**** BEGIN LOGGING AT Thu May 10 03:00:02 2018 May 10 04:19:34 So I've barked up a lot of wrong trees trying to netboot (over USB) a completely blank beaglebone clone from scratch May 10 04:20:22 The goal is rapid deployment, so an installer that starts by netbooting over USB seems to be the best option: plug in, come back later, it's done. May 10 04:21:03 I've gotten as far as getting the board to get it's address and files for various stages of loading via bootp/tftp May 10 04:21:23 it'll load u-boot, but I can't get u-boot to load a kernel May 10 04:21:52 I see there's some sort of magical header I need to add to a kernel to get u-boot to like it, using mkimage, and I've tried that May 10 04:22:36 Can anyone recommend somewhere good to start reading to make progress at this stage? May 10 04:23:51 Also, I'm not wholly sure how to tell the kernel where to get its rootfs from. I'm guessing another set of dhcp options conditional on vendor-class-identifier May 10 04:26:58 My u-boot and SPL built from git.denx.de seems to work fine, but my kernel from kernel.org, my kernel from https://github.com/beagleboard/kernel, and my kernel from buildroot do not seem to boot May 10 04:28:45 I guess I'm kinda getting lost in the weeds here and am looking for a sort of canonical guide May 10 04:33:54 The only guides I've found involve inserting an SD card somewhere along the way, and that's the whole thing I'm trying to avoid. May 10 04:48:55 * n8vi discovers that http://processors.wiki.ti.com/index.php/Ubuntu_12.04_Set_Up_to_Network_Boot_an_AM335x_Based_Platform somehow got lost from his list of useful URLs. May 10 08:05:50 whoa, wall of text May 10 08:09:01 n8vi: might it not be easier to use an initramfs that mounts the root filesystem rather than having the kernel use nfsroot? then you have much better facilities at your disposal and you should be able to test it more easily May 10 08:10:04 what is the the currently preffered way of loading an overlay (want to enable can) I was following http://www.thomas-wedemeyer.de/beaglebone-canbus-python.html (but 90% of the stuff is already done e.g. the CAN0 overlay is present) but the overlay documentation no longer match. /boot/uEnv.txt is also very confusing about eeprom override stuff. Elinux.org is even more confusing about different type of overlay May 10 08:10:10 s and deprecate stuff May 10 08:10:14 n8vi: if the system you want to run is sufficiently compact you might even be able to run entirely from initramfs May 10 08:13:41 keesj: if you're using a recent system in its default configuration (i.e. with cape-universal enabled) then all you have to do is configure the pins to can using config-pin May 10 08:14:26 e.g. config-pin P9.24 can && config-pin P9.26 can May 10 08:14:28 A.... I did manage to load the dtb but I was also funny that even without the dtb the can drivers where already loaded May 10 08:14:43 just no output May 10 08:14:52 "load the dtb" ? May 10 08:14:57 what dtb? May 10 08:15:30 device tree binary May 10 08:15:40 loading an overlay for CAN is not necessary when using cape-universal, and in fact incompatible with it May 10 08:15:47 ##Overide capes with eeprom May 10 08:15:49 uboot_overlay_addr0=/lib/firmware/BB-CAN1-00A0.dtbo May 10 08:15:49 yes I know what it stands for, I meant what dtb are you talking about May 10 08:15:52 okay, that one May 10 08:16:08 9970640 bytes read in 876 ms (10.9 MiB/s) May 10 08:16:10 uboot_overlays: [uboot_base_dtb=am335x-boneblack-uboot.dtb] ... May 10 08:16:51 I think that *should* still work too, and should implicitly disable cape-universal (i.e. config-pin will no longer work) May 10 08:16:56 but so .. I will undo that change and try the pin config May 10 08:17:43 I want... as simple/reliable as possibe hence .. guessing config-pin is here to say May 10 08:18:14 I think both mechanisms are here to say. the only thing that's really deprecated is loading overlays at runtime May 10 08:19:58 the output of u-boot gets lost (when I don't have a serial) hence I think .. runtime is better May 10 08:20:49 yeah, it would be nice if u-boot could buffer its logs in ram and somehow pass that to the kernel May 10 09:23:56 also different from the blog post it is can1 that is mapped to pins 24 and 26 May 10 09:29:20 correct, if it's claiming otherwise then it's plain wrong May 10 09:30:04 either that or it's relying on only can1 being enabled and ending up being (sequentially numbered by the kernel) can0 instead May 10 09:30:38 it is possible to ensure can1 is always named can1 even in the absence of can0, but I don't know if that has been taken care of May 10 09:32:22 there are two can interfaces, I did not probe anything just the ip up just https://pastebin.com/kmtiH6Dn May 10 09:32:41 yes I know cape-universal enables both of them May 10 09:33:09 but that wouldn't be the case when using an overlay to enable can1 May 10 09:34:06 why are you first setting the interface up with ip link and then again with ifconfig? May 10 09:34:28 ifconfig is obsolete, it's wholly replaced by ip May 10 09:34:55 not sure what you meant with "I did not probe anything" May 10 09:39:22 just following http://www.thomas-wedemeyer.de/beaglebone-canbus-python.html . May 10 09:40:24 well. I mean I did not do "modprobe can" or similar that might have affected the order of devices May 10 09:42:04 manually modprobing is unnecessary for nearly every kernel driver, and its presence on that page is mostly indicative that this website should not be trusted too much May 10 09:42:36 kernel modules for drivers are normally loaded automatically by the presence of devices in the DT May 10 09:43:07 ah this dude is also doing the pointless use of both ip and ifconfig May 10 09:46:41 I can understand people who still use ifconfig instead of ip ... either they're simply used to it and can't be bothered to get used to new tools, or are just following information that has been handed down from ancient times May 10 09:47:03 I have no idea however how someone ends up using both ip and ifconfig in the same setup instructions May 10 10:26:35 the ip already does the up May 10 10:29:22 I once bought me a beagle bone black board and now I have another one, but I am not sure which is which. How can I see the revision of each boards May 10 10:32:02 unless it's really old and has a 2GB eMMC, they will be almost indestinguishable. Possibly a newer eMMC might be a faster model. May 10 10:34:51 the one has a small marking under the lacker "PCB Revb6" and the oither has "PCD RevB6" so still clueless... https://beagleboard.org/black does not list these revisions May 10 10:35:14 zmatt: but strictly speaking. you are the weird person with lots of knowledge May 10 10:37:45 tverrbjelke: either boot them and check the size of the eMMC or look up the part number printed on the emmc. Also is there no sticker on one of the cape ports? May 10 10:37:45 "PCB RevB6" has "AM3358BZCZ100"- while "PCB RevB4", has XAM3359AZCZ100" - I like to find out which is which revisiopn! May 10 10:39:13 the XAM3358AZ part is *probably* on the onlder board. Is the eMMC 2GB? May 10 10:39:51 Yes, they have a sticker: Rev B4" has on sticker "A5C" - while pCb Rev & has sticker with sevaral more (small ) numbers). Where I see the eMMC chip? May 10 10:40:33 A5C is the older one May 10 10:40:43 B4 is also old, but newer May 10 10:40:47 both are 2GB May 10 10:41:05 OK now the one is Revision A5c then? May 10 10:41:21 come again? May 10 10:41:41 sticker with "A5C"it has "PCB REv4" mareking under the lacker May 10 10:42:31 no clue about the PCB revisions. The relevant information is the device (as a whole) and its revision. May 10 10:43:16 nowadays you can only buy Rev C boards with 4GB eMMC May 10 10:43:20 and "PCB RevB6" has smaĺl Sticker with 4 rows: "400524 / 1301 / 000883 / 1806" and a small "2D ID-icon" May 10 10:44:22 so what are you still trying to answer? It's clear which one is older. May 10 10:44:33 I find if VERY concerning, that I cannot find any revision marking on the boards May 10 10:44:46 I want to know which revision is the other May 10 10:45:06 (newer) May 10 10:45:40 where can I see the eMMC chip andhow much it it? May 10 10:46:21 is it the Kingston emccO4G-M627? May 10 10:47:01 the eMMC should be on the bottom of the board May 10 10:47:19 it sounds like a 4GB part number, that would mean it's a Rev C board May 10 10:49:23 http://d2r5da613aq50s.cloudfront.net/wp-content/uploads/466770.image1.jpg May 10 10:49:32 so this picture points out the eMMC May 10 10:50:03 OK. So that now helps me! May 10 10:50:06 thx May 10 10:53:08 the emcc datasheet (PDF file) https://www.jetone.com.tw/uploadfiles/327/datasheet/KSI/ksi-emmc-upsi.pdf confirms the newer board has 4G eMMC! May 10 10:53:51 tbr: I think rev C has a pcb marking with B4 May 10 10:54:08 since rev C is just a BOM change, not a pcb change May 10 10:54:54 on my 4GB eMMC BBB it has a "PCB RevB6" marking May 10 10:55:01 eh B6 sorry May 10 10:55:12 pcb revisions are separate anyway, I don't really know them May 10 10:55:19 4GB means rev C May 10 10:55:26 and the older one is A5C May 10 10:56:26 if I want to run debian on it as irc / jabber server , and have it run 24/7 - must I carfe about /var/ and /tmp mounted into ram? To avoid eMMC and SD-Card wearing? May 10 10:57:28 it's a good idea to avoid unnecessary writes to eMMC or sd yes May 10 10:57:28 Or is the SD card the best place to mount this logging stuff? May 10 10:57:34 sd is easier to replace, but wears out much faster May 10 10:58:24 if you don't need too much storage, you can also configure eMMC into SLC mode. this sacrifices 50% of the capacity in exchange while increasing lifetime about 10x May 10 10:58:40 ehh, sentence fail May 10 10:59:04 I only want to put my OS (debian or gentoo) onto it. So 50% of 2GB would still be OK? May 10 10:59:37 what I do personally is: make /tmp a tmpfs, remove rsyslog, and remove /var/log/journal to make systemd-journald keep a runtime journal (in ram) only May 10 11:00:08 I don't think I've ever spotted any significant amount of writes to the rest of /var May 10 11:00:31 nice . May 10 11:01:07 So I never actually flashed something into this eMMC. I would use https://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Flashing_eMMC instructions somehow? I May 10 11:01:27 our production beaglebones currently use less than 400 MB of storage, and I haven't tried very hard to trim it down further May 10 11:01:55 So how woudlI proceede to use SLC mode? May 10 11:02:22 I create "normal" SD Image, plug sd in, boot up BBB and then...? May 10 11:02:45 as those instructions say, you uncomment one line in /boot/uEnv.txt on the sd card to turn it into a flasher card May 10 11:03:26 be sure to download the console image (stretch-console) and not the -iot one (let alone -lxqt) May 10 11:03:33 So I dd the debioan image from my linux pc , mount it, edit t /boot/uEnv.txt and THEN plug sd-card into bbb and boot it? May 10 11:03:45 for example yes May 10 11:04:09 and after it flashed (4 user-leds show "allOK") then I could remove sd card? May 10 11:04:23 And Debian stretch-console would be inside eḾCC? May 10 11:04:26 yes. I think it actually powers off these days May 10 11:04:27 yes May 10 11:04:33 That I could do. May 10 11:04:42 bit the SLC mode? May 10 11:04:56 https://github.com/dutchanddutch/mmc-utils/tree/custom May 10 11:04:57 how do I that? May 10 11:05:17 read the commit message of the latest commit May 10 11:06:36 ok I cloned it now and read the log May 10 11:06:50 the actual command is just ./mmc reliable-slc-configuration /dev/mmcblk1 May 10 11:06:50 "fix .gitignore" ?? May 10 11:07:02 you cloned the wrong branch probably? May 10 11:07:05 git co custom May 10 11:08:17 join May 10 11:08:25 tarun: success. May 10 11:08:32 cat May 10 11:08:42 meow May 10 11:08:57 ls /home/ubuntu May 10 11:11:31 git branch --list _> master. OK May 10 11:14:13 tverrbjelke: "git co custom" should automatically create the missing branch (following origin/custom) May 10 11:14:33 you could also have passed -b custom to git clone May 10 11:15:56 I added a custom command for this to mmc-utils since it previously required multiple steps and was error-prone (which is not a great quality when dealing with one-time-programmable settings) May 10 11:16:16 ok the log tells me "so I add reliable-slc-configuration command ..." May 10 11:16:16 i somehow have to cross-compile it and add it to the SD-card, Not edit /boot/uEnv.txt, boot BBB and ssh there, and run this command once? May 10 11:16:16 And then edit /boot/uEnv.txt to flash emcc and then all is done? May 10 11:16:45 for example yes May 10 11:17:03 Or I indetall build-essentiasl inside bbb-debian and compile it there? May 10 11:17:06 or if there's a default image currently on the beaglebone you can also just native-compile it there May 10 11:17:34 you can run this command even when you've booted from eMMC. the actual reconfiguration and erase doesn't happen until you power-cycle May 10 11:18:20 Lets see - to "just fire the bbb up" i connect to usb. and then I dont know if the bebb would have access to network. Would I also connectit to my local lan? There is a dhcp running it ocul d use... May 10 11:18:46 if you connect it to ethernet it'll acquire an ip via dhcp May 10 11:19:16 So USB not requered at all? May 10 11:19:17 and you should be able to find it via avahi (beaglebone.local) ... although, if it's running a really ancient system then I don't know if the latter is still true May 10 11:19:43 usb is not required May 10 11:20:17 Hm. lets see If I cann boot into my bbb... May 10 11:20:48 for my 24/7 jabber /Irc server I rather not install build-essential - or does itnot consume too much spoafce? May 10 11:21:29 that's why I mainly suggested it in case it already had an image installed anyway May 10 11:21:41 in which case a compiler will already be installed, and you're going to erase that system anyway May 10 11:22:43 cross-compiling should be pretty easy too, mmc-utils has zero library dependencies May 10 11:23:24 wow, it is now attached to my lan, and powered from pc via usb. It now shows as USB drive May 10 11:23:47 it shows as a composite device by default May 10 11:23:59 including a small read-only usb drive with a readme and some stuff May 10 11:24:40 hm lets see which IP it got... May 10 11:37:33 somehow the user/login is different than debian/temppwd :-( May 10 11:38:03 the A5C might be running something truly ancient like angstrom rather than debian May 10 11:38:09 try root / blank password ? May 10 11:38:46 I just attached the newer one... May 10 11:39:08 omg I actually THINK it oculd be angstrom, indeed on the old one. May 10 11:39:27 Since I want to reset them anyway I just do it with the new one. May 10 11:43:10 the REvC has plenty of tools ready (git etc) but it already has 2.9GB space used! May 10 11:43:27 the default image comes with a desktop environment May 10 11:43:30 so... yeah May 10 11:45:11 I will try to write me some documentation stuff what I am doing ... May 10 11:47:00 sounds good :) May 10 11:50:43 now I somehow need dependencies... make generates this: Makefile:36: recipe for target 'mmc_cmds.o' failed May 10 11:50:57 uhh, what's the actual error? May 10 11:51:14 debian@beaglebone:~/mmc-utils$ make May 10 11:51:14 cc -Wall -Werror -Wuninitialized -Wundef -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 -g -O2 -Wp,-MMD,./.mmc.o.d,-MT,mmc.o -c mmc.c -o mmc.o May 10 11:51:14 cc -Wall -Werror -Wuninitialized -Wundef -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 -g -O2 -Wp,-MMD,./.mmc_cmds.o.d,-MT,mmc_cmds.o -c mmc_cmds.c -o mmc_cmds.o May 10 11:51:14 mmc_cmds.c: In function 'hexdump': May 10 11:51:14 mmc_cmds.c:1306:4: error: 'for' loop initial declarations are only allowed in C99 or C11 mode May 10 11:51:14 for( int i = 0; i < 32; i++ ) { May 10 11:51:15 ^ May 10 11:51:15 mmc_cmds.c:1306:4: note: use option -std=c99, -std=gnu99, -std=c11 or -std=gnu11 to compile your code May 10 11:51:16 mmc_cmds.c:1315:2: error: 'for' loop initial declarations are only allowed in C99 or C11 mode May 10 11:51:16 for( int i = 0; i < len; i++ ) { May 10 11:51:17 ^ May 10 11:51:17 Makefile:36: recipe for target 'mmc_cmds.o' failed May 10 11:51:18 ARGH May 10 11:51:18 make: *** [mmc_cmds.o] Error 1 May 10 11:51:22 don't paste text into IRC! May 10 11:51:26 hm. shall I spill this stuiff here? May 10 11:51:40 :-) yes.... May 10 11:51:50 use a paste service like pastebin.com or hastebin.com or any other May 10 11:52:24 hum that's weird, I should have gotten this error too when compiling May 10 11:53:12 shall I add -std=gnu11 to FLAGS in makefile? May 10 11:53:15 and those aren't looks I introuced May 10 11:53:20 *loops I introduced May 10 11:53:34 yeah May 10 11:53:41 maybe the gcc default changed over time May 10 11:54:22 so actually my efforts here are gonna improve the PRoject!" ? May 10 11:54:27 * tverrbjelke jumps around for joy May 10 11:55:02 * tverrbjelke sees the mmc binary done May 10 11:55:42 ah yes, the default has change over time. for gcc 7 the default is gnu11 for C and gnu++14 for c++ May 10 11:59:48 so now i need to make install, or can I just run it fro the place there? May 10 12:00:47 you can run it right there May 10 12:01:48 so I am a bit nervous now. just "sudo ./mmc reliable-slc-configuration"? May 10 12:02:27 followed by the block device, i.e. /dev/mmcblk1 just to be extra sure, you've read the commit message (specifically the part about this being irreversible) May 10 12:02:30 ? May 10 12:03:31 /dev/mmcblk0 /dev/mmcblk0boot0 /dev/mmcblk0boot1 /dev/mmcblk0p1 ... which ? May 10 12:03:43 oh right, ancient system May 10 12:03:46 mmcblk0 then May 10 12:04:03 on current systems, sd card is always mmcblk0 and emmc is always mmcblk1 May 10 12:04:14 on old ones, they used to be sequentially numbered in order of appearance May 10 12:04:38 i.e. eMMC would be mmcblk0 or mmcblk1 depending on whether an sd card was present at boot or not May 10 12:05:09 wow - I am now in the Rev C board! May 10 12:05:21 that doesn't mean you're not running an old system ;) May 10 12:05:24 cat /etc/dogtag May 10 12:05:50 BeagleBoard.org Debian Image 2016-01-24 May 10 12:06:06 see, early 2016... that's basically stone-age May 10 12:06:06 ;) May 10 12:07:17 my dear,ö that other bnoard is from 2014 or so May 10 12:07:35 I just rsynced the binary to my pc so I can do that on th eolder one, too May 10 12:07:56 btw: I did not attack any sdcard to it May 10 12:08:30 hm. how oculd I myself check which device to use? May 10 12:08:42 only eMMC has the boot0/1 subdevices May 10 12:09:19 also, the mmc utility would simply fail without causing any harm if attempted on an sd card May 10 12:10:02 so I could blindly do that on all major mmcblk[0|1] devices safely? May 10 12:10:48 should be safe yes May 10 12:14:53 "Confirm that PARTITION_SETTING_COMPLETED bit is set using 'extcsd read' after power cycle" -> howto power cycle now? shutdown now? May 10 12:15:20 poweroff, or just unplug it May 10 12:15:31 nice May 10 12:15:55 SO I now create me sd-card with recent debian 9 image? May 10 12:16:15 yes, grab the stretch-console image May 10 12:17:15 if you want to prepare multiple beaglebones in some way (i.e. have certain software installed), you can also first boot from the sd card and do whatever preparations before turning it into a flasher by uncommenting the appropriate line in /boot/uEnv.txt May 10 12:17:52 since the flasher works by making eMMC a copy of the system that's on the sd card May 10 12:20:13 Ah. oK. since I have 2 BBB here, I rather do this. Esp. since I would also do the " make /tmp a tmpfs, remove rsyslog, and remove /var/log/journal to make systemd-journald keep a runtime journal (in ram) only" May 10 12:21:03 I am looking at http://beagleboard.org/latest-images I only see "Stretch IoT (without graphical desktop)" fro BBB May 10 12:21:11 is that what you mean? May 10 12:21:25 other stuff I personally do is improve boot time by disabling some pointless services and remove initramfs-tools May 10 12:21:28 no, not iot May 10 12:21:34 https://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Stretch_Snapshot_console May 10 12:22:12 what is the differnce? May 10 12:22:31 iot has a ton of crap installed, I don't know if it would even fit 2GB, let alone 1GB May 10 12:32:35 dd to sd or to sd1? May 10 12:32:55 just plain sd I asume? May 10 12:34:31 yes, it's an image for the full card, not just for the rootfs May 10 12:36:39 are there any hints how properly " make /tmp a tmpfs, remove rsyslog, and remove /var/log/journal " May 10 12:36:43 ? May 10 12:36:59 Or I just google and puzzle that for myself? May 10 12:37:04 /tmp might actually already be a tmpfs, not sure May 10 12:37:11 otherwise it's a matter of adding one line to fstab May 10 12:37:21 remove rsyslog with: sudo apt-get purge rsyslog May 10 12:37:37 remove /var/log/journal with sudo rm -rf May 10 12:38:20 which programm will then do the logging stuff? a bit lofrotating in ram could be fine... May 10 12:38:34 journald will keep a runtime journal May 10 12:39:42 Ah. And put its logs into /var/log? or somewhere in /tmp? or in /var/tmp? May 10 12:39:42 it normally starts by keeping only a runtime journal and will then flush it to a persistent journal in /var/log/journal, but if it sees that this directory is missing it will just keep using the runtime journal instead May 10 12:40:01 (with auto-rotation based on limits configured in /etc/systemd/journald.conf ) May 10 12:40:52 I am thinking... will e.g. jabber server and irc server need some sys logging service? (I have not checked all this) May 10 12:40:57 the runtime journal is in /run/log/journal but you don't really need to care, you view journal using journalctl regardless of whether runtime or persistent journal is used May 10 12:41:27 if they're started as services and log to stdout/stderr or syslog then it will automatically end up in journal May 10 12:41:53 ok. I still think I could mount a sd card and use it for /tmp/ and /var/tmp... May 10 12:42:29 seems like a pointless complication unless you strongly value having persistent logs May 10 12:42:55 actually I have NO experience with lgging onto a sd card. (space consumed depends on traffic and how it rotates sure) butr wearing the sd card ... I am inexpereiennced May 10 12:43:49 well actually poweroff and be able to recover the data of irc /jabber seems desirable... but I just play around with it right now May 10 12:44:04 wear-out of NAND flash is measured in erase cycles, and total amount of data you can write before wearout is then disk size * erase cycles May 10 12:44:34 TLC flash (like most sd cards) can do a few hundred erase cycles May 10 12:44:37 hm. SO I should buy me some cheakl 8GB cards or use 32GB ones and have PLENTY of head room? May 10 12:44:42 MLC flash (eMMC in default state) a few thousand May 10 12:44:49 SLC flash a few tens of thousands May 10 12:45:24 considering power consumption: usb ssd drive would consume "alot" I guess, copmpared to simple sd cards May 10 12:45:55 (you can actually buy SLC sd cards for industrial purposes, but they're not cheap) May 10 12:46:02 I presume? not sure really May 10 12:47:07 ok I think I just grab what I have here (32GTB sd card) and save the data onto it - when configuring irc proxy server May 10 12:47:15 32GB May 10 12:47:29 at the very least peak power consumption will scale with data transfer speed I'd assume May 10 12:48:19 I mostly would like to have me irc logged in, even when my compi is off. SO I thought May 10 12:48:31 so I dont miss relevant irc chatter May 10 12:48:50 it may of course be perfectly safe to use eMMC, it all depends on how much data we're talking about here and how many years of operational lifetime you're hoping for May 10 12:49:36 Haha! It's just my little home and maybe later dynamic dns so I can login from outside, too May 10 12:49:41 just for me! May 10 12:50:26 we've managed to wear out eMMC on a few devices, but this was caused by a little bug in a logging module that was causing something like 16GB of data to be written per hour May 10 12:50:32 When I have done this BBB irc server I later will try the same with my ODROID C2 and see if that is different at all May 10 12:51:14 I wouldprefer to use my bbb to control some robot devices - its IO capabilities are so nice May 10 12:51:34 but right noiw I sem to need some low power server running here May 10 12:52:01 indeed. for just an arm-based linux system there are plenty of options available, and probably many better than the bbb, but its i/o is where it really shines May 10 12:52:02 btw: it is ajust a normal debian, right? May 10 12:52:32 (secuiring linux server tutorials would work just fine?) May 10 12:52:44 the console image is fairly normal, although all of these images are customized for beagleboards and beaglebones to some extent May 10 12:52:49 yeah you should be able to May 10 12:52:59 the only thing slightly unusual maybe is that connman is the default network manager May 10 12:53:16 but it's not much hassle to replace it with something more usual. I personally use systemd-networkd myself May 10 12:53:17 someone suggested me to just install this armdebian and that seems a bit closed-source May 10 12:54:03 btw, here's a small perl script I wrote that checks how much data has been written to eMMC: https://pastebin.com/N2ps2Fjn May 10 12:54:12 armbian.com May 10 12:54:17 never heard of it May 10 12:56:02 that script could be useful! Is pearl installed on this debian-console version? (I myself ojnly know a bit python) May 10 12:56:49 "slurp" ? May 10 12:57:00 perl-base is a required debian package May 10 12:58:05 hehe so pearl wont ever never shade away May 10 12:58:17 it's perl, not pearl :P May 10 12:58:30 i am innocent, sir! May 10 12:58:54 you can also attribute eMMC usage to specific services by setting DefaultBlockIOAccounting=yes in /etc/systemd/system.conf May 10 13:01:51 systemd - the mighty tower of babylon! I am still only to init... :-( May 10 13:03:07 I find more and more things need to learn May 10 13:04:44 It boots! And df tells me /dev/mmcblk0p1 817M 327M 433M 44% / May 10 13:05:17 yes the image doesn't automatically expand to cover the sd card May 10 13:06:17 now mmcblk0 is sd having mmcblk0p1 and mmcblk1 is the emmc. what is the mmcblk1boot0 mmcblk1boot1 for? May 10 13:06:38 SO I am confident I can also do this with my old bbb! May 10 13:07:24 they're small hardware partitions that every eMMC has. its intended use is for storing the bootloader (two copies so you can do atomic upgrades), but the AM335x doesn't support it May 10 13:10:15 ah! ok. May 10 13:10:27 tmpfs on /run/user/1000 - but in fstab i only see two lones May 10 13:10:55 fstab also holds "debugfs /sys/kernel/debug debugfs defaults 0 0" -> what is htis? May 10 13:11:05 systemd takes care of /run May 10 13:11:30 that line is useless, /sys/kernel/debug is also mounted automatically May 10 13:11:35 oh. I dont know /run... what is iit for? May 10 13:11:58 so I can delete it safely? May 10 13:12:20 /run is for stuff that only exists at runtime, and it's guaranteed to be a tmpfs May 10 13:12:33 especially stuff like unix domain sockets May 10 13:12:54 /var/run is a symlink to /run May 10 13:15:31 I just see this 1st time!!! May 10 13:16:17 it's been around for quite a while May 10 13:18:09 The BBB is getting a bit warm while idling. Will I need to do some coolung? May 10 13:18:27 highly unlikely May 10 13:18:41 :-) May 10 13:19:13 can you touch the processor without hurting your hand? => no cooling needed May 10 13:21:49 ok I am now scanning /etc/systemd/journald.conf -> I am n ot sure how to change the uncommented values to something useful... May 10 13:22:50 #SystemMaxUse= #SystemKeepFree= #SystemMaxFileSize= and same with runtimexxx - what to set this? or just leave`? May 10 13:23:18 RuntimeMaxUse is how much memory it is allowed to use for the runtime journal May 10 13:36:26 also, in case it's useful: here's another perl script to check eMMC writes, but this one can attribute which service or slice is responsible for it: https://pastebin.com/NrZRpSn3 May 10 13:39:35 so you put them in /root/bin? May 10 13:39:35 (although in practice I frequently see the number in the root being much bigger than the sum of system.slice and user.slice. I'm not sure why) May 10 13:39:51 uhh, you put them whereever you want? May 10 13:40:08 you don't need to be root to run them May 10 13:40:58 used beaglebone for two projects in the past, is it still a good choice, considering TI is not active in ARM silicons these days May 10 13:41:16 not active? what are you talking about? May 10 13:41:45 not too many new chips are designed comparing to other companies May 10 13:42:24 TI seems backing out of high-end chips, more on mcu and analog these days, I used to use TI8168 but they since stopped developing them May 10 13:42:52 quite the contrary, they have some flagship ARM-based SoC for automotive vision applications (ADAS) May 10 13:43:01 the chips on beagle are more of a left-over instead of anything promising for the next decade May 10 13:43:40 there's no EOL for the AM335x even remotely in sight May 10 13:43:59 you're right that the DM81xx SoCs got ditched in a bad way May 10 13:44:06 they're crushed by hisilicon(huawei) at least surveillance chips, TI used to be sole and largest surveillance chip verdor, now it is totally irrelevant May 10 13:44:08 which is really unfortunate May 10 13:44:43 TI SoCs seem mostly focussed on industrial and automotive these days May 10 13:45:22 and I wouldn't know of any real competitor to the beaglebone with regard to I/O (though I haven't really searched either) May 10 13:45:45 I am currently looking into these ODROID things. May 10 13:46:06 yes i love beaglebone which is why i'm asking, there are alternatives to it though, many actually May 10 13:46:35 i'm just worrying that while beagle still good, but for future projects it might be the right one anymore May 10 13:46:37 depends on what you're looking for. if you're just looking for any arm-based SBC then sure May 10 13:47:06 So mainly what my little research popped oput was ODROID C2 or this UX4 if yxou need CPU power. Raspie3 maybe, too May 10 13:47:11 I don't see any cause for worry currently, and we use them in products we sell May 10 13:47:27 i.mx series are great, from consumer to military, sadly they do not have a popular sbc format like beaglebone, not to mention communities etc May 10 13:47:37 maybe they don't care about this market segment May 10 13:47:54 ausjke: There is the Hummingboard for i.mx6 May 10 13:48:21 Humpelstilzchen: yes there are ,but not as well adopted for whatever reason. May 10 13:48:55 another relevant factor is that the beaglebone is an open design and consists of parts you can actually buy in small quantity. so even though making a customized beaglebone-derivative is still not simple, the option is available May 10 13:48:56 ausjke: probably because it costs > 100$ May 10 13:49:06 Hummingbird looks like a SOM-based design, no wonder more expensive May 10 13:49:22 it is May 10 13:49:23 ON the other hand - this realtime ahrtware in the TI chipset in the BBB - I have not yet made use of it, but it sounds really cool. May 10 13:49:32 it is cool May 10 13:50:10 a nice showcase for it is BeagleLogic, which is a piece of software that uses PRU to turn a beaglebone into a 14-channel 100Msps logic analyzer May 10 13:50:30 so when it comes to real audio processing with low latency - I could ask you fo rthis realtime aspekt? May 10 13:50:57 Freescale laid off all its i.MX team a few years back, until it realized i.MX6 is so popular, so they have to build a new team to produce i.MX8 and so on, I just wish they can supply a same price-range beaglebone as NXP is still seriously developing many ARM chips May 10 13:50:58 pru is not that useful for audio, it isn't a DSP May 10 13:52:09 Allwinnder is producing industrial parts with upgraded once-super-popular A20, now called A40 May 10 13:52:22 ausjke: anyway, one aspect of TI's focus on industrial and automotive is that they're more developing in depth than producing new SoCs at a high rate like you'd see in consumer-oriented manufacturers (which used to include TI back in the OMAP days) May 10 13:52:42 reliability and long-term support are more important than the latest fanciest ARM core May 10 13:52:52 I sedont that! May 10 13:52:54 d May 10 13:52:56 zmatt: yeah i just want to get some confidence to adopt bbb for future projects May 10 13:53:09 I would LIKE to see a new bbb with more ram May 10 13:53:23 as i know not even TI CEO can promise for anything, it's all market driven, and TI is losing in ARM A8+ market May 10 13:53:27 tverrbjelke: the beaglebone enhanced has twice the RAM (and gigabit ethernet) May 10 13:53:58 hm... also the similar hardware gpio support? May 10 13:54:00 ausjke: what does "ARM A8+ market" mean? how is that a market? May 10 13:54:20 zmatt: does TI have a long-term supply contract in public terms? as what Freescale did for many of its parts, I think it's 10-year-promise-of-supply May 10 13:54:51 ausjke: I don't know, maybe jkridner[m] does May 10 13:55:46 zmatt: i mean newer ARM socs development at TI, am355x might be a cash cow now, they may kill the cow when the milk is all squeezed, anyways May 10 13:56:01 ausjke: Phytec has a guarantee they can sell their AM335x SoM for 10 years May 10 13:56:05 just a bit worried, nothing else May 10 13:56:24 Humpelstilzchen: good to know, they must have stocked it fully in the warehouse :) May 10 13:56:25 industrial users would not be amused if TI were to suddenly EOL the am335x May 10 13:56:40 nor easily forgive I would imagine May 10 13:56:49 ausjke: or they have the same from TI May 10 13:57:39 generally industrial means long development cycle (to ensure a product really works like it should) and long support cycle (you don't mess with something that works) May 10 13:57:47 i have seen more clients designing board around SOMs these days, good for companies like Phytec May 10 13:58:27 the BBE seems nice May 10 14:00:20 what's so special about BBE? May 10 14:00:46 it's a beaglebone derivative with twice the ram, gigabit ethernet, and various other stuff May 10 14:00:48 recommended BBGW for a client last year and they adopted it May 10 14:00:55 eww May 10 14:01:04 the BBGW is the only beaglebone variant I'd strongly recommend against May 10 14:01:10 zmatt: why? May 10 14:01:49 their idiotic pin usage. instead of reusing the pins normally used for ethernet for wifi (like the BBBW does) they sacrificed a large pile of expansion header pins for it May 10 14:01:59 while leaving the pins normally used for ethernet unused May 10 14:02:58 thought BBBW and BBGW are the same design :( May 10 14:03:14 i could not find detailed comparison then as I can recall May 10 14:03:26 no, the BBGW was incompetently designed in my opinion May 10 14:03:51 oops, BBGW was designed by seeedstudio correct? May 10 14:03:58 yes May 10 14:04:15 the earlier BBG also contained dubious design choices, but they weren't quite as harmful May 10 14:04:31 i saw their pdf schematic then, could not find BBBW documents a while ago, i since moved to something else May 10 14:04:50 who designed BBBW, hope they have some doc in public domain now May 10 14:05:00 I could get me this BBE Wifi for 88$ ( I htink they sit in UK and still demand USD?) May 10 14:05:16 BBBW is from beagleboard.org, and yes you can find the schematic on github May 10 14:05:38 so the main different is the iomux especially the ethernet/wifi sharing pins? May 10 14:05:54 good to know BBGW is not good, will double check when time comes May 10 14:06:07 maybe someone should write a blog about BBBW vs BBGW May 10 14:07:20 I just checked, you lose 11 expansion header pins because of this on the BBGW May 10 14:07:33 (and probably compatibility with a variety of CAPEs as a result) May 10 14:07:52 if that's the case BBGW should be avoided at all costs May 10 14:08:13 zmatt: thanks for bringing this up, I'm sure I will need this sometime soon May 10 14:08:47 P8 pins 11, 12, 14-18, 26, and P9 pins 12, 15, and 41 May 10 14:09:18 PocketBeagle looks NICE and Tiny! May 10 14:09:37 Beagleboard/beaglebone was really the leader in OSS low-cost embedded board, then we all know RPi stole the thunder, in a big way...I heard RPi's SOM was used in large amounts, probably eclipse BBB totally :( May 10 14:09:40 oh and P9.28-31 May 10 14:09:57 these days there are too many SBCs to choose, all in BBB/RPi size May 10 14:10:42 well, to each their own... I personally wouldn't want to use a basically undocumented SoC that doesn't even have a product page on the manufacturer's website May 10 14:12:30 (this in contrast with the AM335x whose 5000-page technical reference manual you can download at ti.com without even having to register) May 10 14:12:40 agree, but in chip industry only volume speaks, i have yet to use RPi/SOM for any projects though, but who knows May 10 14:13:24 TI and Freescale are the most OSS friendly companies on earth, 6000+ pages of reference manuals online is common for them, not anybody else May 10 14:14:05 zmatt: maybe you can have a loot at this: https://pastebin.com/ri7vipeJ May 10 14:16:06 "Can't locate experimental.pm in @INC " May 10 14:17:08 o.O May 10 14:17:32 maybe experimental.pm isn't in perl-base but requires the main perl package? May 10 14:17:56 yep, that's it May 10 14:18:19 its in perl-modules according to packages.d.o May 10 14:19:09 so I could rewrite the script to python or install this (which) package? May 10 14:19:21 lets see... May 10 14:19:27 sudo apt-get install perl-modules May 10 14:19:53 18,1 M May 10 14:20:31 or it can be easily changed to not need experimental May 10 14:20:50 weird - it states perl-abse replaces it and it wont install it May 10 14:21:10 how to change the scrtips? May 10 14:21:18 tverrbjelke: which debian is this? stretch? jessie? May 10 14:21:23 stretch May 10 14:21:38 https://pastebin.com/8ukakUUV May 10 14:21:39 tverrbjelke: try perl-modules-5.24 May 10 14:21:41 I have the installed "-console" May 10 14:21:52 ah yeah, or apt-get install perl May 10 14:21:56 grr apt-serach is not installed May 10 14:21:58 or just use the modified script May 10 14:24:10 ok so both scripty use experimental May 10 14:24:22 how can I modyfy them? May 10 14:24:50 (not to use experimental) May 10 14:24:54 same change: remove the parameter list from the 'sub' line and add a my( ... ) = @_; assignment as the first statement May 10 14:25:36 sub slurp { my( $path ) = @_; May 10 14:25:52 "my( ... ) = @_; " And I comment the "use experimental"? May 10 14:26:06 and remove the use experimental line yes May 10 14:28:33 so I have this scrtipt now... https://pastebin.com/e6zfXi3s May 10 14:29:15 what wait? I gave a correct replacement for this script already May 10 14:29:27 the sub slurp is for the other one.. which actually has a sub slurp :P May 10 14:29:48 you just changed the name of the subroutine, which isn't going to work obviously May 10 14:29:55 I dont perl! at all May 10 14:30:14 'sub' in perl is like 'def' in python May 10 14:30:39 slurp for me e.g. just sounds like someone is eatung soup... hm SO what have I done fwrong? May 10 14:31:06 slurping is a somewhat common name for reading the whole contents of a file into memory May 10 14:31:17 :-) May 10 14:31:50 still I dont know what I have done wrong May 10 14:32:03 you renamed 'sub check' to 'sub slurp' May 10 14:33:06 and did trhis fix the experimental ting? I obviously dont understand your fix it by "xxx" hints May 10 14:33:27 fortunately I already gave you a fixed version earlier, https://pastebin.com/8ukakUUV May 10 14:33:32 * tverrbjelke woinders If NOW to learn perl May 10 14:34:58 the experimental feature I used is including a parameter list in the subroutine declaration (sub foo ( $arg1, $arg2 )) instead of receiving the arguments via the argument array (@_) and then e.g. destructuring that array (my( $arg1, $arg2 ) = @_;) May 10 14:36:57 hm I think I now understand that a bit more. May 10 14:38:50 no. So the argument to pass in is the emmc device? May 10 14:38:59 no May 10 14:39:10 the script itself doesn't take any arguments May 10 14:40:47 https://pastebin.com/AJyKCQNA here's the fixed version of the other script May 10 14:40:55 hm. SOK I will now enable this "copy sd to emmc" line... and reboot and then I see what the script is doing...? It currently barks at me because of missing llocales May 10 14:41:05 oh crap, autodie is no good either May 10 14:41:38 https://pastebin.com/MQ4653Tt here's the hopefully fixed version to work with perl-base May 10 14:42:01 missing locales? hum May 10 14:42:29 is some environment variable setting a locale other than 'C' without having the 'locales' package installed maybe? May 10 14:47:32 env only has this locales defined: LANG=de_DE.UTF-8 May 10 14:47:49 which is a locale that's not C May 10 14:48:07 presumably it got it from your ssh client... I'd turn that feature off May 10 14:48:41 so I could install locales? May 10 14:48:42 in /etc/ssh/sshd_config comment out the "AcceptEnv LANG LC_*" line May 10 14:48:55 or you could install locales if you want May 10 14:50:30 omg I am so used to aptitude... locales is pure virtual and I dont dinf the relevant package May 10 14:50:53 what? locales isn't virtual May 10 14:51:22 you can install aptitude if you want but it's sloooooooow May 10 14:51:31 apt-get install locales -> E: Package 'locales' has no installation candidate May 10 14:51:44 have you run apt-get update ? May 10 14:51:49 nooo May 10 14:51:58 (sniff) May 10 14:52:20 but i have set the ssh var as you suggested. doing update now May 10 14:52:28 now I have to eat some cake... May 10 14:52:29 "time aptitude show locales" ... 13 seconds May 10 14:54:38 ok thx fro your help. I will activate the "copy from Sd->emmc" and reboot... lets see. But first some cake... thx! May 10 15:00:39 Okay, still trying to boot entirely via USB ethernet gadget on a blank beaglebone clone ... I get uboot trying to load the kernel three seperate times then aborting. I've tried this with both zImage and uImage, and get the same results. The first time it loads (at 0x82000000) it says "missing environment variable: pxeuuid" and tries to load a bunch of pxe config files, the second time (at 0x80000000) it says "executing script" then "wrong image format" (ob May 10 15:01:16 n8vi: huh, I've never seen or heard of such issues May 10 15:01:40 I've never needed to mess with stuff like that, it loaded zImage just fine May 10 15:02:29 I'm not sure if I wrote the eeprom on this board, perhaps I need to do that May 10 15:02:45 I'm pretty sure I can do that from a u-boot script if necessary May 10 15:02:52 I'll check that May 10 15:02:57 if eeprom contained invalid data, u-boot would refuse to load entirely May 10 15:03:05 it's a modified u-boot May 10 15:03:11 ew May 10 15:03:13 patched to assume beaglebone black May 10 15:03:31 well, then there's also no need to program the eeprom May 10 15:03:39 it doesn't have any further consequence on boot May 10 15:03:43 okay, so the kernel doesn't care then? May 10 15:03:52 it sounds like you're not even getting to the kernel May 10 15:04:09 and no the kernel doesn't care about eeprom, it just cares about the device tree passed to it by u-boot May 10 15:04:25 I can get to a u-boot prompt without issues May 10 15:04:54 Lemme bootm it and see what it says, I've forgotten. I think the error message when I do that is different May 10 15:05:24 for me that was the tricky part for netboot. after that it loaded the kernel just fine (dhcp && bootm or something like that? I don't quite remember) May 10 15:05:31 => bootm 0x82000000 May 10 15:05:32 Wrong Image Format for bootm command May 10 15:05:32 ERROR: can't get kernel image! May 10 15:05:45 why are you manually specifying an address? I've never needed to do that May 10 15:05:51 I think May 10 15:05:57 I don't recall doing so anyway May 10 15:06:09 ah May 10 15:06:16 it probably was an environment var May 10 15:06:42 also it would be bootz for zImage May 10 15:06:58 "bootz $loadaddr - $fdtaddr" is in the usual bootscript May 10 15:07:30 You, I keep crashing it, when it comes up I'm going to try a straight bootz May 10 15:07:48 er, yep, not you May 10 15:08:14 I get "Starting kernel ..." a long pause, then uboot restarts May 10 15:08:46 did you pass a device tree? May 10 15:09:31 note that for this purpose a FIT (flattened image tree) may be useful. it combines the kernel, DT, initramfs (optional), and kernel parameters into a single file May 10 15:09:54 it's grabbing one that I have on the TFTP server sometimes May 10 15:10:07 "dtb/am335x-boneblack.dtb" May 10 15:10:19 be back in a bit, I'm getting pulled into a meeting May 10 15:10:46 ah it looks like that's what I used in my tests May 10 15:14:25 n8vi: https://pastebin.com/BuRDr26P this is what I found in my irc logs. the patch I'm referencing is just for netboot via ethernet, not relevant for usb netboot. so I actually remembered the "dhcp && bootm" right, and it worked because I was using a FIT evidently (based on my dnsmasq.conf serving a "linux.itb" file) May 10 15:22:30 at least I don't think it's needed for usb netboot? not 100% sure. here's a bit more context, including discussion of the patch: https://pastebin.com/K2TWuPPf May 10 15:22:51 maybe usbnet defaults to picking a random ethaddr if none is provided May 10 15:35:11 as for creating a FIT, here's what I use: https://pastebin.com/35D4RzNe May 10 16:19:36 thanks, zmatt, I'll take a look at that stuff and report back May 10 16:54:13 so, looking at your makefile fragment, what's "linux.its"? May 10 16:54:51 image tree source, I included it in my pastebin May 10 16:55:02 below the makefile May 10 16:55:28 oh May 10 17:32:31 Hi, gstreamer kmssink plugin is back after dist-upgrade to debian buster , but when testing it gives me an error: May 10 17:32:57 ERROR: from element /GstPipeline:pipeline0/GstKMSSink:kmssink0: Could not open DRM module (NULL) May 10 17:33:13 Additional debug info: May 10 17:33:13 gstkmssink.c(709): gst_kms_sink_start (): /GstPipeline:pipeline0/GstKMSSink:kmssink0: May 10 17:33:13 reason: No such file or directory (2) May 10 17:33:59 perhaps beaglebone DRM not recognized ? May 10 17:35:45 it looks like it has a hardcoded list of supported drivers for some odd reason. I've seen that before and don't understand why they don't simply query the device to confirm it supports the "dumb buffer api" May 10 17:35:58 https://github.com/GStreamer/gst-plugins-bad/blob/master/sys/kms/gstkmssink.c#L169-L173 May 10 17:36:28 that should include tilcdc and omapdrm May 10 17:36:36 or better yet, don't have a hardcoded list at all :-/ May 10 17:38:54 you mean that kmssink works for a restricted number of drivers/hardware ? May 10 17:39:02 nope, they just require the dumb buffer api May 10 17:39:10 so it'll work fine May 10 17:39:19 any workaround ? May 10 17:39:19 if it weren't for this stupidly written open function May 10 17:42:23 sudo perl -pi -e 's/radeon/tilcdc/gaa' /usr/lib/arm-linux-gnueabihf/gstreamer-1.0/libgstkms.so May 10 17:42:28 maybe make a backup of the .so first May 10 17:52:16 thanks, a step further , now: May 10 17:52:50 ERROR: from element /GstPipeline:pipeline0/GstKMSSink:kmssink0: GStreamer encountered a general resource error. May 10 17:52:50 Additional debug info: May 10 17:52:50 gstkmssink.c(1529): gst_kms_sink_show_frame (): /GstPipeline:pipeline0/GstKMSSink:kmssink0: May 10 17:52:50 drmModeSetPlane failed: Invalid argument (-22) May 10 17:52:50 ERROR: pipeline doesn't want to preroll. May 10 17:53:37 hard to say without knowing what it tried to do May 10 17:54:59 maybe try strace -o strace.log -f -- followed by the command, and see if it has more details about what went wrong... drmModeSetPlane() would show up as an ioctl() call, and with a bit of luck strace knows how to decode it May 10 17:55:10 I'm testing with a simple : gst-launch-1.0 videotestsrc ! kmssink May 10 17:56:50 it may be worth checking if it's possible to force the resolution and format May 10 17:57:36 actually, videotestsrc presumably has no way of knowing which resolution to use, and kmssink won't be able to scale the video May 10 17:57:54 so you should probably tell videotestsrc to use the appropriate resolution May 10 17:58:11 but maybe I'm wrong, I don't know anything about gstreamer really May 10 17:59:49 oh what the hell May 10 18:00:43 they didn't bother supporting any 16-bit format May 10 18:01:15 none is included in their table mapping between drm formats and gstreamer formats: https://github.com/GStreamer/gst-plugins-bad/blob/master/sys/kms/gstkmsutils.c#L43-L68 May 10 18:02:19 at least they do support using both RGB and BGR, so using 24 or 32-bit will work, but that's just a pointless waste of memory bandwidth May 10 18:04:50 so that's already two bug reports for you to file so far ;) May 10 18:04:57 does it have to do with previous error ? May 10 18:05:53 no May 10 18:05:59 I have strace.log , what should I looking for ? May 10 18:06:10 an ioctl() returning EINVAL May 10 18:06:20 probably near the end of the file May 10 18:07:10 2954 ioctl(7, DRM_IOCTL_MODE_SETPLANE, 0xb68b672c) = -1 EINVAL (Invalid argument) May 10 18:07:11 2954 ioctl(7, DRM_IOCTL_MODE_SETPLANE, 0xb68b672c) = -1 EINVAL (Invalid argument) May 10 18:07:18 crap, it doesn't show any details May 10 18:08:27 try passing -v to strace May 10 18:08:53 actually that probably doesn't help May 10 18:10:26 no, more than twice in length but same EINVAL rows May 10 18:11:16 there should be a way to enable debug info from gstreamer May 10 18:11:56 just set the GST_DEBUG var May 10 18:11:59 try setting GST_DEBUG=7 in environment May 10 18:12:00 yeah May 10 18:12:53 7 will give quite a lot output :) May 10 18:14:11 yes, but it's presumably needed to see this: https://github.com/GStreamer/gst-plugins-bad/blob/master/sys/kms/gstkmssink.c#L1482-L1484 May 10 18:15:22 but gstreamer log where is located ? May 10 18:17:00 my current guess is still wrong resolution May 10 18:17:00 lcdc has no flexibility whatsoever, it can't do scaling, nor padding, nor even cropping May 10 18:17:08 it doesn't log to stderr? May 10 18:17:42 it usually does May 10 18:19:41 rusted.....how to visualize sterr output ? May 10 18:19:47 stderr May 10 18:20:00 by opening your eyes and looking at the screen May 10 18:20:15 no , nothing then... May 10 18:20:25 then something isn't right May 10 18:20:38 are you sure you set the environment variable correctly? May 10 18:22:57 export GST_DEBUG=7 May 10 18:26:47 just realized I was wrong in setting variable....:-(( May 10 18:27:24 laval 7 gives me a ton of rows.... May 10 18:28:27 kids asking me for dinner......have to go, I'll take a look in 1-2 hours May 10 18:28:45 thank you very much for now May 10 18:29:30 maybe filter it by doing gst-launch-1.0 videotestsrc ! kmssink 2>&1 | grep -P 'Created BO plane|drmModeSetPlane' May 10 18:47:37 zmatt: it nicely worked out. this mmc tool - does it work only fro bb or is it more a general thing? May 10 18:47:53 (e.g. also similar with this odroid or other boards? May 10 18:48:08 any emmc May 10 18:48:28 as far as I know May 10 18:50:23 okay, zmatt, what should the fdt look like? It looks like everything is working except after it says loading kernel it hangs for a while then reboots u-boot May 10 18:51:16 the fdt should simply be a valid for the beaglebone (for the kernel used) May 10 18:51:34 for example am335x-boneblack.dtb May 10 18:51:52 okay, thats what I copied over was am335x-boneblack.dtb May 10 18:52:02 although if you're getting no kernel output at all, that actually suggests something else is wrong May 10 18:52:18 actually no, it could still be the dtb May 10 18:52:31 except apparently you're already using one that should work May 10 18:53:14 no kernel output means either it doesn't have a working console (e.g. wrong 'console' kernel parameter) or it crashes before the console can be initialized May 10 18:53:25 I'm using initramfs compiled into the kernel (and thus removed the initrd stuff from the linux.fts, but I wouldn't think that'd matter May 10 18:53:30 hmmm May 10 18:53:45 I think I enabled early printk too, but I should check May 10 18:53:46 that sounds very dubious, why would you use that? May 10 18:53:53 early printk? May 10 18:54:11 Apparently an option to allow printk to work earlier in the boot process May 10 18:54:31 just checking, have you first tested whether your oddball kernel boots at all? May 10 18:54:40 e.g. from sd card May 10 18:55:13 I guess I could just put the actual file on an sdcard and boot from that May 10 18:55:22 yes I know what early printk is, but I hope you configured it right since otherwise enabling it is more an "early crash" option May 10 18:55:36 okay, I can remove that May 10 18:55:45 I added it because this was happening May 10 18:56:18 be sure to use a kernel + dtb + initramfs that is known good, otherwise you have no idea what you're debugging May 10 18:56:34 and embedding the initramfs sounds like a really weird thing to do May 10 18:56:44 really? May 10 18:56:47 See I have no idea May 10 18:57:04 who wants to have to recompile the kernel just to update the initramfs? May 10 18:57:07 On the embedded and particularly the device tree side I'm scrambling in the dark here May 10 18:57:46 I'm just trying to make an installer. After it works I non longer care. It can go fetch it's install media via some other network protocol, copy it to emmc and reboot May 10 18:57:47 DT is just a data structure describing the hardware to the kernel, passed to it by u-boot May 10 18:57:54 in that regard it's a bit like ACPI I guess May 10 18:58:56 If I could figurte out how to write an image from TFTP or even HTTP to EMMC in a u-boot script, I'd be doing that instead May 10 18:59:50 But I don't think u-boot is quite that sophisticated May 10 19:00:20 I know how to do it via usb, although u-boot's drivers are pretty slow so you're probably still better off booting a small linux image May 10 19:00:57 yeah, it'd have to be usb, my board has no 802.3 May 10 19:01:07 ums 0 mmc 1 May 10 19:01:24 turns eMMC into an usb mass storage block device May 10 19:01:45 where do I put that? May 10 19:01:51 it's an u-boot command May 10 19:02:07 I'm not clear on that May 10 19:02:08 => ums 0 mmc 1 May 10 19:02:08 Unknown command 'ums' - try 'help' May 10 19:02:24 I guess for some reason your u-boot doesn't have ums compiled in? May 10 19:02:48 okay, I can recompile, lemme look at menuconfig and see what's missing May 10 19:02:52 I used the defaults mainly May 10 19:02:59 I thought it was enabled by default May 10 19:03:46 I know I can spew stuff into eeprom from a u-boot script, so this would possibly solve the other half of my problem ... hell, without having to run dhcpd and tftpd May 10 19:05:24 and then I could just do udev tricks on the host you plug the board into May 10 19:11:21 I've never used dhcpd and tftpd actually, only dnsmasq May 10 19:11:34 I see an "MUSB gadget mode support" option and a "USB Mass Storage support" option, but those're both checked May 10 19:15:56 hmmm, this is the precompiled u-boot I tested for a while, let me try switching back to the one I compiled myself May 10 19:16:12 although both are version blah.2018.blah May 10 19:20:33 nope May 10 19:21:11 Also my prompt is different than a lot of examples I find online, mine's "=>" whereas I see examples with "UBoot#" all over the place May 10 19:22:32 perhaps that's since the change to hush shell or something May 10 19:26:22 Also,there seem to be two sources for u-boot, https://github.com/u-boot/u-boot and http://git.denx.de/u-boot.git/. I'll try the other one to see if they're different. May 10 19:29:53 Hm I now try to boot the other (older A5C) BBB and cant get it running. root / blank password wont work either. May 10 19:30:15 when I boot without sd-card it seems responding to ssh, but I dont know any login data May 10 19:30:32 when i boot via SD-card (that I just generated) it seems stuck May 10 19:32:02 tverrbjelke: try powering on with the S2 button held down May 10 19:32:18 this bypasses the ancient u-boot on eMMC and uses the one on sd card instead May 10 19:32:25 brb May 10 19:32:56 ok, try it May 10 19:35:33 it now behaves differently, but is no loinger visible in my lan (nmap) May 10 19:36:10 (D2 & D4 blinking, D1(PWR) on May 10 19:44:31 NOW it worked. not sure what is differently done... tried 6-8 times, pressing s2 for 20s 1 min or sd out boot angström, then again ... May 10 19:50:42 the only correct sequence is: remove power, press S2 button, apply power, release S2 button May 10 19:51:52 the only possible reasons for that not to work is 1. you didn't actually hold down the little fidgety button 2. if it boots but isn't on lan, it might be the rare occasion where the ethernet phy bugs out (recognizable also by no or erratic link/activity indication) May 10 19:54:33 in any case you can let go of the S2 button once the power led turns on (or technically a few miliseconds after, but whatever :P ) May 10 19:58:30 zmatt: OK. Now I have booted from SD card and tried this: " ./mmc reliable-slc-configuration /dev/mmcblk1" but it results only in "Floating point exception" May 10 19:58:38 o.O May 10 19:58:51 what on earth May 10 19:59:03 are you really sure you booted from sd card? May 10 19:59:08 what does uname -r say? May 10 19:59:22 I think. ssh login was nice and clean this time May 10 19:59:32 4.9.88-ti-r111 May 10 19:59:40 oookay May 10 20:00:23 maybe there's something really weird with this ancient eMMC and some missing or untested error handling case results in that error May 10 20:00:24 i did tarball the whole mmc-utils, including the compioled binary from th eother bbb onto tzhe host and now back to the A5C one May 10 20:00:46 so it should be the native bbb compilerd one May 10 20:01:24 hceg_sectors=0 \ hcwpg_sectors=0 \ user_sectors=0 \ Floating point exception May 10 20:01:30 thats precisely whatr happened May 10 20:01:41 uhh what May 10 20:02:36 (that 4 lines were the output of "./mmc reliable-slc-configuration /dev/mmcblk") May 10 20:02:38 hmmmm, okay so maybe my code was never designed to deal with the case of a 2GB eMMC May 10 20:03:44 hmmm... you earlier mentioned, that the tool was designed to handlöe all kinda eMMC= May 10 20:03:50 X-| May 10 20:04:21 could it be worth that I play the unit test for this usecase? May 10 20:04:45 well, yeah, sorta, except I forgot it isn't due to changes in the eMMC spec to deal with large capacities May 10 20:04:54 can you pastebin the output of: May 10 20:05:48 is all statically linked into the binary? maybe I simply dont have installed a dependency/lib? Since I compiled it on the "default-factory" debian bbb May 10 20:05:59 sudo ./mmc extcsd read /dev/mmcblk1 May 10 20:06:12 it has no library dependencies, just libc May 10 20:06:47 also, if not then you'd get an error from the dynamic linker, not a floating point exception May 10 20:07:20 which is caused by a division-by-zero caused by hcwpg_sectors=0 May 10 20:07:25 https://pastebin.com/6SNwLPNr May 10 20:08:29 lol, the extcsd read command does the same fuckup May 10 20:08:40 "Max Enhanced Area Size [MAX_ENH_SIZE_MULT]: 0x0001ca i.e. 0 KiB" May 10 20:08:44 that's not 0 KiB you sillybeans May 10 20:10:02 I'll need to dig a bit how to get the correct info for low-capacity devices May 10 20:10:34 yess! I would in the meantime install me git and build-essential on the A5C BBB May 10 20:10:56 so I could just pull/push or whatever your code May 10 20:11:27 folks, natively hacking on bbb!!! May 10 20:18:19 * tverrbjelke pushes some kids into bed in the meantime. May 10 20:25:46 What is the maximum amperage the built in lipo charger will put out? Also, will it convert the Lipo's voltage to 5v? May 10 20:27:06 beagle_noob: the pmic does not have boost converter to generate 5v May 10 20:27:54 beagle_noob: also, running the beaglebone directly from lipo has a bunch of caveats, and isn't really safe for the processor unless a hardware patch is done May 10 20:28:37 see https://elinux.org/BeagleBone_Power_Management#Battery May 10 20:29:23 Dang... thank you zmatt! :) A bit unfortunate as it's the most powerful SBC that fits in an Altoids tin AND has a battery charger May 10 20:29:44 it *can* work, mind you... https://nurdspace.nl/BeagleBone May 10 20:29:53 uhh, I meant to link https://nurdspace.nl/images/0/0a/Bbb-zmattified.jpg May 10 20:56:00 huh, I cannot get the ums command to show up, weird. I enabled CONFIG_USB_FUNCTION_MASS_STORAGE and CONFIG_CMD_USB_MASS_STORAGE to no avail. May 10 20:57:18 also CONFIG_USB_MUSB_GADGET May 10 20:59:14 perhaps it needs a filesystem first May 10 20:59:26 uh, no May 10 20:59:41 I don't get it May 10 20:59:59 where did you get u-boot wherein it has this handy command? May 10 21:00:11 lemme try a preloaded beagle May 10 21:00:23 I'm pretty sure I didn't have to do anything special for it, it's always been there for me May 10 21:01:27 nope, not on this preloaded beagle either May 10 21:01:39 weird May 10 21:01:51 are you sure the ums command is available on this platform? Sure it wasn't some other board? May 10 21:02:00 100% May 10 21:02:10 what the actual hell May 10 21:02:16 I've used it often enough May 10 21:02:46 A beagle I put the latest beaglebone.org image on: May 10 21:02:48 Press SPACE to abort autoboot in 2 seconds May 10 21:02:49 [here I hit space] May 10 21:02:49 => ums May 10 21:02:49 Unknown command 'ums' - try 'help' May 10 21:02:49 => May 10 21:03:18 and if you invoke it with the right args? ums 0 mmc 1 May 10 21:03:28 just in case the error handling of the command parser is being stupid May 10 21:03:31 => ums 0 mmc 1 May 10 21:03:31 Unknown command 'ums' - try 'help' May 10 21:03:31 => May 10 21:03:34 ok May 10 21:04:13 well then... ¯\_(ツ)_/¯ May 10 21:04:13 zmatt: https://pastebin.com/MiLFHFwP May 10 21:05:06 fred__tv: yep, it's trying nonsense resolutions May 10 21:05:18 it looks like your display is actually 1280 pixels wide? May 10 21:06:40 no it's a LCD7 clone with a 640x480 display hardware (resolution is 800x640 however, like classic LCD7) May 10 21:06:51 huh May 10 21:07:45 then I have no idea what the hell it's doing allocating a buffer with stride 1280 May 10 21:08:37 you might need to ask gstreamer people about this May 10 21:09:06 tell them: the driver requires a contiguous framebuffer, so the stride (in pixels) needs to be equal to the display width May 10 21:09:26 hopefully someone knows how to convince gstreamer to do that May 10 21:10:58 but...I think I can give resolution options in gstreamer command... May 10 21:11:53 * tverrbjelke now back May 10 21:12:49 like: gst-launch-1.0 -v videotestsrc ! video/x-raw,width=1280,height=720 ! kmssink May 10 21:13:08 or something else that way.. May 10 21:13:21 zmatt: just asking cautiously: the problem here with the oolder revision bbb - something to quickfix or something that will/would take you longer? Could I provide you with more info? May 10 21:13:41 fred__tv: I thought you just said it wasn't that resolution? May 10 21:14:02 tverrbjelke: yeah I just checked quickly, it looks like it would be a fair headache May 10 21:14:30 oh crap. So I better go without slc my emcmc? May 10 21:14:42 video/x-raw,width=800,height=480 May 10 21:15:33 GOT IT !!! May 10 21:16:33 gst-launch-1.0 -v videotestsrc ! video/x-raw,width=800,height=480 ! kmssink May 10 21:16:40 i vahe set that pastebin only to "1 day". In case you going to work on debugging - maybe copy the text out for you May 10 21:16:46 I have test pattern displayed ! May 10 21:18:26 nothing else than wrong resolution... May 10 21:20:22 zmatt big thx for your time! May 10 21:20:34 well the resolution of the display is configured in the device tree overlay May 10 21:21:02 if it's not correct for the display then the DT needs to be fixed May 10 21:21:11 (the overlay for this cape) May 10 21:23:33 so, I have to modify dtb called in uEnv.txt ? May 10 21:26:12 well if it's an actual cape (with cape identification eeprom) then the correct overlay should get loaded automatically May 10 21:26:32 if that overlay isn't right, then the overlay needs to be fixed May 10 21:27:18 if you're forcing a specific overlay to be loaded then it is evidently not a correct one, and you need to find or make one that is May 10 21:32:01 .....I've done this months ago , I remember no more , I found something wrong in LCD7 dt fixed together with rcn May 10 21:33:07 wait, why /sys/devices/platform/bone_capemgr/slots exixsts no more ?? May 10 21:33:22 loading overlays at runtime is deprecated May 10 21:33:38 they're loaded by u-boot now May 10 21:33:41 how does it work now ? May 10 21:33:45 ahhh May 10 21:34:01 u-boot applies the overlays to the main DT before passing it to the kernel May 10 21:34:26 at boot ? May 10 21:34:32 yes May 10 21:35:11 so how to check which overlay is loaded ? May 10 21:39:48 uhh, u-boot's output on the serial console? May 10 21:40:02 it would indeed be nice if it left some notes about this somewhere May 10 21:40:42 if I actually cared about overlays myself I might have added some code to u-boot for that ;) May 10 21:46:11 to fix my display resolution I used to modify BB-BONE-LCD7-01-00A3.dts, then create the new dtbo, what to do now ? May 10 21:47:36 Hello May 10 21:48:52 FYI I cannot get the beaglebone black IoT image to connect to wifi May 11 02:29:01 Hello... May 11 02:29:47 How can I set Python3 on the BBB as my default Python? May 11 02:30:32 I am currently trying to download digi-xbee but it is only compatible w/ Python3. I used sudo apt install python3-pip. May 11 02:31:02 I then used sudo pip install digi-xbee. May 11 02:31:04 ... May 11 02:31:06 It did not work. May 11 02:31:18 Are there other people using the Python3 version of pip in here? May 11 02:33:10 I do not get it. Do I need to set the PATH of Python3 to my working directory? May 11 02:41:14 pip3: command not found <<<< this is what my terminal prints out May 11 02:41:39 Is this an issue w/ the BBB or just some software like Python? May 11 02:55:50 Forget it. I think the people on digi-xbee's side of things made a mistake. May 11 02:56:07 Who knows? Blah! **** ENDING LOGGING AT Fri May 11 03:00:03 2018