**** BEGIN LOGGING AT Thu Mar 17 02:59:58 2016 Mar 17 04:13:06 Running embedded BB Black on a robotic research platform. Robot all of a sudden went offline (after running strong almost 24/7 for nearly a year). Cycled power, power LED blinks once. Been a while since a backup and I'm afraid I've completely lost a year of meticulous OS formatting, any chance of recovery, repair? Mar 17 04:13:40 sounds like a short somewhere Mar 17 04:13:48 a single flash is often the PMIC saying WTF Mar 17 04:14:34 I ran the board under a microscope. No visible signs of damage Mar 17 04:14:58 I am aware of the PMIC issue, but support basically says "youre screwed" Mar 17 04:15:35 Board had a stable PSU, no overvoltage, only occasional hard shutdowns Mar 17 04:16:21 not really Mar 17 04:16:34 the eMMC is probally okay so... Mar 17 04:16:49 dike off the am335x, access eMMC via expansion header Mar 17 04:17:44 it'd be good hot air rework practice Mar 17 04:18:08 Okay, that actually sounds doable, just wanted to make sure that was my only option Mar 17 04:18:19 may want to inspect the PCB layout, maybe you're lucky and can isolate eMMC from the am335x by cutting traces on the surface Mar 17 04:19:39 there's also the pull-ups and power supply to consider Mar 17 04:20:36 Yeah, when my head stops throbbing I'll hit the traces with a scope and do some tests, but it's the data that is the crucial part, the boards just a bummer Mar 17 04:21:05 lemme see if the PC upstairs is still logged in (it usually is), I can take a peek at the pcb layout and see how headachy it looks Mar 17 04:22:33 I mean, there's also a free allegro viewer, but I've tried that thing and it's horrible. not to mention the 3d view was unusably slow Mar 17 04:23:24 Yeah, this might give me a reason to look through the reference manual Mar 17 04:24:38 no pcb layout in there Mar 17 04:29:15 question is... I downloaded the Altium version of the BBB, but that's rev A5B. I don't know if the pcb changes since then are relevant, but I've also never tried importing an Allegro project (which apparently Altium can do) Mar 17 04:30:49 Any tips on where to start testing? I have the files, I'll do some research on it tomorrow because it's getting really late here. Mar 17 04:31:47 ewwwwwwwwwwwwwwwwww Altium Mar 17 04:32:08 Haha, and what would you prefer? Mar 17 04:32:37 just pop the emmc off and put it on a emmc to SD carrier board Mar 17 04:32:44 some of them even come with ZIFs Mar 17 04:33:28 ds2: my experience with schematic/pcb sw is limited but has typically been one filled with horror... so far my (admittedly very limited) experience with it is positive, and afaik my collegues who *do* use it daily like it a lot also Mar 17 04:33:37 removing the eMMC seems like a risky move to me Mar 17 04:33:51 eMMC will be my LAST resort Mar 17 04:35:00 damn, apparently this is an E14 Rev B6... but it has a 4Gb storage Mar 17 04:35:18 I thought rev B's all had 2 Mar 17 04:36:51 Thank you guys anyway, total help! Mar 17 04:38:59 well the only difference between B6 and C is the eMMC Mar 17 04:39:23 hence the pcb may very well be marked B6 Mar 17 04:39:47 also, probably needless to say, but consider making backups in the future :P Mar 17 04:48:57 sweet, you can't import Allegro files unless they've been saved in ascii format (they haven't) or you have Allegro installed (we haven't) Mar 17 04:50:32 where'd he go? Mar 17 04:58:53 <_< Mar 17 04:58:54 >_> Mar 17 04:58:56 * Ferris shrugs Mar 17 05:44:58 Hi everyone! Someone worked with T/R bytes via GPIO (from PRU)? Mar 17 09:16:25 grrr, my code is crashing g++ Mar 17 10:23:48 Anyone sent a single byte from PRU0 to PRU1 via GPIO? Mar 17 12:40:17 pru docs being great again... iloop being "interruptible based on an external break signal" Mar 17 12:40:43 no mention anywhere of what that break signal might be, doesn't seem to be anything r31-related like "slp 1" Mar 17 12:40:46 -.- Mar 17 12:43:30 zmatt: why doesn't that surprise me!? Mar 17 12:47:16 also, pasm uses the mnemonic "iloop" while the new crappy assembler that goes with the C compiler uses "loopi", how convenient Mar 17 12:48:20 Sounds pretty loopy to me. :D Mar 17 12:56:51 Hmm Mar 17 12:57:05 Okay, what causes colour inversion on HDMI on a BBB? Mar 17 12:57:14 Red and Blue channels seem to be swapped Mar 17 12:57:35 lcdc somehow got configured into 24-bit mode instead of 16-bit ? Mar 17 12:57:56 possible Mar 17 12:57:59 Hwo do I check? Mar 17 13:00:21 hmm, maybe my lcd-util outputs it ( http://gerbil.xs4all.nl/lcd-util.tgz ) Mar 17 13:00:27 I don't remember if I include that in the dumped info Mar 17 13:01:14 there's presumably also a place in sysfs that can tell you Mar 17 13:02:14 Seems it's not in the output. no Mar 17 13:02:35 lots of timings and clocks Mar 17 13:02:42 do I include the frame buffer pointers by any chance? Mar 17 13:02:50 dma fb 0: 0x9f200000 - 0x9f5cc000 Mar 17 13:02:50 dma fb 1: 0x9f200000 - 0x9f5cc000 Mar 17 13:03:20 take framebuffer size, divide by number of pixels :) Mar 17 13:03:37 haha Mar 17 13:03:45 Might be easier to just parse the Xorg log then? Mar 17 13:03:52 what's your resolution? Mar 17 13:04:07 *shrug* ;p Mar 17 13:04:15 (that *is* in the output) Mar 17 13:04:18 Looking at the log. Mar 17 13:04:30 active lines, active pixels Mar 17 13:04:38 bbb is only 16 bit color buffer, unlesss they've upgraded it Mar 17 13:04:51 1152x 864 by the look Mar 17 13:04:57 Which is a weird resolution to use Mar 17 13:05:27 echo $(( (0x9f5cc000 - 0x9f200000)/1152/864 )) Mar 17 13:05:29 4 Mar 17 13:05:43 4 bytes per pixel... that means lcdc is configured in 24-bit color mode indeed Mar 17 13:05:46 the internals can do 24 bit, but only 16 bits are actually wired to hdmi Mar 17 13:05:47 mmmm. Mar 17 13:06:33 which would normally just have been a waste of memory and time, except there's a known issue where red and green are swapped in 24-bit mode compared to 16-bit mode Mar 17 13:06:37 So, is it as simple as to beat Xorg.conf with a 16bit hammer, or do I do something more interesting? Mar 17 13:06:47 yeah I recall that's exactly what the default images do Mar 17 13:06:53 Hmm Mar 17 13:06:58 *downloads default image to see * Mar 17 13:07:24 I can also just show you the /etc/X11/xorg.conf Mar 17 13:07:58 True. :) Mar 17 13:08:07 http://pastebin.com/GC2Zw7Xb Mar 17 13:08:49 Mmm Mar 17 13:09:02 the relevant line being "DefaultDepth 16" obviously... dunno if the rest is obligatory scaffolding or whether you can just have a screen section Mar 17 13:09:15 https://xkcd.com/963/ Mar 17 13:09:15 So, split the screen section and toss it into /etc/X11/xorg.conf.d/ Mar 17 13:09:24 hahaha Mar 17 13:14:13 *deploys config* Mar 17 13:14:16 Hmmm Mar 17 13:14:26 I've got to check if I'm using the watchdog properly here. Mar 17 13:16:25 Thanks, that worked. My kiosks are now in the correct colour settings Mar 17 13:52:38 okay. what's the "/opt/scripts/boot/generic-startup.sh" tool doing that I _really_ need in day-to-day? Mar 17 13:52:49 absolutely nothing Mar 17 13:54:31 So why is it eating 6+ seconds of my boot time? *prepares to stab* Mar 17 13:55:04 systemctl disable is your friend Mar 17 13:55:47 Yeah Mar 17 13:55:53 But I'm looking to see what it's doing first ;) Mar 17 13:57:14 ... Mar 17 13:57:31 WTF is it manually editing /etc/issue rather than use templates? *STAB* Mar 17 13:59:19 don't just do ip |grep |grpe to get stuff, just do 6{usb0} 4{usb0} in /etc/issue and it will automatically fill in your ip-address. Mar 17 14:00:23 he probably doesn't know that... I didn't know that until right now either Mar 17 14:00:49 then again i prefer my /etc/issue rather minimal Mar 17 14:00:57 probably not. is there a repo where I can add an issue? Mar 17 14:01:35 `man getty` contains all the funky ones Mar 17 14:14:18 Note to self: IF you ever find the UI designer for HP switches web frontends, buy them drinks until they can't walk and dump them in a back alley to be eaten by rats. Mar 17 14:15:34 lol Mar 17 14:16:12 To get a list of which VLAN's are active on a port.. Mar 17 14:16:23 You have to select each VLAN from the dropdown and then find the port in the list Mar 17 14:16:29 Repeat that for _Every_ vlan Mar 17 14:16:52 If your infrastructure has a few VLAN's you end up with...... Mar 17 14:19:19 In this case I'm about to migrate away from a switch, and I've got a _lot_ of VLAN's active and just a few ports in use, so to get an inventory there's a fuckton of click, scroll-copy, Mar 17 14:19:33 no commandline? Mar 17 14:19:46 of course not Mar 17 14:19:48 or backup/dump thing? Mar 17 14:19:57 Yeah, binary format. Mar 17 14:20:07 ok that does suck Mar 17 14:20:40 hmm Mar 17 14:20:48 Now that I'm thinking, there might be something in SNMP... Mar 17 14:20:50 *goes to look* Mar 17 14:21:14 otherwise WWW::Mechanize to the rescue? Mar 17 14:21:23 yea Mar 17 14:21:38 The question is if it's worth it vs. just do the manual for a one-off task ;P Mar 17 14:23:18 there, updated lcd-util to include the color depth :P Mar 17 14:36:53 Hi everyone Mar 17 14:37:50 I'm looking into buying a Beagleboard XM and I wanted to know if there was still active support for this product since I just got burned with the panda board. Mar 17 14:38:27 that is pretty dead too, just like the Panda Mar 17 14:38:39 I'm not even sure if XMs are still made Mar 17 14:39:13 you might want to check if one of the many versions of the Beaglebone Black/Green/... works for you Mar 17 14:39:21 or explain your use case / requirements Mar 17 14:40:04 I'd love a BBB with SATA :) Mar 17 14:40:28 bitbang it on the PRU ;) Mar 17 14:40:28 I need it to be compatible with Halcon Embedded Mar 17 14:41:11 that looks like commercial software Mar 17 14:41:19 Haha Mar 17 14:41:35 The XM is listed in the "Succesfully tested with" category Mar 17 14:42:10 Its a Library for image analysis Mar 17 14:43:03 does it work on the CPU or does it try to do fancy DSP hardware things? Mar 17 14:44:23 I'm 99.9% sure it runs on the CPU Mar 17 14:45:10 So the the beagleboard is dead but the beaglebone is still active? Mar 17 14:45:25 Also werent you on the other IRC? Mar 17 14:45:33 The pandaboard one Mar 17 14:45:48 yes... Mar 17 14:46:15 so what are you doing with a image processing library that doesn't use hardware acceleration? Mar 17 14:48:43 I does image analysis like barcodes, shape recognition and such. Although now that you mention it i'm not so sure its purely CPU based. I'll ask one of my superiors (I'm just an intern. I have litteraly no idea what im doing)) Mar 17 14:49:40 Hmmm Mar 17 14:49:45 if it uses any of the hw features it'll be fairly bound to a particular SoC Mar 17 14:49:58 sounds like a fancy openCV with a price-tag Mar 17 14:50:14 my impression is that the omap4 is actually more dead than the omap3 ? Mar 17 14:51:04 yes, sort of Mar 17 14:51:15 ah, well omap3 e2e forum has been closed too Mar 17 14:51:23 My boss sais its CPU only Mar 17 14:51:26 omap3 had enough hardware and upstream to stay afloat on the community side Mar 17 14:51:35 http://www.halcon.com/embedded/ - claims it to be DSP on the bottom Mar 17 14:52:00 yeah and TI marketed it for non-cellphone applications also Mar 17 14:52:13 in the form of AM.... yeah Mar 17 14:52:29 AM37xx / DM37xx Mar 17 14:52:30 He sais it can be configured for DSP but that its out of the box its only CPU Mar 17 14:52:45 and do you need the DSP performance? Mar 17 14:52:49 Nope Mar 17 14:52:55 is this a one off thing like a demo or such? Mar 17 14:53:06 or is that a product study / prototype sort of thing? Mar 17 14:53:16 Its a prototype Mar 17 14:53:25 for new stuff I wouldn't go for an omap3 Mar 17 14:53:44 ANd yes halcon is essentially a pricey OpenCV Mar 17 14:53:55 then you might want to use a SoC that you can productize on Mar 17 14:54:30 though unfortunately the SoC series you'd probably want for new development, the AM57xx, are still a bit lacking in boards :P Mar 17 14:54:47 unless the BBB suffices for the job Mar 17 14:54:51 what sort of camera or whatever are you connecting this to? Mar 17 14:54:58 but video isn't exactly its strong point Mar 17 14:55:04 Sorry if my lack experience shines through but what does that even mean? "a SoC that you can productize on" Mar 17 14:55:44 it means a System-on-a-Chip that will be still available and supported by its manufacturer tomorrow and maybe a year or two or however long you need Mar 17 14:56:01 We have several cameras available but if I can get Halcon to work the objective is GiGE camera Mar 17 14:56:04 neither OMAP3 nor OMAP4 are fit for this purpose IMO Mar 17 14:56:31 and how do you connect to a "GiGE camera"? Mar 17 14:56:34 the omaps were cellphone chips... manufactured today = obsolete tomorrow Mar 17 14:56:48 Ethernet cable Mar 17 14:57:04 then a BBB wouldn't actually be that bad an idea Mar 17 14:57:14 do you need a display? HDMI? LCD? Mar 17 14:57:24 the am335x and am57xx (and same chip under other partnames) target industrial and automotive applications... those tend to be a bit more long-term Mar 17 14:57:55 To develop it yes, but we'll probably let it run headless once its setup. Mar 17 14:57:57 AlexEXA: note the BBB doesn't have gigE Mar 17 14:58:01 though the BBB enhanced will Mar 17 14:58:11 http://beagleboard.org/enhanced Mar 17 14:58:52 Halcon comes with a GiGE driver Mar 17 14:59:14 zmatt: I think that's just a product name, not a network requirement Mar 17 14:59:18 But we can probably forego GiGE if need be Mar 17 14:59:21 oh Mar 17 14:59:36 It is Mar 17 14:59:50 so yeah, get a BBB and get cracking Mar 17 15:00:01 When you say BBB do you mean beagle bone black? Mar 17 15:00:15 AlexEXA: the am335x itself actually has dual gigabit ethernet... but the BBB has only one pinned out, and it's connected to a 100 Mbit PHY Mar 17 15:00:20 yes Mar 17 15:00:24 if it works out for you performance wise, then you can either use a BBB as a SoM or get a proper SoM and design a carrier or... Mar 17 15:00:51 SoM? Mar 17 15:00:56 System on Module Mar 17 15:01:25 small board with CPU and a few necessities and a lot of connectivity pins to a 'proper' product board Mar 17 15:01:30 if you do need gigabit ethernet, and don't want to wait for the enhanced, there's other boards such as http://www.ti.com/tool/TMDSSK3358 Mar 17 15:02:03 will the AM335x be able to digest a >100M data stream though? :) Mar 17 15:02:42 Ok Mar 17 15:04:01 So when the page says "Ubuntu" do they mean an actual full Ubuntu distro or do tey say that the same way Microsoft says Win 10 runs on RPi2? Mar 17 15:04:32 tbr: you're saying that as if it's actually much data? Mar 17 15:05:47 AlexEXA: you should go with the default debian images on the BBB. they should be "close enough" to whatever they mean by 'ubuntu' Mar 17 15:06:11 What I mean is that the Win 10 distro for raspberry is essentially just a shell with 0 front end that only serves to mount programs written on an actual computer. Mar 17 15:06:22 Its not really what id call an OS Mar 17 15:07:03 "It's an OS, but not as you know it" to paraphrase Star Trek Mar 17 15:07:24 sounds appealing to me Mar 17 15:07:36 if you insist, you can run a BBB as a _desktop_, but be prepared for performance hits Mar 17 15:07:56 Not I dont mind command line Mar 17 15:08:13 But thats exactly the problem i had with win10 Mar 17 15:08:25 its doesnt even have a command line Mar 17 15:09:13 Just a static screen while it waits for a computer with visual studio to mount a progam on it Mar 17 15:09:38 I have a problem that systemd-udevd is taking up all CPU while and even after I stop my program. Toggling many leds on i2c io-expander via sysfs, seems like udev is catching all these changes, is it possible to make it not catch these events, if these could be the problem? Mar 17 15:12:27 bbs_: could be.. an io-expander generally has a very simple register interface so you may want to consider ditching the kernel driver for it and controlling it directly via i2c Mar 17 15:13:37 zmatt: that actually sounds much better, would I need to remove it from the device tree you think? Mar 17 15:13:43 yes Mar 17 15:13:46 ty Mar 17 15:14:19 if you apt-get install i2c-tools you get some cmdline utils for directly poking i2c devices Mar 17 15:14:27 it's also not hard to do directly from C Mar 17 15:15:06 zmatt: yea I think I go with C, else I need to spawn that tool for every change Mar 17 15:15:42 the tool is still useful for "playing" with the chip first a bit Mar 17 15:16:16 the datasheet of the io-expander we're using is not exactly the most excellent documentation on earth :P Mar 17 15:17:09 zmatt: yeah hehe, using the nxp pca9555 for now, havn't been looking at the datasheet for that since the kernel did a good job with it already :p it just worked Mar 17 15:17:45 sysfs is a pretty horrid interface for gpio though Mar 17 15:18:11 having to write a file for every pin change (instead of generally being able to update all pins at once) Mar 17 15:19:21 yeah, lesson learned, worked until I was changing ~32 / second Mar 17 15:20:00 of course an io-expander will still not be very fast compared to the BBB's own gpio Mar 17 15:20:20 was thinking having an always open file descriptor would make things better, and it did, but hit new limits Mar 17 15:21:01 yeah I know, but it was too many leds to make place for on the bbb's own pins Mar 17 15:21:25 use multiplexing? Mar 17 15:22:54 could have been an option, but our custom board has filled up all pins for other stuff Mar 17 15:23:06 (or if you really want to go nuts https://en.wikipedia.org/wiki/Charlieplexing .. ) Mar 17 15:23:07 [WIKIPEDIA] Charlieplexing | "Charlieplexing is a technique for driving a multiplexed display in which relatively few I/O pins on a microcontroller are used to drive an array of LEDs.The method uses the tri-state logic capabilities of microcontrollers in order to gain efficiency over traditional multiplexing. Although it is more..." Mar 17 15:23:11 wtf Mar 17 15:23:22 since when do we have a spammy bot? Mar 17 15:23:39 no idea :p Mar 17 15:24:20 interesting idea, might be something for the future Mar 17 15:24:34 no time for redesigning now :) Mar 17 15:25:10 yeah and you'd need to produce periodic output on gpios... I'd just setup dma from a circular buffer (yes, from userspace) but I'm nuts ;) Mar 17 15:25:32 haha Mar 17 15:26:57 brb Mar 17 15:27:01 ... Mar 17 15:27:12 I just got rate limited by a _power_ mater Mar 17 15:27:14 meter Mar 17 15:27:19 For energy efficiency.. Mar 17 15:27:22 FUCK YOU SCADA. Mar 17 15:43:07 zmatt: there's been a couple tested here lately .. Mar 17 15:43:15 zmatt: puh good thing the other pins which were connected to the same io-expander only needs to be used after the program is running Mar 17 15:43:20 complain to the @s Mar 17 15:44:14 ^ oh the irony :p Mar 17 16:09:31 Hey its me again Mar 17 16:10:06 I just realised I forgot to thank tbr and zmatt for all the help Mar 17 16:10:11 Thanks guys Mar 17 16:11:21 ++AlexEXA Mar 17 16:15:17 np, glad to help Mar 17 16:17:58 I love hardware. I go to replace an ethernet cable and of course I come back bleeding. Mar 17 16:18:29 I'm just happy I don't have to get u-scsi working, it's so hard to find living goats for sacrifice Mar 17 16:21:24 Hello I want to build project Beagleboy Mar 17 16:21:47 But I dont know how start Mar 17 16:28:40 hi Mar 17 16:29:24 I have some ideas for the GSoC, should I just post them on the ideas page? Mar 17 16:45:38 Spidler: That's good. I like that. Mar 17 16:45:45 * Nevyn|Work laughs at his desk Mar 17 16:46:53 * Spidler eyes Nevyn Mar 17 16:47:00 Shouldn't you be off dreaming somewhere? ;-) Mar 17 16:47:42 I'm in the middle of an online ITIL training class. IRC *IS* where I'm off dreaming at. Mar 17 16:47:51 * Spidler laughs Mar 17 16:48:11 Damned, I know the name but I can't remember which books it was from, and I'm too honest to google it ;P Mar 17 16:48:17 Titles escape me. *sigh* Mar 17 16:48:20 I worked for a company in 2011 that still had an actively used SCO Unix server. We claimed it ran on voodoo magic and chicken blood. Mar 17 16:48:29 Hahaha Mar 17 16:48:39 Nevyn, Master of the Aethyr, long of Deverry! Mar 17 16:49:07 That it was, the dreamer, the fuckup looking for his messed up lovelife, chasing after people _way_ too young for him. Mar 17 16:49:08 Spidler: I can't tell you the name(s) of the books, but they were written by Katherine Kerr. Mar 17 16:49:17 Indeed. Mar 17 16:50:02 I don't know that I've EVER ran into someone who recognized it from the source - And I've used some variation on it for 15 years. Mar 17 16:50:26 Haha Mar 17 16:50:38 I once ran into a guy who used "Asmodean" as his handle - a Wheel of Time reference. Mar 17 16:50:46 Yeah Mar 17 16:50:52 simmondscd: that's one option .. or we can bug jkridner too :D Mar 17 16:50:57 I've got a friend who still uses "Aginor" as a nick .:) Mar 17 16:51:16 Hang on. I used to know a guy some.. .15 years ago or so that used Nevyn as a nick... Mar 17 16:51:32 Don't tell me you're from somewhere in Sweden originally? Mar 17 16:51:52 Probably not me. I started out using "aethyrmaster" - Oh, DEFINITELY not me. Mar 17 16:52:48 Ah, Definitely not then. :) Mar 17 16:52:56 Daggerspell. That was the book. Mar 17 16:54:11 Hi anyone available? Mar 17 16:54:24 Ah right, it was Mar 17 16:54:42 Omar: We're here, if that's what you're asking, but I'm not available, sorry. I've already got a partner. Mar 17 16:54:50 ( Hang on, that wasn't a dating question, was it? ) Mar 17 16:55:10 How start with project Beagleboy? Mar 17 16:55:46 veremit: you know I don't read scrollback buffers. :-) Mar 17 16:55:49 Beagleboy? No clue. Mar 17 16:56:17 jkridner: that's fine .. my client highlights pings :P you need me to reiterate? Mar 17 16:56:33 Sorry Omar Mar 17 16:56:58 Thanks same You are great person Mar 17 16:56:58 jkridner: simmondscd was asking about GSoC project ideas .. Mar 17 16:59:10 Alguien habla español? Mar 17 16:59:10 UnicodeEncodeError: 'ascii' codec can't encode character u'\xf1' in position 18: ordinal not in range(128) (file "/home/ahiam/.sopel/modules/gitter_bridge.py", line 52, in ircToGitter) Mar 17 16:59:24 Alguien habla espaniol Mar 17 16:59:26 ? Mar 17 16:59:28 Un poco. Mar 17 17:00:42 https://github.com/CapelessCrusaders/BeagleBoy este es el link del proyecto Mar 17 17:00:51 Por donde puedo empezar? Mar 17 17:01:28 Just use Beaglebone for read and write I/Os Mar 17 17:03:10 Sadly, I still do not know. Mar 17 17:03:46 veremit: ah! did you send him toward #beagle-gsoc? Mar 17 17:03:59 Gaming Capes have started shipping. Mar 17 17:04:06 jkridner: didn't think of that ... Mar 17 17:07:27 jkridner: whilst you're near the keyboard .. and rcn-ee is absent .. any ideas what the rational behind the default B* images now containing RT kernels .. or is that accidental?! Mar 17 17:07:43 definitely not accidental. Mar 17 17:08:03 is there a possibility of getting both RT and non-RT options built? Mar 17 17:08:23 I know there's a small bias towards robotics now ;P Mar 17 17:09:22 the rational is that real-time performance is a key differentiator for BeagleBone. Mar 17 17:09:26 jkridner: was disappointed to see RN's collection of sw images disappear from the elinux Beaglebone:Debian page :( Mar 17 17:09:32 veremit: they are both built... just an apt-get away. Mar 17 17:09:57 ah cool .. is there a way .. let me guess .. apt-cache search? Mar 17 17:10:21 Hmmmm. has anyone got a CentOS image setup built and tested? Mar 17 17:10:31 * veremit needs to push the gentoo-embedded folks to get a BBB page going. Mar 17 17:10:49 ere nerdboy .. :P lol Mar 17 17:10:56 I'd need to wiki around. Mar 17 17:13:02 jkridner: except RT kernels only get you anything if you actually set real-time priorities on processes designed to treat those privileges with respect Mar 17 17:13:23 otherwise they're just less efficient and buggier kernels Mar 17 17:13:52 zmatt, jkridner: backs up my research on such Mar 17 17:14:29 zmatt: think its documented as such in the kernel docs, no? Mar 17 17:14:43 yea Mar 17 17:14:49 I know I read it somewhere Mar 17 17:14:58 Peak and bulk performance are quite a bit down on -rt kernels Mar 17 17:14:59 notably, I once noticed that ping-flooding a BBB running a non-RT kernel caused hiccups in the GUI, doing the same on the equivalent RT kernel completely froze the GUI Mar 17 17:15:14 Mmm Mar 17 17:15:24 It's pretty easy to cause starvation on RT systems. Mar 17 17:15:26 I would say that's quite undesierable Mar 17 17:15:30 * Spidler goes off to kill his cache layer. Mar 17 17:15:35 erm. Undesireable Mar 17 17:15:44 I should however note that on an RT kernel I could actually completely solve that Mar 17 17:15:55 by demoting the network irq handlers to SCHED_OTHER Mar 17 17:16:10 that's something you can't do on an RT kernel Mar 17 17:16:26 on an non-rt you mean? Mar 17 17:16:34 ^^ that Mar 17 17:16:35 lol Mar 17 17:16:39 non-RT Mar 17 17:16:40 yes Mar 17 17:17:09 but it does highlight an important point: RT kernels give you the knobs, you still need to manually tune them Mar 17 17:17:12 the wiki stuff for embedded was just updated Mar 17 17:17:16 Mmm Mar 17 17:17:27 *gentoo wiki even Mar 17 17:17:29 and they have considerable overhead Mar 17 17:17:32 Even though I do 'embedded' stuff I don't bother with RT kernels Mar 17 17:17:41 I prefer bulk & Batch :) Mar 17 17:18:26 basically, on a non-SMP non-RT kernel every "spinlock" is actually just "disable irqs" Mar 17 17:18:31 and they're around quite a bit Mar 17 17:18:40 Mmmm Mar 17 17:18:42 an RT kernel turns those into priority-inheritance mutexes Mar 17 17:18:49 and every irq handler into a kernel thread Mar 17 17:18:59 Which has it's own problems. Mar 17 17:18:59 (since otherwise they wouldn't be able to take a mutex) Mar 17 17:19:57 ( Sidenote. I hate debian's handling of locales. ) Mar 17 17:20:33 so, RT kernels can be pretty neat if you really need them Mar 17 17:20:55 But don't expect better performance. Mar 17 17:21:22 don't expect anything positive to happen by magic Mar 17 17:21:36 they just give you tools... you still need to put in hard work to use them Mar 17 17:22:16 Magic [| ] More magic. don't flip the switch. Mar 17 17:22:22 hehehe Mar 17 17:22:36 it was left on more-magic I'm pretty sure Mar 17 17:23:32 but basically, it's pretty powerful to be able to prioritize a userspace thread above most of the kernel code Mar 17 17:24:06 but then you do need to make sure that thread does not directly or indirectly make any use of the kernel functionality it's being prioritized over Mar 17 17:29:12 Mmmm. Mar 17 17:30:39 so given that I know most people will not need or actually use RT functionality, and the fact they adversely impact performance and stability, I actually recommend replacing rcn-ee's default RT kernel by a non-RT one as one of the first things to do after a fresh install Mar 17 17:31:22 I'm in complete agreement Mar 17 17:31:43 Currently I'd recommend the 4.1 -bone series Mar 17 17:31:54 As most people might not be interested in upgrading the kernel at all time. Mar 17 17:31:55 I use 4.4-bone actually Mar 17 17:31:59 Same here Mar 17 17:32:12 well, custom-compiled kernel based on 4.4-bone Mar 17 17:32:50 Just upstream 4.4 :) Mar 17 17:33:17 Slimming it down would be nice, but is currently not worth the effort. Mar 17 17:33:21 4.1-ti might still be preferred for some purposes... like I'm not sure whether all the power management stuff is there yet Mar 17 17:33:48 I think it would be good with a nice and conscise page for beginners that explain the differences in the kernels Mar 17 17:34:08 Now Mar 17 17:34:22 I shall go AFK and edit fucking excel files Mar 17 17:34:43 recent kernels finally introduced using gpios as spi chip selects Mar 17 17:34:51 still buggy, but hey it's a start Mar 17 17:35:59 Could an RT kernel be 'split' to simultaneously use a PRU to drive the RT parts of the OS, kind of treating the BBB as a dual (or tri-) core CPU? Mar 17 17:36:35 That would require you to flip the switch into | [more magic] Mar 17 17:37:10 Nevyn|Work: that makes no sense.... linux cannot obviously not view the PRU as an SMP core Mar 17 17:37:35 they have 8 KB of instruction memory :P Mar 17 17:37:37 zmatt: Well no - it'd be AMP, not SMP Mar 17 17:37:51 zmatt: but they can reference out to the main DDR3L mem. Mar 17 17:37:58 not code Mar 17 17:38:30 ah. Mar 17 17:38:43 * Nevyn|Work #MerelyCurious #HasNoIdea Mar 17 17:39:13 and asymmetric multiprocessing is exactly right: if you can manage your real-time needs on them, you avoid the need for an RT linux kernel Mar 17 17:39:33 (and PRU will get you more real-time then an RT linux kernel will ever get you) Mar 17 17:40:40 like, on an RT linux kernel I measured the latency of an irq-triggered process that toggled a gpio in response to that Mar 17 17:40:55 latency was about 40-80 microseconds Mar 17 17:41:45 (gpio-triggered irq) Mar 17 17:41:51 Ah. Thanks for giving an answer; I know I ask weird questions but how does one test limits without doing so? Or how does one expand knowledge? Mar 17 17:42:10 doing the same on a PRU that's waiting for a GPIO gives a latency of about 25-30 nanoseconds Mar 17 17:43:01 so basically a 2000 point improvement .. Mar 17 17:44:47 one could argue that the comparison is unfair since that PRU core would be dedicated to the task... but having a high-RT-priority task waiting for an event is the best you can do on an RT linux system, so it is fair in the sense of showing the best-case limits of both Mar 17 17:45:18 Nevyn|Work: I'm not sure I follow your question Mar 17 17:45:31 I suspect it was rhetorical :) Mar 17 17:46:15 I expanded my knowledge by writing a test program and hooking up a scope to see the delay between input edge and output edge ;) Mar 17 17:46:27 zmatt: It was rhetorical. Was merely a "thank you for not telling me to shut up and leave and calling me a stupid troll for asking a weird question that has a seemingly obvious answer if you're into coding" Mar 17 17:47:06 ah sorry for some reason my eyes missed part of what you said and misparsed the rest Mar 17 17:47:21 I blame fonts. :) Mar 17 17:47:30 * Nevyn|Work gives zmatt thumbs-up Mar 17 17:49:12 the PRU cores are pretty awesome engineering though Mar 17 17:50:20 being completely non-pipelined, even though they have instructions like 32-bit compare-and-branch Mar 17 17:51:17 Oh, nifty. Mar 17 17:51:29 I haven't looked more than compiling a hello world for the pru Mar 17 17:51:39 Mostly because I couldn't do anything useful with them for my setup ;) Mar 17 17:51:59 I've got all this IO and I'm not allowed to use it. Mar 17 17:52:10 ( Too easy to fry the boards. ) Mar 17 17:52:31 If I were a code/programmer guy, I would TOTALLY fork this one 3d printer firmware for the PRU's...but I don't have the skills/XP to do that. Mar 17 17:53:02 jkridner: are you reading our review of RT vs non-RT kernels?! :P Mar 17 17:53:14 pretty sure there is already a project for that, Nevyn|Work Mar 17 17:53:31 so that means: *clock* program counter becomes available to instruction ram, the instruction is looked up, decoded, two operands are muxed out of its register file (32x 32-bit), unsigned comparison done, this decides whether the next program counter will be current pc+1 or pc+branch_offset Mar 17 17:53:39 just in time for *clock* Mar 17 17:53:44 jkridner has had a development printer at i3Detroit a few times, working on it. Mar 17 17:53:47 that's 5 nanoseconds of time to do all that Mar 17 17:54:02 agmlego: There's a couple that use the PRU's, (BeBoPr++,CRAMPS, Replicape), but I'd like to see how Smoothieware runs on it. Mar 17 17:54:48 ah. Mar 17 17:55:03 I've seen several using arduino/Atmel targets .. they don't need barely any 'processing' Mar 17 17:56:05 depends on the robot. Mar 17 17:56:36 deltas are a hell of a lot smoother on a 32-bit controller. Mar 17 17:57:14 and with the bone, maybe the slicing could be done onboard. Mar 17 17:57:15 the pru cores even have a 32x32 -> 64 bit multiplier Mar 17 17:57:37 zmatt: at that level, they bloody well should .. it ain't an 8 bit micro ffs. Mar 17 17:58:25 veremit: well they weren't originally designed for doing any serious arithmetic afaik Mar 17 17:58:42 but bit twiddling Mar 17 17:58:54 I should imagine there's one in the cortex-a7?! :P Mar 17 17:59:04 agmlego: That's kind of what I'm thinking: the main CPU for network interface and slicing, and then offload to PRU for printing. Mar 17 17:59:15 ehh, a cortex-a can't be compared to a PRU core Mar 17 18:00:07 zmatt: true .. but even the m has something, no? Mar 17 18:00:47 1. the m is 3-stage pipelined Mar 17 18:01:12 2. on a cortex-m3, a 32x32->64 multiply takes 3-7 cycles Mar 17 18:01:44 3. maybe they exist, but I don't recall having seen a cortex-m3 at 200 MHz ? Mar 17 18:02:31 not at 200mhz .. certainly not1ghz Yet. Mar 17 18:02:43 ok, they do exist according to google Mar 17 18:02:47 haha Mar 17 18:02:48 but I had to google to know that :P Mar 17 18:03:00 yeah you're good :P lol Mar 17 18:03:00 and again, 3-stage pipeline Mar 17 18:03:14 that means the work of an instruction is spread across 3 cycles Mar 17 18:03:18 PRU needs to do it in 1 Mar 17 18:03:20 pru doesn't do it single instruction? Mar 17 18:03:24 get away Mar 17 18:04:00 all PRU instructions are single-cycle except external bus access Mar 17 18:04:15 at 200mhz? a single-cycle multiply? Mar 17 18:05:35 there are also three "scratchpads" in pruss of 30 x 32 bits... a pru core can copy or exchange any contiguous byte-range of its 30 gp registers (r0-r29) to/from one of the pads in a single cycle Mar 17 18:06:56 it can also use its register file as a tiny RAM by doing indexed access in it, e.g. mvid r3, *r1.b0++ Mar 17 18:07:22 (where r1.b0 contains a byte-offset into the register file) Mar 17 18:08:14 somebody spent some time actualyl designing that .. rather than gluing a few gates together Mar 17 18:08:18 the only annoying thing about the multiplier is that it's hooked up via the broadside interface (same as the scratchpads) and uses fixed source/destination registers Mar 17 18:08:57 but in non-accumulate mode it samples the source regs and performs the multiply every cycle, so you just need to gets the args in there and XIN the result Mar 17 18:11:14 but yes, this thing has been seriously designed Mar 17 18:11:21 it does crazy shit in single cycles Mar 17 18:11:26 at fucking 200 MHz Mar 17 18:11:28 :D Mar 17 18:12:33 I still don't know where it comes from originally... i.e. where PRU core v0 can be found Mar 17 18:13:03 I did notice Keystone processors have TONS of 'em, under the alias "PDSP" (Packed Data Structure Processor) Mar 17 18:14:05 e.g. the K2E/K2L packet accelerator has 15 cores to inspect, classify, manipulate, and route packets Mar 17 18:15:42 I just heard the story some months ago, but... I don't remember =) Mar 17 18:16:11 it has had many names Mar 17 18:16:40 tomba: the keystone cores seem to be v2 already, but "PDSP" appears to be an older name than PRU Mar 17 18:16:54 ("PDSP" can be found in PRU docs, not other way around) Mar 17 18:18:21 maybe the older communications processors also had some around... but not much publicly known Mar 17 18:19:50 the TCI6498 apparently did have a CPINTC, which is the interrupt controller also found in PRUSS (and instantiated in a different configuration as the keystone Chip Interrupt Controller) Mar 17 18:20:31 I've learned more about those PRU's from reading this chat than from reading the documentation :) Mar 17 18:20:51 the TCI6498 apparently doesn't exist however, other than in some sysbios packages and such :P Mar 17 18:21:29 zmatt: nsekhar told me the story, but he doesn't seem to be on this channel Mar 17 18:21:52 ok, this is a weird commit... http://git.ti.com/keystone-rtos/edma3_lld/commit/bc53585317e513b425e771856efed4f6ffe40515/diffs Mar 17 18:23:55 tomba: I'm always curious about the history of things Mar 17 18:24:40 often you can spot peripherals or modules that have had a history elsewhere Mar 17 18:25:26 like the PWMSS peripherals are clearly foreign to both the OMAP and the C6x lineages Mar 17 18:25:38 (and in fact they're actually C2000 peripherals) Mar 17 18:27:17 Or in case like the piece of shit I'm trying to implement right now... Mar 17 18:27:33 time formats: 192:27:33(h:m:s) =37498 Mar 17 18:28:13 2740=off V/mA , 2750=1V/1mA ..... Mar 17 18:28:53 OF course. 2740 is a _perfect_ value for "off" . If you know the history that 2740 is factor 10 off Degrees Kelvin elsewhere in this controller. Mar 17 18:29:18 Which means that 2740 is converted to 0 Degrees C, which might suitably be useful for "Off" . Mar 17 18:29:24 lol Mar 17 18:29:53 Of course you express Volts _or_ amperes in a unit scaled up for Temperatures. Mar 17 18:30:09 one scale to rule them all Mar 17 18:30:14 Oh, no Mar 17 18:30:24 Other things it expresses in 0-10 with a factor 100 Mar 17 18:30:35 Especially Voltages in other places. Mar 17 18:31:10 Except when you _enter_ a voltage, then you enter a value between 0-32k that matches a digitalized input/ouput of a PT100 sensor to express a value between 0-10V Mar 17 18:31:38 Now please execute me, becuase there's not enough alcohol in the world to cope with this shit. Mar 17 18:31:53 btw, "I've learned more about those PRU's from reading this chat than from reading the documentation :)" ... that's probably in no small part because PRU documentation consists of a random bunch of snippets scattered in random places Mar 17 18:31:57 :P Mar 17 18:32:08 Mmm Mar 17 18:32:19 Now, I have the spec sheets for 6 different of these pieces of shit Mar 17 18:32:22 Different programming Mar 17 18:32:26 Different unit conversions Mar 17 18:32:31 Everything sucks and it's all horrible. Mar 17 18:32:47 And it's all fucking legacy and layers of mud Mar 17 18:32:50 zmatt: you should know by now .. there *is* no documentation for the am335x .. or for the BBB .. etc. Mar 17 18:33:01 Spidler: welcome to engineering! :D Mar 17 18:33:09 veremit: well, there's a 5000-page TRM, which isn't "no documentation" Mar 17 18:33:13 there's also a PRU chapter even Mar 17 18:33:22 zmatt: bog roll or firelighters? Mar 17 18:33:23 in fact, three versions thereof, each having a different subset of info Mar 17 18:33:40 hmm .. I'm not sure you can burn a PDF -sigh- :/ Mar 17 18:33:59 no, I'll be fair, while I can complain plenty about TI's docs, I know that in fact they're really quite good relatively speaking Mar 17 18:34:16 shades of grey :) Mar 17 18:34:21 I mean, have you ever seen that joke that rpi developers have to make do with? Mar 17 18:34:33 oh you mean "broadcom documentation" Mar 17 18:34:35 (and the wiki-maintained errata since broadcom doesn't publish any official ones) Mar 17 18:34:36 that's all locked away Mar 17 18:34:43 and rightly so Mar 17 18:34:47 stupid f*in eben. Mar 17 18:34:55 just cos he works for the co. Mar 17 18:35:03 Because reading it would cause Cthulhu to rise up? Mar 17 18:35:06 Well seriously Mar 17 18:35:19 "Here, have a binary blob bootloader" Mar 17 18:35:19 he's been lucky they sold so many .. broadcom now take him seriously .. because they can flog some chips. Mar 17 18:35:40 still no docs though Mar 17 18:35:58 they're Applications processors ffs though Mar 17 18:36:20 who *needs* docs .. except if you're trying to glue on an FOS project onto the side of them. Mar 17 18:36:35 Again .. he's been Lucky. Mar 17 18:36:38 the addendum on the quad-a7 is funny though... "I couldn't find what the speed of this timer is supposed to be" or something like that is actually written in there Mar 17 18:38:02 In my old job we had some early engineering samples of the Opteron cores Mar 17 18:38:17 so TI puts the datasheet (a real one, not a 4-page glossy advertisement) as the first link at the top of the product page, and the errata as the second link Mar 17 18:38:19 There was some seriously funky shit going on there, as they weren't quite documented yet ;) Mar 17 18:38:21 I really do respect them for that Mar 17 18:40:28 I can understand and (usually) forgive chip bugs... they're complicated things and made by humans. but if I'd end up wasting a lot of time due to an erratum that's known but the manufacturer didn't bother to publish I'd be pissed :P Mar 17 18:41:25 I'm sure the people who found out the hard way that the processor on the rpi randomly corrupts data if an i2c slave uses clock stretching had loads of fun Mar 17 18:42:05 hello everyone, I am Pranjal Bhor and I want to apply to Google Summer of Code (GSoc) this year. Can anyone help me regarding it? Mar 17 18:43:44 psmlbhor: The channel is #Beagle-gsoc Mar 17 18:44:15 oh! sorry. Thank you Mar 17 18:44:23 No worries Mar 17 18:49:44 veremit: I wonder if I'll hear back from the PRU experts... https://e2e.ti.com/support/arm/sitara_arm/f/791/t/499000 Mar 17 18:53:46 zmatt: nice Mar 17 19:47:08 The external break signal comes from an alien civilization poised to attack Earth, watch out Mar 17 20:07:48 ..... Mar 17 20:08:05 "Default: 0" "Min value": 45 "Max value": 100 .... Mar 17 20:08:22 Who the Fuck in their right brain would _ever_ configure something where the default is not in the legal range?! Mar 17 20:08:33 Someone left handed Mar 17 20:08:47 But yeah, that made me 1/2 smile 1/2 cringe Mar 17 20:09:31 Got a few other ones here in the same file Mar 17 20:09:37 legal range: 0- 9 Mar 17 20:09:41 default: 258 Mar 17 20:27:00 zmatt, freescale/nxp also has "full" datasheets and reference manuals (at least for the imx6solox) publicly available, so thats also nice of them, but the devboards obviously can't compete with a bbb or raspi from a price point for the hobby market since they are all >250euro Mar 17 20:28:02 and as far as i'm heard from some hw guys here, the imx6 chips themselves are more expensive as well Mar 17 20:29:22 but ye, i also hate it that the people think rasbpi is super open and everything, but as soon as you want to deep digger you're out of luck Mar 17 20:29:34 *dig deeper Mar 17 20:36:38 open, *cough* Mar 17 20:36:58 they still haven't even released the schematics of the rpi3 Mar 17 20:37:25 not that they're terribly useful, since you won't be able to make a similar design since you simply can't buy the processor Mar 17 20:38:00 are the rpi1 schematics even available? Mar 17 20:38:26 https://www.raspberrypi.org/documentation/hardware/raspberrypi/schematics/README.md Mar 17 20:40:23 ahh ok :D but ye, i read somewhere that for own designs you should use the compute module ... which costs ~40$ Mar 17 20:42:35 which gets you the BCM2835 with its lovely arm11 cpu and broken glue logic onto the bcm2708 interconnect Mar 17 20:42:59 yes and still no datasheet Mar 17 20:43:29 actually i'm really curious if someone is using the compute module for real products, my guess would be no Mar 17 20:43:49 the source code dump broadcom once did is informative though... contains quite uncensored headers Mar 17 20:46:48 the glue logic bug is hilarious though... the arm11 can issue up to two transactions with the same tag. AXI allows this and requires them to be responded to in-order. this was a dumb move from arm I'll admit, but it's still broadcom's fault for not taking that info account when glueing it onto a typical interconnect which doesn't guarantee order of responses from different targets Mar 17 20:47:24 the net result is that if you do int x = (read from peripheral A); int y = (read from peripheral B); Mar 17 20:47:40 and peripheral B responds before peripheral A Mar 17 20:47:53 and the response from peripheral B ends up in x, the response from peripheral A ends up in y Mar 17 20:48:33 so any code that interacts with more than one target needs to insert memory barriers to prevent this Mar 17 20:48:39 jolly fun XD Mar 17 20:51:42 so you need a patched gcc for this soc? Mar 17 20:52:04 no, gcc has no way of knowing when this is necessary Mar 17 20:52:15 kernel devs simply have to take it into account Mar 17 20:52:24 oh even more fun Mar 17 20:52:26 fortunately most driver code only interacts with a single peripheral Mar 17 20:52:40 and irq entry/exit already acts like a barrier Mar 17 20:54:06 * Spidler sighs Mar 17 20:54:08 Meeep Mar 17 20:54:21 filt3r: the solid run boards are a bit less then 250 Mar 17 20:54:27 Being told "hi, you need support secure boot soon" is perfectly what I want to hear. Mar 17 20:54:53 "on what fucking hardware" " dunno, you find something" ... <.< *shifty eyes* Mar 17 20:55:18 Spidler: imx5 Mar 17 20:55:42 or almost any microcontroller, but I'm assuming that's not what you meant Mar 17 20:55:45 Mmmm Mar 17 20:56:10 I don't really look forwards to adding a new hardware target to my build systems and distribution maintenance. Mar 17 20:56:19 That's not the kind of thing that makes me feel good about life Mar 17 20:57:30 Spidler, oh the solidrun boards look nice and not that pricey, also at leas imx6 has top notch yocto integration, in case you are using yocto Mar 17 20:57:32 the stupid thing is, any SoC could easily support secure boot using a dynamic root of trust... it sure as hell doesn't require public key crypto in ROM or hw, or customization other than a randomly generated key per chip Mar 17 20:58:09 instead, TI stuck a whole fucking kernel of some sort into secrom Mar 17 20:58:37 Yeah. Secure boot is just a hash store and verify Mar 17 20:58:38 and if you want a HS version, you'd better want a lot of them Mar 17 20:58:39 Hardly magic Mar 17 20:59:24 And while I _do_ want something with hardware key storage /generation for _other_ uses, that's not really magic either. Mar 17 20:59:29 hmmm, not quite since that doesn't get you a root of trust, and "hash store" suggests you'd need to program the hash into the chip which is wholly unnecessary Mar 17 20:59:38 filt3r: Sounds good, and no. not yocto. Mar 17 20:59:53 I have a device tree issue, I don't know what the pinmux offset of GPIO0_19 is. It's not listed at https://github.com/jadonk/bonescript/blob/64732854d4c296de8792d5d4c49639ae5f5ecede/src/bone.js Mar 17 21:00:06 zmatt: True, but I really just want a measured boot where the software can say "fail, die now" . Mar 17 21:00:28 I need to configure GPIO0_19 as gpio so I can turn off the 12 MHz hdmi framer clock that's coming out of it. Mar 17 21:00:31 Anyhow, that's adiscussion for when I'm not suicidal after dealing iwth stupid fucking FAN controllers all evening Mar 17 21:00:48 Right now, it's 22, and I haven't eaten since lunch and there's no bloody beer in the house. Mar 17 21:01:35 Spidler: the trick is: do the measurement and then execute the code providing HMAC(devicekey, measurements) as argument Mar 17 21:01:48 it can then use that as key for its private shit Mar 17 21:01:56 Yeaps. Mar 17 21:02:07 feel free to load any other code, but it won't be able to decrypt it Mar 17 21:02:15 that + dm_verfiy Mar 17 21:02:57 But honestly. I don't care if the drive is encrypted or not, it only matters if it matches the pre-calculated signature or not Mar 17 21:03:16 I guess it depends on what you mean by "secure boot" Mar 17 21:03:16 The usecase is integrity, not making it peek-proof. Mar 17 21:05:17 without knowing more context I can't say if that makes sense, but it certainly sounds more complicated Mar 17 21:05:40 (not the checking what's loaded onto the cpu part, but the broader framework of trust you expect you achieve by that) Mar 17 21:05:49 Mmmm Mar 17 21:06:17 like, if I can peek at what it does, I can yank it out and replace it by something similar but that does something subtly different Mar 17 21:06:28 Yeap Mar 17 21:07:00 The goal is a hash-chain, where each part is matched by the part before it in the chain Mar 17 21:07:12 Heading off now Mar 17 21:07:44 GPIO0_10 has a pinmux register offset of 0x0d8. Then GPIO0_11, 12, 13, 14, 15 each have an offset that increases by 0x004 from there. Mar 17 21:07:49 ecdhe: https://goo.gl/Jkcg0w Mar 17 21:08:15 ecdhe: the Pins tab is general for the CPU, the BBB tab includes BBB-specific info Mar 17 21:08:54 there are filter views (menu Data -> Filter Views) for sorting e.g. by GPIO Mar 17 21:09:30 the pin number is the index into the pinmux array Mar 17 21:09:50 (i.e. multiply by 4 to get the offset) Mar 17 21:11:15 Gnite all! Mar 17 21:12:26 zmatt, this looks useful. Mar 17 21:15:13 zmatt, from https://github.com/jadonk/bonescript/blob/64732854d4c296de8792d5d4c49639ae5f5ecede/src/bone.js : GPIO0_22 has a pinmux register offset of 0x020. Mar 17 21:17:03 So you're exactly right. GPIO0_22 is pin 8, and 8 times 4 is 32 (dec) or 0x20. Mar 17 21:17:16 Thanks!!!! Mar 17 21:17:38 filt3r: wandboard.org Mar 17 21:20:57 yeye, i take it back, there are reasonably priced imx devboards :D Mar 17 21:30:02 they're really quite good. Mar 17 21:30:41 interesting Mar 17 21:31:12 although not suited to everything Mar 17 21:31:56 I don't -think- they suffer usb issues .. I've *never* had any .. and the onboard wifi of the dual+ is a real bonus. Mar 17 21:32:20 native sata on the quad is cool too :D Mar 17 21:32:54 only other boards with that are allwinner soc's. eg. banana-pi and orange-pi Mar 17 21:33:10 and the upcoming pine64 Mar 17 21:33:15 I got pcduino3 Mar 17 21:33:33 But I ont like the allwinners violation of gpl Mar 17 21:33:45 so wont buy that again Mar 17 21:34:24 https://linux-sunxi.org/GPL_Violations if you are interested Mar 17 21:35:39 yep the wandboard looks quite nice feature wise Mar 17 21:36:03 yeah the allwinner stuff is a bit meh. its cheap, I'll give it that. Mar 17 21:36:31 io ain't quite so easy on the wandboard .. but if you design your own carrier board (edm) you've got a LOT of flexibility Mar 17 21:36:42 the edm standard is quite good Mar 17 21:38:13 I had *miserable* luck with an rtl8188 wifi chip on an olimex wifi board. drivers are *** .. I may be no fan of broadcom .. but their wifi does *work* Mar 17 21:39:03 you've even got a low-power sound codec on the bottom-end wandboards :) Mar 17 21:39:23 http://acmesystems.it/arietta I got this one, with the wifi module on Mar 17 21:39:44 so that's +working-usb +0.5gb ethernet +sound on all boards, +wifi on dual-core, +sata on quad. Mar 17 21:41:23 epipe: definitely small .. but how have you found it for usefulness? Mar 17 21:42:18 I needed a small linux device, that would act like a usb storage device Mar 17 21:43:15 a machine that would normally output data to a printer or usb. Now the arietta puts it right to the database Mar 17 21:43:51 Also maybe I will use the 3 serial connections in other projekt Mar 17 21:44:07 But I just got a beaglebone Mar 17 21:44:17 yeah the bbb got io + 3/4 uarts .. Mar 17 21:44:36 its not the black Mar 17 21:44:49 I want to lash one up to test some patches for rs485 .. then I got an ethernet->rs485 bridge :D Mar 17 21:44:58 oh .. one of the expensive ones?! :P Mar 17 21:45:01 or the green Mar 17 21:45:13 its white. I got it used from ebay Mar 17 21:45:18 ah ok Mar 17 21:45:35 is it more expensive than the black? and why? Mar 17 21:45:45 seems like its older version Mar 17 21:45:58 black should be around $40-50 .. its the latest one .. with 4GB eMMC onboard Mar 17 21:46:04 A3 it says on the ethernet Mar 17 21:46:30 its £30 here .. be stupid-money on ebay .. because there have been supply problems until element14 got involved Mar 17 21:46:30 UnicodeEncodeError: 'ascii' codec can't encode character u'\xa3' in position 4: ordinal not in range(128) (file "/home/ahiam/.sopel/modules/gitter_bridge.py", line 52, in ircToGitter) Mar 17 21:46:49 !whois ahaim Mar 17 21:46:52 gcl-bot, :D Mar 17 21:47:08 !whois ahiam ** Mar 17 21:47:27 aha I can see him ... ~~~~~> Mar 17 21:47:53 one of the 260 lurkers Mar 17 21:48:01 or 250 bots .. I dunno haha Mar 17 21:48:12 £ Mar 17 21:48:13 UnicodeEncodeError: 'ascii' codec can't encode character u'\xa3' in position 0: ordinal not in range(128) (file "/home/ahiam/.sopel/modules/gitter_bridge.py", line 52, in ircToGitter) Mar 17 21:48:17 heh :D Mar 17 21:48:23 yeah I figured it was that :D Mar 17 21:48:39  Mar 17 21:48:46 oo what else can I find .. haha Mar 17 21:48:53 ³ Mar 17 21:48:53 UnicodeEncodeError: 'ascii' codec can't encode character u'\xb3' in position 0: ordinal not in range(128) (file "/home/ahiam/.sopel/modules/gitter_bridge.py", line 52, in ircToGitter) Mar 17 21:48:54 but the arietta looks nice, for projects where a bbb or a raspi is too big and cumbersome Mar 17 21:49:15 yeah that's a fairly grunty processor Mar 17 21:49:20 yea. also raspi does not have usb device Mar 17 21:49:35 if you need that for some reason Mar 17 21:49:40 device mode Mar 17 21:49:41 regular pi does .. its just off a hub :( Mar 17 21:49:49 oh for otg/client mode .. no Mar 17 21:49:56 yea sry Mar 17 21:50:00 otg I meant Mar 17 21:50:35 I have no use for otg. Whatsoever. Ever. Mar 17 21:50:58 unlesss *everything* else breaks .. and I can't even boot through uart and upload new code :P Mar 17 21:51:35 and at that point .. I think I'd probably buy a new board anyway lol Mar 17 21:51:50 only thing I needed it for, was this old machine that would output data to a usb device. Mar 17 21:51:59 :D Mar 17 21:52:34 ah .. so what did you use .. a console or a hid or something on the bb? Mar 17 21:52:53 the arietta Mar 17 21:52:57 oh heh nvm Mar 17 21:53:05 it's got usb .. Mar 17 21:53:11 oh yeah .. like Lots of it .. haha. Mar 17 21:53:16 but bb also got otg Mar 17 21:53:30 at first I used pcduino3 for this Mar 17 21:53:47 cool Mar 17 21:54:36 so the machine thinks its a usb device, and saves the files like it normally would. The arietta detects changes on the partition, and saves the data to the database on a server Mar 17 21:54:58 tru its wifi Mar 17 21:56:22 the bb I bought for another projekt. Experimenting with alot of nmea signals Mar 17 21:56:27 on a boat Mar 17 21:56:51 the multiple serial connections are interesting here Mar 17 21:57:20 ah cool yea Mar 17 21:59:11 I need some digital isolators first Mar 17 21:59:29 some of the signals are 10V Mar 17 22:02:11 http://www.mouser.com/ds/2/368/Si87xx-54602.pdf Mar 17 22:02:18 maybe this is suitable Mar 17 22:04:10 instead of optocouplers Mar 17 22:04:45 http://gerbil.xs4all.nl/show-pins-v3.pl <-- updated version that also shows the state of any gpios actively used (as far as the kernel knows, i.e. for its own purposes or by being exported) Mar 17 22:15:34 bbl Mar 18 02:22:03 Does anybody know how to debug the beaglebone black without the JTAG **** ENDING LOGGING AT Fri Mar 18 02:59:58 2016