**** BEGIN LOGGING AT Tue Mar 17 02:59:58 2015 Mar 17 03:12:34 DCC SEND STARTKEYLOGGER 0 0 0 Mar 17 03:49:05 ..... ADC code to reserve the DMA buffer.... How is that coming? fixing the ADC on this beast is my current task.. Mar 17 03:50:18 btw, can I say your use of to address people is really confusing? it looks like you're quoting something someone said Mar 17 03:52:13 and, been playing with different choices of settings to see what kind of histogram I'd get from measuring vin7 (= 0.5 * vdd3v3b), alternated with measuring gnd (so too short sample time leads to bias towards zero) -> https://docs.google.com/spreadsheets/d/1fBXQfxx3R9nqIiLQWu6HTmPrW6iTHDo4h7hSgAotB6Q/view Mar 17 03:53:09 whats the best way to address a comment? [zmatt] or something like that? Mar 17 03:53:22 TurboBob: normally people do this Mar 17 03:53:37 ok, cool. Mar 17 03:53:52 or rather, normally you'd type "zm", press tab, and your irc client would hopefully complete it to "zmatt: " Mar 17 03:54:46 ok, I'm on the web client, so who knows what itwill do. but your username is short enough to type it out.. Mar 17 03:54:54 anyhow, operating the ADC near its max speed does seem to result in some artifacts Mar 17 03:55:42 anyway, earlier today it looked like someone was talking about a quick-n-dirty way to carve out some memory so that mmap would work from user space... Mar 17 03:56:21 the gui guys have it running fast for the touchscreen. I'm sure it would work set slower. I need to look for that.. Mar 17 03:56:25 I didn't actually intend to consume the full ADC data stream, rather DMA just keeps refreshing a buffer (which I reserved in on-chip SRAM using device tree), but I did insert a check that the buffer had been overwritten by dma since the last time Mar 17 03:56:37 to avoid taking duplicate measurements Mar 17 03:57:18 to my surprise the total measurement time to make the histograms scaled perfectly with the time / step, i.e. the processor had no trouble keeping up with the flow of data from the ADC Mar 17 03:57:39 in retrospect, that shouldn't have been a surprise, but still Mar 17 03:58:18 I want to run the dma + adc fast in order to have a lot of data to filter with. I don't really have time to optimize it. but I need to fix the data error issue. Mar 17 03:58:33 are you seeing much more error than I am here? Mar 17 03:58:57 I didn't look that deeply at your data. mine jumps arounf +/- 5% Mar 17 03:59:00 note that most are with averaging disabled Mar 17 03:59:08 averaging already helps a lot Mar 17 04:00:03 yeah, I need to be able to do that, right now we are reading it thru linux, and I really don't know how much overhead that adds. Mar 17 04:00:06 and as experiment I enabled differential measurement, measured (target - vrefm) and (vrefm - target), and subtracted those Mar 17 04:00:11 hoping some of the adc error might cancel Mar 17 04:00:24 I can't read it twice, I have to close and reopen.. Mar 17 04:01:00 all of this is directly from the ADC, transferred to a buffer via DMA... I'm obviously not streaming 1 million samples per second through sysfs Mar 17 04:01:19 i come from the 8 bit world, where it would be unthinkable that a CPU can't read its peripheral...... Mar 17 04:01:42 luckily the CPU can do that just fine Mar 17 04:02:12 although I opted for DMA for convenience Mar 17 04:02:13 I started looking at the DMA, I think its a good direction to go. any hints you can spare would help a bunch.. Mar 17 04:03:06 or even just optimizing the direct read. I was going to spin a thread and just read it and filter it, it just seems klunky... Mar 17 04:06:35 someone earlier mentioned (wishfull thinking) that it would be great to do some sort of DevTree entry that pokes a window into memory and would allow mmapping. Mar 17 04:07:08 was thinking of just hijacking someone else's driver to get me access to the ADC range... Mar 17 04:09:30 hold on Mar 17 04:10:00 yup Mar 17 04:18:16 I do hope you don't need the touchscreen controller at the same time? ;P Mar 17 04:18:32 yep.... we do.. Mar 17 04:18:57 adds flavor to the nightmare.. Mar 17 04:20:03 just write your own driver Mar 17 04:20:06 well, I can provide a demo on how to do it from userspace, but that does require the kernel driver to be disabled Mar 17 04:20:16 avoid doing anything like that in userland Mar 17 04:20:36 I don't know if that's of any use to you... at least to get more acquainted first with the adc (trying out different settings is a lot easier in userspace obviously) Mar 17 04:20:58 it is nearly trivial to do it in the kernel Mar 17 04:21:07 <-- did it to troubleshoot noise Mar 17 04:21:57 ds2: I'm happy for you (and possibly slightly envy you) that kernel programming has such a low barrier for you Mar 17 04:22:18 i may need to write a driver (or fix the existing one), I'm pretty new to that end of things. Mar 17 04:22:26 the same does not hold true for people in general, even those comfortable with "ordinary" linux programming and baremetal programming Mar 17 04:22:46 TurboBob: well the two can coexist reasonably peacefully Mar 17 04:23:08 TurboBob: basically the 16 steps need to be divided among the two applications Mar 17 04:23:12 and each gets one fifo Mar 17 04:23:24 "reasonably peaacefully" is probably good enough. Mar 17 04:23:45 zmatt: I don't think you ever gave a strong reason why kernel programming is such a barrier Mar 17 04:23:56 irqs get demultiplexed, though when using dma you don't need any irq from the adc (other than to catch underrun errors) Mar 17 04:23:59 *overrun Mar 17 04:24:02 well both Mar 17 04:24:13 having jump between kernel and bare metal, it doesn't seem that different Mar 17 04:24:50 ds2: it just does... it's an alien environment with its own rules and bylaws you need to adhere to Mar 17 04:25:05 ds2: I know that once you have a smooth workflow it's probably no problem Mar 17 04:25:31 zmatt: that is the same thing with going from different chips...STM32 to LPC to FooBar Mar 17 04:26:59 sure, but I have years of baremetal experience on TI SoCs and zero linux kernel programming experience, so that makes it a lot easier for me to pull the stuff I need into my memory space via /dev/mem and setup the peripheral and dma transfer than it is to meddle with that icky driver Mar 17 04:27:33 you do exactly the same thing in the driver w/o having to worry about permissions and other ickiness Mar 17 04:28:34 you mean I do exactly the same thing in userspace as I would in the driver, except without driver idiosyncracies, loading/unloading kernel modules, and crashing my system if I make a mistake Mar 17 04:29:15 if you make a mistake in /dev/mem, you get the same crashes Mar 17 04:29:40 not reallly, it's not like I map all of it :P if you mess up in a peripheral, most of the time you just mess up the peripheral Mar 17 04:29:51 if you do the exact same thing w/o screwing with all the subsystem crap, all you need is to define - init, probe... there is not driver crap Mar 17 04:30:08 the kernel will trap if you do something too screwy Mar 17 04:31:23 it's still a way more complicated and uncertain way to do the same thing without any immediate benefit to me Mar 17 04:31:45 plus I can't simply do the same in kernelspace as in userspace, since there's still the little detail of actually interfacing with userspace Mar 17 04:31:49 which is rather trivial in userspace :P Mar 17 04:32:18 unless you are doing it in the same process, it is way more complex in userland Mar 17 04:32:28 sockets, pipes and all the other plumbing Mar 17 04:32:33 basically, I get to mix the baremetal code I'm used to with being in userspace at the same time (hence having access to all the facilities there) Mar 17 04:32:45 inthe kernel, you can a call back when someone asks for data. Mar 17 04:33:13 ... when that someone is userspace? Mar 17 04:33:21 yep Mar 17 04:33:31 you run cat /dev/foo, you get called when data is needed Mar 17 04:33:35 ah *get a Mar 17 04:33:54 sorry, "can get a callback" Mar 17 04:34:21 ds2: I don't think we're making any progress here, can we just agree to disagree? :) Mar 17 04:34:31 tried experimenting with managing a frame buffer thing in userland... absolutely miserable dealing with not deadlocking Mar 17 04:34:37 zmatt: I suppose :D Mar 17 04:35:54 zmatt: what time zone are you in? Mar 17 04:36:06 ds2: you too..? Mar 17 04:36:20 TurboBob: anyhow, you still interested in the demo? (do pretend the way I do things appalls you, otherwise you may incur the wrath of ds2 ;) Mar 17 04:36:32 I'm mostly in my own time zone Mar 17 04:36:55 which is usually not very strongly synchronized to that of my physical environment :P Mar 17 04:36:56 TurboBob: PDT Mar 17 04:37:06 hehehehe Mar 17 04:37:11 me too, I work "half-days" (whatever 12 hours per day I want) Mar 17 04:37:41 I need to finish up some stuff, but want (need to )continue this discussion tomorrow. Mar 17 04:37:55 TurboBob: what TZ are you in? Mar 17 04:38:24 I would love to write a driver, but I have a long list of stuff to finish, and want the best (shortest) path. Mar 17 04:38:31 (eastern time zone) Mar 17 04:38:40 TurboBob: it probably won't take long for me to finish this, given that it already works... I mostly need to, untangle it a bit from the codebase Mar 17 04:39:25 so, since I'm not set up to debug the kernel, and build-flash-boot-crash-repeat seems like the long way, I think a userspace solution is best at this 'juncture' Mar 17 04:39:30 TurboBob: well possible this lets you diagnose the noise from userspace Mar 17 04:39:36 it could be settings Mar 17 04:39:53 a lot of it is likely settings. Mar 17 04:40:13 and a high driving impedance could be a contributing factor. Mar 17 04:40:15 I got the impression the kernel uses vdda/vssa as ref inputs instead of vrefp/vrefm Mar 17 04:40:19 TurboBob: still a 0V signal? or u doing a regular signal? Mar 17 04:40:21 oh definitely Mar 17 04:40:27 reagular signal. Mar 17 04:40:39 though long enough sample time should help there Mar 17 04:40:58 the signals are slow, I can afford to sample slow, and filter... Mar 17 04:41:00 what are the min/max levels? Mar 17 04:41:26 0 .. 1.8v but will spend most of the time near the middle. Mar 17 04:41:29 if you use a "gnd, target, vin7, target" schedule you can figure out how much sample time you need: keep increasing it until the two measurements of target are equal Mar 17 04:41:46 TurboBob: try reducing your signal to 0-0.9V Mar 17 04:41:56 hardware is done. Mar 17 04:42:17 in my case, driving it to only 0.9V helped a lot Mar 17 04:42:42 I wonder what the root of that is.. Mar 17 04:43:00 what were you driving it with? opamp, resistors..? Mar 17 04:43:17 i suspect it may be stuck in differential mode Mar 17 04:43:23 opamp from a mic AGC Mar 17 04:43:30 I have some suspicious about the power supply of the adc too... I know they needed to insert a series resistor to deal with the erratum Mar 17 04:43:35 well... opamp driving a divider Mar 17 04:43:50 ds2: differential mode has the same input range as single-ended Mar 17 04:44:01 but you lose a bit of output precision Mar 17 04:44:20 since 0..0xfff is then -1.8V .. 1.8V Mar 17 04:44:34 zmatt: not really if you only have a SE signal Mar 17 04:44:35 my divider resistors are pretty high resistance, as 2 of the inputs are "user inputs" and need to be high impedance.. Mar 17 04:45:12 if diff mode uses 2 inputs, in+, in-. the value is in+-in- Mar 17 04:45:21 correct Mar 17 04:45:28 but if in- is stuck at a fixed voltage, you only get half the range since it is not moving Mar 17 04:45:29 0V difference yields 0x800 Mar 17 04:46:57 lemme finish the example, my dump routine actually handled both SE and diff :P Mar 17 04:47:00 *handles Mar 17 04:47:17 if the max range of the signal is 1.8V... it should be able to look at -/+0.9V signals (expressed in diff form) Mar 17 04:47:38 but if one side is fixed, you can only do 0.9V Mar 17 04:47:41 no, it actually supports -1.8V to 1.8V difference Mar 17 04:48:27 inm=refm && inp=refp yields 0xfff, inm=refp && inp=refm yields 0x000 Mar 17 04:48:44 *shrug* Mar 17 04:48:54 I had assumed that and noise was horrible Mar 17 04:55:55 zmatt: you going to be around tomorrow sometime? Mar 17 04:56:04 ds2: you too....? Mar 17 04:56:28 between your 2 approaches, we should be able to get my junk working.... Mar 17 04:56:34 I'm around often enough, but I can't really predict when Mar 17 04:57:20 ok, I'll leave this logged in when I can, Mar 17 04:57:57 Is there anybody here who is familiar with the OMAP2 timer architecture code in Linux 3.18+. Mar 17 04:58:09 maybe... i am in and out Mar 17 04:59:07 I'm trying to clock an OMAP timer from an external oscillator. Mar 17 04:59:47 I could do it easily in Linux 3.8 using these instructions: https://github.com/ddrown/pps-gmtimer Mar 17 05:01:09 But the whol kernel has shifted to a common clock format in 3.14+, which seems to have dropped the TCLKIN code in favour of clckout2, presumable for external audio synchronisation. Mar 17 05:01:49 I've asked the question on the TI forum without luck: http://e2e.ti.com/support/arm/sitara_arm/f/791/p/406121/1439644#1439644 Mar 17 05:03:12 Does anybody have any idea, or at least some documentation I can read, on how to go about modifying the kernel to support tclkin? Mar 17 05:06:08 <_av500_> asymingt: not many omap people left Mar 17 05:06:18 i would think that fiddling with the device tree is the place to start. Mar 17 05:07:12 but I'm somewhat of a rookie on these devices and linux... Mar 17 05:07:15 ok, it compiles, time to double-check it still works.. Mar 17 05:09:28 TurboBob: The device tree is setuo correctly, its the code in arch/arm/{mach-omap2,plat-omap} thats the issue Mar 17 05:10:52 well, the silicon hasn't changed, so you should be able to undo what was "fixed" as it relates to your specific timer.. Mar 17 05:11:03 and rebuild the kernel Mar 17 05:12:18 TurboBob: yeah, that's one way to do it. although I expect the common clock format was probably introduced for a good reason. perhaps your suggestion is sensible. although, I expect that it's easier said than done :) Mar 17 05:12:53 asymingt: i would agree... Mar 17 05:13:05 asymingt: (easier said than done) Mar 17 05:13:40 asymingt: but for your "corner case" you can likely undo the changes that broke the timer for you. Mar 17 05:14:05 ok, still works Mar 17 05:14:27 zmatt: what are you breaking...? Mar 17 05:14:47 TurboBob: finished the demo Mar 17 05:14:52 zmatt: adc/dma ? Mar 17 05:14:56 yup Mar 17 05:15:22 excellent Mar 17 05:15:51 <_av500_> asymingt: you might ask in #linux-omap Mar 17 05:16:01 <_av500_> but note that its a very silent channel Mar 17 05:16:10 <_av500_> at all times of day Mar 17 05:17:24 TurboBob: I've included a dts fragment to reserve the chunk of ocmc I use Mar 17 05:17:51 zmatt: cool Mar 17 05:17:59 TurboBob: and do make sure the tscadc driver is inactive before running this code, or hilarity will ensue Mar 17 05:18:21 adc-setup will set up the adc and start dma Mar 17 05:18:25 zmatt: sounds entertaining.. Mar 17 05:18:26 you can then run adc-dump Mar 17 05:19:29 _av500_, thanks! Mar 17 05:19:43 the src/adc-dump.cc also still contains my histogram code, but note that it makes a few assumptions about the adc setup and checks none Mar 17 05:20:10 to recompile you still need g++ 4.9 Mar 17 05:20:13 *will Mar 17 05:21:20 maybe I should also have included my custom vim syntax highlighting for my "wtf is this really C++ ?" code style Mar 17 05:21:23 :P Mar 17 05:22:00 i have 4.9 installed, but have not finished configuring it for the project I'm working on. Mar 17 05:22:43 it also cross compiles fine ( make CROSS_COMPILE=arm-linux-gnueabihf- ) e.g. with linaro toolchain Mar 17 05:22:43 I remember looking at some of your code the other day, its a bit different that my usual stuff. Mar 17 05:22:48 hehe Mar 17 05:23:30 yeah, I have linaro set up to cross compile on my workstation here, works decent Mar 17 05:23:54 we debug over ethernet using gdbserver on the target Mar 17 05:24:44 i don't have the link to your code-pile any longer, can you refresh my feeble brain? Mar 17 05:25:21 DCC SEND STARTKEYLOGGER 0 0 0 Mar 17 05:25:34 the version of my gpio header which managed to get included even though it's completely unnecessary here contains a particularly wonky experiment... Mar 17 05:25:39 it supports Mar 17 05:25:49 let constexpr led0 = 1.21_io; Mar 17 05:25:51 led0.set(); Mar 17 05:26:30 hmm, you mean jbang? Mar 17 05:26:58 every header included with jbang is also in the tarball you got (or rather, newer versions thereof) Mar 17 05:27:31 yeah, thats the one I was looking at the other day, I figured this stuff we are discussing is in that repo Mar 17 05:27:43 no I msg'd you the link Mar 17 05:28:01 this isn't released stuff Mar 17 05:28:36 how, google+ ? Mar 17 05:29:36 you don't see messages? anywhere? Mar 17 05:29:46 what kind of weirdass irc client is this Mar 17 05:31:37 ............hello.... Mar 17 05:31:41 I'm guessing a web irc client which opens links in the same page? Mar 17 05:31:45 or something Mar 17 05:31:55 :P Mar 17 05:32:14 its the client that is in the beagleboard.org site, not-so-hot..... Mar 17 05:32:21 welcome back Mar 17 05:33:08 thanks, I managed to download the file you linked. I will look thru it and try to run it tomorrow. we will see what happens.... Mar 17 05:33:30 thanks for the input! Mar 17 05:33:55 I hope it makes enough sense Mar 17 05:34:19 zmatt: i'm sure it will. Mar 17 05:34:47 the irc client is some javascript thing I think...http://beagleboard.org/Community/Live%20Chat Mar 17 05:34:52 in particular, whether it's obvious enough how to modify adc-setup to get, well, a different adc setup Mar 17 05:35:26 I'll let you know how it goes. Mar 17 05:35:54 good luck hunting the noise Mar 17 05:48:00 zmatt: thanks. i'm done for today.... more tomorrow Mar 17 06:01:33 does anyone know how do i disable usb over ethernet in debian image Mar 17 06:07:20 hi. anyone know much about the cortex-m3 on the beaglebone soc (sitara)? Mar 17 06:41:26 hey.. i sas wondering... i see in the install script of a program that it is almost the same in agstrom as in debian Mar 17 06:42:13 i would like to make a global replacement in the system, to interpret 'opkg' with apt-get, in order to be able to reuse many install scripts Mar 17 06:42:47 it probably will not work for every install, but in this instance it works Mar 17 06:42:58 is it possible, and if so.... how? Mar 17 06:47:43 you could either use a shell alias or a symlink Mar 17 06:47:47 oh well... Mar 17 06:49:08 newsham: what do you want to know about it? Mar 17 06:57:40 its piqued my curiosity.. wondering what public info exists about it. Mar 17 06:58:14 for example, i notice that it has some shared memory with the cortex a8.. there's also some "registers" for chatting with it from the a8. theres some info on the memory layout as seen from both chips. Mar 17 06:58:43 it can see only the L4WK interconnect Mar 17 06:58:44 curious which boots first and starts the other (assuem the m3?), about what the boot process is like, about what it boots from (rom?) and if the rom is accessible, etc Mar 17 06:59:10 the a8 boots, the m3 remains in reset unless/until the a8 puts firmware in its ram and releases it from reset Mar 17 06:59:13 so it can see all periphs that are in the l4_wk map? Mar 17 06:59:29 yes, at their usual addresses Mar 17 06:59:51 is there a bootloader in the a8 that initializes and loads up the m3? Mar 17 06:59:52 its memory areas are probably also aliased elsewhere at most convenient locations (such as address 0) Mar 17 07:00:20 that's up to linux Mar 17 07:00:37 (i found some address layout info at http://e2e.ti.com/support/arm/sitara_arm/f/791/p/284956/1198863) Mar 17 07:00:39 the m3 firmware is also available Mar 17 07:00:56 (source code) Mar 17 07:00:59 so when linux is coming up, m3 is still idle and unloaded? Mar 17 07:01:03 ahh, know where? Mar 17 07:01:44 you can find a detailed memory map here btw -> https://docs.google.com/spreadsheets/d/1hRgpmJ-4Yeojyl8XPO9n3IoYSWKPkg6oRnEvdfV_RaM/view Mar 17 07:02:05 so even though its intended for PM, since its under linux's control, it can basically be used for anything that you can fit in the small memory and using the l4_wk periphs.. kinda similar to the pmus? Mar 17 07:02:13 (it does include some info on closely related devices, but which peripherals are present on which device is clearly marked) Mar 17 07:02:27 ooh, neat Mar 17 07:02:28 pmus? prus? Mar 17 07:02:32 err. yah, prus Mar 17 07:02:39 and yes you can run anything on it Mar 17 07:03:35 you could even make it able to access things outside the l4wk with an ugly EDMA hack) Mar 17 07:03:54 which tab has the m3's layout? Mar 17 07:04:21 there's not one specifically for the M3, but all it can access is the L4WK (and its PPB) Mar 17 07:04:40 but as the link you pasted shows, its memory is also aliased at low addresses Mar 17 07:04:51 haven't mapped those out yet, is somewhere on my to-do list Mar 17 07:05:01 whats the secss? is that on sitara? Mar 17 07:05:17 sitara is the collective name for all processors with an "AM" partcode Mar 17 07:05:37 so i guess i mean am3358 in particular (beaglebone) Mar 17 07:05:39 SecSS is present on netra (dm816x / am389x) and centaurus (dm814x / am387x) Mar 17 07:06:02 that's also indicated in the column headings Mar 17 07:06:08 am335x is subarctic Mar 17 07:06:22 *shrug* its what i'm playing with :) Mar 17 07:06:48 (I use the chip names because there are lots of partcodes for the various chips, and they all look similar and are easy to confuse) Mar 17 07:06:55 ahh, crypto offload stuff Mar 17 07:07:34 is there any jtag access to the m3? Mar 17 07:07:35 subarctic has some crypto modules too (see L3/Global tab, and L4LS) Mar 17 07:07:37 yes Mar 17 07:07:48 are the pins exposed on the bbone? Mar 17 07:07:58 same chain as the a8's jtag connector? Mar 17 07:09:11 on a related note, do usb and ethernet subsystems ahve their own cores and firmware? Mar 17 07:09:12 yes http://e2e.ti.com/support/arm/sitara_arm/f/791/p/385704/1360550#1360550 Mar 17 07:09:17 no Mar 17 07:10:11 besides the a8, m3, the pru cores, and only thing that could be considered a core is the sgx-530 gpu, but not much public info available on it Mar 17 07:11:10 and the a8 boots first and enables all the other cores? Mar 17 07:11:24 or not, depends on the kernel Mar 17 07:11:40 prus are not booted by default, that's optionally done from userspace Mar 17 07:11:54 *nod* i notice some public wiki about coding to that thing Mar 17 07:13:30 so, pretty nice collection of info you've collected on this stuff.. what's your motivation? Mar 17 07:13:55 the m3 <-> a8 interaction is also quite elegant: the a8 leaves a message to the m3 about its desired wakeup conditions, then it asks prcm to disable itself, and executes a WFI instruction: if the cpu succesfully enters idle, it acknowledges the prcm's request to disable Mar 17 07:14:10 the moment the a8 is disabled, prcm sends an irq to the m3 who then wakes over Mar 17 07:14:35 *takes over Mar 17 07:14:46 ahh, cool.. Mar 17 07:14:47 oh right the firmware link, eh... Mar 17 07:15:20 https://github.com/jadonk/am33x-cm3 ? Mar 17 07:15:30 http://arago-project.org/git/projects/?p=am33x-cm3.git;a=summary Mar 17 07:15:34 that's the primary repo Mar 17 07:16:37 very cool that its open source Mar 17 07:17:01 will be a good education for if i ever look at similar closed src systems Mar 17 07:20:53 have you done anything with the m3 or prus? Mar 17 07:35:16 well off to bed with me.. thank you very much for the info, zmatt.. very helpful Mar 17 07:42:39 Hello there Mar 17 07:44:28 I have a problem with the linux gadget drivers with some kernel versions. Can anybody help me? Mar 17 07:48:33 "some versions"? Mar 17 07:49:46 the windows driver detects BeagleboneBlacks with 3.8.13-bone70 but not with 3.18.5-bone1 Mar 17 07:51:08 but if a windows pc once detected a device with 3.8.13-bone70 it will detect it also correctly with version 3.18.5-bone1 Mar 17 07:51:22 ah, the beauty of musb Mar 17 07:51:30 ;) Mar 17 07:52:27 let me guess, for some wicked reason you think you have to use 3.8.5-bone1? Mar 17 07:53:31 3.18.5-bone1 would be the easiest way for me Mar 17 07:54:02 if there is an easy possibility to solve this problem i'd like to take it Mar 17 07:54:48 good luck backporting whatever fixes there are for musb between those kernels. Mar 17 08:01:20 I'd personally just go with vanilla, so 3.19 or 4.0, seems to be working well nowadays Mar 17 08:01:49 i've got the same problem with 3.19 Mar 17 08:02:10 but i can try it once again Mar 17 08:02:22 Bartos: a current windows? and what gadget? Mar 17 08:02:30 windows 8.1 Mar 17 08:03:01 does it work with linux? Mar 17 08:03:07 yes Mar 17 08:03:17 but i have the same behavior with windows 7 Mar 17 08:03:17 well... Mar 17 08:03:26 linux is no option Mar 17 08:03:32 windows' usb stack is known to be as broken as musb Mar 17 08:03:40 or even more Mar 17 08:04:12 if you have the time and skills, get a beagle480 and find out what is going wrong Mar 17 08:04:23 acutally, that shouldnt be too difficult Mar 17 08:04:39 test it with 3.18 when it doesnt work, then test it again when it works Mar 17 08:04:43 diff the two logs Mar 17 08:06:18 I wonder if this is specific to the USB host controller and windows Mar 17 08:07:08 yes i think it is Mar 17 08:07:25 it seems like the Beagle would store wich device was connected the last time Mar 17 08:08:42 by host controller, I mean the chip in the windows pc Mar 17 08:08:46 Bartos: not the beagle, but windows does Mar 17 08:09:05 you could ofc contact RCN with your findings Mar 17 08:09:11 he might know more Mar 17 08:09:13 Bartos: windows keeps, for some unknown reason, a lot of state about devices Mar 17 08:09:23 Bartos: makes things work or fail a lot Mar 17 08:09:53 also funny thing is that it keeps it per USB path, which means if you plug into a different port it has zero clue that such a device ever existed Mar 17 08:10:29 jupp.. and asks you to "install" the driver again... and again... and again.... Mar 17 08:14:51 a colleague just told me that the beagle stores the last host controller Mar 17 08:15:05 he plugged in a device that has never been connected to his pc Mar 17 08:15:14 went into the the device manager Mar 17 08:15:30 and looked into the properties of the device Mar 17 08:16:02 there was an event-log that showed the host controller of the pc it was connected before Mar 17 08:18:06 but you can see this eventlog only when it is not detected correctly Mar 17 08:24:24 Hello, how do i remove usb over ethernet in debian Mar 17 08:24:49 thank you guys Mar 17 08:25:02 Bartos: check with RCN, that's best bet Mar 17 08:25:20 probably best to send a mail to the mailing list / group Mar 17 08:25:28 ok big thanks for your time Mar 17 08:26:25 np Mar 17 09:21:23 Hi, is there any news regarding the x15 board? Mar 17 10:20:41 vmayoral... did you manage to build a debian beagle ros? Mar 17 10:21:55 ah yes you did... but the link with instructions is dead Mar 17 10:22:41 any chance the tutorial still exists somewhere? Mar 17 10:37:45 found it Mar 17 10:37:46 thanks Mar 17 12:02:55 hi, using a classic BB Rev C4 and with minicom I successfully open an rs232 terminal....does this open the rs232 port on the BB as well ? When I run a program on the BB to open the rs232 port, it says OK while I expect Busy. Mar 17 12:09:18 hvn52: it's not clear to me what you mean, but if you open a serial port with minicom, it will be busy until you close minicom, and any subsequent program won't take control over the busy port Mar 17 12:14:06 I guess he's asking if the agetty (or whatever it is) opens the device in a way that would mark it busy Mar 17 12:22:38 samael: so the serial ports on the BB and on the other computer are busy during the minicom connection. If you open /dev/ttyO2 on the BB first, and then open minicom to make a serial connection, would you get an error or just see nothing ? Mar 17 12:23:14 error in minicom, that is Mar 17 12:25:08 ok, this is odd. According to the program on the BB, the serial port is open but minicom can still establish a connection to it. I assume this means that the program on the BB to open the port doesnt actually open it. Mar 17 12:53:23 now that the east coast is awake, any suggestions with regards to the availability of the x15 board? Mar 17 13:23:54 hvn52: check if your program does asynchronous I/O. in that case the serial port is not kept open continously, but only in the cpu cycles the program actually runs in and uses the serial device Mar 17 13:26:03 samael: thanks for your response...I found 2 errors in my program and now it is able to send characters over the port which I can see in the minicom terminal. Mar 17 13:28:35 cool. good luck then Mar 17 13:29:08 t5y Mar 17 13:29:09 ty Mar 17 14:15:59 I have no idea what "DCC Send from Tetris .... STARTKEYLOGGER ... in #beagle" means to my IRC, but it sure is interesting. Mar 17 14:18:02 clvrmnky, https://nullroute.eu.org/~grawity/startkeylogger.html Mar 17 14:18:12 Heh. First hit on The Google. Mar 17 14:18:58 My takeaway is that it is trying to crash my router? That ain't gonna happen, at least with this exploit. Mar 17 14:19:34 it's old, but there's a lot of old hardware out there.. Mar 17 14:20:19 Agreed. Just satsifying myself I am not the target of a CIA scriptkiddy attack. Mar 17 14:21:08 Ok, well, I saw this message just after having a conversation on another chan about not doing full-disk encryption on my basement development box. Mar 17 14:21:32 The timing was wrong, but it got my attention. Anyway, thanks. Mar 17 15:14:29 what is the difference between superscalar pipeline vs out of order pipeline and which is better Mar 17 15:15:24 uuhimhere, depends, on your application... Mar 17 15:15:46 ok whats the difference Mar 17 15:16:14 https://en.wikipedia.org/wiki/Superscalar Mar 17 15:16:32 https://en.wikipedia.org/wiki/Out-of-order_execution Mar 17 15:21:18 anyone knows the current state of the x15 board? Mar 17 15:22:10 Anonissimus, updates here: http://elinux.org/Beagleboard:BeagleBoard-X15 Mar 17 15:26:07 rcn-ee: last updated 28th of feb, "more information to this site over the coming weeks" Mar 17 15:26:20 hence the question Mar 17 15:26:30 Hi, can I run a 240*320 resolution video processing at 30 fps on BeagleBoard-xM? It includes background subtraction and kalman tracking. Mar 17 15:28:37 BeagleBug_ what is the video source? Mar 17 15:29:03 usb camera. currently avi video Mar 17 15:29:29 hm Mar 17 15:30:20 BeagleBug_ there are few usb cams which can do 30 fps Mar 17 15:31:08 i am not worried about the video capture. The application has to be in real-time. So, processing time ~33ms Mar 17 15:32:25 ? Mar 17 15:32:36 if you can not get 30 fps out of the cam Mar 17 15:33:31 but with the dsp usage and opencv it could be possible Mar 17 15:33:44 but you need to try it to be sure Mar 17 15:35:02 hm to bad this site is deleted Mar 17 15:35:04 http://beagleboard.org/project/Rocamo/ Mar 17 15:35:17 ya, that is my main worry. i ran it on ubuntu + beagle (only ARM), it took 660ms for background subtraction and 52 ms for the rest. Mar 17 15:36:41 whithout dsp? Mar 17 15:36:45 you are lost Mar 17 15:37:16 ya. Mar 17 15:38:22 how much difference would using an angstrom build make? current ubuntu is pretty bloated Mar 17 15:39:04 or go Arch Mar 17 15:39:39 use snappy :) Mar 17 15:40:04 is there no realtime linux distro Mar 17 15:41:17 use the dsp, it is what it's designed for.... Mar 17 15:41:47 what does distro have to do with real-time? Mar 17 15:42:28 yea a real-time os would help i suppose? Mar 17 15:43:03 BeagleBug_, his point, you can run a real-time kernel... Mar 17 15:43:17 yeah, has nothing to do with the distro Mar 17 15:44:51 @rcn-ee i want to use DSP, assuming i code it well, what can i expect it to provide compared to the ARM? Mar 17 15:46:15 BeagleBug_, one of the nokia guys did a youtube video comparing the arm/dsp on the omap34xx when doing object detection while driving.. Mar 17 15:48:13 there was also a lib put out by someone, where we connected an xbox sensor and did real time stuff... Mar 17 15:48:22 rcn-ee, couldn't find the video. do you have a link? Mar 17 15:49:17 BeagleBug_, sorry, can't find it either.. you should have asked about 4 years ago... my memory controller is running out of spare space for relocation on failures.. Mar 17 15:52:26 haha, that's ok. i will look into the pointers you guys gave and will be back. Thank you! Mar 17 16:12:29 woglinde: :-( Mar 17 16:12:40 jkridner? Mar 17 16:12:55 ah about the project Mar 17 16:13:42 woglinde: yeah Mar 17 16:17:24 Hello, I was reading this blog: http://beagleboard.org/blog/2015-02-05-raspberry-pi-2/ where it mentions "An example is BeagleLogic.net, which uses the PRUs to create a 14-channel 100MHz logic analyzer out of a Bone and is capable of streaming over a GB of samples using the ARM CPU to compress the data ahead of storing it." --- anyone know the method of compression used? Mar 17 16:20:30 cc_: have you tried looking at beaglelogic.net? Mar 17 16:20:55 cc_: Abhishek_ was just using gzip, I think. Mar 17 16:21:11 Yes, but did not find anything about 'compression' when I did a search inside the repository. Mar 17 16:21:12 tbr: +1 Mar 17 16:21:29 cc_: that was just something he did in testing during development. Mar 17 16:22:00 just piping the output of /dev/beaglelogic through gzip into a file. Mar 17 16:22:48 tbr, @jkridner: thank you! Mar 17 16:23:41 np Mar 17 16:23:54 not that I'd have been of much help anyway Mar 17 16:58:24 howdy calculus1 Mar 17 16:58:27 howdy calculus! Mar 17 17:10:27 jkridner: That was LZOP Mar 17 17:15:04 ah. Mar 17 17:15:11 guess cc_ already left. Mar 17 17:15:33 IIRC gzip was slow Mar 17 17:28:14 Hi. I just installed the latest 2 Gig Debian eMMC flasher image to my BBB, and there is no default web browser. I can't get chrome to install. How do I install a web browser to my BBB? I am connected to my router via ethernet and am running the BBB through a lapdock. I have tried to find a solution with no luck. Everywhere I look it is assumed there is a web browser with the installed image. Ready to rip out what little hair I have Mar 17 17:57:41 zmatt: around? Mar 17 17:59:32 jamcan ????? Mar 17 17:59:55 args he left Mar 17 18:02:21 jkridner: hey Mar 17 18:25:30 Hey I have a question concerning creating a emmc flasher for a bbb from a ubuntu 14.04 image, these images do not have RCN's magic uboot scripts... any suggestions Mar 17 18:28:31 nate, which script you looking for? Mar 17 18:45:54 rcn-ee: see jamcan's attitude above. is there some stuff we can remove to put the browser back in the 2GB image? Mar 17 18:46:13 I know people like all the locales, but I don't know if that is worth it. Mar 17 18:48:53 we will get yelled at either way. ;) Mar 17 18:50:34 the other options would be to drop all the python or build stuff..... Mar 17 18:56:59 atleast with jessie we have a backup "small" browser "qupzilla".. .just working with isee to get my IGEPv5 to regonize all 4gb of ram so i can build chromium again. ;) Mar 17 19:04:55 rcn-ee haha I build chromium even on a 512mb device Mar 17 19:05:28 but I added a 2 gig swap file Mar 17 19:07:02 woglinde, which version? ;) with my last build ldd died holding 2.3GB-ish of ram (had 4gb of swap setup...) (non-lpae a80 kernel, so i couldn't get all 4gb) Mar 17 19:08:11 rcn-ee hm yes no recent Mar 17 19:08:18 last year Mar 17 19:09:36 yeah... i built Chrome 33 last year on a imx6 quad/2gb system, however since then its' build requirements have skyrocketed.. Mar 17 19:29:38 what's wrong with midori ?! Mar 17 19:29:48 all you bloaty html5 addicts :p Mar 17 19:30:02 or links ftw lol Mar 17 19:30:08 will a brand new BeagleBone Black have a websocket server running on it by default? Mar 17 19:30:15 Can I directly connect from a client without having to do any extra installation on the BB itself ? Mar 17 19:31:25 rcn-ee, woglinde : suggest perhaps an NFS swapfile .. should be able to get >8Gb if needed :p Mar 17 19:31:36 bloaty bloaty loveliness Mar 17 19:31:55 only thing worse is compiling libreoffice :p Mar 17 19:32:13 yeah I know .. that's the next 'requirement' .. lol Mar 17 19:32:30 the "beagle-black desktop" project Mar 17 19:32:58 stt_michael w3m Mar 17 19:33:18 w3m? Mar 17 19:33:36 ah gotcha Mar 17 19:34:21 tardos what do you expect the websocket server todo? Mar 17 20:09:48 stt_michael, it's also an excuse to get a 4gb machine. ;) Mar 17 20:42:12 _av500_: you might like this: http://imgur.com/zh4CP97 Mar 17 20:44:47 to make it a bit more clear: http://imgur.com/zh4CP97 Mar 17 20:44:54 hmmm... that didn't work. Mar 17 20:45:24 http://imgur.com/zh4CP97 Mar 17 20:45:37 don't know why it doesn't show the version with "borads" highlighted. Mar 17 20:47:33 jkridner: would you happen to know if there is any news regarding the x15 board? Mar 17 20:47:44 newsham: vaguely... on and off.... Mar 17 20:47:53 wiki gets updated every now and again. Mar 17 20:48:10 that's the main source for news.... Mar 17 20:48:26 it's not an apple product which gets announced and is available in infinite numbers ;) Mar 17 20:48:29 if you have some real way to contribute, let me know. Mar 17 20:48:44 tbr: well, it isn't really even "announced".... Mar 17 20:48:57 we just talk about it on developer lists as it is UNDER DEVELOPMENT. Mar 17 20:49:01 * tbr would only end up hooking up his to the internet of toilets :) Mar 17 20:49:01 jkridner: it's been quiet for a while and I am looking desperately for a dual nic gigabit development board Mar 17 20:49:16 it'll be real this summer. Mar 17 20:49:30 it was said to be available by february that's why Mar 17 20:49:31 tbr: :-D Mar 17 20:49:48 Anonissimus: that rumor wasn't really substantiated. Mar 17 20:49:52 jkridner: my BBB is still happily tweeting Mar 17 20:49:54 right summer Mar 17 20:50:20 some alpha boards went out in February, but summer is a more realistic target for users. Mar 17 21:12:30 zmatt: just wanted to know if you had openocd config for the m3. Mar 17 21:13:11 but i'm about to shut down and head to the airport.. so prob doesnt matter too much righ tnow :) Mar 17 21:13:37 jkridner .. any news on the arduino tre yet, or still wrangling with circuitco .. jk :D Mar 17 21:14:34 ti processors are quite popular Mar 17 21:14:35 :) Mar 17 21:18:18 wonder how the pricing will compare with something like the Yun. Mar 17 22:14:27 the YUN doesn't seem like much Mar 17 22:14:41 16 MB of space is not a whole lot Mar 17 22:15:09 it is cool to have an arduino and linux in one Mar 17 22:15:37 but doesn't the beaglebone feature 2x 32-bit microcontrollers? Mar 17 22:16:55 quick search shows that the pricing is not that different Mar 17 22:17:07 about 6 euro's more Mar 17 22:17:38 seems kind of lame.... unless i don't understand it's target or purpose Mar 17 23:21:51 YUN is wifi . Mar 17 23:22:07 no hubs, no power supply, no dongle, no crap. Single board. Mar 17 23:22:58 its lightweight sure .. but look what people are trying to accomplish with the Beagle and Pi ... Mar 17 23:39:02 I have a uart device that I can successfully read from via screen/cat. I would like to create a cpp program to monitor the output and do something if it matches. What would be a good place to start? **** ENDING LOGGING AT Wed Mar 18 02:59:58 2015