**** BEGIN LOGGING AT Sat Mar 31 02:59:59 2012 Mar 31 03:14:48 mdp: btw i can confirm whatever the green + blue tabs are i can't see them Mar 31 03:17:15 rlrosa: interesting question, can you elaborate? Mar 31 03:29:40 jay6981: sure. the main loop is given in https://github.com/rlrosa/uquad/blob/master/src/main/main.c . basically, i read from a imu connected to uart2, and when i have data, i process it, and send out commands via kernel msg to another program, that is continuosly sending i2c commands to drive 4 escs Mar 31 03:30:36 the read-process-send loop normally takes 10ms, but occasionaly takes several seconds, and much less than 10ms for a while after the choke Mar 31 03:31:24 i think this "fast" runs after the choke are because the uart has a bunch of data ready, so sampling is much faster than when i have to wait for fresh data, which comes at a 10ms rate (from the imu) Mar 31 03:31:29 is it blocking on something or is it not being scheduled during that time? Mar 31 03:32:38 i'm not sure i understand what you say, but top marks it as "D". if it were running it would be R, and if it were off the schedule it would get a S (right?) Mar 31 03:33:28 i googled and found that it might be I/O trouble, and top shows mccqd being in the middle of things when by program is kicked out Mar 31 03:33:49 if i log a bunch of data (to the SD card) then things get much worse (up to 7s stuck) Mar 31 03:34:41 D is uninterruptible sleep Mar 31 03:35:17 so all signals are being queued until your process runs again Mar 31 03:35:47 what is mccqd? Mar 31 03:37:08 so your process is blocked on i/o when the mmc bus is busy Mar 31 03:37:14 sorry, it's mmcqd. w/r manager for mem card (according to some blog) Mar 31 03:37:21 yes Mar 31 03:37:35 that fits the fact that it gets worse if i want to log data Mar 31 03:37:49 i suggest logging to memory then have another thread write to storage Mar 31 03:38:32 i don't think RT kernel is going to help you here Mar 31 03:39:39 ok, the thread thing sounds smart. but won't it have the same problem? or will the "logger" get stuck every once in a while, without bothering my control loop? Mar 31 03:39:48 that's the idea Mar 31 03:39:53 nice Mar 31 03:39:55 let the logger block on i/o Mar 31 03:40:03 one more Mar 31 03:40:04 while your worker process does work Mar 31 03:40:14 can logging errors to stderr also contribute to this? Mar 31 03:40:23 well logging in general is expensive Mar 31 03:40:37 where does stderr go? Mar 31 03:40:44 mmm...? Mar 31 03:41:01 not sure, i'll look it up. if it's mem card, then i'll send it to the logger Mar 31 03:41:11 i use fprintf(stderr... Mar 31 03:41:13 is stderr routed to a file, pty, serial port, ...? Mar 31 03:41:31 i think pty, the terminal where i call from Mar 31 03:41:36 does that make sense? Mar 31 03:41:53 ok and the pty is fed to a ssh session? Mar 31 03:41:58 yes Mar 31 03:42:11 ok, so you're incurring network io on each log write Mar 31 03:42:19 which will busy the kernel too Mar 31 03:42:39 good. so the logger solution again :) Mar 31 03:43:03 the network is probably less of a bottleneck than mmc but i'm not certain Mar 31 03:43:21 but yeah, those fprintf calls will block for some time Mar 31 03:43:27 and the rest of your code will have to wait Mar 31 03:44:04 ok, first task tomorrow is the logger. how should i comm with it? a pipe? Mar 31 03:44:25 i use kernel messages for comm between my main and the motor driver Mar 31 03:44:37 that should be ok.. mmap would be the most efficient afaik Mar 31 03:44:37 doesn't seem adequate for large chunks of data Mar 31 03:45:16 cool. i'll read about it and try to get it running Mar 31 03:45:21 thank you very much for your help Mar 31 03:45:23 :) Mar 31 03:45:32 sure. sounds like you'll get it sorted out Mar 31 03:46:07 hope so :) thanks! Mar 31 03:47:21 * mranostay checks out Pachube Mar 31 04:18:15 need help : cannot connect to beagle bone , got Permission denied, please try again. Mar 31 04:26:47 I am having trouble getting angstrom to load on my beagleboard mx Mar 31 04:26:55 I am pretty sure that I don't have enough current over USB Mar 31 04:27:37 I just got the board and reinstalled angstrom and now instead of getting kernel panic erorrs, I have a loading bar that just hangs Mar 31 04:30:01 do most people just power them with the barrel connectors? Mar 31 07:02:33 greetings, I have question regarding u-boot I compile for my BBrevC5. I've compiled u-boot.img and MLO but when I flash the NAND with it, the boot process stops with a message "Card did not respond to voltage select!". However if I put the MLO and u-boot.img files onto the SD card and boot with "USER" button -works flawlessly. I tried to switch cards but results are the same. Mar 31 07:05:22 When I flash the NAND with different MLO (obtained as in http://elinux.org/BeagleBoardUbuntu#MLO_and_U-Boot), my u-boot.img works properly so I think that I screwed up my MLO somehow. Mar 31 07:06:39 u-boot sources I use are marked as 2011.12 Mar 31 07:08:09 Right now I just use different MLO but I'm looking for the thing that I'm doing wrong with MLO compilation. Mar 31 07:18:48 just for information: I asked some time ago about serial ports on 3.2.8 stock kernel - the issue was that I couldn't obtain 1MBit/sec speed, all we get was around 1.23MBit/sec with 1MBit settings. The issue is that in /drivers/tty/serial/omap-serial.c there is simple section of the divider 13 vs. 16 but we discovered that it's not properly implemented. Someone took the datasheet and set the condition dependent only on the bitrate reques Mar 31 07:19:30 That is simple implementation of the example given in the datasheet but it souldn't be so. There should be calculation for the minimum error and then selection 13 vs 16. Mar 31 07:20:25 However it's now beyond my experience to do this so I just hardcoded divider to 16 - that fixed my problem and I could get 1MBit/sec. Mar 31 07:50:35 as for my MLO vs u-boot "Card did not respond to voltage select!" issue - I've tested my MLO and u-boot.img on BBrevC4 - the same issue. I've noticed that only this combination of my MLO and u-boot gives this failure. When I use my MLO and some other u-boot - board properly boots from SD. Mar 31 11:08:58 before loading the image on the SD card, how do I need to format the SD card? Mar 31 14:26:46 I've just got my BeagleBoard xM running with angstrom and x11 gui. I'd love to install a browser and understand epiphany has the best performance, but I don't seem to be able to use the package manager to install it Mar 31 14:26:58 any tips would be enormously appreciated Mar 31 14:27:16 what seems to be the problem with the pkg manager? Mar 31 14:27:47 I don't think I've worked out where to install it from properly Mar 31 14:28:47 I've been using this URL http://www.angstrom-distribution.org/feeds/core/ipk/eglibc/armv7a/base/epiphany_2.30.6-r0_armv7a.ipk Mar 31 14:29:09 but it doesn't like it (I don't really know how to use a package manager) Mar 31 14:29:43 I haven't used ipkg in a while Mar 31 14:30:06 but try sthg like 'ipkg update' and 'ipkg install epiphany' Mar 31 14:30:16 if in doubt 'ipkg --help' Mar 31 14:30:34 ipkg Mar 31 14:30:52 I think I have to use opkg? Mar 31 14:31:08 ipkg isn't a command for me Mar 31 14:31:14 I keep confusing those Mar 31 14:31:18 it's been a while Mar 31 14:33:08 opkg update is not valid, and okpg install epiphany returns "not a valid package? Mar 31 14:34:00 opkg is a command now - wasn't using su - Mar 31 14:34:24 amazing ! Mar 31 14:34:27 solved Mar 31 14:34:31 thanks Mar 31 14:35:30 :) Mar 31 16:15:03 hello Mar 31 16:19:58 I am trying to build a custom angstrom kernel for the beaglebone, under open embedded, but I have difficulty in getting started Mar 31 16:20:24 I can build a systemd-image and make it run Mar 31 16:22:19 that sounds good Mar 31 16:24:20 I used bitbake following standard build recommandations Mar 31 16:25:23 I ultimately need to interface with a rotary encoder (for which a driver exists in angstrom) Mar 31 16:26:42 I have identified the pins on the board on which I want to connect it. A person told me to edit the board-am335xevm.c to get what I want. Mar 31 16:28:33 Looking at board_am335xevm.c files floating on internet, I can understand what the different structures are used for. Mar 31 16:29:42 However, I don't have a "base" board_am335xevm.c file in my openembedded workspace to work on, and I have difficulty to understand how to get one, for starters. Mar 31 16:38:51 erwin_schaefer - disable cleanup in your conf/local.conf (comment out INHERIT += "rm_work"), then bitbake virtual/kernel -c compile -f Mar 31 16:39:35 you will have the kernel source in build/tmp*/work/beagle*/linux*/git Mar 31 16:49:10 W1N9Zr0 - Thanks a lot. It looks good, since I can find the board-am335xevm.c in (...) /linux*/git/arch/arm/mach-omap2. Mar 31 17:22:02 howdy! i was wondering if anyone had any suggestions about batteries for the beagle bone? Mar 31 17:29:12 ftPeter wasnt there some stuff which has battery and miniusb connector? Mar 31 17:30:55 http://wiki.daviddarts.com/PirateBox_DIY_OpenWrt#Enclosure_and_Power http://wiki.daviddarts.com/PirateBox_DIY_Plug_Computer#System_Enclosure_and_Power Mar 31 17:31:22 not sure how much i'd trust a lithium charger/power converter off ebay though Mar 31 17:38:58 W1N9Zr0 thanks for the info. Seed studio has a similar type of device that does LiPo to USB host 5v. I'm looking for something that doesn't ship from China though. Mar 31 17:39:25 lol Mar 31 17:39:32 all stuff is shipped from china Mar 31 17:40:47 a lipo-battery, is that a battery based on fat? Mar 31 17:41:16 not THAT would solve the US energy problems Mar 31 17:42:02 opening a new market for fight club guy Mar 31 17:42:49 http://www.forbes.com/2008/12/21/fat-fuel-biodiesel-tech-sciences-cz_pcb_1222fatfuel.html Mar 31 17:44:56 is anyone using usb wifi dongles on beagleboard on angstrom? i can't get mine out of rfkill soft blocked state Mar 31 17:45:12 it worked on the ubuntu image Mar 31 18:55:16 Does anybody know why I would get this error when trying to compile a kernel module (a device driver .ko)? "make: *** No rule to make target `modules'. Stop." Mar 31 18:55:31 how can that target not exist? Mar 31 18:56:20 my exact command is long but I'll paste it here, I hope is not too much Mar 31 18:56:25 make CROSS_COMPILE=arm-angstrom-linux-gnueabi- ARCH=arm -C ~/setup-scripts/build/tmp-angstrom_2010_x-eglibc/work/beaglebone-angstrom-linux-gnueabi/linux-ti33x-psp-3.2-r6a+gitr09e9651bcf2ee8d86685f2a8075bc6557b1d3b91/git/kernel M=`pwd` modules Mar 31 18:57:13 that command should work Mar 31 19:00:36 OrlandoT: That's a lot of stuff Mar 31 19:00:50 I don't like the -C Mar 31 19:01:01 lol I know Mar 31 19:01:14 the path suck Mar 31 19:01:17 can you build it directly from the kernel directory? Mar 31 19:01:46 make CROSS_COMPILE=arm-angstrom-linux-gnueabi- ARCH=arm M=path/to/your/module's/dir modules Mar 31 19:01:47 ? Mar 31 19:01:47 I prefer to be in another directory plus I will put that command in a Makefile and not have to type it everytime Mar 31 19:02:21 by prefer to be in another directory..... do you mean you don't want the executable/intermediate files to go in with the source? Mar 31 19:02:21 i thought it was the path to the kernel dir not to the module dir Mar 31 19:02:25 no Mar 31 19:02:35 make -C foo should be equivalent to cd foo && make Mar 31 19:02:41 M= is relative to the kernel source Mar 31 19:03:01 alan_o: not if M is an absolute path Mar 31 19:03:11 * alan_o stands corrected Mar 31 19:03:26 and that should work Mar 31 19:03:33 but, if you want the generated files to go somewhere else, you can try O= Mar 31 19:03:53 it's a while since I built an out of tree module, but absolute paths did work then Mar 31 19:04:09 i'm getting really confused now, lets start simple Mar 31 19:04:24 I am going to try to compile right from the kernel's path Mar 31 19:05:06 Yeah, that's how I'd start; do it the obvious way and add one complication at a time. Mar 31 19:05:19 ok Mar 31 19:05:47 so If i got to my kernel path and type "make modules" it should recognize that target right? Mar 31 19:05:52 mru says that your basic idea should work, so you probably have some detail of it not exactly right (like a path), or maybe something in your path (like the +) might need to be escaped or something Mar 31 19:06:14 digging out some old scripts, that command line looks sane Mar 31 19:06:21 well, as sane as that kernel dir is Mar 31 19:06:54 yes, I used to compile my modules without any issues doing this command below Mar 31 19:06:57 make CROSS_COMPILE=arm-angstrom-linux-gnueabi- ARCH=arm -C ~/setup-scripts/build/tmp-angstrom_2010_x-eglibc/sysroots/beaglebone/kernel M=`pwd` modules Mar 31 19:07:10 I saved the command in a file so this is a copy and paste Mar 31 19:07:25 it worked flawlessly so now I have only changed the path Mar 31 19:07:32 make a symlink to that path, and try it with the symlink Mar 31 19:07:45 that'll tell you if your path (or some chars in it) are the problem Mar 31 19:08:34 well, I went to the kernel directory and typed "make modules" Mar 31 19:08:35 I've done my out-of-tree stuff differently, but now that I understand what you're doing, I like this. Mar 31 19:08:36 and it says Mar 31 19:08:37 make: *** No rule to make target `modules'. Stop. Mar 31 19:08:55 do you have a file named modules in the directory? Mar 31 19:09:33 (I had that one time before with an ./install script I had put in there. Couldn't figure out why make install failed) Mar 31 19:09:49 no file called modules Mar 31 19:09:58 it's a target in a makefile Mar 31 19:10:09 sure Mar 31 19:10:37 but if you have a file called modules, and it's not declared .PHONY in the makefile, it'll mess that target up. Mar 31 19:10:49 oh sure Mar 31 19:10:51 anyway, it's not that then.... Mar 31 19:10:59 but I have not touched that dir Mar 31 19:11:11 this is weird Mar 31 19:11:35 you have module support enabled right? Mar 31 19:11:45 not sure what happens if you don't and then do a "make modules" Mar 31 19:11:49 never tried it :) Mar 31 19:12:03 having a file called modules would not make it fail with that message Mar 31 19:12:19 then it would just do nothing and possibly print "nothing to be done for 'modules'" Mar 31 19:12:23 I imagine since the only thing I did was "MACHINE=beaglebone ./oebb.sh config beaglebone MACHINE=beaglebone ./oebb.sh update MACHINE=beaglebone ./oebb.sh bitbake systemd-image" Mar 31 19:12:38 mru: ahh, yes, that's the one. It's been a while since I made that mistake. Mar 31 19:13:40 so I have not touched any configuration or source file or anything at all. I just have what open embedded dumped in that directory Mar 31 19:13:46 OrlandoT: Do you have a proper .config and all that stuff. I'm a little curious about "haven't touched that dir," and I'm not sure how the OE stuff would handle that. mru probably knows better Mar 31 19:14:21 it has to be configure of course Mar 31 19:14:24 +d Mar 31 19:14:34 MACHINE=beaglebone ./oebb.sh config beaglebone Mar 31 19:14:38 as I did above Mar 31 19:14:50 meaning some make fooconfig must have been completed Mar 31 19:14:57 everything was automated by OE Mar 31 19:15:04 module support enabled? Mar 31 19:15:08 mru: true, but does OE put a .config _in_ the kernel directory, or is it somehow building it with a .config outside that or something. Mar 31 19:15:19 good point Mar 31 19:15:26 maybe it needs an O= argument Mar 31 19:15:27 module support is enabled because I do a lsmod in my device and it shows modules Mar 31 19:15:41 set to the same value the main kernel build uses Mar 31 19:37:40 ok it was a stupid mistake on the path Mar 31 19:38:01 the path should have ended at /git instead of /git/kernel Mar 31 19:39:36 so the exact command that works is Mar 31 19:39:41 make CROSS_COMPILE=arm-angstrom-linux-gnueabi- ARCH=arm -C ~/setup-scripts/build/tmp-angstrom_2010_x-eglibc/work/beaglebone-angstrom-linux-gnueabi/linux-ti33x-psp-3.2-r6a+gitr09e9651bcf2ee8d86685f2a8075bc6557b1d3b91/git M=`pwd` modules Mar 31 19:41:07 thank you guys for the help Mar 31 19:45:04 I finally have my system fully setup to do some actual development Mar 31 19:49:16 OrlandoT: It seems sometimes that's the hardest part :) Mar 31 19:51:26 I knowwww Mar 31 19:51:29 crazy Mar 31 19:51:51 there was a learning curve specially knowing where to work from Mar 31 19:52:51 I was trying to work from /sysroots/beaglebone/kernel Mar 31 19:53:05 instead of from /work directory Mar 31 19:53:35 it tooks weeks before Koen told me about this link http://www.slimlogic.co.uk/2011/05/openembeddedangstrom-kernel-workflow/ Mar 31 20:06:17 * mranostay yawns Mar 31 21:02:34 I formated the SD card then load the image again to the SD card. However, the password that I set before the format has not changed Mar 31 21:02:49 I followed the same instruction that is written on angstroms dist. webpage Mar 31 21:02:57 isnt it weird? Mar 31 21:26:38 angs: are you doing it wrong then Mar 31 21:39:37 I executed the same instructions that is written on http://www.angstrom-distribution.org/demo/beaglebone/ Mar 31 21:40:07 I dont know what is wrong Mar 31 21:41:06 first I formated the SD card by "sh mkcard.txt /dev/sdX" Mar 31 21:41:16 I had two partitions Mar 31 21:41:47 then copied MLO, u-boot.img and uImage to the partition that is written "angstrom" Mar 31 21:42:16 then $ sudo tar -xjv -C /media/rootfs -f /path/to/Angstrom-BeagleBone-demo-image.rootfs.tar.bz2 Mar 31 21:42:43 of course I corrected the path and names regarding to my bone Mar 31 21:45:16 installing the angstrom to beaglebone and beagleboard is almost the same, right? Mar 31 21:50:15 when I connected to the bone through ssh, RSA key was changed after the format and image load Mar 31 21:50:50 and the files that I loaded to the beaglebone were deleted Mar 31 21:51:23 it seems like it is formated without problem but the password setting has not change, it is weird. Mar 31 21:51:35 is there anyway to understand if the loaded image is fine? Mar 31 21:51:43 if it works fine? Apr 01 00:07:50 is doing an opkg update; opkg upgrade a reasonable thing to do on my beagle bone or is that a bad idea? Apr 01 00:38:09 ftPeter: do a 'umount /tmp' to avoid running out of space on tmpfs Apr 01 01:04:25 koen: and run out of space on rootfs instead? Apr 01 01:04:55 that would be harder, but yes Apr 01 01:05:05 depends on the size of it, no? Apr 01 02:01:08 * koen discovers the chocolate cookie from lunch in his jacket pocket while putting in another cookie Apr 01 02:01:25 so 2 cookies to go with this coffee Apr 01 02:10:34 heh Apr 01 02:10:50 koen, now you know about evenings on the left coast Apr 01 02:11:07 mru, what is the deal with Ne10 Apr 01 02:17:36 is that the neon thing from arm? Apr 01 02:18:00 Crofton|work: california still weirds me out Apr 01 02:26:21 Crofton|work: are you attending collab? Apr 01 02:27:52 heh... Parallel -> HDMI -> Parallel is a good way to burn 500-600mA ! Apr 01 02:30:37 why does it weird you out? Apr 01 02:30:49 the dirty smelly hippies? Apr 01 02:42:59 rofl Apr 01 02:43:06 yeah flying tomorrow at 0800 Apr 01 02:43:37 I'm not saying ALL hippies are dirty and smelly. Apr 01 02:43:44 Some of them might just be dirty but not smell. Apr 01 02:44:03 bed time here Apr 01 02:53:51 does beagle board have a AREF input , analog reference voltage Apr 01 02:59:57 lurk mode, ENGAGED. **** ENDING LOGGING AT Sun Apr 01 02:59:58 2012