependant Apr 16 13:16:08 there is. it's just that TI parts choose not to use it...yet Apr 16 13:16:20 ah ok Apr 16 13:16:48 are you aware of any examples of dma on the bone I failed to find any on the google Apr 16 13:20:36 I don't know of any bone specific examples. However, since the edma block is also on the davinci parts….you can see how drivers/spi/spi-davinci.c uses it directly, for example Apr 16 13:21:05 mdp, ah excellent, thank you very much Apr 16 13:21:26 my best recommendation is to steer clear of the soon to be deprecated sdma2edma hack around the old arch-specific dma calls Apr 16 13:21:45 I think it'll be easier to convert from direct edma api to dmaengine when the time comes Apr 16 13:22:26 jsabeaudry: dmaengine api has only been around for 6+ years so it's too new for TI parts to adopt :) Apr 16 13:22:58 mdp, will the edma handle scatter gather lists? Apr 16 13:23:07 yes Apr 16 13:23:56 edma makes sdma look like a child's toy too, btw…it's crazy flexible Apr 16 13:24:35 however, the implementation doesn't take advantage of everything in the h/w…that would be a retirement project for 1+ people Apr 16 13:27:00 jsabeaudry: dunno what your linux driver experience is on other platforms so I just want to remind you to read and understand the DMA API stuff so you are sure to use it properly before handing off buffers to the edma and before you touch buffers provided by the edma. Apr 16 13:27:32 that's a generic API so use is well-documented Apr 16 13:28:09 mdp, my linux experience with DMA is reading the DMA chapter in LDD 15 minutes ago :) Apr 16 13:28:23 you are all set, go! Apr 16 13:28:25 lol Apr 16 13:28:28 ok, then Apr 16 13:28:36 I'll take extra care Apr 16 13:31:31 to put it in perspective, "The Generic DMA Layer" section is what lines up with what I called the "DMA API"…source documentation is in Documentation/DMA-API.txt Apr 16 13:31:56 in short, that's just how you manage the lifecycle of dma buffers in Linux Apr 16 13:32:35 they get born, used and abused and die young Apr 16 13:32:38 the edma api and/or dmaengine api are APIs for accessing a specific general purpose DMA h/w block Apr 16 13:32:55 av500, exactly…but they get cleaned and flushed down the toilet along the way :P Apr 16 13:33:15 * av500 must not make foxconn worker remark Apr 16 13:34:49 jsabeaudry, so just keep in mind that the stuff in the chapter about "set_dma_addr()" is all legacy PeeCee ISA world stuff and ignore it Apr 16 13:35:30 * av500 throws a DMA channel jumper at mdp Apr 16 13:35:57 lol.I was just dreaming of ISA irq/dma channel jumpers/switches Apr 16 13:36:29 * koen moves mdps soundblaster pro(compatible) to irq 11 Apr 16 13:36:58 mdp, I was thinking of starting with http://lxr.free-electrons.com/source/samples/kfifo/dma-example.c, then I should just have to insert a few EDMA API call where they say "put here your code to execute the dma operation" Apr 16 13:37:04 just wait until the beaglebone ISA slot cape is a hot seller! Apr 16 13:37:37 I really like the idea of an ISA cape Apr 16 13:37:46 provided it comes with a proper ISA kernel driver Apr 16 13:38:01 koen: +1 Apr 16 13:38:13 koen, what? s/w support? Apr 16 13:38:29 koen, I don't understand your crazy expectations Apr 16 13:39:11 I can always run devmem2 from inside nodejs Apr 16 13:39:20 * mdp chokes and sputters Apr 16 13:39:33 you do enjoy punishing me with those statements Apr 16 13:39:37 I do Apr 16 13:40:32 ISA should be easy by mapping GPMC into user space Apr 16 13:40:33 I've learned that instead of saying "this is bad engineering" you just say "we hardened and productized it" Apr 16 13:41:12 hey , fast question, im not finding it in the docs. can we power the beaglebone throught the expansion headers Apr 16 13:41:25 battery cape says yes Apr 16 13:41:25 pastjean, yes you can Apr 16 13:41:42 nuclear cape says YES Apr 16 13:41:45 pastjean, you'll have to be extra smart to run at 720 MHz however Apr 16 13:42:06 *I am smart :) Apr 16 13:42:13 jsabeaudry: cpufreq-set -f 720000 Apr 16 13:42:20 jsabeaudry: ahh thank you Apr 16 13:42:54 av500: I don't argue with nuke capes or SilicaGel's DeathCape Apr 16 13:43:14 DEATHCAPE Apr 16 13:43:18 RRRAAARRRRRR Apr 16 13:43:39 jsabeaudry: AHHH thank you i just found it, i went over that page without noticing Apr 16 13:43:41 SilicaGel: sorry, I thought it had StudlyCaps, my bad Apr 16 13:45:08 koen, thanks for that command, how can I validate that it worked? Apr 16 13:45:31 jsabeaudry: cpufreq-info Apr 16 13:46:53 koen, works beautifully, thanks Apr 16 13:47:02 http://pastebin.com/1cc9VH6w Apr 16 13:47:27 the hw guys will be extra happy not to have to cheat the board into thinking its powered from the plug Apr 16 13:48:32 jsabeaudry: with the 3.2.14 kernel 'ondemand' is the default governor, which will scale up to 720 on load, you can select 'performance' if you want to keep it at 720 Apr 16 13:49:00 koen, even on USB ? Apr 16 13:50:01 yes Apr 16 13:50:25 I have a bone running at 720 and powering the lcd+backlight from usb as well Apr 16 13:50:28 it seems to work Apr 16 13:51:17 when people started mentioning cpufreq I said "that is probably going to ignore the usb/dc-plug status" Apr 16 13:51:35 TI responded with "No, it will check the status off that" Apr 16 13:51:43 fast forward 4 months Apr 16 13:51:50 it does NOT check the status Apr 16 13:52:22 Hi, anyone who can tell me how I can define the mode of a gpio bank in CCS using Starterware? Apr 16 13:53:25 ...there are 8 modes [0 7] to choose from Apr 16 13:53:47 and gpio is usually 7 Apr 16 13:54:30 8 modes, binary search should be quick.... Apr 16 13:54:37 yes I see that. I need to route McASP out to the pins Apr 16 13:55:08 I thought these pins all go by the name "GPIO".. Apr 16 13:55:54 McASP belongs to mode 3 and I need to inforce this mode in CCS...but don't know how Apr 16 13:56:36 read register, edit mode bits, write register back Apr 16 13:57:28 aaah ok! I started using Starteware today so any hints on how to follow your recipe would be great Apr 16 13:58:22 ..I mean I understand what to do but I'm not familiar with the syntax yet Apr 16 14:20:18 morning Apr 16 14:23:37 My CCS disassembly window shows wrong info (it does not match the memory image in the memiry view), how can I fix this? Apr 16 14:24:15 well, you could simply stop using ccs Apr 16 14:24:34 what's the alternative? Apr 16 14:24:44 what are you trying to do? Apr 16 14:25:08 initialize DDR on AM3517 with own code Apr 16 14:25:27 how does that need ccs? Apr 16 14:25:43 single step my asm code? Apr 16 14:25:54 what good does single-stepping do? Apr 16 14:26:06 hoping to sneak past the bugs by going slowly? Apr 16 14:27:26 running it does now work, the timing is not usable, looking at the chip does not really help Apr 16 14:27:55 mru: what do you use jtag debug for? Apr 16 14:28:04 I don't Apr 16 14:28:17 there's rarely any need for it one these chips Apr 16 14:28:20 *on Apr 16 14:29:11 I don't see how you can then do testing on timings, unless you can write bug free asm code from scratch Apr 16 14:30:12 you can't do that? Apr 16 14:30:53 No, and the CCS disasm guys can't either. Can you? Apr 16 14:31:10 usually Apr 16 14:31:50 you're invited to write the asm ddr code :-) Apr 16 14:32:41 why don't you just take it from the standard mlo, whatever that is? Apr 16 14:33:11 that's the one that does not work on this board (no surprise). Apr 16 14:41:19 What does OE mean? Apr 16 14:41:29 louisbob, Open Embedded Apr 16 14:41:33 Output Enable Apr 16 14:41:36 lol Apr 16 14:41:51 outlook express Apr 16 14:42:47 mail like a boss Apr 16 14:45:02 XorA: http://www.thingiverse.com/thing:21346 Apr 16 14:46:40 koen: sweet Apr 16 14:47:39 those worms are pretty damn tough, remember back at school we had a challenge to lift max weight using lego motor, all other teams used normal cogs to gear down, we used 2 worms. We ran out of weights, everyone else stripped teeth Apr 16 14:51:18 koen: I'm trying to compile a Kernel, and I saw that you have done a list of patches for the kernel(http://elinux.org/BeagleBoard#Linux_kernel). What do they bring to the board? Apr 16 14:52:02 that wiki is horribly out of date Apr 16 14:52:12 but basically: power management and making it actually work Apr 16 15:22:17 guys, I'd crashed my board. No screen signal, no serial signal... The SD card's LED are lighting up, so I guess the beagle is booting. What can I do? Apr 16 15:22:58 if both lights light up, it's stuck in uboot Apr 16 15:26:17 ok what can I do? Apr 16 15:26:27 http://c.learncodethehardway.org/book/learn-c-the-hard-waych3.html#x8-190003 Apr 16 15:30:29 av500: heh classic Apr 16 15:32:45 koen: I just tested with another sdcard that has working distrib installed on, and it didn't boot... Apr 16 15:32:57 +a Apr 16 15:33:38 +b Apr 16 15:33:54 +c Apr 16 15:34:12 +d Apr 16 15:34:28 +e Apr 16 15:36:30 my mentor tell me off, he thinks that I've burned the board... Fucking bad day Apr 16 15:39:48 any help? this board make me crazy... Apr 16 15:43:55 the board is very hot Apr 16 15:44:02 not good Apr 16 15:44:11 :s Apr 16 15:44:25 hotter than a bread box? Apr 16 15:44:50 yes Apr 16 15:45:57 is it possible to burn a board only with GPIO manipulations? Apr 16 15:46:43 burn a board? Apr 16 15:46:59 is it in a microwave? Apr 16 15:47:13 no the chip on the beagle board Apr 16 15:47:29 come on ^^ I'm in trouble Apr 16 15:50:32 * mdp joins everybody in taking a step back leaving prpplague by himself in front Apr 16 15:50:54 hehe Apr 16 15:52:37 av500, thanks for that link. I've given it a prominent bookmark so I can frequently cite it Apr 16 15:53:36 * LetoThe2nd likes burning things. Apr 16 15:53:41 av500, I'm afraid that Exercise 15 will kill most javascript programmers Apr 16 16:02:45 mdp: ah yes Apr 16 16:06:27 Hi does beagleboard framebuffer use rgb ? (/dev/fb) Apr 16 16:07:08 yes Apr 16 16:07:28 ok thx Apr 16 16:10:44 I remade the SD card but always no screen and no serial. And chip is still hot. what can I do now? Buy a new card? Apr 16 16:15:01 hmm, die hard, that book uses i++ and '\0'. Apr 16 16:15:28 what's wrong with i++? Apr 16 16:15:37 '\0' is a useless construct Apr 16 16:15:47 usually indicative of lack of clue Apr 16 16:16:01 or cargo cult Apr 16 16:16:11 which indicates lack of clue Apr 16 16:16:15 ++i, just in case one wwant to advance to c++ Apr 16 16:16:29 huh? Apr 16 16:16:46 prefix or postfix increment DOES NOT MATTER with any remotely modern compiler Apr 16 16:16:51 i++ uses a temporary for complex stuff, ++i not Apr 16 16:17:02 complex stuff? Apr 16 16:17:06 and EVEN WITH MODERN COMPILERS Apr 16 16:17:23 aholler, another increment and you jump straight to perfe^H^H^H^H^HJava Apr 16 16:17:24 classes Apr 16 16:17:29 5+7i Apr 16 16:17:33 complex Apr 16 16:17:51 you can solve that if you take it to another dimension…then bring that answer back here Apr 16 16:17:52 aholler: show me a case where a sane C compiler generates less efficient code with postfix increment Apr 16 16:18:12 mru: teach c++ yourself Apr 16 16:18:23 mru: unpossible, I use post and my code is perfect Apr 16 16:18:26 we're talking about C here Apr 16 16:18:43 [18:16] < aholler> ++i, just in case one wwant to advance to c++ Apr 16 16:18:50 if you care about efficiency, you don't use c++ in the first place Apr 16 16:19:00 you won Apr 16 16:19:05 go asm Apr 16 16:19:25 and teaching bad habits in C because of is stupid Apr 16 16:19:43 I wonder why javascript is promoted for beaglebone…it must be more efficient than C Apr 16 16:19:45 what's next, typecast the return value of malloc? Apr 16 16:19:58 because c++ requires that Apr 16 16:20:23 c and c++ are two different languages with some similarities in syntax Apr 16 16:20:54 * prpplague has been trying to teach his son C programming from K&R Apr 16 16:21:13 now that's a sensible thing to do Apr 16 16:21:49 mru: i am little disappointed how they introduce variables, but other than that it seems to be doing well Apr 16 16:22:10 prpplague: smart…the link above is useful for those with a 5 minute attention span..like python and javascript script kiddies Apr 16 16:22:39 sorry, I lack the brain to understand why using ++i a bad habbit Apr 16 16:22:48 yours must belarger Apr 16 16:22:59 sometimes i++ makes for more readable code Apr 16 16:23:11 belarger ... nice word :) Apr 16 16:23:25 always use whichever form gives the nicest-looking code Apr 16 16:23:48 ogra_, that's up there with "embiggen" Apr 16 16:23:56 haha Apr 16 16:24:15 that thread about stable must have had teach some people Apr 16 16:31:46 Hi, I can't get usb working on the beagle at all... Apr 16 16:33:30 I'm running 2.6.34 kernel. And it's only showing usb1 in sysfs. Apr 16 16:34:01 I have the ehci driver loaded, I have gpio_vbus compiled in, what could I possibly be missing? Apr 16 16:34:17 pastebin full boot log Apr 16 16:37:40 Hello! I've beaglebaord revision B4 and I would like to install ubuntu-11.10-preinstalled-desktop-armel+omap.img on it. Could anyone link me to instructions which explains why and how to do so? Apr 16 16:38:27 oragsy, the directions are usually at the place that pointed to the download.. Apr 16 16:38:35 av500, sorry, just recompiled and now it won't boot, so I'll be about 2 minutes... Apr 16 16:38:41 and as for "why", I have no idea why you would wont that.... Apr 16 16:38:51 want* Apr 16 16:39:36 oragsy: http://elinux.org/BeagleBoardUbuntu Apr 16 16:39:55 is the best ubuntu support for the BB Apr 16 16:40:05 rcn-ee: You mean http://cdimages.ubuntu.com/releases/oneiric/release/ ? Apr 16 16:40:11 smplman: Yes, thanks. I'm on it Apr 16 16:40:33 oragsy, "For OMAP3 boards. See ARM/OMAP for detailed installation information." ;) Apr 16 16:40:36 smplman: I was also looking for one which describes about connections for beagleboard Apr 16 16:40:52 oragsy: connections? Apr 16 16:42:17 smplman: I mean ports and cables I need to connect... different types of adapters required etc ... Apr 16 16:43:00 oragsy: ahh i see Apr 16 16:43:45 smplman: I'm new to it. So, I'm not sure what more things are required ... Apr 16 16:43:58 the ports are pretty standard Apr 16 16:44:05 what are you trying to do? Apr 16 16:46:20 http://pastebin.com/hU3J05ve <-- there's my boot info Apr 16 16:46:44 There is one IDC10 to DB9 adapter which I'll try to get it by tomorrow. .. I meant to ask about basic connections required after copying the image. I could find - https://wiki.ubuntu.com/ARM/Server/Install?action=show&redirect=ARM%2FOMAPHeadlessInstall- after rcn-ee pointed that out.. So, I'll try the connections now Apr 16 16:47:28 hi Apr 16 16:49:01 Thanks smplman and rcn-ee Apr 16 16:49:14 np, keep on keeping on Apr 16 16:49:32 run is the man Apr 16 16:49:36 rcn** Apr 16 16:50:38 is there any news about the OpenElec project? Apr 16 16:50:56 why would you get such news here? Apr 16 16:51:03 dont they have an irc channel? Apr 16 16:51:17 or a mailing list? or a forum Apr 16 16:53:38 av500, can you see anything in the boot? Apr 16 16:53:47 or, can you see anything that isnt in the boot Apr 16 16:53:48 lol Apr 16 16:54:34 niro, which board is that? Apr 16 16:54:43 its the bb rev b7 Apr 16 16:54:58 ah, so your trying to get musb to work on it then.. Apr 16 16:55:22 *shrug*? Apr 16 16:55:28 maybe? Apr 16 16:55:32 what's musb? Apr 16 16:55:41 niro: that is not a full boot log Apr 16 16:55:42 the otg usb adapter, kernel driver.. Apr 16 16:55:52 nightmare.. Apr 16 16:56:24 I have some musb stuff compiled in... Apr 16 16:56:56 config_usb_musb_host is not set? Apr 16 16:57:04 is that something the b7 needs? Apr 16 16:57:20 if you want musb in host mode yes Apr 16 16:58:16 I didn't include it because it said, omap 343x high speed support... Apr 16 16:58:25 niro, havent touched 2.6.34, since aug of 2010.. but take a look at this defconfig as reference.. http://rcn-ee.homeip.net:81/dl/mirrors/deb/squeeze/v2.6.34.4-x2/ Apr 16 16:59:02 niro: also, why are not running something more recent? Apr 16 17:01:12 Currently the target device is running 2.6.34, so that's what I need to work with as I don't have the target to test on Apr 16 17:01:48 rcn-ee, thanks Apr 16 17:07:03 When I started out, I could get pictures from my webcam, now I have a kernel that doesn't even recognise the camera is there Apr 16 17:07:10 I feel like I may be going in the wrong direction Apr 16 17:07:25 so revert to the old build? Apr 16 17:08:28 the problem with the old build was that the camera would only take pictures up to vga, after that it failed... Apr 16 17:09:01 what was the value of "dmesg | grep fifo" Apr 16 17:09:25 setup fifo mode 4 Apr 16 17:09:37 musb_hdrc: setup fifo_mode 4 Apr 16 17:09:44 well, that is the full result Apr 16 17:10:49 mode 4, doesn't work to well on the original omap3, (missing the 2nd half of 16kb).. retry with this bootarg: musb_hdrc.fifo_mode=5 (hopefully that mode option was pulled in 2.6.34).. Apr 16 17:12:23 ok, rebooting... Apr 16 17:12:49 what bus is the usb controller connected to btw? Apr 16 17:14:07 yeah, it's still not detecting anything Apr 16 17:14:13 no bus Apr 16 17:14:16 universal serial bus Apr 16 17:14:21 well, the L3 or so Apr 16 17:17:25 ok, I've checkout'd v2.6.32, the original kernel I had that was working... Apr 16 17:22:10 A bit confused with all the devices, is the dev_t inside a struct cdev actually a struct device ? Apr 16 17:28:44 Somehow I need a struct device to dma_map_sg but LDD only talks about struct device in the context of sysfs it seems Apr 16 17:36:25 yay, still no usb Apr 16 17:50:15 jsabeaudry: struct device goes with the internal Linux kernel driver model (Documentation/driver-model/device.txt) whereas dev_t is a POSIX API Apr 16 17:54:00 mdp, Oh, and the reason I need that is is LDD : "There are times when the device model leaks out from behind the other layers; for example, the generic DMA code" Apr 16 17:54:52 hmmm…leaking layers Apr 16 18:21:34 jsabeaudry: that's like if you want non-coherent memory..the api requires a struct device Apr 16 18:22:14 jsabeaudry: since you are creating a device driver…presumably a platform driver will be registered…you will be instantiating a platform device which contains a struct device Apr 16 18:23:57 mdp, I was under the impression that it was the case only if I used sysfs Apr 16 18:25:38 void *dma_alloc_noncoherent(struct device *dev, . . .) Apr 16 18:25:56 just an example, I have nfc if you are using that api or just the streaming dma apis... Apr 16 18:28:22 but streaming also requires it, of course Apr 16 18:29:25 the reason for this is that the architecture dependent implementation may do special things based on the device..*may* Apr 16 18:30:33 jsabeaudry: so anyway, I recommend you register this as a platform driver…then register a complementary platform device in your board file to go with it Apr 16 18:35:38 mdp, I really try to do the least stuff possible in the board file as the source control management is a bit complicated Apr 16 18:36:13 ok, you it in your .dts :) Apr 16 18:36:20 s/you/do Apr 16 18:59:36 not sure what a .dts is, is the reason I need to do that stuff in the board file "unknown symbol: device_register" ? Apr 16 19:00:39 .dts is for device tree Apr 16 19:01:22 e.g. arch/arm/boot/dts/omap3-beagle.dts Apr 16 19:01:50 but I assume it was meant as a joke ;) Apr 16 19:02:22 sorry, must enable sartalics Apr 16 19:05:10 I did fdisk on the image I'll be using, it says - http://pastebin.com/1BfefgW8 - so does it mean I will not have to partition the SD card? It already got FAT and ext3/4 ? Apr 16 19:05:11 jsabeaudry, althought it doesn't maintain the intendended platform<->driver separation, you could register both platform device and driver in your module init Apr 16 19:08:20 mdp, the correct way to to register_device in the board file and driver_register in my module init? Apr 16 19:09:17 that's the correct way in absence of DT support for your platform Apr 16 19:09:21 today Apr 16 19:09:40 oragsy, that is my impression too Apr 16 19:09:56 if you want to avoid touching the board file, you can implement what many would consider to be a hack..both in module init Apr 16 19:10:02 the board init should create a platform device for which the module registers a driver Apr 16 19:10:06 jsabeaudry: Ok, thanks.. Apr 16 19:10:26 the platform device can be created manually or through DT Apr 16 19:11:07 oragsy: yes, therefor it's an image Apr 16 19:11:16 of course DT only works fully on a few rare boards Apr 16 19:11:19 mdp, I tried to device_register in my module init but the module wont load because of unknown symbol device_register Apr 16 19:11:19 since he's working on beagle*, I deemphasize the DT path since it's not practical Apr 16 19:11:22 mru, exactly Apr 16 19:11:34 jsabeaudry, platform_device_register() Apr 16 19:11:53 aholler: Otherwise it would have been some .iso? Apr 16 19:12:09 aren't images .jpg? Apr 16 19:12:21 jsabeaudry, and platform_driver_register. you want to refer to Documentation/driver-model/platform.txt Apr 16 19:12:23 mdp, ah interesting, LDD doesnt mention that one :) I'll give it a try Apr 16 19:12:43 oragsy: otherwise an archive (e.g. tar.*) would made sense Apr 16 19:12:58 LDD is old and unfortunately doesn't tell the complete story :-/ Apr 16 19:13:27 aholler: Well, it was ubuntu-11.10-preinstalled-desktop-armel+omap.img.gz initially Apr 16 19:13:32 jsabeaudry: I usually look at some actively maintained driver with pretty code Apr 16 19:14:00 if the code in general looks nice, it's probably doing things properly Apr 16 19:14:11 oragsy: they make images to avoid problems with people which are unable to format a sd-card Apr 16 19:14:55 aholler: So that means I need not follow - http://elinux.org/BeagleBoardUbuntu#U-Boot_uImage_and_uInitrd - when using the above image? Apr 16 19:15:19 oragsy: don't know, didn't read that Apr 16 19:15:50 aholler: That section is under PArtition SD card... so was wondering Apr 16 19:19:48 I also get unknown symbol when using platform_device_register, board file time I guess Apr 16 19:20:02 oragsy: some images write to the SD with two partitions, others you have to build and fill the two partitions. Depends on what images you are using. Apr 16 19:20:17 really a pain that the dma layer leaks like that Apr 16 19:21:31 smplman: I did fdisk-l on the image I'm using. It said - http://pastebin.com/1BfefgW8 So I think I need not format the SD or follow http://elinux.org/BeagleBoardUbuntu#U-Boot_uImage_and_uInitrd . Could you confirm me on that? Apr 16 19:23:34 jsabeaudry: #include , yes? Apr 16 19:23:38 oragsy: https://wiki.ubuntu.com/ARM/OmapNetbook#ARM.2BAC8-SDInstall.content Apr 16 19:24:01 that image you have does have both partitions contained in it Apr 16 19:24:28 you will have to change the name of the image file also Apr 16 19:24:53 I followed - https://wiki.ubuntu.com/ARM/Server/Install?action=show&redirect=ARM%2FOMAPHeadlessInstall#Copy_the_image_to_SD_card . Yea, so no formatting and the u-boot ulimage and initrd part etc Apr 16 19:25:29 jsabeaudry: just for fun, I noted that there *are* upstream mainline drivers actually registering both driver/device in their init code Apr 16 19:25:30 mdp, Did not have the include, but the module compiles fine, it just won't insmod, tried with the include, same result Apr 16 19:25:53 I'll try cleaning Apr 16 19:26:06 ahh, ok, you know, I think you need platform_add_device()..maybe Apr 16 19:26:27 * mdp goes back to look at what is available after all the initcalls Apr 16 19:26:53 smplman: Er sorry. I meant to say that I need not follow http://code.google.com/p/beagleboard/wiki/LinuxBootDiskFormat - ? Apr 16 19:27:31 oragsy: if you are using the stock ubuntu images then yes you will need to copy the uImage. If you use the images for rcn-ee you don't need to. Apr 16 19:28:07 after you write your ubuntu image you should have two partitions Apr 16 19:28:19 you are going to modify the first with uImage Apr 16 19:28:30 jsabeaudry: hrmph..dunno why the problem..you should be able to do that late... Apr 16 19:29:35 mdp, ah found it, usually the error is more explicit, it would seem that these calls are gpl only Apr 16 19:29:36 is anybody familiar with bitbake opencv ? Apr 16 19:30:29 <_av500_> ask a real question please Apr 16 19:31:11 jsabeaudry: heh, I was just about to ask about your module license Apr 16 19:31:31 it should mention the taint..but won't tell you that's why the symbols are unknown Apr 16 19:31:31 I am geeting errors in bitbake opecv, need help Apr 16 19:31:59 <_av500_> still no question Apr 16 19:32:44 if you have an error log post it to a pastebin site Apr 16 19:32:45 there's an opencv accelerated version of bitbake? Apr 16 19:33:19 okay, I will do that Apr 16 19:34:47 So basically, DMA is available for GPL code? Apr 16 19:36:26 depending on your legal interpretation, all kernel code might possibly be required to be GPL Apr 16 19:36:32 that's between you and your lawyers Apr 16 19:37:09 Well, I was thinking in terms of EXPORT Apr 16 19:37:10 smplman: Sorry, pinged out. I'm using ubuntu-11.10-preinstalled-desktop-armel+omap.img.gz from http://es.archive.ubuntu.com/cdimage/releases/11.10/release/ . And I copied the image following the instructions I linked above. And now I can see two partitions on SD card as - http://pastebin.com/DJ4cXPD0 Apr 16 19:37:19 a driver lifted from, say, bsd and hammered into the linux device model ought not be required to be gpl Apr 16 19:37:44 jsabeaudry: implicitly you might say that…but it's specific to a kernel module... Apr 16 19:38:41 there's nothing stopping you from having a UIO-like driver that has a small gpl shim to do the things you want to do in the kernel..and keep your proprietary goop in userspace Apr 16 19:39:08 it's been done many many times by those that feel that they don't want the legal risk of binary kernel modules Apr 16 19:39:18 smplman: I didn't quite get when I'll have to use ulimage... Apr 16 19:40:24 jsabeaudry, let's just solve the technical issue though because legal stuff is fscking waste of time Apr 16 19:40:49 set your license to GPL…make it work..then sort out what makes sense before you distribute the code. Apr 16 19:40:50 Agreed Apr 16 19:41:45 smplman: I probably made a mistake. The image size is 2GB, so dd would require 2GB card not 8GB ... Apr 16 19:44:21 mdp, thanks for all the help btw, it is greatly appreciated Apr 16 19:44:55 np Apr 16 19:57:31 What exactly is DCHMAP as referred in here: http://e2e.ti.com/support/embedded/linux/f/354/t/105149.aspx#370135 Apr 16 19:59:30 Oh probably the DMA Channel Mapping registers Apr 16 19:59:44 wierd that he had to play with that directly Apr 16 20:09:49 I am following Opencv-DSP Accelration project. If I already have working Opencv 2.3. still Do I need to bitbake opencv ? as it is one of the steps in tutorial Apr 16 20:18:37 In order to use EDMA3 in a very simple way, is it correct to simply set ACNT to a value bigger than my biggest transfert and not worry about BCNT AND CCNT? Apr 16 20:44:41 what could possibly go wrong if i put 3.3V on the adc of the beaglebone Apr 16 20:49:48 nvm Apr 16 20:50:10 in linux ain go from 1-8 and in doc from 0-7 Apr 16 21:18:35 Yay! Got some data coming from the DMA, it's mostly garbage and the DMA status is CC_ERROR but it's firing! :D Apr 16 21:28:23 Oh, bad news, it seems I only get the callback to trigger because i edma_start twice, :( Apr 16 21:33:24 There are some good news, the scope says there is action on the bus Apr 16 22:02:18 Oh wow, the only way to write options is to use edma_write_slot it seems? Apr 16 23:01:15 I'm doing the exact example of the TRM section 11.3.19.1, I set the exact same parameters and still I cannot get a DMA transfert complete IRQ, I can only get error IRQ by calling edma_start twice Apr 16 23:01:42 what could possibly cause the TC irq not to be called? Apr 17 00:43:23 jsabeaudry: still around? Apr 17 00:46:27 jsabeaudry: you might look at http://processors.wiki.ti.com/index.php/EDMA_sample_test_application Apr 17 00:46:39 it's on our wiki so it must be true Apr 17 00:48:23 I cannot get internet on my beagle board and it is killing me Apr 17 00:48:37 I can scp files to it but I cannot ping google Apr 17 00:49:17 it is hooked up via Ethernet to my computer Apr 17 00:49:37 and your computer is connected to the internet? Apr 17 00:49:44 haha, yes Apr 17 00:50:11 how is the beagle connected? Apr 17 00:50:20 to a separate network port on the pc? Apr 17 00:50:40 directly to the computer via ethernet Apr 17 00:51:04 and how is the computer connected to the internet? Apr 17 00:51:11 wi-fi Apr 17 00:51:25 so the computer has two network interfaces Apr 17 00:51:32 one for the internet and one for the beagle Apr 17 00:51:36 yes Apr 17 00:51:45 what OS on the pc? Apr 17 00:51:55 debian Apr 17 00:52:14 you need to enable packet forwarding at the very least Apr 17 00:52:22 I did Apr 17 00:52:29 how? Apr 17 00:52:31 somebody named gandalf should be able to get this rockin' Apr 17 00:53:06 scromp: irc nicks don't work like that Apr 17 00:53:12 quite the opposite, in fact Apr 17 00:53:13 damn it all Apr 17 00:54:09 I can check like this: sudo cat /proc/sys/net/ipv4/ip_forward Apr 17 00:54:13 returns 1 Apr 17 00:54:32 you don't need sudo just to look, but sure, that seems correct Apr 17 00:55:22 have you configured the pc as gateway on the beagle? Apr 17 00:56:31 hmmm, maybe not, not sure. How is that done? Apr 17 00:56:40 depends Apr 17 00:56:53 static configuration or dhcp? Apr 17 00:57:16 static Apr 17 01:14:37 koen: ping Apr 17 01:33:42 what does "cdc_acm: Unknown symbol usb_kill_urb" mean? Apr 17 01:34:02 [ 234.893707] cdc_acm: Unknown symbol usb_kill_urb [ 234.894256] cdc_acm: disagrees about version of symbol usb_buffer_free [ 234.894256] cdc_acm: Unknown symbol usb_buffer_free [ 234.894836] cdc_acm: Unknown symbol sub_preempt_count [ 234.895355] cdc_acm: disagrees about version of symbol dev_set_drvdata [ 234.895385] cdc_acm: Unknown symbol dev_set_drvdata [ 234.895690] cdc_acm: disagrees about version of symbol usb_buffer Apr 17 01:37:00 gandalf__: looks like you are using different versions of the kernel modules Apr 17 01:37:48 that would explain why none of my drivers work Apr 17 01:37:54 how do you fix it? Apr 17 01:38:36 get the right versions? Apr 17 01:38:57 i am guessing you probably have some mix up of root filesystem and kernel builds Apr 17 01:39:04 probably build separately Apr 17 01:39:11 I got this with opkg install kernel-module-cdc-acm Apr 17 01:39:47 wouldn't the package manager know to download the right one? Apr 17 01:41:09 gandalf__: probably not if you built your kernel separately Apr 17 01:41:16 gandalf__: or if you have the wrong build info Apr 17 01:43:17 I did not build the kernel. It might be possible I installed the wrong filesystem. How do you change the build info? Apr 17 01:54:12 ah whatever, I'll just install a different os Apr 17 01:57:58 where'd you get the kernel? Apr 17 02:08:34 hope he finds a better OS Apr 17 02:55:38 heh **** ENDING LOGGING AT Tue Apr 17 02:59:59 2012