**** BEGIN LOGGING AT Thu Jul 26 02:59:58 2012 Jul 26 04:11:25 Sweet, just got the mt9p031 sensor working on the BBxM with angstrom 3.2.18 Jul 26 06:31:03 spacecolonyone: https://github.com/beagleboard/kernel/tree/beagleboard-3.2/patches/camera Jul 26 06:37:53 <_av500_> mranostay: just dont hit him over the nose Jul 26 06:43:54 _av500_: no i may need him alive Jul 26 09:21:39 Could anyone please tell me if Leopard Imagining composite video input board (LI-VI365) is compatibile with BeagleBoard by any chance? Jul 26 14:35:20 koen: So I got your new stuff booting. I was even able to do NFS-root over the OTG port. Jul 26 14:37:21 koen: There's of course no SPI yet. Do you think I could just add that to the DTS, or is there more groundwork that needs to happen first? Jul 26 14:40:24 koen: One of the things I haven't gotten my head around yet with respect to device tree is the muxing. The board-amx335x.c had a lot of pin muxing stuff in it, but the DT doesn't and omap_hwmod_33xx_data.c doesn't seem to either. Jul 26 14:42:06 There is a small amount of pinmux in the .dts for the LEDs. I wonder if the muxing I need for SPI would just go here... Jul 26 14:47:20 alan_o: there was SPI stuff inthe dtsi, but that got removed Jul 26 14:47:27 alan_o: SPI needs dma as well, so no-go Jul 26 14:47:47 and yes, pinmuxing suck in DT land because we don't have nice helper constants Jul 26 14:48:36 So DMA is still not in then? I know you said that at one point, I assumed it must have made it in because I assumed USB must use it. I'm surprised USB doesn't use DMA. Jul 26 14:48:53 usb doesn't even compile if you enable dma Jul 26 14:49:12 So DMA is optional for the USB controller then I guess Jul 26 14:50:03 Since there's GPIO, I wonder if I can set up one of those SPI-over-GPIO deals. Never had reason to try that before.... Jul 26 14:50:07 hi @ all Jul 26 14:50:13 you could Jul 26 14:50:41 alan_o: but be warned: https://plus.google.com/100242854243155306943/posts/AHjSoGBHNue Jul 26 14:50:41 it would go nicely with my NFS over USB g_ether :) Jul 26 14:51:34 * koen tries to locate his usb2ethernet dongle Jul 26 14:51:42 i have Beagleboard XM rev c which ubuntu image i need omap or omap4 ?? Jul 26 14:51:50 I hope the cpsw guy posts his work soon Jul 26 14:53:23 koen: Indeed, I wouldn't think of bit-banging if I had a port available. Although, it's really nice that in this case, Linux will do the bitbanging for me. Jul 26 14:53:24 cpsw, mmc and da8xx-fb are the big ones Jul 26 14:53:35 (mmc needs dma as well, no idea about fb) Jul 26 14:56:47 so as far as DMA goes, what needs to happen there? Groundwork from Vaibhav? Rebasing stuff from arago onto linux-omap? Jul 26 15:01:46 mdp is doing a driver from scratch Jul 26 15:03:01 well, a am18xx edma driver, but that maps to am335x Jul 26 15:04:38 * koen checks progress on workstation update and dinner preparations Jul 26 15:04:50 wow, that's cool. So am I to assume that this branch will at some point become official, breaking away from arago? Jul 26 15:05:05 alan_o: I'm going to test linus' tree in an hour or so and base it on that Jul 26 15:05:13 alan_o: no arago in there at all Jul 26 15:06:17 so OE will at some point be with arago then. Interesting Jul 26 15:06:25 guest23: A BeagleBoard is omap3, so you don't want omap4 Jul 26 15:06:44 guest23: I haven't run Ubuntu on Beagle, so I may not know much more than that :) Jul 26 15:07:34 alan_o: the experiences with beaglebone have shown that using TI kernels (which is what you mean with 'arago' I guess) is just not worth it Jul 26 15:07:44 no way to get patches upstream Jul 26 15:07:52 need to redo everything pretty much everyweek Jul 26 15:07:59 no updates from the stable tree Jul 26 15:08:25 only works with one magic config, everything else isn't tested and hence broken Jul 26 15:10:38 and it's only ticking the marketing checkboxes, not making it work well Jul 26 15:10:42 koen: yeah, definitely. (and I get the arago name from the domain arago-project.org in the git URL in the recipe. I don't know any more about arago than that, except that it's the repo for TI's OSS stuff). Jul 26 15:11:36 arago is like yocto Jul 26 15:11:48 the same name used for every $*#()%&@*%(*@ part of the project Jul 26 15:11:53 koen: and it's now the double-whammy of having all "new" stuff going upstream having to be device tree. I guess that's good long-term. Jul 26 15:12:20 if you're attending ELC-E drop by my "the broken promis of devicetree" task Jul 26 15:12:23 (if it gets accepted) Jul 26 15:12:35 ah. yeah, one day I'll understand yocto. I try at ELC to get yocto people to sell me on it, and none have made a convincing case. Jul 26 15:13:07 I am planning on being at ELC-E, and I will definitely check it out (pending approval). I'm not completely sold on DT either. Jul 26 15:13:21 There are still a lot of things which I don't see how it can really do. Jul 26 15:13:58 I think there's a layer about dt needed for things like beaglebone Jul 26 15:14:07 and on top of that I'm always suspicious whenever people want to turn complex code into what often turns out to be even more complex config files. Jul 26 15:14:08 probe the hw, generate a dt, (re)boot Jul 26 15:14:24 I'd love something like vala for DT Jul 26 15:14:34 so I can say "mux pin foo as gpio" Jul 26 15:14:49 instead of "foo: 0x32 0x4343 0x422" Jul 26 15:14:53 yeah, but in that case, how is it better than a board-* file, where all that stuff can live in code, and work automatically? Jul 26 15:15:06 no need to recompile Jul 26 15:15:08 yeah, that, the hex stuff that has to go in the .dts Jul 26 15:15:15 in principle no need to compile Jul 26 15:15:20 right :) Jul 26 15:15:21 but Jul 26 15:16:03 WHo is going to mess with a device tree that isn't capable of compiling? How are they going to build the blob without kernel source (I'm sure it can be done, but isn't it going to be just as hard, or close to it?) Jul 26 15:16:23 aiui the dtc doesn't need the kernel Jul 26 15:16:29 uboot could compile the dts Jul 26 15:16:48 I have long forgotten how to use DT on ppc405 Jul 26 15:17:23 Well if it's built into u-boot...... I suppose someone could make a mod, use u-boot to compile it, then have u-boot boot from it. Jul 26 15:18:04 uboot used to be ppcboot, so it knows about DT Jul 26 15:18:05 That's somewhat easier than building a kernel I guess. Jul 26 15:18:11 but I need to look into what tricks it can do Jul 26 15:18:50 at work the hw guy drives me mad with moving the gpio-keys pins around a lot on lcd cape designs Jul 26 15:19:01 I saw an email of yours from months back about having u-boot detect capes and then re-build a DT to support. That's a cool idea. Jul 26 15:19:23 with DT I can just pass a new dt file Jul 26 15:19:31 instead of hack kernel, compile, boot Jul 26 15:19:59 right now simple changes like gpio-key changes take me at least an hour to integrate Jul 26 15:20:00 Yeah, I mean there are advantages, no doubt. Jul 26 15:20:35 but then again, I need to rebuild the kernel everything there's a new stable kernel release (every 2-3 weeks) and patches to fix drivers Jul 26 15:20:54 so I guess I'll go from building a kernel 10 times a day to 7 times a day :) Jul 26 15:21:33 right. There's definitely a use case difference between users and hackers. You want any system to "just work" for users, but be easy to modify and/or easy to do the hard things for hackers. Jul 26 15:22:04 you can see on the mailinglist how adverse people are to rebuilding the kernel Jul 26 15:22:04 DT has some of both, but also makes it a bit harder for both in other ways. Jul 26 15:22:24 yeah, well most people don't have the skills for it. It's not something they've done yet. Jul 26 15:22:46 Stuff like the Beagle* products and the RPi attract two kinds of people, from what I gather Jul 26 15:22:51 one group thinks it's a mini-PC Jul 26 15:22:56 a few years ago I said to jkridner___ "if people need to rebuild the kernel, we have failed" Jul 26 15:23:05 the other group thinks it's a large microcontroller. Jul 26 15:23:57 Each of those two groups, if they don't have other knowledge will be somewhat confused and disappointed, because it's both, and at the same time, it's neither. Jul 26 15:24:25 At the end of the day it's an embedded linux board, and embedded linux requires building a kernel :) Jul 26 15:24:27 imo, anyway. Jul 26 15:25:02 Cloud9 and things like that make it a lot better for the 2nd group, and in that sense, you're right in your comment to jkridner___ . Jul 26 15:25:20 I think the largest part of the beagle users doesn]'t know or want to rebuild a kernel Jul 26 15:25:31 I think that's true. Jul 26 15:26:30 but, unless they can do it from u-boot or some other way that's really easy, they won't want to build a new DT either. So I guess that's the goal, make DT easy enough for most users to modify to suit their needs. Jul 26 15:27:40 because even though people don't want to build kernels, not everyone has the same uses. With all the pinmuxing, it's hard to make it one-size-fits all. Many users will have to change _something_ Jul 26 15:28:12 some wanting SPI, some wanting i2c, some wanting none of that but all those pins set to GPIO Jul 26 15:29:37 the kernel can autopinmux based on eeprom contents nowadays Jul 26 15:29:53 so in theory people only needs to do cat foo.bin > eeprom ; reboot Jul 26 15:30:40 I didn't know about that. That's pretty cool. Jul 26 15:31:04 It's so hard to keep up with everything :( Jul 26 15:31:30 indeed Jul 26 15:50:58 is twitter down? Jul 26 15:51:27 another ddos? O-o Jul 26 16:01:22 alan_o: fwiw mainline + patches works Jul 26 16:05:47 koen: very cool! Jul 26 16:08:23 koen, link for patches? assume you mean am33xx? (I read quite a bit above, but...) Jul 26 16:08:44 * bradfa would love a mainline kernel! Jul 26 16:08:48 :) Jul 26 16:09:14 bradfa: We're talking about https://github.com/beagleboard/kernel/tree/beaglebone-3.6 Jul 26 16:09:28 alan_o, thanks! will look Jul 26 16:09:29 So DMA-related things don't work yet Jul 26 16:10:04 no ethernet, SPI, Jul 26 16:10:11 that's the gist I got from reading above Jul 26 16:10:12 other things no doubt too Jul 26 16:10:14 frame buffer Jul 26 16:10:17 but USB works Jul 26 16:10:17 mmc? Jul 26 16:10:18 ok Jul 26 16:10:20 no mmc Jul 26 16:10:25 afaik Jul 26 16:10:39 I have rootfs on a usb stick Jul 26 16:10:39 ah, ok, so not that much progress since my last attempt at annoying everyone about mainline... Jul 26 16:10:41 but you can boot off a USB memory stick Jul 26 16:10:58 and you can boot off NFS via g_ether Jul 26 16:11:08 or usb2ethernet dongle :) Jul 26 16:11:11 * bradfa needs to boot off mmc1 Jul 26 16:11:26 custom cape coming, soon... Jul 26 16:11:36 * koen looks at the eMMC cape on his desk Jul 26 16:11:48 need to add the code for that after dinner Jul 26 16:12:02 * koen was stupid enough to accept a meeting invite without doing the timezone math Jul 26 16:12:13 1pm PST is 10pm here Jul 26 16:12:17 ouch Jul 26 16:12:22 I always do the timezone math wrong. Jul 26 16:12:28 which emmc device? Jul 26 16:12:38 er, chip? Jul 26 16:19:12 koen, I'll take some world peace, pony, and ti public git tree, too (if you have any spare wishes) :) Jul 26 16:49:32 What steps should I follow to submit a patch for adding support for the LI-5M03 to the BBxM with angstrom? I take it I should: git format-patch against the 3.2.18 kernel (or possibly against 3.2.18 AFTER it is patched with all the current patches in the 3.2.18 recipe??), then submit it to a mailing list, right? Jul 26 17:35:39 Any ideas how to make it so I don't need to issue 'udhcpc -i eth1' after every boot up to pull an ip? I'm running bitbaked angstrom 3.2.18 slightly customized systemd-image Jul 26 17:36:33 the onboard eth0 face is connected, but not to an active network. eth1 is a USB to wired adapter on the real network Jul 26 17:37:54 does /etc/network/interfaces exist? Jul 26 17:40:42 Someone might be interested in this: I have a patch for Openembedded's meta-ti that makes it build BeagleBone's kernel with RT_PREEMPT, and I've run timing tests compared to the regular kernel Jul 26 18:05:42 bradfa: a micron emmc Jul 26 18:06:47 koen: how much storage? Jul 26 18:07:47 koen, part numbers? Jul 26 18:13:38 bradfa: 21a18 jw864 Jul 26 18:13:49 bradfa: I'll know more after I have booted the board :) Jul 26 18:15:38 jkridner___: estimated deadline is july 31st for the new image, do you think you can have bonescript master in a decent state by then? Jul 26 18:15:51 jkridner___: I haven't checked the current state, it could very well be decent already :) Jul 26 18:22:34 koen, I assume that's the markings on the part, not the part number, if so, PN is MTFC2GMTEA-0F WT Jul 26 18:23:11 it's indeed the markings Jul 26 18:23:29 micron's fun like that Jul 26 18:35:09 dm8tbr: yes, but i though conman doesn't respect it or even care Jul 26 18:35:16 I think I may have figured out what was going on with the serial stuff on my openbsd box Jul 26 18:35:26 currently uftdi can only handle one serial endpoint Jul 26 18:35:32 the beaglebone appears to have two Jul 26 18:42:53 spacecolonyone: https://github.com/beagleboard/kernel/tree/beagleboard-3.2/patches/camera Jul 26 18:44:48 koen: and I was so proud of my first contributeable patch! :( Jul 26 18:44:57 beat me to it Jul 26 18:45:25 spacecolonyone: it still needs to get detected in u-boot, then properly configured in linux Jul 26 18:45:28 much more work to do :) Jul 26 18:47:03 I don't follow. I just applied a few patches (mostly copied from Max Galemin's buildroot work) removed a duplicate initialization of the i2c-2 from the -camera.c file and it just works. Jul 26 18:47:29 * spacecolonyone works as in I can use media-ctl, yavta, & mplayer to get an image Jul 26 18:50:59 koen: I didn't need to patch u-boot at all. I added camera=mt9p031 in uenv.txt. What am I missing? Do you mean add u-boot code so that the camera= is detected automatically? Jul 26 18:56:24 yes Jul 26 18:58:21 so much for discourse :p Jul 26 18:58:48 Looking for help with an error compiling mysqldb libraries. Jul 26 18:58:57 Got this on the compile: "this linker was not configured to use sysroots" Jul 26 18:59:57 More of the error: "arm-angstrom-linux-gnueabi/bin/ld: this linker was not configured to use sysroots" Jul 26 19:00:53 anyway, I'm probably not going to pursue this camera too much further. I bought it more to check that I got OMAP3ISP support and all was working. Now I'm on to using the mt9[031 driver as a base for a custom FPGA based camera I'm building. Jul 26 19:02:01 Oh, Koen: do you know why OMAP_IOMMU and IOVMM were disabled by default in the defconfig for the beagleboard? Jul 26 19:26:00 bmercer: That's true, the first one is the JTAG, the second is the console. Too bad about the openbsd. I should have thought of that yesterday. Jul 26 19:29:59 what about openbsd? Jul 26 19:36:41 alan_o: ok, that explains it! Jul 26 19:37:48 mranostay: bmercer says the openbsd ftdi driver can't support multiple serial interfaces. Jul 26 19:37:56 nice Jul 26 19:43:57 Here's a good example graph for RTLinux-3.2.18 on beaglebone made from cyclictest and simulating load with fping and calibrator: http://i.imgur.com/OeAga.png Jul 26 20:28:02 bradfa: Jul 26 20:28:04 [ 0.750427] mmc1: new high speed MMC card at address 0001 Jul 26 20:28:04 [ 0.751037] mmcblk1: mmc1:0001 MMC02G 1.79 GiB Jul 26 20:28:04 [ 0.751708] mmcblk1boot0: mmc1:0001 MMC02G partition 1 1.00 MiB Jul 26 20:28:05 [ 0.751922] mmcblk1boot1: mmc1:0001 MMC02G partition 2 1.00 MiB Jul 26 20:37:23 bradfa: http://dominion.thruhere.net/koen/angstrom/beaglebone/0001-beaglebone-add-rudimentary-support-for-eMMC-cape.patch Jul 26 20:38:07 that makes it detect the emmc, reading it seems to fail Jul 26 21:10:05 koen, pushing u-boot emmc patches to https://github.com/joelagnel/u-boot/tree/AM335XPSP_04.06.00.08 Jul 27 00:56:53 Someone to help me building, or sending a ubuntu .img for the beaglebone....? Jul 27 00:57:12 im planing to write it down to a SD card. on windows 7 Jul 27 00:57:13 * mranostay groans Jul 27 01:00:14 * ka6sox-away hopes the git gods are merciful tonight... Jul 27 01:03:01 any1? Jul 27 01:03:21 Need help with a ubuntu .img for the beaglebone.... Jul 27 01:03:36 patience will get you there... Jul 27 01:56:21 jkridner: ping **** ENDING LOGGING AT Fri Jul 27 02:59:58 2012