**** BEGIN LOGGING AT Wed Oct 31 02:59:58 2012 Oct 31 03:14:18 i refuse! Oct 31 03:22:46 mranostay, I'll bet I can get you to say it over beer...in fact, I am highly confident you will weaken your stance Oct 31 03:23:53 with what free samples? Oct 31 03:24:09 only if you say the name Oct 31 03:24:22 in a praising manner...with witnesses Oct 31 03:25:18 is this a cult? Oct 31 03:25:34 more of a lifestyle ;) Oct 31 03:28:33 so disney plans to pump out Star Wars movies? Oct 31 03:30:06 mdp, is the A10 *not* documented? Oct 31 03:30:15 sorry..I missed that reference. Oct 31 03:30:33 ka6sox, yes Oct 31 03:30:52 let me find the link I jokingly referred to as the A10 TRM one day Oct 31 03:31:33 http://rhombus-tech.net/allwinner_a10/A10_register_guide/ Oct 31 03:32:05 brings new meaning to "Use the source!" Oct 31 03:33:35 no kidding Oct 31 03:40:07 mdp: nice.... Oct 31 03:41:31 that's more than exynos has publicly ;) Oct 31 03:41:48 though to be fair, they have good kernel support Oct 31 03:52:40 mdp, so the headerfiles are the documentation...got it. Oct 31 03:53:29 yep, with allwinner, that's from their vendor code tree for android, of course Oct 31 04:01:08 mdp: who thinks of these names? Oct 31 04:04:09 mranostay, Allwinner= China, Exynos = Korea.... Oct 31 04:06:53 they can't have the one token native speaker on the team? :) Oct 31 04:06:59 we have new synonyms for design disaster? Oct 31 04:07:31 ds2: usually a bad work followed by 'up' Oct 31 04:08:08 bad? isn't that a description of a the mamillian way of making more? Oct 31 04:08:40 *word Oct 31 04:08:51 heh Oct 31 04:12:22 mranostay: you going to santa clara this week? Oct 31 04:15:46 for? Oct 31 04:15:56 arm techcon Oct 31 04:16:20 my day job isn't ARM :) Oct 31 04:16:45 ARM is the world. you will encounter it like it or not ;) Oct 31 04:18:03 tell that to some intel fanboys i will not name :) Oct 31 04:18:28 even intel made (still makes?) ARM parts Oct 31 04:18:50 shhh, that's not to be talked about Oct 31 04:18:54 still xscale? Oct 31 04:19:20 isn't that basically dead? :) Oct 31 04:19:20 :) Oct 31 04:19:26 tell that to Marvell Oct 31 04:19:30 and the iXP users Oct 31 04:19:48 mranostay, not the comic book folks ;) Oct 31 04:19:59 hello. I am new to beagleboard xm. but love it so far. I have it hooked up to my 42" panasonic tv but I cannot get the entire screen to display. I was looking at the bootargs with default omapfb and it seems like I need to modify the parameters. Has anyone done this? Oct 31 04:20:05 i really don't care what platform i have to us as long as it ISN'T X86!!! Oct 31 04:20:07 May I get some suggestion? Oct 31 04:21:03 bulbage: 1080p? Oct 31 04:22:29 720 Oct 31 04:23:34 mdp: cute Oct 31 04:25:37 mranostay: to be exact: Panasonic Viera TH-42PX80U 42-Inch 720p Plasma HDTV Oct 31 04:25:56 oi, so it looks like we are losing a ARM kernel developer, Linus Walleij apparently must go http://lkml.indiana.edu/hypermail/linux/kernel/1210.3/02834.html Oct 31 04:26:45 *gulp* Oct 31 04:26:47 I think av500 suggested he could change to Legolas Oct 31 04:28:25 * mranostay thinks after 5 years he'd remember to find his glasses first before taking out his contacts Oct 31 04:28:36 I hate it when that happens Oct 31 04:38:20 Russ: are you doing techcon? Oct 31 04:38:48 anyone has any sugesstions on the display issue? Oct 31 04:39:36 hadn't even heard about it Oct 31 04:39:59 oh hahahah Oct 31 04:44:02 ds2: get your LDO to fit? Oct 31 04:44:20 emeb_mac: yep Oct 31 04:44:27 cool Oct 31 04:44:34 ordered the Processor and RAM too Oct 31 04:44:45 going to sit on it for an hour more before releasing to fab Oct 31 04:44:50 DK has thousands of those PSRAMs - should be plenty. Oct 31 04:45:20 digikey is $$$$$$$$ Oct 31 04:45:25 ordering elsewhere Oct 31 04:47:15 understood - but that's a good sign that they're not likely to disappear from all sources like the other. Oct 31 04:47:44 the other had plenty available a few weeks ago :( Oct 31 04:48:23 oh well - now you've got lower power. Oct 31 04:49:00 that's the glass is half full way of describing it :) Oct 31 04:49:11 still prefer the other part if I can get it later. made it an option on the board Oct 31 05:08:06 good idea Oct 31 05:32:06 mdp: with spi there are docs that say the memory has to be DMA memory Oct 31 05:32:16 mdp: but I see code all over that doesn't do that. Oct 31 05:32:20 even uses stack memory Oct 31 05:32:25 any thoughts on that? Oct 31 05:32:33 show me Oct 31 05:32:39 which? Oct 31 05:32:44 since I broke my promise and didn't crash yet Oct 31 05:33:04 (I haven't read the backlog yet) Oct 31 05:33:13 well, here's the deal...when you pass a buffer to the spi subsystem...you have no idea whether the backing master driver will use pio or dma Oct 31 05:33:22 so you need to assume dma Oct 31 05:33:38 and buffers on the stack are an absolute no-no Oct 31 05:34:01 IIRC, the SPI driver looks at xfer size to decide Oct 31 05:34:04 hmm.... Oct 31 05:34:13 ds2, that's specific to mcspi, yes Oct 31 05:34:24 and you are using mcspi, right? :D Oct 31 05:34:29 I suppose that's an appropriate assumption for this channel Oct 31 05:34:31 I hope so Oct 31 05:34:35 unless you are bitbanging!!!! Oct 31 05:34:42 bitbang, w00t! Oct 31 05:34:46 dammit Oct 31 05:35:03 then you can make all the sloppy buffer allocations you want Oct 31 05:35:15 yeah, since I made this driver with spi-gpio...... Oct 31 05:35:38 alan_o: how are the flames? Oct 31 05:35:39 so I was expecting you to say, "yeah, but all memory is DMA on omap... or something" Oct 31 05:35:53 alan_o, it doesn't work that way Oct 31 05:36:05 it's about how you allocate the memory Oct 31 05:36:25 you have to be compliant with the DMA mapping rules in linux Oct 31 05:36:25 what would be a non DMAable memory on this platform? Oct 31 05:36:39 yeah, I have some of it that's DMA, then I saw code examples that did otherwise and I stopped doing it... Oct 31 05:36:41 unless you are transfering > 1 page Oct 31 05:36:58 ds2, from a kernel context, that would be a buffer allocated from vmalloc space (non-contiguous) Oct 31 05:37:18 mdp: don't see how that matters Oct 31 05:37:21 or a buffer allocated from the stack (char my_xfer_buf[16]) Oct 31 05:37:31 actually.... I didn't convert this to spi-gpio until I got on the new kernels. I originally did this stuff with the TI (arago) 3.2 kernel using the real SPI controller. Not sure if that one uses DMA or just PIO Oct 31 05:37:33 contiguous only matters if you are > 4K Oct 31 05:37:45 and unless your stack is spanning page boundaries.... Oct 31 05:38:14 ds2, so long as we stick to 4k stacks, we'll be ok! Oct 31 05:38:15 ds2, the dma mapping apis are used by the spi master driver...and they are invalid on anything except for kmalloc allocated buffers or dma_* allocated buffers Oct 31 05:38:51 ds2, also, your buffer on the stack...when the backing invalidate/clean executes...it can corrupt adjacent stack variables Oct 31 05:39:00 Russ: didn't we have this discussion like 5 years ago? :) Oct 31 05:39:05 that's why allocation from the stack is not allowed Oct 31 05:39:21 the old thing is the new thing Oct 31 05:39:53 ds2, that was a problem back in like 2002 era and has mostly been wiped out a long time ago Oct 31 05:40:21 that is not a DMA issue. it is a caching/page property thing Oct 31 05:40:32 so anyway, the point is, those dma mapping apis only work on certain memory allocation types Oct 31 05:40:59 ds2, that's why I told him it's not a "dma-able memory thing" Oct 31 05:41:11 at a high-level that's what it is Oct 31 05:41:21 Okay Oct 31 05:41:24 makes more sense now Oct 31 05:41:26 because the operative apis fall under dma-mapping Oct 31 05:41:39 usual DMA issues are more to do with the way the underlying HW is wired up Oct 31 05:41:48 and strictly speaking, there isn't that issue here Oct 31 05:42:04 the only way all these driver work across all these architectures is that people follow the dma-mapping rules in drivers Oct 31 05:42:17 the backend to those apis get very complicated when iommus are in play Oct 31 05:42:29 yes they do Oct 31 05:42:46 but this is more of a sw issue then a memory thing Oct 31 05:42:51 ds2, there is an issue if he allocates memory from the stack Oct 31 05:42:57 yes, a sw issue Oct 31 05:43:25 I suspect it can be made moot by clever kernel configuration Oct 31 05:43:32 I think maybe the confusion came because people often thing there might be something like the old PC ISA GFP_DMA on other architectures Oct 31 05:43:36 nevertheless, I see what you are saying Oct 31 05:43:38 s/thing/think/ Oct 31 05:44:01 PPCs and PCI buses have other DMA issues Oct 31 05:44:11 some memory is just not directly visible Oct 31 05:44:29 yes, I live that world for a long time ;) Oct 31 05:44:43 before ARM had real processors Oct 31 05:44:45 just saying.. that is a HW reason Oct 31 05:45:17 PPC (classics, at least) also had snooping where they could be lazy like x86 Oct 31 05:45:19 hmm... I'm passing in a buffer that I get passed in an skb.... Oct 31 05:45:27 I'm pretty sure that's not been allocated with DMA Oct 31 05:45:44 which is one reason it took the dma mapping api a while to get fixed up for everybody Oct 31 05:45:44 I really have to memcpy that? Oct 31 05:45:52 as only low-end processors didn't have snooping Oct 31 05:45:56 doesn't seem right..... Oct 31 05:46:11 skbs are ok Oct 31 05:46:32 so MIPS are low end processors? ;) Oct 31 05:46:41 yes ;) Oct 31 05:47:00 just telling you what the perception and push-back was historically Oct 31 05:47:09 *nod* Oct 31 05:47:15 it was horrid to get any traction to have this stuff sane back then Oct 31 05:47:42 the 20lb hammer solution to all this is to turn off the cache Oct 31 05:48:17 alan_o, I highly recommend reading Documentation/DMA-API.txt Oct 31 05:48:28 ok, I'll check it out Oct 31 05:48:32 then all you are left with is address space reachability issues Oct 31 05:49:00 so you said skb's are ok... but even an skb allocated with just gfp_kernel? Oct 31 05:49:45 Hey, just found this comment: /* TODO: Is DMA really required for SPI? */ Oct 31 05:49:58 DMA is never required Oct 31 05:50:00 yes, ignore the gfp flags for now, unimportant Oct 31 05:50:18 right, it's always an option Oct 31 05:50:30 unless you are using omap_hsmmc.c Oct 31 05:50:38 then it's absolutely required..due to driver design Oct 31 05:50:43 wait... what>? Oct 31 05:51:08 isn't that in the same category as musb*.c as far as talking about it? :D Oct 31 05:51:43 alan_o, that's somebody's comment in a driver where they have a very tiny perspective on how spi is used broadly Oct 31 05:51:53 must be morning i see _av500_ posts Oct 31 05:52:10 alan_o, it's required in many use cases, of course Oct 31 05:52:20 That's _my_ comment :) Oct 31 05:52:29 there you go Oct 31 05:52:50 exactly, that's why it has a TODO. So you just said yes, ignore the gfp flags for now, unimportant Oct 31 05:53:09 <_av500_> yo Oct 31 05:53:13 I told you that because your gfp_kernel flag has nothing to do with it Oct 31 05:53:47 ok yeah, I mean an skb that had _only_ gfp_kernel (and not DMA) Oct 31 05:53:52 alan_o, allocate the skb (it uses kmalloc, and therefore is legal) Oct 31 05:53:57 ok Oct 31 05:54:09 alan_o, gfp_dma means nothing in our world Oct 31 05:54:34 oh... and that's the stuff that's in DMA-SPI.txt then I assume Oct 31 05:54:50 alan_o, keep in mind the driver will make a decision based on xfer length as to whether it will bother with setting up a dma transfer Oct 31 05:55:15 DMA-API.txt Oct 31 05:55:16 yes Oct 31 05:55:29 <_av500_> mranostay: ill stop posting to confuse you Oct 31 05:55:47 read about streaming mappings, it mentions the restrictions on the memory Oct 31 05:55:49 my stack variables for SPI are only done with very small (2 bytes) buffers. Oct 31 05:55:55 no way that's going through DMA Oct 31 05:56:03 you can make that assumption yes. Oct 31 05:56:05 below the magic threshold (160-something iirc) Oct 31 05:56:20 I can tell you that it's non-portable though Oct 31 05:56:29 go to another platform and that driver will explode Oct 31 05:56:32 *boom* Oct 31 05:56:33 yes, definitely. Oct 31 05:56:39 I'll fix it. Oct 31 05:56:42 rock you like sandy the hurricane Oct 31 05:56:49 but probably not tonight :) Oct 31 05:57:22 so if that's not my problem then.... I still don't know why I drop so many packets... :( Oct 31 05:57:32 through some pointer in your private structure and devm kzalloc away...or equiv Oct 31 05:57:48 yeah, I actually do that. Oct 31 05:57:49 _av500_: not cool dude Oct 31 05:57:52 well, with that, alan_o, I must go to slumber ;) Oct 31 05:57:53 cya Oct 31 05:58:00 later, thanks Oct 31 05:59:07 I leave you in the capable hands of the master troll, av500 Oct 31 05:59:15 hehe Oct 31 05:59:22 mru may take exception to that :) Oct 31 06:01:31 I have Windows 7 in 64 bit machine + I have installed virtualbox from oracle : is this machine setup okay for beagle board setup Oct 31 06:01:49 alan_o: be careful you know we have to deal with these trolls next week :) Oct 31 06:02:28 just follow the vendor precedent Oct 31 06:02:36 put your drivers in arch/arm/mach-omap2 and be done with Oct 31 06:02:37 :D Oct 31 06:02:59 yeah lets now Oct 31 06:03:02 *not Oct 31 06:03:11 I am worried because some time back I tried using NXP ARM9 but the setup was meant only for windows XP Oct 31 06:03:23 surely, the silicon makers can't be wrong Oct 31 06:03:28 so everything went as failure Oct 31 06:03:36 embedded_deba: what? Oct 31 06:04:12 the screen size/resolution problem seems to follow beagleboard. I tried another hdtv and the screen edge is not lining up. I tried changing the boot arg in sd card from hd720 to 1680x1050MR-16@60 and screen did not work at all. Going back to hd720 the screen is back but size issue is back too Oct 31 06:04:22 embedded_deba: nxp arm9? The beagle boards are neither of those Oct 31 06:04:27 repeat ad inf - Thou shalt not run windows. Oct 31 06:04:36 yeah, don't run windows Oct 31 06:04:50 @alan : I am intending to buy a beagle board from an authorized vendor : but they said they does not help with the setup or the software Oct 31 06:05:03 It's a linux board, use Linux to work with it. You'll be happy you did. Spend 30 minutes installing Ubuntu and be done. Everything will be easier. Oct 31 06:05:07 @alan : I know Oct 31 06:05:50 Ubuntu in virtual box - does it work? or I need to install i nthe main hardware itself? Oct 31 06:06:07 bulbage: Beagleboard won't do full HD at 60Hz Oct 31 06:06:09 @alan : Ubuntu in virtual box - does it work? or I need to install i nthe main hardware itself? Oct 31 06:06:23 dispose of windows entirely Oct 31 06:06:30 ^^ what he said Oct 31 06:06:34 if you must use it, run that in a VM and firewall it liberally Oct 31 06:06:49 hehe, that's what I do.. man what a pain. Oct 31 06:06:51 alan_o: thank you for your reply. what should i set it to? Oct 31 06:06:57 30Hz? Oct 31 06:07:00 24Hz? Oct 31 06:07:12 I think it will do 1920x1080 at 30Hz.... Oct 31 06:07:15 It's about the pixel clock. Oct 31 06:07:26 I saw a mailing list post from mru about it some time ago Oct 31 06:08:21 so the Hz setting will eliminate the screen not coming back? but I should keep changing the resolution until I get the screen edges aligned? Oct 31 06:08:46 you need to make sure the high speed PLL is enabled to get a suitable pclk Oct 31 06:08:46 bulbage: I don't know other than mess-with-it-until-it-works Oct 31 06:08:57 stick a scope in it and measure the numbers Oct 31 06:09:06 otherwise, it gets painful Oct 31 06:09:09 bulbage: what I do know is that certain screens refuse to do certain resolutions. Oct 31 06:09:22 ds2: I remember it being painful even so :) Oct 31 06:09:32 ok Oct 31 06:09:48 thank you alan_o. I will play with it. Oct 31 06:14:06 @alan_o : I am still not clear Oct 31 06:14:28 embedded_deba: exact steps follow...... Oct 31 06:14:33 1. get a box Oct 31 06:14:36 2. install Ubuntu on it Oct 31 06:14:40 @alan_o : ok Oct 31 06:14:45 :) Oct 31 06:15:12 @alan_o : by box you mean a virtual box or a real machine? Oct 31 06:15:23 asldhglsadhg;lsadkgh;lahsg Oct 31 06:15:30 you really want to do this VM thing...... Oct 31 06:15:48 @alan_o : I want to know if its possible Oct 31 06:15:52 It could be done. In fact, no reason it couldn't..... Oct 31 06:16:04 What are you trying to do anyway? Oct 31 06:16:19 but.... Oct 31 06:16:29 when youre trying to format your memory card, etc..... Oct 31 06:16:34 @alan_o : I have Windows 7 64 bit already installed Oct 31 06:17:00 and can't get it to attach to your VM.... you probably won't get much sympathy in this channel, if you know what I mean. Oct 31 06:17:12 @alan_o : ok Oct 31 06:18:51 @alan_o : but once the build is ready as a file it can be flashed from Windows 7 ..am I right? Oct 31 06:19:16 @alan_o : is it ok..the right wawy? Oct 31 06:19:44 I'm not sure if there are exact steps online for making up an SD card from Windows.. Maybe there are Oct 31 06:20:14 @alan_o : I see..but thats not the standard procedure I guess Oct 31 06:20:37 most people who do embedded Linux do it from Linux Oct 31 06:21:50 alan_o: heh Oct 31 06:21:52 @alan_o : thanks alan..let me see any work around..if none I have to follow the standard way Oct 31 06:23:33 @mranostay : ?? Oct 31 06:25:42 alan_o: I tried a few settings none seem to work. I will continue to play with it but want to make sure I have the right command. In boot->uEnv.txt, I changed dvimode="hd720 omapfb.vram=0:8M,1:4M,2:4M" to dvimode="1280x1024MR-16@24 omapfb.vram=0:8M,1:4M,2:4M" Oct 31 06:27:26 bulbage: I've told you everything I know, unfortunately.... I played with those settings like a year ago, and got something to work. I know it will take the hd720 as a shortcut, but I don't remember what all the MR and whatnot means off hand Oct 31 06:29:01 so here's one for the WTF folder...... if I just send packets to my device (mrf24j40 802.15.4 module on bone), a lot of them get dropped. But... if on my bone, I run cat /dev/zero and run the CPU up to 100, I don't lose any. Oct 31 06:29:49 seriously, lost 1 out of 6573 Oct 31 06:29:52 ok. You said some mailing list post might have some possible answers. I refreshed my webchat and lost the history. :( Oct 31 06:29:58 can you repost? Oct 31 06:30:12 the name so I can perhaps google it Oct 31 06:30:13 uhh... look for Mans Rullgard pixel clock beagle board. Oct 31 06:31:39 define send Oct 31 06:31:53 send as in ping flood or one packet per unit of time or ? Oct 31 06:31:59 and what size? Oct 31 06:32:07 ping -i 0.3 Oct 31 06:32:18 and actually..... this is strange Oct 31 06:32:30 what's ping -i 1 do? Oct 31 06:32:55 cat /dev/zero doesn't even show up high on the cpu list.... cpu shows at 0% Oct 31 06:33:04 ping -i is interval Oct 31 06:33:07 oh, sorry Oct 31 06:33:10 ping -i 0.03 Oct 31 06:33:40 i know Oct 31 06:33:45 sorry :) Oct 31 06:33:49 I am saying, what does it do if you bump up the time Oct 31 06:33:55 ping -i 1 gets most all of them Oct 31 06:34:03 falls apart around 0.3 Oct 31 06:34:04 Hmmmmm Oct 31 06:34:12 0.3 or 0.03? Oct 31 06:34:13 I think 0.2 it loses 20% Oct 31 06:34:22 okay, do this - Oct 31 06:34:32 so without this hack, it comes apart at 0.2 or so Oct 31 06:34:40 ping -i 0.03 .... > /dev/null & Oct 31 06:34:45 but as long as I'm catting /dev/zero, I can Oct 31 06:34:55 then look for the entry in the output of netstat Oct 31 06:35:31 are the numbers in the 2nd and 3rd column relatively large? Oct 31 06:35:40 interrupt...... Oct 31 06:35:42 Recv-Q/Send-Q Oct 31 06:35:43 shared interrupt.. Oct 31 06:35:48 I bet I'm on a shared interrupt Oct 31 06:36:06 ok, stand by, checking netstat.... Oct 31 06:38:56 gar, now I seem to have lost everything..... Oct 31 06:39:40 are you leaking skbs? Oct 31 06:39:56 I hope not :) Oct 31 06:41:42 I think something got hosed on the PC side Oct 31 06:45:19 oh... had to reset the hardware (PC side: Econotag) Oct 31 06:47:03 ugghh... something is holding the serial port open Oct 31 06:47:19 ftdi = arrrrrgh Oct 31 06:47:44 http://www.redwirellc.com/store/node/1 Oct 31 06:48:28 It's a cool board. I just wish Freescale made one that had USB and 802.15.4 on the same chip. That'd be sweet.. until then, we're stuck with ftdi, which is great right up until it isn't. Oct 31 06:49:42 ok rebooting Oct 31 06:49:47 brb Oct 31 06:53:18 only at 3am does it want to fsck on bootup... never on Sunday afternoon. Oct 31 06:58:35 so ds2, what do you want me to look at? Oct 31 06:58:44 is ICMP even in netstat? Oct 31 06:58:59 should be Oct 31 06:59:05 and here's another one.... NOHZ: local_softirq_pending 08 Oct 31 06:59:16 I seem to remember looking that up one time.... Oct 31 06:59:17 look at the send and recv queues when you are pinging and seeing drops Oct 31 06:59:24 See I'm thinking this is interrupt related Oct 31 07:00:03 and that somehow running cat /dev/zero is pushing on something that's causing my interrupts to actually get delivered, where it's being clogged up otherwise Oct 31 07:00:24 that cat would cause a lot of context switches Oct 31 07:00:32 read(), write(), read(), write.... Oct 31 07:00:43 miaow Oct 31 07:00:53 could... but for whatever reason it's not showing any actual CPU usage Oct 31 07:00:59 (the CPU LED comes on though) Oct 31 07:01:04 but it's not in top Oct 31 07:01:37 Even at ping -i 1, without the cat running, ping times are ~200 Oct 31 07:01:43 with cat running , drops to 20 Oct 31 07:01:59 Hmmmmmmmmm Oct 31 07:02:01 (and 20 is the theoretical min because I have a sleep in the driver on the PC side) Oct 31 07:02:14 tickless enabled? Oct 31 07:02:47 I can check... that one message I pasted says NOHZ Oct 31 07:03:19 see if disabling it works around the problem Oct 31 07:03:37 NO_HZ = y Oct 31 07:03:49 you can disable it on the commandline Oct 31 07:03:59 forgot what it is offhand Oct 31 07:04:00 bootup command line? Oct 31 07:04:03 I'll look it up Oct 31 07:08:20 I can't find it... I'll just rebuild it Oct 31 07:09:17 nohz=off Oct 31 07:12:53 works Oct 31 07:13:03 Hmmmmm Oct 31 07:13:16 * alan_o bows to the master :) Oct 31 07:13:21 you must not be setting something right to let the kernel know it needs to stuff Oct 31 07:13:41 yeah, the why is certainly interesting..... Oct 31 07:13:51 I'm registering for an interrupt Oct 31 07:14:01 I get the number from gpio_to_irq() Oct 31 07:14:11 now... I'm not using device tree Oct 31 07:14:31 I mean there _is_ a device tree, but I load my driver from arch/arm/mach-omap2/board-alan.c Oct 31 07:14:42 (in the spirit of vendor code... :) ) Oct 31 07:15:04 I followed what the other drivers did with the 3.2 TI bone kernel Oct 31 07:15:19 not drivers, I mean what the other hardware config sections did Oct 31 07:15:55 * alan_o wonders whether he can take out the sleep in the pc-side driver and get some real performance..... Oct 31 07:16:33 What should tickless have to do with the interrupt controller though Oct 31 07:17:56 http://www.indiegogo.com/%20open-amlogic-box Oct 31 07:18:03 so these days we "buy" GPL code Oct 31 07:18:21 tickless uses certain hueristics to determine how long to sleep the kernel Oct 31 07:18:28 dont tell the chinese, they might make a business out of it Oct 31 07:18:42 av500: I had a guy call me this week wanting to buy rights to HIDAPI, which is BSD Oct 31 07:18:50 mabye it was last week... Oct 31 07:19:10 I told him I can't sell it because I don't own it all. Oct 31 07:19:19 you did wrong Oct 31 07:19:23 put it on kickstarter Oct 31 07:19:48 He wanted to not have to have a BSD license file hanging around Oct 31 07:19:56 because he wants to sell his company Oct 31 07:20:10 oh Oct 31 07:20:11 and convincing lawyers that it's all up-and-up is expensive Oct 31 07:20:49 ds2: but even if it sleeps.... the GPIO is a hard interrupt line Oct 31 07:21:16 well.... we hope, unless that functionality isn't implemented Oct 31 07:21:30 I mean implemented in kernel Oct 31 07:22:01 but your "interrupt" is not really running in a true interrupt context Oct 31 07:22:16 the real one runs, it queues, then when it gets scheduled, you run Oct 31 07:22:20 but it should be Oct 31 07:22:24 oh Oct 31 07:22:25 really? Oct 31 07:22:29 not if you are doing spi Oct 31 07:22:43 SPI shouldn't enter into it Oct 31 07:22:48 threaded/top bottom half stuff. Oct 31 07:23:00 you accessing SPI directly from interrupt or you using a work queue? Oct 31 07:23:02 That's the workqueue though Oct 31 07:23:08 yeah, workqueue Oct 31 07:23:26 oh I see what you mean... the isr runs, workqueue doesn't get called right away Oct 31 07:23:31 but Oct 31 07:23:47 I was still seeing 1:1 isr and workqueue and didn't see a lot of delay between them Oct 31 07:24:37 so if an isr runs... calls up a workqueue, that makes another task runnable, right? so before it goes back to sleep it should know that it should run the workqueue task Oct 31 07:25:57 IIRC, tickless would delay it til the timer fires Oct 31 07:26:33 but that's busted..... why would it sleep when there are runnable tasks? Oct 31 07:27:14 it comes down to the scheduler Oct 31 07:27:59 The way the kernel is designed, and I mean by the way they tell us to write interrupt handlers, and interrupt handler firing is nearly _always_ going to mean there's a task becoming runnable, right? Oct 31 07:28:36 not really Oct 31 07:28:45 gpio-keys w/o a process sitting on it Oct 31 07:32:50 i am out Oct 31 07:32:57 ok man Oct 31 07:33:03 thanks for your help today. Oct 31 07:33:20 I think I'm finally able to start working on the application for this demo :) Oct 31 07:33:28 as soon as I finish up my slides. Oct 31 07:33:30 later Oct 31 09:15:30 arf Oct 31 09:15:59 av500: not ard? Oct 31 09:16:27 * av500 glares at unitymedia Oct 31 09:16:41 * LetoThe2nd doesn't think that will help. Oct 31 09:16:54 I switched routes to VDSL Oct 31 09:17:30 <_av500_> or not Oct 31 09:20:27 having both helps Oct 31 09:26:41 <__av500__> hmm, this was the wrong DSL Oct 31 09:26:43 <__av500__> -ETOOMANY Oct 31 09:29:11 gm av500 Oct 31 09:29:15 gm Oct 31 09:29:28 * LetoThe2nd thinks av500 must be looking like http://allcommsolutions.com/images/cable_confusion.jpg Oct 31 09:30:07 yes Oct 31 09:30:13 we have 2 dsl and 1 cable Oct 31 09:31:34 at home? Oct 31 09:31:57 <_av500_> work Oct 31 09:32:06 the admin is ill? Oct 31 09:32:13 unitymedia is ill Oct 31 09:32:45 why? Oct 31 09:32:54 fix the bonding Oct 31 10:23:00 av500: for your next #caturday http://funnie.biz/58648 Oct 31 10:24:57 ha, but I dont do #caturday Oct 31 10:42:48 av500: just as well as you don't do animated gifs Oct 31 10:42:55 right Oct 31 10:43:27 grah, now I might fly from BCN to ORY Oct 31 10:43:58 av500: ORYgano? Oct 31 10:44:07 orl Oct 31 10:44:09 orly Oct 31 12:24:16 prpplague: when are we supposed to send in our slides for elc-e? Oct 31 13:10:07 hi vipcount... Oct 31 13:10:18 hello Oct 31 13:10:23 I try not to answer support questions individually... Oct 31 13:10:26 the os is Ubuntu 11.10 minimal , and i cannot use the audo device Oct 31 13:10:28 it would eat up all my time. Oct 31 13:10:46 you are trying to install Ubuntu 11.10 minimal on BeagleBoard-xM and the audio device isn't working? Oct 31 13:10:56 which kernel are you using? Oct 31 13:11:08 1 moment Oct 31 13:11:24 3.1-XXX Oct 31 13:11:27 For your audio system, are you trying ALSA directly, or are you using phonon or any of that sort of crazy stuff? Oct 31 13:11:32 :) Oct 31 13:11:43 i download from http://www.brianhensley.net/2012/03/beagleboard-xm-ubuntu-1110-with-gui.html Oct 31 13:11:44 vipcount: where did you get the kernel from? Oct 31 13:12:20 i need to use audio input and video to stream Oct 31 13:12:38 and what does not work? Oct 31 13:12:47 what have you connected? Oct 31 13:13:15 no audio device Oct 31 13:13:26 i use this image : http://rcn-ee.net/deb/rootfs/oneiric/ubuntu-11.10-r14-minimal-armel-2012-09-27.tar.xz Oct 31 13:13:33 right... Oct 31 13:13:45 looks like a prebuilt kernel from rcn-ee... Oct 31 13:13:54 but, going back to av500's question, what is actually failing? Oct 31 13:14:16 what inputs have you tried with what applications? Oct 31 13:14:35 for example, feeding in data from you iPod and capturing with arecord? Oct 31 13:14:40 alsamixer not found audio device Oct 31 13:15:00 yes i use ffserver to stream vide and audio Oct 31 13:15:38 do you can reccomend beagleboard linux os Oct 31 13:15:46 i test debian opensuse ... Oct 31 13:16:05 but all OS no audio device Oct 31 13:16:33 vipcount: I use Angstrom the most and would recommend the latest image at http://www.angstrom-distribution.org/demo/beagleboard Oct 31 13:16:45 if you like Ubuntu and know Ubuntu, I'd recommend you use that. Oct 31 13:17:07 latest Ubuntu should be at http://elinux.org/BeagleBoardUbuntu (http://beagleboard.org/project/ubuntu) Oct 31 13:17:38 if you don't mind moving up to the latest image, there is probably better support there. Oct 31 13:17:54 ok :) thanks Oct 31 13:17:58 but, it is odd that you aren't seeing an audio device in any of the distros you tried... Oct 31 13:18:14 which makes me think there might be some other type of issue, such as how you are using alsamixer. Oct 31 13:18:21 are you seeing /dev/snd/... ? Oct 31 13:18:32 yes Oct 31 13:19:00 if you have /dev/snd/pcmXXXXc, then I suspect you have a working audio driver. Oct 31 13:19:01 in /dev/snd dont have anyting Oct 31 13:19:13 so, just /dev/snd, but nothing in it? Oct 31 13:19:19 yes Oct 31 13:19:24 just timer Oct 31 13:19:42 what does dmesg tell you? Oct 31 13:20:02 1 moment i copy form terminal Oct 31 13:20:19 be sure to use pastebin Oct 31 13:22:22 the dmesg can see into a.vpcount.com/dmesg.txt Oct 31 13:26:53 That's a much newer kernel! Oct 31 13:27:24 i not download newer kernel Oct 31 13:27:54 what you sent is output from a 3.6.2 kernel... Oct 31 13:27:57 perhaps rcn-ee updated it. Oct 31 13:28:06 but i downlaod angstrom liux 2 days back and not audio device Oct 31 13:28:27 alsamixer say annot open mixer: No such file or directory Oct 31 13:28:52 in /dev/snd have only seq timer Oct 31 13:29:35 I'd guess it is probably coming off of this development tree: https://github.com/beagleboard/kernel/tree/beagleboard-3.6 Oct 31 13:29:52 there might have been some audio regressions to inform rcn-ee. Oct 31 13:30:31 ok Oct 31 13:32:16 looks like rcn-ee is updating his kernel images: http://elinux.org/BeagleBoardUbuntu#Precise_12.04_armhf Oct 31 13:32:45 probably best to submit your dmesg log and info of failure to discuss@beagleboard.org, since rcn-ee isn't logged in. Oct 31 13:32:51 now i patch bb ubuntu Oct 31 13:36:00 I suspect you could move back to https://github.com/beagleboard/kernel/tree/beagleboard-3.2, but if you can help rcn-ee fix 3.6, that would be better. Otherwise, you'd need to go back to a fully tested older image. Oct 31 13:37:43 I believe http://www.alwaysinnovating.com/beagleboard/ had working audio capture. Oct 31 14:06:15 Hello every one Oct 31 14:06:47 Hello! Oct 31 14:07:44 i am using a beagleboard with a canbus cape and i'm trying to make it communicate with an armadeus board via the can bus. Oct 31 14:07:54 hey, that sounds fun Oct 31 14:08:30 however i read online that there were a bug with the canbus cape's buffer that blocked outgoing signals Oct 31 14:08:32 terminate properly? Oct 31 14:08:44 hmm Oct 31 14:08:47 sounds unfortunate Oct 31 14:09:01 that's what the doc says Oct 31 14:09:07 what doc? Oct 31 14:09:08 where? Oct 31 14:09:19 on the beagleboartoys wiki Oct 31 14:09:23 So the Tower cape has an mcp2515 on it, but theres' also on-board CAN on the beagle bone Oct 31 14:09:37 YOME: link please! Oct 31 14:09:48 so YOME have you tried _both_ ? Oct 31 14:09:53 onboard and 2515? Oct 31 14:10:10 the Tower cape routes both to connectors iirc Oct 31 14:10:33 (I did a project with the 2515 on another ARM board, and it does work good) Oct 31 14:10:35 here is the link http://beagleboardtoys.com/wiki/index.php?title=BeagleBone_CANBus Oct 31 14:10:59 i used the beagleboard alone to read frames from a can bus and it works great Oct 31 14:11:18 ok, cool, so what your'e saying is that you're done then. Oct 31 14:11:22 and that it works Oct 31 14:11:23 case closed Oct 31 14:11:26 no Oct 31 14:12:08 i'm trying to send can frames from an armadeus board to a beagleboard Oct 31 14:12:23 ok Oct 31 14:12:37 do you have an analyzer Oct 31 14:12:38 but i can't achieve this so first i tested the beagleboard alone on an existing working can bus Oct 31 14:12:53 beagle on existing CAN netwokr. Oct 31 14:12:57 works? Oct 31 14:12:57 yeah Oct 31 14:13:02 send/receive? Oct 31 14:13:10 works in receive Oct 31 14:13:15 not in send? Oct 31 14:13:20 i didn't test sending Oct 31 14:13:29 So what doesn't work? Oct 31 14:13:34 rx from Armadeus ? Oct 31 14:13:35 (and i starting to think that i should have) Oct 31 14:14:01 i can't receive from the armadeus when these two boards are connected Oct 31 14:14:23 ok..... so can the armadeus send/receive to the existing CAN network/ Oct 31 14:14:29 but you said you can receive Oct 31 14:14:33 yes it can Oct 31 14:14:37 friday Oct 31 14:15:23 to be more precise when i trys to send from the armadeus to the beagle board they are the only two devices on the network Oct 31 14:15:37 ok Oct 31 14:15:52 Sound like you want 4 things to work... RX/TX from both boards Oct 31 14:15:55 right? Oct 31 14:15:59 you have an existing network.... Oct 31 14:16:01 yes Oct 31 14:16:11 test each separately with the existing network Oct 31 14:16:14 figure out which doesn't work Oct 31 14:16:18 and then fix that one Oct 31 14:16:35 I'm not sure what you're looking for here...... Oct 31 14:16:59 have each one tx and rx with the analyzer too Oct 31 14:17:38 incremental steps... write the result of each in your engineering notebook. Make sure to log the exact steps you did for each test. Oct 31 14:18:25 well i'll try tx from the beagleboard (the only thing i didn't test) Oct 31 14:18:51 So the Armadeus can talk to the existing network? Oct 31 14:18:57 yes it can Oct 31 14:19:09 but not to the bone Oct 31 14:19:27 check your network parameters, particularly your oscillator Oct 31 14:19:28 yes exactly Oct 31 14:19:52 SocketCAN will silently just give you out-of-spec timings if your oscillator can't match what you want. Oct 31 14:20:15 I shouldn't say SocketCAN, but at least SOME drivers (like the 2515 for example) will. Oct 31 14:20:48 what do you mean by your oscillator (i'm not a hardware person) Oct 31 14:20:49 Get that AN754 document and figure out what the timings should be in the registers for your network timings. Oct 31 14:21:11 Theres' a crystal attached to the 2515 (and of coruse to other CAN controllers as well) Oct 31 14:21:17 timing is based off that crystal Oct 31 14:21:31 if you have a 4MHz crystal, you can't generate 325ns, for example. Oct 31 14:21:46 (can only generate multiples of 500) Oct 31 14:21:53 hmm ok Oct 31 14:22:13 (It uses 2x cycles) Oct 31 14:22:22 Check that on all parts. Oct 31 14:22:29 does that have something to do with the bitrate Oct 31 14:22:31 also try _BOTH_ CAN interfaces on the Tower cape Oct 31 14:22:37 Yes :) Oct 31 14:22:42 That _is_ the bit rate Oct 31 14:22:47 there's more than bit rate Oct 31 14:23:03 Read AN754 from Microchip Oct 31 14:23:12 It's more complicated than you currently think to set up the bit timings Oct 31 14:23:16 but obviously the bitrate is ok since i configure it manually Oct 31 14:23:53 It's not just a bit rate Oct 31 14:23:58 there are a few parameters Oct 31 14:24:04 bit timing are computed by the linux kernel Oct 31 14:24:10 PS1, PS2, TQ, SJW, and the one at the beginning..... Oct 31 14:24:25 you have to set them when bringing up your socketCAN interface Oct 31 14:24:30 they're passed by the driver to the hardware Oct 31 14:24:53 (in your case a 2515 on the TowerCAN cape, and to the internal CAN controller on the bone) Oct 31 14:25:23 yeah but my kernel is configured to compute them automatically but i see if they're right Oct 31 14:25:34 http://www.kernel.org/doc/Documentation/networking/can.txt Oct 31 14:25:54 6.5.2 Setting the can bit-timing Oct 31 14:26:11 tq needs to be something that will actually work with the hardware/oscillator Oct 31 14:27:03 it's "automatic" but it needs to be something that's compatible, or else you'll get something that's close, and close isn't good enough. Oct 31 14:27:43 ok bb in a few.... Oct 31 14:28:09 but why would that be a problem only when the two boards are connected together Oct 31 14:28:16 good to know alan_o knows about canbus Oct 31 14:42:52 woglinde: hehe.... I know some... I'm not an expert. I've only done raw CAN for an industrial application, but not any of the higher layer protocols (3 and up) used in automotive. Oct 31 14:43:36 in automotive, so I understand, they tell you exactly which software stack to use, because it's all safety-critical. Oct 31 14:43:50 woglinde: alan_o can can Oct 31 14:44:03 alan_o: yes Oct 31 14:44:16 I guess they also tell you what compiler version to use Oct 31 14:44:53 yeah, probably Oct 31 14:45:01 hm not so good Oct 31 14:45:20 yes good Oct 31 14:45:31 yes, good Oct 31 14:45:32 compiler version Oct 31 14:45:40 we will see Oct 31 14:45:52 its some time till I will get involved Oct 31 14:45:53 otherwise car crashes and burns Oct 31 14:45:56 with it Oct 31 14:46:16 av500 thats because of shitty accu Oct 31 14:46:30 thats US e-car Oct 31 14:46:35 here we drive real cars Oct 31 14:46:43 because we CAN! Oct 31 14:46:52 That's right... but BMW has their own bus Oct 31 14:47:01 ByteFlight Oct 31 14:47:15 but WV microbus is much better! Oct 31 14:47:18 er, VW Oct 31 14:47:22 hehe Oct 31 14:48:03 Then there's this bus.. http://www.linuxjournal.com/files/linuxjournal.com/linuxjournal/issues/130/cover130.small.jpg Oct 31 14:48:21 argh Oct 31 14:48:29 no rickrolling on this channel please! Oct 31 14:48:50 ??? don't like D-Bus? Oct 31 14:49:09 * av500 does not like anything invented in the past 10 years :) Oct 31 14:49:29 fair enough... I guess you're not a systemd guy then :) Oct 31 14:49:31 micro-bus from VW Oct 31 14:49:59 alway see those in flames on the roadside here. too hot for them I guess. Oct 31 14:50:04 I love the systemd documentation.. "all you have to do is..." and then 70 or so characters of nonsense to type Oct 31 14:50:19 "it's easy!" Oct 31 14:50:24 140 chars is the new limit :) Oct 31 14:50:37 yeah, man.... sysvinit is too easy Oct 31 14:50:45 too easily deterministic..... Oct 31 14:51:13 It uses shell, therefore it's bad, because we all know that shell is not webscale. Oct 31 14:51:23 determinism is for small minds. Oct 31 14:51:40 "remember sequential code.. you know code you can read?" Oct 31 14:52:03 mdp: I'm still laughing out loud at those videos..... Oct 31 14:52:28 l? Oct 31 14:52:32 http://www.youtube.com/watch?v=bzkRVzciAZg Oct 31 14:53:02 I think that's the one you sent. Maybe that's one I got from "related" to one you sent Oct 31 14:53:08 yes ;) Oct 31 14:54:05 alan_o, "this is where I typically stab myself repeatedly..." Oct 31 14:54:16 yeah..... Oct 31 14:54:29 "All the complexities of assembler with the speed of javascript" Oct 31 14:54:57 http://26-26-54.hardwarebug.org/128 Oct 31 14:55:13 [15:53:28] could someone please point me in the right direction for web-application-framework development ? Oct 31 14:55:14 hah Oct 31 14:55:22 mdp: your helpdesk is needed Oct 31 14:55:40 maybe I should propose node.js? Oct 31 14:55:41 reminds me that I haven't seen the troll today Oct 31 14:55:53 av500, bad ass rockstar technology! Oct 31 14:55:58 checked under the bridges? Oct 31 14:56:13 mdp: do you know the "Brogrammer" video? Oct 31 14:56:16 I just need to know what t-shirt the troll is wearing Oct 31 14:56:17 checked in from under the bridge? Oct 31 14:56:19 alan_o, no! Oct 31 14:56:31 mdp: that's what I'm wondering Oct 31 14:56:51 I need to bring my circa-linuxexpo timeframe LinuxCare coffee cup to ELC-E Oct 31 14:57:09 * mdp becomes nostalgic for simpler times Oct 31 14:57:28 * av500 will bring his slackware floppies Oct 31 14:57:49 and, why exactly must we have pipelines and cache on these damned processor cores? Oct 31 14:58:05 * dm8tbr would bring his Amiga Workbench 1.3.2 floppy... Oct 31 14:58:05 inconceivable Oct 31 14:58:20 mdp: we have cache to not overheat the RAM chips Oct 31 14:59:15 av500, don't be one-uped by somebody bringing HJ Lu's boot/root floppies Oct 31 14:59:25 what about them? Oct 31 14:59:50 it's a contest to see who can bring the most antique item to a conference Oct 31 15:00:33 I think that's limited by luggage allowance Oct 31 15:00:41 ;) Oct 31 15:00:57 this is the first time I wished I had save some antiques Oct 31 15:00:57 it would make for a nice #essentialtools photo shoot Oct 31 15:01:21 extra points for having them on 5.25" Oct 31 15:01:33 mdp: I dont think there was a digicam with 5.25" Oct 31 15:01:37 3.5" yes Oct 31 15:01:42 the mavicas Oct 31 15:01:45 sony did that? Oct 31 15:01:50 yeah Oct 31 15:01:51 hehe Oct 31 15:02:03 there was also that kodak thing that stored video stills Oct 31 15:02:04 epic Oct 31 15:02:12 on a rotating VHS disc Oct 31 15:02:35 and yet nothing still compares to a classic polaroid Oct 31 15:02:48 * av500 has ~1000 of them Oct 31 15:02:52 waiting to be scanned Oct 31 15:02:52 mdp: http://www.youtube.com/watch?v=Qi_AAqi0RZM Oct 31 15:03:08 av500, I was thinking of that age tested with the pen and the cassette tape... Oct 31 15:03:13 mdp: there are some great lines. some of his slides are hilarious. Oct 31 15:03:23 mdp: yep Oct 31 15:03:35 and how another one would be a guy waving a snapshot rapidly (motion lines) with a polaroid sitting next to it...explain the relation Oct 31 15:04:02 "I'm at a party and we need to hand that drunken photo around quick!" Oct 31 15:04:07 mdp: fun thing is, I had some many polaroids of people looking at polaroids Oct 31 15:04:12 lol Oct 31 15:04:28 I got tons of EOL film for free Oct 31 15:04:47 used to take snaps at parties a lot Oct 31 15:04:51 they were great if you ran the local school fake ID business Oct 31 15:04:56 important stuff in .us Oct 31 15:04:59 * dm8tbr remembers av500's pola party webpage Oct 31 15:05:06 oh Oct 31 15:05:36 dm8tbr: that was just a small part of it Oct 31 15:05:45 I expected no less Oct 31 15:05:46 I need an intern to scan all my polas Oct 31 15:05:52 yes, the intern idea Oct 31 15:06:02 isnt there an app for that Oct 31 15:06:14 I guess ill build a frame and write some imagemagic Oct 31 15:06:19 and use the DSLR Oct 31 15:06:29 Can mechanical turk help here? ;) Oct 31 15:06:47 * prpplague grumbles at his todo list Oct 31 15:06:52 you just scan them all and then have somebody scan them for you... Oct 31 15:06:53 if they are willing to travel Oct 31 15:06:54 er, nm Oct 31 15:06:57 :) Oct 31 15:07:05 mdp: scan, print and then scan! Oct 31 15:07:26 sung to the tune of, "Can DT help here?" Oct 31 15:07:40 av500: at TUD you could just get a mechatronik student to build you a scanning rig. Oct 31 15:08:12 just like that one guy built for his Amiga floppy collection. 3.5" autofeeder which included a disk photo Oct 31 15:08:53 that is easy Oct 31 15:08:56 polas are hard Oct 31 15:08:59 they wobble Oct 31 15:09:03 and come in 2 sizes Oct 31 15:11:35 ha: http://linuxgazette.net/issue32/tag_libc5.html Oct 31 15:15:58 ...I've heard complaints that glibc takes up significantly more disk space and run-time core (RAM)... Oct 31 15:17:32 so mdp.... Oct 31 15:17:50 I've been corresponding with a guy about my mrf24j40 driver. I think he's user #1. Oct 31 15:17:59 so he says... well let me paste it Oct 31 15:18:56 http://pastebin.com/JMJ7u2gJ Oct 31 15:19:56 so he's saying that putting a tx_buf and rx_buf on a transfer at the same time won't work with some SPI controllers Oct 31 15:20:05 I told him that I think that's a bug on the SPI controller side Oct 31 15:20:27 because there's lots of examples of full-duplex spi transfers in drivers... Oct 31 15:21:03 TI $70 BOM tablet, how does it compete with a pretty solid 7" A8 tablet sold as $55 single unit from shenzhen? Oct 31 15:21:17 xxiao: url? Oct 31 15:21:21 eetimes Oct 31 15:21:24 wait Oct 31 15:21:41 http://www.eetimes.com/electronics-news/4400230/TI-shows-embedded-tablet-with--70-BoM Oct 31 15:22:20 lol Oct 31 15:22:22 am335x Oct 31 15:22:30 aint that a bone? Oct 31 15:22:41 maybe... Oct 31 15:22:45 tablet is inaccurate Oct 31 15:22:46 its one year to late, china is 1.x GHz A9 now Oct 31 15:22:58 how will that am335x decode hd? Oct 31 15:23:07 again, tablet is inaccurate Oct 31 15:23:26 mdp: shut up, we are trying to be negative here Oct 31 15:23:29 eetimes, despite having EE in the name is that stupid as phoronix by nature of being the press Oct 31 15:23:32 TI is leaving tablet, i.MX is being killed at FSL, it seems all the low to mid-range tablet sockets will be left to A8/A9 from China Oct 31 15:23:55 xxiao: killed, really? Oct 31 15:24:06 right, all the commodity tablets that run mainstream android for consumer general purpose use Oct 31 15:24:27 being killed as rumored, FSL is realinging businesses and I saw no futhre roadmap on i.MX, hope I missed something Oct 31 15:24:49 xxiao: so, that am335x speciality tablet is ok for that purpose I guess Oct 31 15:24:58 xxiao, companies realign away from unprofitable businesses regularly Oct 31 15:25:11 having it from TI and not china might reasure some industrial customers Oct 31 15:25:28 as in long time buy Oct 31 15:25:54 well, on that part, it has the ability to be a control panel UI + industrial process control device Oct 31 15:26:01 that's true, tablet chips from china most do not have industrial temp ranges, it's purely focusing on tablets where volumes rule Oct 31 15:26:07 due to the peripherals for IA on the part Oct 31 15:26:25 the parts from china only have peripherals purpose built for commodity consumer tablets Oct 31 15:26:30 mdp: right Oct 31 15:26:43 but as a control panel, the actualy control would be elsewhere in the machine Oct 31 15:26:52 it depends Oct 31 15:26:58 right Oct 31 15:27:00 always :) Oct 31 15:27:30 the point is, TI/FSL is too slow in consumer chips, they're kind _forced_ into embedded markets where stuff moves much slower, sadly Oct 31 15:27:32 this is why I keep looking at these other cheap boards... Oct 31 15:27:54 and I find myself constantly only coming back to TI/FSL as those who make documented parts with peripherals that are useful for my projects Oct 31 15:28:47 xxiao, it's also inevitable. I want you to count how many competitors for the exact same handset/tablet vertical there are Oct 31 15:29:02 the parts are all almost feature identical Oct 31 15:29:10 true Oct 31 15:29:15 now count the oems that make stuff for these other markets Oct 31 15:29:27 as admiral ackbar once said, "It's a trap!" Oct 31 15:29:57 mdp: I remember talk of a cost down OMAP4, the omap4428 Oct 31 15:30:03 that was fun Oct 31 15:30:07 Nvidia will die in that market and if you know the right people...there's rumors there as well Oct 31 15:31:06 an omap4 with DDR3 could have competed Oct 31 15:31:43 but alas, 'twasn't to be had Oct 31 15:31:55 it's a shame there's some gaps that could be filled easily on "my TI" roadmap Oct 31 15:32:06 Hello again Oct 31 15:32:19 such as a real AM part with ducati...but oh well Oct 31 15:32:37 it's still pretty hard to find arm with 1G eth, when powerpc is doing 2x10G these days, ARM's IO throughput sucks Oct 31 15:32:54 yep, I agree...arm sucks there Oct 31 15:34:07 i just hope ppc/mips do not give up too soon, not before my retiring age... Oct 31 15:34:59 I've seen much faster throughput and pps figures out of a ppc440gp/gx from a long time ago than on the TI network interfaces with a core clocked at half the speed Oct 31 15:35:12 maybe AMD can help here? Oct 31 15:35:39 I just suspect the designers have no experience with network performance Oct 31 15:36:10 cpu is no longer the problem, IO is Oct 31 15:36:13 FSL has an advantage with their move from powerpc->arm as they get it Oct 31 15:36:37 mdp: who need 10GigE on a mobile phone? Oct 31 15:36:46 or a chromebook Oct 31 15:37:22 exactly, big change after investing all your efforts into designing for a market where you have 7 competitors all with the same part and only a couple significant customers for design wins Oct 31 15:37:42 10G is huge for SAN/NAS Oct 31 15:38:22 for media from cellphones i don't even know why we need 11ac, unless we are going to do uncompressed av Oct 31 15:39:05 av500, that's exactly why samsung, nvidia, allwinner, etc. parts are 100% uninteresting in my world Oct 31 15:39:28 and omap4/5/6 for that matter Oct 31 15:40:21 I shouldn't be such a meanie to those parts...they might feel bad Oct 31 15:42:19 * av500 whipers vayu Oct 31 15:42:31 heh Oct 31 15:44:49 koen: in your 3.7 readme when you say "ADC: capebus has ADC support....." what exactly does that mean? Oct 31 15:45:09 koen: I'm not seeing any adc in sysfs, at least not in the same spot it used to be Oct 31 15:54:01 maybe panto is the better one to ask on this one Oct 31 15:56:24 alan_o, what do you want to know? Oct 31 15:57:20 panto: I really just need to read adc from userspace, similar to the way it used to work in sysfs Oct 31 15:57:24 I'm reading the capebus code Oct 31 15:57:51 I'm thinking I'll have to put something in the dt, but I still don't have a full understanding of what I'm looking at yet Oct 31 15:58:06 ok, gimme a minute to look at it Oct 31 16:00:55 hmm, the driver is definitely not going to export in the same way that it used to Oct 31 16:01:00 it's now an IIO ADC driver Oct 31 16:01:11 hmm Oct 31 16:02:08 so I've not used iio before. Is there a dev/ioctl for that stuff? I suppose I can look it up Oct 31 16:02:43 it's very new... drivers/staging/iio/Documentation Oct 31 16:03:05 I definitely used the in-kernel client so I know it works Oct 31 16:03:12 the tsc driver doesn't yet Oct 31 16:04:11 hmm, let me reboot once more to check Oct 31 16:04:50 hmm... I have nothing in /dev/iio* Oct 31 16:04:56 wonder if I need to turn that on Oct 31 16:05:19 well, if you don't have an activated cape node with a ti-tscadc-dt node you won't Oct 31 16:05:34 oh... Oct 31 16:05:45 what tree are you using? Oct 31 16:05:58 https://github.com/beagleboard/kernel/tree/3.7 Oct 31 16:08:47 alan_o, gimme a few minutes Oct 31 16:09:16 ok, no problem. I'm grateful for the help :) Oct 31 16:10:07 you're welcome... guinea pigs^W^Wusers are always welcome :) Oct 31 16:14:34 alan_o, hmm, IIO doesn't put out sysfs node for user-space usage automatically Oct 31 16:14:52 alan_o, good catch; I'll try to find out how to put them out (if possible) Oct 31 16:15:01 I see IIO_SYSFS_TRIGGER in the menuconfig Oct 31 16:15:13 but that's turned off Oct 31 16:15:46 try turning it on Oct 31 16:15:46 also, I'm looking at the iio_event_monitor.c in staging/iio/Documentation Oct 31 16:15:48 yeah Oct 31 16:16:39 but, I'm still going to need something in the dt, right/ Oct 31 16:16:45 ? Oct 31 16:17:06 alan_o, yes Oct 31 16:17:39 tscadc { Oct 31 16:17:39 compatible = "ti-tscadc-dt"; Oct 31 16:17:39 ti,hwmods = "adc_tsc"; Oct 31 16:17:39 adc-channels = <8>; Oct 31 16:17:39 }; Oct 31 16:18:04 ok, I see that text under stuff like the geiger cape Oct 31 16:18:04 something like that in a cape definition Oct 31 16:18:22 iio is kinda rough Oct 31 16:18:25 and can I turn that cape on automatically? Oct 31 16:18:41 without eeprom, etc? Oct 31 16:20:18 yes, look at the override section Oct 31 16:20:35 overrides: no eeprom (proto) Oct 31 16:20:40 yep Oct 31 16:20:43 oh, nice Oct 31 16:20:52 recent capebus kernels can do that on runtime Oct 31 16:21:13 yeah, that's what I've heard/read... that's cool stuff. Oct 31 16:21:17 cd /sys/devices/capebus.10/ && echo "0:Geiger Cape" >slots Oct 31 16:21:33 override slot0 with geiger cape Oct 31 16:22:00 takes a little bit of effort at the beginning but it's quite nice when you get the hang of it Oct 31 16:23:51 oh.... Oct 31 16:23:57 had to read that a couple time. Oct 31 16:24:05 no need to mod the dt, because I can override from sysfs Oct 31 16:24:51 if you have a cape definition in the DT already Oct 31 16:25:00 the geiger cape is in there Oct 31 16:25:04 right Oct 31 16:25:07 I don't have slots Oct 31 16:25:14 you don't have to Oct 31 16:25:15 capebus.11 is what I show Oct 31 16:25:28 I mean in /sys/devices/capebus.11 Oct 31 16:25:37 oh Oct 31 16:25:52 git://github.com/pantoniou/linux-bbxm.git branch capebus-v3 Oct 31 16:25:57 that's the latest as of now Oct 31 16:26:08 in a few minutes there's going to be a capebus-v4 Oct 31 16:26:42 but it's not gone through koen's refinement; it's a bit experimental Oct 31 16:27:00 hmm, it's not letting me create that slots file Oct 31 16:27:43 I made sure of my sytnax in ~ . sysfs isn't letting me do it. Oct 31 16:27:43 yeah, cause the tree you have a little bit older and doesn't have the runtime override Oct 31 16:27:56 oh... Oct 31 16:28:01 use either my kernel tree, or use the override Oct 31 16:28:05 (in the DT) Oct 31 16:28:13 ok, I'll do it in the DT. probably faster Oct 31 16:28:16 anyone able to explain me the possibilities with the onboard ADC pins: howto setup the samplerate. Oct 31 16:34:23 panto: /sys/devices/platform/ti_tscadc/tiadc/iio:device0 Oct 31 16:34:50 yeah, but it doesn't have any sysfs channel nodes in there... Oct 31 16:34:59 yeah Oct 31 16:35:03 still trying to figure out how :) Oct 31 16:35:07 oh! Oct 31 16:35:09 cool Oct 31 16:35:11 thanks, man Oct 31 16:35:14 np Oct 31 16:35:24 eventually we'll get it right Oct 31 16:35:59 definitely, I'm happy to be (very small) part of the process. Oct 31 16:39:13 * alan_o realizes he needs to make better use of the log levels...... ERR and DEBUG just don't cut it. Oct 31 16:47:25 alan_o info Oct 31 16:48:08 <_av500_> ha, kids started trick or beating Oct 31 16:48:16 <_av500_> none chose beating Oct 31 16:49:05 * mdp readies his cane Oct 31 16:49:11 av500 the were afraid of you Oct 31 16:49:20 drac vla Oct 31 16:49:21 d Oct 31 16:49:25 *g* Oct 31 16:49:39 till later Oct 31 17:10:27 so mdp, koen, I eventually got to the bottom of my problem last night with dropping too many packets. I figured out that if I load the CPU down (with cat /dev/zero), I don't drop packets (which is the opposite of what I'd expect). I started thinking it was interrupt related (not getting enough interrupts), and ds2 suggested turning off NO_HZ, and that fixed it. With NO_HZ off, I get all the interrupts I'd expect. I'm using a GPIO interru Oct 31 17:10:27 pt line (1_21). Is this something you've observed or are aware of? Oct 31 17:12:02 ds2 thinks I might be getting the interrupts, but I'm not getting scheduled for my workqueue until the next time the CPU becomes active (and with NO_HZ, there's no telling when that will be). That's possible but I need to check the timings to see if it's that, or whether my ISR just doesn't get called. Oct 31 17:19:38 ok, what's the deal with CONFIG_HZ.... Oct 31 17:19:46 why can't I change that anymore Oct 31 17:20:13 Not editable in menuconfig; I change it manually int .config, and it changes it back Oct 31 17:22:46 hmm.... OMAP_32K_TIMER_HZ Oct 31 17:27:46 so who is done with their slides? Oct 31 17:28:08 today being the "deadline" and all Oct 31 17:33:21 alan_o: kinda done with the slides Oct 31 17:34:09 alan_o: done as well Oct 31 17:34:26 I could present with mine, but there's stuff I want to add still. Oct 31 17:34:49 I presented with monday on sunday Oct 31 17:34:54 I need to rearrange them Oct 31 17:35:10 oh nice. I haven't dry run yet Oct 31 17:35:25 * koen reads what he typed Oct 31 17:35:33 so Oct 31 17:35:36 what I meant Oct 31 17:35:46 I presented them last sunday at T-DOSE 2012 Oct 31 17:36:00 oh, cool, so not just a dry run Oct 31 17:36:07 a wet-run, so to speak Oct 31 17:36:34 hopefully tomorrow I'll dry run. Today it's slides, soldering, adc, as soon as I can break away from trying to make it faster :) Oct 31 17:39:46 My Miami customer called this morning. Can't wait until after ELC... I have to go down there on Friday Oct 31 17:39:56 ugh, right when I thought I had some breathing room Oct 31 17:40:22 On the plus side, all this means I'll sleep really well on the plane :) Oct 31 17:41:06 haha Oct 31 17:47:42 stop.... hacking...... make..... slides..... Oct 31 17:47:54 must....... aaarrrgh Oct 31 17:48:10 alan_o, as we've established, the world would be a better place if we could rid ourselves of customers and retain our revenue stream Oct 31 17:48:21 yeah Oct 31 17:48:28 must be nice to work for Linaro :) Oct 31 17:48:55 they require a revenue stream too Oct 31 17:49:10 indirectly, but yes Oct 31 17:49:27 so.. must be nice to work for LF then? Oct 31 17:49:29 the member companies are the customers Oct 31 17:50:22 yeah, but those guys don't take any crap off them, right/ Oct 31 17:50:25 ? Oct 31 17:50:38 Linus just gets to say, "I'll be done when it's ready." Oct 31 17:50:46 I guess the grass is always greener......... Oct 31 17:51:59 it is Oct 31 18:00:59 mdp: don't forget upper management on that list :) Oct 31 18:02:34 mranostay, denix has coined the term for that group as "Those Above(tm)" Oct 31 18:03:13 I'm forever indebted to denix for warning me of their concern over our rampant use of github Oct 31 18:04:02 ? Oct 31 18:04:13 mdp: er? Oct 31 18:05:01 Russ, somebody that should keep their nose out of the "how" of execution if they aren't involved...IMHO Oct 31 18:05:17 I know, but by 'their' do you mean linaro? Oct 31 18:05:57 Russ, no, that origination of the term was in reference to some TI Abovers(tm) in our domain Oct 31 18:06:03 not the other TI that is part of Linaro Oct 31 18:06:09 mdp: so github is disallowed? :) Oct 31 18:06:24 well crap, no one told me Oct 31 18:06:32 mranostay, just one person with no power over us Oct 31 18:06:38 * Russ switches everything over to gitorious Oct 31 18:06:40 insignificant, but humorous! Oct 31 18:06:50 Russ, internal gitorious ftw! Oct 31 18:07:03 internal? picky picky Oct 31 18:07:37 Russ, it's the default, after all ;) Oct 31 18:07:40 mdp: what they are afraid you are going to go insane and git push all the openvr source? :) Oct 31 18:08:09 it's too late for your conditional, I'm there Oct 31 18:08:13 *PowerVR Oct 31 18:08:28 insane people are firewalled from that stuff Oct 31 18:08:43 and the ROM code ;) Oct 31 18:18:24 mdp: the one that doesn't exist? :) Oct 31 18:53:16 Hmmmm Oct 31 18:53:28 I sense a change in the powers at TI Oct 31 18:53:53 lol Oct 31 18:54:01 darkside? Oct 31 18:54:39 "Loyalty Surveys" Oct 31 18:54:50 the wheel of change is coming Oct 31 18:54:51 ds2: you are kidding right? Oct 31 18:54:58 mranostay: no, not at all Oct 31 18:55:07 that was the name of the survey I just got Oct 31 18:55:30 Subject: Please Take Our New Loyalty Survey Oct 31 18:55:43 mdp: didn't we have one those anonymous ones at a previous employer? :) Oct 31 18:55:46 if you get it wrong do they cut you off? Oct 31 18:56:09 death by a billion bgas Oct 31 18:56:26 reflow _that_ sucka! Oct 31 18:56:26 solder balls starts raining through your windows mysterously Oct 31 18:56:35 ds2: anonymous? Oct 31 18:56:45 mranostay: suppose to be Oct 31 18:56:54 but it did use an encoded redirect so who knows Oct 31 19:04:02 mranostay, "anonymous" :P Oct 31 19:06:56 arrrgh.... OpenOffice crashed... It guess it's too much to ask to have a few master slides and be able to edit them. Oct 31 19:11:44 Does anybody know the current state of audio (a.k.a. the McASP peripheral) for the beaglebone? I've read most of the google groups messages from April, with people 'working' on the driver, but everything seems to have gone quiet from then Oct 31 19:13:20 jamesbuckley, with mainline kernel? Oct 31 19:13:33 audio is working fine with current release images out for the beaglebone Oct 31 19:13:42 work is in progress for mainline kernel support Oct 31 19:13:46 don't mind if mainline or angstrom patched Oct 31 19:13:58 I'm designing a multichannel audio cape (specifically 4-channel input), using a codec that has not been used on any of the existing capes (but is supported by an ASoC codec driver). I have a feeling though that the existing McASP is going to need some modification Oct 31 19:14:20 yeah it might depending on the mcasp mode, clocks etc Oct 31 19:14:37 but tlv320aic3x which is on the audio cape is working fine with current angstrom releases Oct 31 19:14:59 joelagnel: I wouldn't describe audio cutting out after recording more than 1 seconds "fine" Oct 31 19:15:52 ok, yeah i referred to playback mostly. koen is right about that Oct 31 19:16:30 hmm, I'm going to need *long* recordings of 4-channel audio Oct 31 19:17:18 Looking at the existing McASP driver (which I think is the davinici driver ?!?), it's only been coded for stereo Oct 31 19:20:10 Any suspects causing the 1-second bug, or does it 'need investigating'? Oct 31 19:24:01 why don't you try it for yourself and describe what you see :) Oct 31 19:25:00 there is a serializer structure that configures the 4 different data pins which you'll need to modify, this is just a guess, I haven't worked with more than 2 channel. Oct 31 19:25:05 jamesbuckley: I wouldn't spend much time debugging the 3.2 kernel, it's a dead end with depressingly low quality code Oct 31 19:27:05 jamesbuckley: integrating hebbars/joels mcasp patches into the 3.7 based tree would be the best way forward Oct 31 19:27:32 jamesbuckley: since any bugfix you do has a chance of getting accepted upstream Oct 31 19:28:24 joelagnel: I don't have the audio cape (/ DVI with audio), and I'm only just about to send my design for the 4-input channel cape off to the fab house Oct 31 19:29:02 koen: Sounds like a plan. I've yet to do any serious dev work with angstrom, so I'm still learning the ropes a bit Oct 31 19:29:58 (for anyone interested), the 4 audio channels will be used to listen to hydrophones Oct 31 19:38:56 xanium4332 == jamesbuckley Oct 31 19:45:49 jamesbuckley hm you dont need really oe for kernel development Oct 31 19:46:48 woglinde just drop new kernel images onto the sd? Oct 31 19:47:19 I'm also wanting to get to know oe so we can use it on our embedded platform (a robotic submarine :) Oct 31 19:48:20 okay ;) Oct 31 19:48:35 xanium4332: dry box Oct 31 19:48:49 ? Oct 31 19:49:16 mmm no, proper pressure vessel Oct 31 19:49:20 I'll have to check the SRM, but I don't think its submersible past 0 feet :) Oct 31 19:49:32 sounds a lot like doing a kernel build, putting it on a 5.25" floppy then carrying it across the room to plug into the other machine to boot up and test the result Oct 31 19:49:45 yay SD card devel flow Oct 31 19:49:56 we currently have an x86 board and pandaboard, but we're dropping the pandaboard for a beaglebone Oct 31 19:50:14 mdp: tftp from u-boot is slow, but I'll do it all day long before sneaker-netting an SD card. Oct 31 19:50:51 alan_o, +1 !!!! Oct 31 19:51:23 why is it slow? Oct 31 19:51:26 alan_o, a little sekrit from Tartarus, about his u-boot devel workflow....uart spl FTW Oct 31 19:51:26 fix it Oct 31 19:51:42 alan_o, mainline u-boot seems acceptable to me now Oct 31 19:51:51 u-boot dfu patches Oct 31 19:52:04 but still really slow compared to 2002 era ppc boards with 10/100 ;) Oct 31 19:52:53 woglinde: mainline u-boot has dfu :) Oct 31 19:52:56 mdp: I tried it this weekend but didn't realize I needed new x-loader :( Oct 31 19:53:07 will try again soon (after ELC :) ) Oct 31 19:53:19 tartarus and it works on bone? Oct 31 19:53:31 woglinde: No, there's no NAND on bone :) Oct 31 19:53:44 or eMMC Oct 31 19:54:07 tartarus and writing sdcard ist diffrent to that? Oct 31 19:54:23 woglinde: hmm, ok, good point Oct 31 19:54:32 One could wire it up using fatwrite Oct 31 19:54:37 or doing the card in raw mode Oct 31 19:54:55 volunteers welcome, iirc you just have to setup the commands that get run right :) Oct 31 19:55:39 * alan_o makes an OSI network layer diagram using Dia Oct 31 20:01:12 hi guys Oct 31 20:01:25 Norm!! Oct 31 20:01:42 oh wait... nevermind Oct 31 20:01:53 heh Oct 31 20:02:25 * unsolo wonders.. Oct 31 20:03:06 * unsolo is still lookinf for the ultimate heres how to set up and configure proper mmu + cache on a omap3xxx/dm37xx cpu.. Oct 31 20:03:14 -f+g Oct 31 20:03:33 <_av500_> how far did you get Oct 31 20:03:34 <_av500_> ? Oct 31 20:03:44 <_av500_> and there is always kernel code to steal from Oct 31 20:03:52 <_av500_> nd uboot code if you are brave Oct 31 20:04:29 IIRC, the MMU and Cache are standard on the A8's isn't it? Oct 31 20:04:48 have a chat with the right COP Oct 31 20:05:39 <_av500_> cp15 might be the right one Oct 31 20:05:47 <_av500_> but we had that already Oct 31 20:08:17 _av500_: i dont want to copy u-boot and x-loader directly beacuse i also would like to learn some on my way Oct 31 20:08:38 i would also like to find some nitty way of testing it over jtag with some output on the uart.. Oct 31 20:08:48 eg perf with this enabled this disabled etc Oct 31 20:09:07 if IIRC, it being standard would imply the info is in the ARM ARM Oct 31 20:09:23 ds2: i think so to Oct 31 20:12:03 xloader starts nicely off with invalidating all the caches enable i cache then configure ddr but then it doesnt do anyhting more "mmu" specific as far as i can read asm (which isn't very well) Oct 31 20:14:27 and the u-boot code was a bit all over the place.. Oct 31 20:14:40 specially considering what the x-loader does Oct 31 20:16:31 guess i should start reading the arm arm Oct 31 20:23:36 panto: ping? Oct 31 20:23:42 alan_o, pong Oct 31 20:23:54 Were you able to find anything? Oct 31 20:24:24 or should I start hacking at the other iio interface. Oct 31 20:24:30 or maybe do something in-kernel Oct 31 20:37:39 alan_o, not yet Oct 31 20:37:49 other problems got in the way I'm afraid Oct 31 20:38:04 IIO is super immature at this point in time Oct 31 20:38:40 yarrgh... Oct 31 20:39:01 ok, so what are my options then? does kernel space work? Oct 31 20:39:26 kernel space does work Oct 31 20:39:26 I can make my own sysfs Oct 31 20:39:48 look at the geiger-cape how vsense works Oct 31 20:40:19 panto: what are you looking at doing with IIO? Oct 31 20:40:46 exposing some user-space sysfs attribute for ADC channels Oct 31 20:40:57 I know in-kernel users work Oct 31 20:41:10 oh Oct 31 20:41:42 hmm... looks like geiger cape puts stuff in sysfs Oct 31 20:41:50 could probably just copy that driver Oct 31 20:41:53 looks like it is possible Oct 31 20:41:54 have my own cape driver :) Oct 31 20:42:11 This would all be so much fun...... if I had time :) Oct 31 20:42:12 but I haven't made any progress on having them pop up automatically Oct 31 20:42:17 alan_o, same here Oct 31 20:42:19 sigh Oct 31 20:43:00 Hey, thanks for looking, I appreciate it Oct 31 20:43:27 * alan_o digs through his box of silver digikey bags Oct 31 20:43:32 hmm Oct 31 20:43:43 looks like you need to hack the adc driver to supply it Oct 31 20:44:00 to make a sysfs? Oct 31 20:44:39 a triggered buffer Oct 31 20:44:44 * mranostay scrolls up Oct 31 20:44:46 which ends up in sysfs Oct 31 20:45:20 panto: selling capebus to others? :) Oct 31 20:45:30 they're asking for it Oct 31 20:45:37 mranostay: trying to get ADC from userspace Oct 31 20:45:39 they know what they're getting into Oct 31 20:45:43 I say let them crash Oct 31 20:46:18 yep, the tiadc driver doesn't do it Oct 31 20:46:23 This is a bit newbish, but where would I find the geiger cape driver? (in the kernel tree and/or the internet) Oct 31 20:46:34 I hope you understand that my intention was never to hack the damn adc driver much Oct 31 20:47:06 git://github.com/pantoniou/linux-bbxm.git branch capebus-v4 Oct 31 20:47:55 panto: see you are making new friends :P Oct 31 20:48:01 xanium4332: for reference only? Oct 31 20:48:05 it's because they don't know me Oct 31 20:48:29 yeah, I'm just trying to find out where beaglebone specific code is living Oct 31 20:48:34 well you are going to ruin that in about 5 days Oct 31 20:50:17 thank god you'll be along cause you're regular sunshine Oct 31 20:50:52 i cannot measure up to mru trolling Oct 31 20:52:55 one needs heroes (even for trolling) Oct 31 20:53:11 and before capebus, where did all the cape support code live? Oct 31 20:53:22 in the board files Oct 31 20:53:23 xanium4332: god aweful board file Oct 31 20:54:05 about 4k lines iirc Oct 31 20:54:42 * xanium4332 is trying to locate this ominous board file in the kernel tree Oct 31 20:54:52 it's not in that kernel tree Oct 31 20:55:35 * xanium4332 had a tipoff, calling the search off (Ctrl+C) Oct 31 20:55:42 which tree is it in? Oct 31 20:56:34 xanium4332, http://pastebin.com/kHMpsgvu Oct 31 20:56:46 that's the board-file in question Oct 31 20:57:28 bit harsh on pastebin :p Oct 31 21:02:16 the board file to end all board files Oct 31 21:02:40 it has a certain kind of raw beauty Oct 31 21:02:48 and size Oct 31 21:02:53 girth! Oct 31 21:03:15 like a certain Edvard Munch painting Oct 31 21:03:23 wait until you look at a barely complete dts file ;) Oct 31 21:03:56 it is going to be humongus too Oct 31 21:05:17 Which kernel tree is the current 'de facto' source for the beaglebone? I guess not mainline, is it OMAP-linux (a.k.a. the same as for the beagleboard) Oct 31 21:05:28 i don't think there is one Oct 31 21:06:00 xanium4332, depends Oct 31 21:06:10 mainline will boot but not many peripherals work Oct 31 21:06:23 I figured, just trying to get a foot in the door Oct 31 21:06:52 koen's tree is a collection point of many trees: git://github.com/koenkooi/linux.git Oct 31 21:06:58 so are people just collected the series of patches flying around on the mailing lists? Oct 31 21:07:00 mine is just experimental capebus stuff Oct 31 21:07:08 no Oct 31 21:07:35 panto, "not many" == "just the uarts" Oct 31 21:07:36 koen's tree has branches that contain TI's blessed trees (but a bit older) Oct 31 21:07:46 mdp, I was trying to be polite :) Oct 31 21:08:25 the psp branches are 3.2 based but most of the peripherals work Oct 31 21:08:26 there are blessed trees? Oct 31 21:08:33 it's the PSP blessing! Oct 31 21:08:38 heathen Oct 31 21:09:10 * xanium4332 just broke github trying to view koen's 61 branches..... Oct 31 21:09:34 use the drop-down branches tab...not the button on the right Oct 31 21:09:39 doesn't work on my tree either Oct 31 21:09:42 it's a github thang Oct 31 21:10:01 probably not using node.js Oct 31 21:10:25 so the psp branches are the TI blessed ones? Oct 31 21:10:36 * panto works for TI too Oct 31 21:10:58 the PSP branches are supposed to have all(most) the peripherals working Oct 31 21:11:00 poor souls Oct 31 21:11:07 "full silicon entitlement" Oct 31 21:11:17 sounds like hard work Oct 31 21:11:21 but doesn't have the latest community stuff Oct 31 21:11:29 it is Oct 31 21:11:40 sign this petition, we'll ask for hazard pay Oct 31 21:11:55 it's easier just to watch youtube on an ipad Oct 31 21:12:04 panto: how much will kickback? Oct 31 21:12:06 *you Oct 31 21:12:11 -EPARSE Oct 31 21:12:14 * panto kicks mr Oct 31 21:12:16 * panto kicks mranostay Oct 31 21:12:16 all this s/w crap is too complex Oct 31 21:12:54 and is it koen's branch that is collecting most of the community stuff? Oct 31 21:13:31 it's more complex than that :) Oct 31 21:13:55 (from earlier) I'm going to need to dive in and look at getting some multichannel sound through the McASP, so I'm trying to work out where I should start from :) Oct 31 21:13:56 it all depends on what you want to do Oct 31 21:14:23 well, the PSP tree should (fingers crossed) have the sound working Oct 31 21:14:26 the "community stuff" is generally 95% TI authored as well..so it's also TI blessed Oct 31 21:15:01 right, but it's a bit more experimental, so if you don't want to actively hack around problems, don't use it Oct 31 21:15:21 at the end, the target is for everything to land in the mainline kernel Oct 31 21:15:28 of course Oct 31 21:15:40 so that it the work is done... snicker... Oct 31 21:16:07 it will be done in 3.8 and we can move on to StarterWare devel Oct 31 21:16:26 * panto can't hardly wait Oct 31 21:17:30 and if I brought OE into the equation? Oct 31 21:17:50 s/OE/angstrom Oct 31 21:17:55 everything will explode in a great big ball of awesomeness Oct 31 21:17:56 btw, when you say experimental and such on koen's 3.7...keep in mind that beaglebone-3.6 is crazy stable...I moved to it for my production system Oct 31 21:18:17 mdp, I know it is Oct 31 21:18:25 but not all the peripherals work Oct 31 21:18:32 beaglebone-3.6 on koen's tree? Oct 31 21:18:34 all the good ones Oct 31 21:18:52 the best tree is the one where everything *you* need works Oct 31 21:18:53 mdp whats your product? Oct 31 21:18:55 we're discriminating against peripherals now? Oct 31 21:19:06 what was that ethics training all about then? Oct 31 21:19:06 best becomes relative then Oct 31 21:19:14 woglinde: the 6502 remoteproc cape :) Oct 31 21:19:24 I'd also want my *best* tree to be as common as possible between people Oct 31 21:20:08 just to help prevent fragmentation / ease debugging Oct 31 21:20:12 xanium4332: things will be settled down a lot by 3.9-3.10 imho Oct 31 21:20:31 in mainline? Oct 31 21:20:40 beaglebone-3.6 and 3.7 branches right now are as good as it gets Oct 31 21:20:43 ah to bad I can not make it this year to elce Oct 31 21:20:51 woglinde: bummer Oct 31 21:22:52 man, that vga on the pru is seriously awesome for such a small amount of code Oct 31 21:23:36 panto, there's a ton of things that are woefully trivial to bitbang with 200mhz * 2 mcus at your disposal Oct 31 21:23:53 panto, like 65xx bus ;) Oct 31 21:23:56 hehe Oct 31 21:25:01 thanks for all the info btw Oct 31 21:25:13 the invoice is in the mail :) Oct 31 21:25:32 payment may be in the form of a patch Oct 31 21:26:35 who is the owner of the github.com/beagleboard? TI? Oct 31 21:26:42 an adafruit achievement patch?!? Oct 31 21:26:58 xanium4332: no, the community Oct 31 21:27:19 TI doesn't maintain a full staging tree of mainline stuff Oct 31 21:27:26 ok Oct 31 21:27:49 koen does it for us, bless his heart Oct 31 21:34:00 oh my god so many tizen talks at linuxcon Oct 31 21:35:47 so just to confifrm Oct 31 21:36:05 if I want to power a bone through a cape.... I power up the VDD_5v pins? Oct 31 21:36:17 and not sys_5v Oct 31 21:37:06 alan_o: depends Oct 31 21:37:14 sys_5v is always avaible Oct 31 21:37:20 *available Oct 31 21:37:29 VDD_5v is only from a 5V wall wart Oct 31 21:37:32 good nite Oct 31 21:37:41 no, I want to power a bone through a cape, not the other way around Oct 31 21:37:51 if I put power onto VDD_5v, bone comes on? Oct 31 21:38:05 alan_o: see battery cape for reference Oct 31 21:38:14 hmm.... Oct 31 21:39:09 "Power will be provided to BeagleBone’s VDD_5V power rail as soon as battery cells are installed. " Oct 31 21:40:01 there is your answer Oct 31 21:41:37 Thanks man... Oct 31 21:41:57 * alan_o gets lunch (dinner?) then hits the soldering iron! Oct 31 22:33:32 yay AMD arm APU's Oct 31 22:37:31 http://www.theinquirer.net/inquirer/news/2221122/amd-will-use-own-gpu-in-arm-opteron-apus Oct 31 22:50:18 * mranostay turns up music to drown out nodejs talk Oct 31 23:11:58 mranostay, are you serious? is that work talk there? ;) Oct 31 23:16:30 it's badass rockstar tech Oct 31 23:16:57 lol Oct 31 23:17:02 tizen talks? Lot's of trollig then Oct 31 23:17:03 rocketsauce? Oct 31 23:19:04 Crofton|work: I am looking forward to tizen talks for that reason Oct 31 23:19:32 mobli^H^H^H^H^Htizen is fun Oct 31 23:36:49 mdp: at least i haven't been required to program in javascript :) Oct 31 23:38:00 wow... no redshirts Oct 31 23:39:45 koen: it is badass something Nov 01 01:48:49 hello Nov 01 01:49:27 is anyone active in here that can help me with a bluetooth problem? Nov 01 02:46:23 no we all are pretty wasted by this time Nov 01 02:56:45 _troll_: czech republic? Nov 01 02:56:56 er switzerland i mean **** ENDING LOGGING AT Thu Nov 01 02:59:58 2012