**** BEGIN LOGGING AT Mon Sep 12 02:59:58 2016 Sep 12 03:14:23 Hey guys. Probably a common question, but I was wondering where I can find the BBB memory mapping diagram. You know, what addresses at boot are mapped to things such as GPIO pins etc. Sep 12 03:21:33 you mean the physical memory map? Sep 12 03:23:51 you can find it in the am335x TRM (technical reference manual), freely downloadable from TI at http://www.ti.com/product/AM3358/technicaldocuments#doctype6 Sep 12 03:25:43 you can find a more complete one at https://goo.gl/UHF2Fy though my spreadsheet covers more than one SoC family so you specifically need to check the "subarctic" column (which is the device family in which the am3358 is located) Sep 12 03:26:51 Perfect! Thanks so much :) Sep 12 04:27:22 hi Sep 12 04:46:51 hi Sep 12 04:46:55 anyone online Sep 12 04:46:56 ?? Sep 12 05:00:01 hinesh: just ask questions and wait patiently for responses Sep 12 05:59:02 My beaglebone is not starting up Sep 12 05:59:15 only usr1 and usr3 led is constant on. Sep 12 05:59:23 I have supplied 5v external supply Sep 12 05:59:39 and I have connected debug cable over debug port( j1 header). Sep 12 06:00:07 on terminal I am not able to see anything. terminal using: teraterm., System : win8 Sep 12 06:00:32 Debug port is connected to my pc usb via one converter. Sep 12 06:01:34 @vagrantc : I have posted my query here Sep 12 07:46:13 tbr: are you around? Sep 12 07:46:36 phschafft: mhm, zzzzzz Sep 12 07:46:57 *g* Sep 12 07:51:13 do you know by chance if there is anything special about the kernel in the debian images you linked me? I got the repo and I would like to build a fresh kernel. Sep 12 07:52:40 depending on which kernel you run, it's very close to vanilla Sep 12 07:52:41 hi Sep 12 07:52:49 hi Sep 12 07:52:53 my beaglebone black is not booting Sep 12 07:52:54 with just a few patches on top Sep 12 07:53:02 from emmc Sep 12 07:53:15 for that I have downloaded flasher image Sep 12 07:53:35 tbr: Linux beaglebone 4.4.9-ti-r25 #1 SMP Thu May 5 23:08:13 UTC 2016 armv7l GNU/Linux Sep 12 07:54:01 that was the one within the image. I also got the rcn-ee repo. Sep 12 07:54:21 ew, ti-rt... Sep 12 07:54:31 that's for some weird reason the default Sep 12 07:54:45 But when i try to write flasher image., Sep 12 07:54:56 its showing this error Sep 12 07:54:57 I'm pretty sure I've pointed you at http://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Kernel_Options Sep 12 07:54:58 1+0 records in 1+0 records out DISK SIZE - 3825205248 bytes CYLINDERS - 465 Checking that no-one is using this disk right now ... BLKRRPART: Device or resource busy sfdisk: This disk is currently in use - repartitioning is probably a bad idea. Umount all file systems, and swapoff all swap partitions on this disk. Use the --no-reread flag to suppress this check. sfdisk: Use the --force flag to overrule all checks. ./mkcard.sh: line Sep 12 07:56:19 phschafft: I'd personally probably go with 4.4 vanilla+patches, non-rt Sep 12 07:56:43 can you help me on this Sep 12 07:56:44 ? Sep 12 07:58:43 tbr: I'm actually patching kernel code. so my usecase is to ensure that it works with the current repo rev this friday. Sep 12 07:59:10 hinesh: you obviously didn't 'umount' the device before writing to it. You MUST make sure, that you select the right device node AND that none of it is mounted. Sep 12 08:00:08 phschafft: that's orthogonal to what I'm saying. There are 5 or six different kernels to choose from in the rcn package repository. Sep 12 08:00:09 Hi tbr Sep 12 08:00:15 thx for reply Sep 12 08:00:22 I did unmount everything Sep 12 08:00:47 It's says :- " line 82 of make.sh , kpartx not c\found Sep 12 08:00:54 --not Found Sep 12 08:01:16 tbr: sure, just wanted to make sure you understand that I'm not looking for the most user friendly thing or for a specific feature. Sep 12 08:01:25 sry, mkcard.sh .,, not make.sh Sep 12 08:01:46 ./mkcard.sh: line 82: kpartx: command not found Sep 12 08:05:32 hinesh: where does that mkcard.sh come from? Sep 12 08:44:25 hi Sep 12 08:44:41 mkcard.sh was there in /build folder Sep 12 08:44:57 I got one flaher image Sep 12 08:45:08 and burned that image over sd card. Sep 12 08:45:39 now I wanted that flaher to be wriiten on emmc Sep 12 08:47:46 hm, menuconfig doesn't show the module I'm interested in. strange. Sep 12 08:54:09 https://ibin.co/2unoxR0WA6XZ.png <-- beagle powered Sep 12 09:12:07 there's a little dog controlling the things? Sep 12 09:12:29 oh, powered, so a hamster-wheel thing but bigger? Sep 12 09:12:50 * tbr reports people involved for animal cruelty Sep 12 09:23:22 hi @tbr Sep 12 09:23:30 what should i do now Sep 12 09:25:34 hinesh: I have no idea which build system you are using and what sort of image that is. You probably should ask the community of that build system/distro. Sep 12 09:26:29 mkcard.sh ... sounds like the TI SDK Sep 12 09:26:52 go bug them on E2E :P Sep 12 09:30:38 Build systems! So much fun for the whole family Sep 12 09:31:36 if you consider a hacky shell script of the "let's hope it doesn't erase my HD" type a build system Sep 12 09:32:06 Aunt Tilly would be proud...... Sep 12 09:32:10 * Spidler mutters Sep 12 09:32:50 "Help, I've got ESR in my build system" Sep 12 09:39:27 image :- Angstrom-Cloud9-IDE-GNOME-eglibc-ipk-v2012.12-beaglebone.rootfs.tar.gz Sep 12 09:39:38 build is latest from elinux.com Sep 12 09:39:46 @tbr Sep 12 09:40:02 angstrom... latest... nope Sep 12 09:41:54 v2012 ? Sep 12 09:41:59 That's... Not quite latest. Sep 12 10:08:47 ok Sep 12 10:08:50 @tbr Sep 12 10:09:19 then if you can guide me the steps or link from where i can go on. Sep 12 10:11:07 hinesh: the question is what you are trying to do over all. what is your goal? Sep 12 10:24:22 @tbr Sep 12 10:24:40 my emmc on beaglebone is not booting., I had kernel panic there. Sep 12 10:25:04 So I wanted to flash emmc with fresh linux kernel Sep 12 10:25:17 @tbr, for that I am working Sep 12 10:25:21 OK and what do you want to have on the eMMC? anything specific? or just the latest and greatest image? Sep 12 10:25:38 just latest image with serial on Sep 12 10:25:48 no specific .. Sep 12 10:25:55 @tbr Sep 12 10:26:16 beagleboard.org/latest-images Sep 12 10:26:16 @tbr: latest image with serial enabled. Sep 12 10:26:28 I have downloaded .. image name with flasher Sep 12 10:27:04 you should download a RECOMMENDED image Sep 12 10:27:17 they can be modified to be flasher images, easily Sep 12 10:27:34 it even says so. RIGHT. ON. THAT. PAGE. Sep 12 10:27:48 But how I flash that recommended over emmc Sep 12 10:28:27 10:27:16< tbr> they can be modified to be flasher images, easily Sep 12 10:28:27 10:27:34< tbr> it even says so. RIGHT. ON. THAT. PAGE. Sep 12 10:29:09 @tbr : Sep 12 10:29:10 ok Sep 12 10:44:33 @tbr.. : I am trying now as shown.. it should worl Sep 12 10:44:36 *work Sep 12 10:44:44 if not ll ping you.. @tbr Sep 12 10:56:13 @tbr Sep 12 10:56:21 that image did not worked Sep 12 10:56:40 I have downloaded recomendded one and after following the instructions.. Sep 12 10:56:50 board didnt booted up Sep 12 11:40:16 * phschafft offers tbr both a strong coffee as well as some chocolate. Sep 12 11:47:27 phschafft: be carefull! Sep 12 11:47:34 phschafft: tbr is used to swiss chocolate Sep 12 11:47:41 phschafft: just "some" chocolate wont do Sep 12 11:50:06 KotH: I doen't like random 'chocolate' myself. But I would never offer him something not premium. Sep 12 11:55:34 phschafft: you underestimate what i mean by swiss chocolate Sep 12 11:55:50 phschafft: even "premium" german chocolate is not good enough Sep 12 11:56:49 and I think you underestimate 'premium' here. Sep 12 11:56:54 lol Sep 12 11:56:58 i live in germany Sep 12 11:57:18 Hi, does anyone know if I can read and write to the PRU0/1 control register from a Linux application? Sep 12 11:57:19 and i am originally from switzerland Sep 12 11:57:50 KotH: so, where in germany do you live? Sep 12 11:58:00 phschafft: arsch der welt Sep 12 11:58:05 phschafft: saarbrücken Sep 12 11:58:21 tommer: from user land? Sep 12 11:58:25 tommer: i dont think so Sep 12 11:58:47 tommer: but i've never done any pru stuff myself Sep 12 11:58:59 tommer: there should be tons of tutorials out there, though Sep 12 11:59:12 KotH: in fact I was in saarbruecken once. Sep 12 11:59:14 heh, there arent that much really. Sep 12 11:59:33 zmatt: what's the recommended pru tutorial these days? Sep 12 11:59:37 there was a small cafe somewhere with a little puddle of water and some ducks. Sep 12 12:00:25 KotH: how should I know? Sep 12 12:01:00 I hate bugs that go away with -DDEBUG Sep 12 12:01:23 zmatt: arent you a master of all that is BBB? Sep 12 12:01:58 I'm familiar enough with PRU, that doesn't mean I know what the "recommended tutorial" is (these days or any previous days) Sep 12 12:04:22 I have set the PRU to use the shift out GPO mode, and it shifts out data from both shadow registers perfectly. However, I would like to control the "Enable shift" from userspace. It would be nice If I could simply write to this register. I could also generate an interrupt on the PRU and in the handler flip the bit, but setting up interrupts is very time consuming! Sep 12 12:07:50 tommer: you cannot directly write form user space into a registr Sep 12 12:08:07 tommer: but you can write a simple kernel module that offers an interface for that Sep 12 12:08:15 One of these days I'll learn what the fuck I'm doing in life. Sep 12 12:08:21 or just use uio Sep 12 12:08:40 Spidler: you sound like a teenager Sep 12 12:09:02 KotH: or a hacker out of his depth looking at an oscilloscope ;P Sep 12 12:09:32 tommer: you can directly write from userspace into a register, via uio (preferred) or /dev/mem (lazy) Sep 12 12:09:32 For some reason I can't reproduce the testcases I had working last friday, and instead I just see garbage and corruption. Sep 12 12:09:51 Spidler: you need a machine "that makes bling" Sep 12 12:10:01 I currently use the UIO Sep 12 12:10:34 but when I try devmem2 to the config register address I dont get the right values Sep 12 12:10:46 Spidler: oh.. sorry "machine that goes bing" Sep 12 12:10:51 KotH: Mmm. Right now I've got a UART that will magically corrupt traffic on the bus if it's attempting to _read_ from it. I think that's pretty amazing, as I can't see shit on the oscilloscope, and I don't have a logic analyzer. Sep 12 12:11:00 KotH: clearly. And a patient to plug it into. Sep 12 12:11:07 A very expensive machine that goes "BING" Sep 12 12:12:47 But yeah, Working with this makes me long for ASN.1 <.< Sep 12 12:13:28 How are you folks, any good success and fun things happening this awesome Monday? Sep 12 12:14:11 tommer: oh, ENABLE_SHIFT is a bit of PRU r30, of course you can't access that from the cortex-a8 (while the PRU core is running anyhow) Sep 12 12:15:29 zmatt: then I should probably go for the interrupt solution? Sending an interrupt from arm to pru and handle the ENABLE_SHIFT in the interrupt routine? Sep 12 12:17:50 yes. (maybe "event" would be a better word than "interrupt" since the interrupt controller outputs to PRU merely show up as two bits of r31 rather than actually interrupting it) Sep 12 12:20:11 Spidler: I'm on the E2E StarterWare forum telling them StarterWare is garbage Sep 12 12:20:32 Spidler: as i've worked most of the weekend... no Sep 12 12:20:42 Spidler: doing easy stuff to waste some time productively Sep 12 12:21:09 zmatt: way to go! Sep 12 12:21:38 well having examples that compile into a deadloop if you compile with -O makes it rather easy Sep 12 12:22:11 and if anyone would ever dare to use -flto you can expect a particularly spectacular bonfire Sep 12 12:28:51 zmatt: how is "event" any different from interrupt? The PRU cannot be interrupted anyways, it has to poll a flag. Or am I not right? Sep 12 12:29:33 that's what I just said, and the reason using the word "interrupt" feels misplaced, since it doesn't :P Sep 12 12:30:47 I agree Sep 12 12:32:45 of course you can use interrupts like that on most processors that do have interrupts, like the cortex-a8. probably overall more efficient too, at the expense of interrupt latency Sep 12 12:35:43 sure, but when running Linux it may take multiple ms before the interrupt is actually handled Sep 12 12:36:23 I measured 44-88 microseconds from gpio line to userspace (via uio_pdrv_genirq) Sep 12 12:37:31 wow! thats very nice Sep 12 12:37:47 not when you consider that's 44000-88000 cpu cycles Sep 12 12:38:02 you can't use the m4 :P cos presumably the pins aren't mapped ... Sep 12 12:38:14 m4? Sep 12 12:38:22 cortex m4 core? Sep 12 12:38:36 thought .. nvm .. I'm delving midway though .. ignore me Sep 12 12:38:38 there's a cortex-m3, which you can use Sep 12 12:38:46 ah one ver out :P Sep 12 12:39:13 but it only connects to the l4wkup and has very limited memory Sep 12 12:39:20 zmatt: but the real question is, how fast can you toggle a io pin from user space using node.js Sep 12 12:39:35 KotH: :P Sep 12 12:39:41 the appropriate response to that is gagging the person asking that question Sep 12 12:39:42 that sounds horrible Sep 12 12:39:48 zmatt++ Sep 12 12:40:03 zmatt++ Sep 12 12:40:12 the question is reasonable .. but the answer is simply "Don't" Sep 12 12:40:35 and/or "Really slowly" :P Sep 12 12:41:49 well, just benchmark the performance of toggling a bit in a buffer object Sep 12 12:42:28 veremit: there are couple of websites out there on which you can find comparisons of gpio toggling with different languages Sep 12 12:42:41 KotH: languages, or lower level? Sep 12 12:42:59 veremit: languages... likey ruby, node.js, python, perl, c.... Sep 12 12:43:05 all userspace Sep 12 12:43:25 because only nation states are able to write drivers Sep 12 12:43:40 the obvious assertion is that you don't do it from userspace :P use the PRU Sep 12 12:44:27 then, assuming you're not crazy enough to run node as root to open /dev/mem, make an uio node for the target GPIO controller of choice (with an udev rule to allow suitable access), and https://www.npmjs.com/package/mmap it Sep 12 12:45:27 possibly using a typedarray instead of a buffer might be even faster Sep 12 12:45:51 mm interesting Sep 12 12:46:05 memory mapping is "scary" though Sep 12 12:46:39 if you're using sysfs, at least don't close and reopen the file every time like way too many people do :P Sep 12 12:47:16 but do/can you just leave it open and come back to it .. you need some discipline for opening/closing and not just randomly writing and reading Sep 12 12:47:43 an if you forget to close it .. does something "clean up" after you ... :P Sep 12 12:48:06 its all common stream stuff ... but most people don't know anything about IO/file access Sep 12 12:48:06 .o0(/dev/exynos-mem) Sep 12 12:48:12 it's a gpio pin, it ain't going anywhere. open it once at program startup Sep 12 12:48:43 zmatt: ... true ... but if you 'leave' it open .. what happens? Sep 12 12:49:02 then nothing happens, you have an open file descriptor? Sep 12 12:49:34 and you restart your app .. and facepalm!? :D Sep 12 12:49:34 but but but... open file descriptors leak! Sep 12 12:49:50 only if you keep opening new ones :P Sep 12 12:50:02 * veremit can see that happening .. Sep 12 12:50:12 veremit: you are aware that upon exit the kernel closes all open file descriptors? Sep 12 12:50:39 well then you run out, and you'll get an error "too many open file descriptors", and you can check in /proc what you've been spending those on Sep 12 12:50:43 KotH: ie. at shutdown, reboot, that makes sense Sep 12 12:50:50 veremit: no on process exit Sep 12 12:50:54 duh Sep 12 12:51:04 zmatt: right. Just sayin .. the levels of moron are unimaginable Sep 12 12:51:46 zmatt: your kernel exits -that- regularly does it?! :D Sep 12 12:52:27 people that stupid shouldn't be let near anywhere hardware, they'll blow it up Sep 12 12:52:34 zmatt: you think? Sep 12 12:52:45 * veremit wonders how many dead beagle's there out there .... Sep 12 12:53:18 we have small cemetary of 'em in the office, and those have actually fallen at the hands of people who mostly know what they're doing :P Sep 12 12:53:40 zmatt: yeah .. extend that to the "uninitiated" ... Sep 12 12:55:34 5 dead, 2 injured so far according to the internal wiki page Sep 12 12:55:47 "7V strike to the chest; instant death." Sep 12 12:57:37 Sorry for the disappearance, I just got lunch(!) Win win! Sep 12 12:57:47 zmatt: All Software Sucks. Sep 12 12:58:00 JS just moreso Sep 12 13:03:21 Also true. Sep 12 13:03:26 .... Sep 12 13:03:40 * Spidler stabs hardware Sep 12 13:04:03 yes, hardware sucks too Sep 12 13:05:04 You're starting to sound like a sysadmin ;) Sep 12 13:05:14 yes, users suck too Sep 12 13:05:37 users and printers. the worst things there are in IT. Sep 12 13:07:18 Printers are in this magical place of advanced software, file format interchange, networking, End users, fine mechanics, heat and power.... Sep 12 13:08:03 Either of them would cause direct and horrible errors. Sep 12 13:08:10 printk(KERN_INFO "lp%d on fire\n", minor); Sep 12 13:08:22 zmatt++ Sep 12 13:12:05 What the fuck is this driver doing? Sep 12 13:12:19 It's flipping RTS when it's _reading_ on a half duplex rs485? Sep 12 13:13:19 maybe it likes to hear itself speak ;) Sep 12 13:14:17 unless you mean it disables the drivers only when you read, in that case its philosophy is "if I'm not interested, there's no reason to let anyone communicate" Sep 12 13:14:40 No.. It's constantly flipping the fucking bit... Sep 12 13:14:58 Causing a ton of random shit to break on the bus. Sep 12 13:47:14 I'm seeing "Failed to start Generic Board Startup" today. Search has so far yielded naught. Sep 12 13:47:46 Ah. Now that I say something I see the next line. Sep 12 13:48:00 ew, xml Sep 12 13:48:21 i am a type. Sep 12 13:48:26 lol And on a *Monday*. Sep 12 13:48:36 I'm so happy for ayjay_t Sep 12 13:48:59 * ayjay_t is undefined Sep 12 13:49:26 * phschafft tries sizeof(ayjay_t). Sep 12 13:49:51 1 Sep 12 13:50:01 * Spidler = None Sep 12 13:50:18 No such file or directory is not as revealing as it could be, when one knows nothing about generic-startup.service. Guess I at least have more search terms. Sep 12 13:51:52 Google beomes less useful every month. Apparently "+" no longer means "MUST contain". Sep 12 13:51:55 hm. I like types that fit into registers. Sep 12 13:52:03 has anyone tried RT_preempt on Beaglebone black? Sep 12 13:52:14 Ragnorok: systemctl cat generic-startup.service Sep 12 13:52:25 Buy ayjay_t might fit into a register. Sep 12 13:52:46 Spidler: Empty Sep 12 13:53:01 Ragnorok: So, that sounds as if you have a symlink in /etc pointing at a file that's been removed. Sep 12 13:53:06 Ragnorok: that's been true for ages already sadly :( @ google Sep 12 13:53:11 Ragnorok: verbatim search helps a lot Sep 12 13:53:14 Ragnorok: find /etc/system -name generic-startup.service Sep 12 13:53:28 Appearantly if a word MUST exist in a search, you can quote it. Sep 12 13:53:38 Doesn't do the same as + used to do, but hey. Sep 12 13:53:38 no that doesn't work reliably either Sep 12 13:53:40 That doesn't work either. Sep 12 13:53:42 Oh fuck Sep 12 13:53:43 only verbatim search does Sep 12 13:53:46 Bah Sep 12 13:53:55 Google is turning into the perl of search Sep 12 13:54:00 Used to be it was easy to narrow search. Now it's heinous. Sep 12 13:54:06 "That might be somewhat close to what you meant" Sep 12 13:54:12 again, verbatim search :/ Sep 12 13:54:39 I have it as default since google search has become pretty much useless without it imo Sep 12 13:54:40 Oh for fucks sake.. getting defunkt sh process's inside fakeroot... Sep 12 13:54:42 *glare* Sep 12 13:55:26 Verbatim will replace "+". Moronic panty stains. Sep 12 13:55:33 without verbatim, adding carefully chosen search terms to find a needle in a haystack will just result in google noticing "hey, if I ignore one of the search terms, I find a whole haystack!!!!" Sep 12 13:55:40 zmatt: Thanks! Sep 12 13:56:10 Right Sep 12 13:56:26 SERP? WTF? Sep 12 13:56:43 iirc generic-startup is a sysv startup file, not a native systemd one, but I might be wrong nowadays Sep 12 13:56:44 I don't see the image they say. Sep 12 13:56:59 it is useless and its primary purpose is significantly increasing boot time Sep 12 13:57:12 The file that's missing appears to /etc/system/generic-startup.service Sep 12 13:57:46 Ragnorok: Empty symlink? Just remove it. :) Sep 12 13:57:53 No symlink. Sep 12 13:58:05 why do you think such a file should exist? Sep 12 13:58:22 Spidler indicated a find for it. Sep 12 13:58:32 I indicated it's a sysv service Sep 12 13:58:57 I defer to zmatt there, the reason I told you to search for it was that your question before mentioned ".service" Sep 12 13:59:02 Roit. Sep 12 13:59:25 So far searching for anything has failed, and I can't see how to get google to do what I want. Sep 12 13:59:28 And I also know by experience that dangling symlinks in /etc/system/ will cause systemd to complain. Sep 12 14:01:12 Ragnorok: grep -r generic.sh /etc/init.d Sep 12 14:01:16 I don't see a URL arg to set verbatim. Do I actually have to click that f'ing button every time?? Sep 12 14:01:33 tbs=li:1 Sep 12 14:01:46 Clever. Thanks. Sep 12 14:02:18 Couldn't make it obvious, like "verbatim", so one could easy enter it every time. Sep 12 14:02:38 grep is empty. Sep 12 14:03:36 hmm maybe it's invoked via an even more indirect route Sep 12 14:04:00 grep -r generic /etc/init.d /etc/systemd/system Sep 12 14:04:07 I really don't care if it's not doing anything important. I mean the board seems to work, I just don't like errors. Sep 12 14:04:32 and rightly so Sep 12 14:05:06 (I do care if it's not doing anything important since it showed up as a major factor in boot time in "systemd-analyze plot") Sep 12 14:05:38 FYI, "Verbatim" also doesn't include hits with "generic-startup.service" in them. Moornic panty stains. Sep 12 14:05:55 Not in them. Sep 12 14:06:02 huh Sep 12 14:06:03 I would expect that term to seriously narrow the hit list. Sep 12 14:07:09 I don't necessarily, but I did expect a hit on generic Sep 12 14:07:16 I repeat. Google is less usable every month. It's increasingly difficult to search effectively, where it USED to be very easy. Sep 12 14:07:20 since it's indirectly invoking /opt/scripts/boot/generic-startup.sh Sep 12 14:07:55 Nada on the find. Sep 12 14:08:01 find or grep? Sep 12 14:08:07 This is a console image. Grep sorry. Sep 12 14:08:16 it's even there on a console image afaik Sep 12 14:08:24 I took out syslog and that's around when this started, afaik. Sep 12 14:08:28 if it weren't there, nothing would be generating an error :P Sep 12 14:08:37 lol True. Sep 12 14:09:11 systemctl | grep generic Sep 12 14:09:11 ? Sep 12 14:09:39 Service is there, loaded failed. Sep 12 14:10:22 ok maybe it is a native systemd service but located in /lib/systemd instead of /etc/systemd Sep 12 14:10:56 Nice. I search for generic-startup.service and get something sort of relevant. So I add "error". All the hits are now about Generic Host Process for windows. W T F. Sep 12 14:11:24 use quotes around generic-startup.service Sep 12 14:11:47 will get you absolutely nothing useful again Sep 12 14:11:51 On hit about Android "startup service". Sep 12 14:12:27 well at least it searches for those three words in the right order instead of merely those three words :P Sep 12 14:12:29 Effing google. What used to be their strong point is now more useless than Bing. Sep 12 14:13:16 don't search the interwebs, search your beagle... systemd didn't pull that service out of its ass Sep 12 14:13:30 Roit. Sep 12 14:14:54 actually, systemctl status generic-startup.service should show the path Sep 12 14:17:05 (on the second line) Sep 12 14:17:51 That's where it says no such file or directory. Sep 12 14:18:06 ok, that is fascinating Sep 12 14:18:26 you removed it but something tried to pull it in as dependency? Sep 12 14:18:34 I'm doing a brute-force find -exec grep on the entire unit. Sep 12 14:19:03 I would suggest adding a -type f unless you like errors Sep 12 14:19:05 I don't recall removing it explicitly. Sep 12 14:19:18 I just 2>1 and don't see them. Sep 12 14:19:24 grep can also take quite a while searching /dev/zero Sep 12 14:19:30 an infinite while to be precise Sep 12 14:19:52 I've done this before and it wasn't infinite, but that doesn't mean today won't be my lucky day. Sep 12 14:20:21 perhaps trying to start the service using strace can reveal the path it is trying to reach? Sep 12 14:20:24 odd, it should be Sep 12 14:20:49 When you say it, it sure seems appropriate. Sep 12 14:21:37 searching /dev/urandom is equally time-consuming :P Sep 12 14:22:56 but /dev/random is fast! Very fast, until it stops ;P Sep 12 14:23:05 Because dev/random is so fucking broken on Linux! Sep 12 14:23:16 //dev/hwrng might also take a bit Sep 12 14:23:37 Today is not my lucky day. I have a command line. Sep 12 14:23:53 command lines are awesome. In the beginning, was the command line. Sep 12 14:24:07 Look like the only ref is in ./etc/systemd/system/multi-user.target.wants/generic-board-startup.service. Sep 12 14:24:49 Ragnorok: and cat on that file will show you that it's doing execStart sh /opt/scripts/boot/generic-startup.sh Sep 12 14:24:57 lol Sep 12 14:25:21 no, cat on that file will give ENOTFOUND Sep 12 14:25:27 Heh Sep 12 14:25:30 It should Sep 12 14:25:35 since if it weren't a dead symlink the earlier grep would have fit Sep 12 14:25:35 That shit is doing stupid things Sep 12 14:25:37 *hit Sep 12 14:25:40 Ah Sep 12 14:25:52 Actually I got what Spidler said, and /opt/scrips/boot/ is also present. Sep 12 14:26:13 huh, but grep -r generic /etc/systemd/system gave no results? Sep 12 14:26:31 yes I mentioned that path earlier too Sep 12 14:26:54 Grep no, yeah I recall. Wanted to follow the trail. Sep 12 14:27:16 that's exactly why I suggested grepping on 'generic' Sep 12 14:27:20 Kinda messed up that grep doesn't see it. Sep 12 14:27:36 Yup. Sep 12 14:28:06 Ah. Caps. -R Sep 12 14:28:12 ah, they fixed that Sep 12 14:28:24 I seem to recall it annoyingly followed symlinks with -r Sep 12 14:28:39 One mystery solved at least. Sep 12 14:28:42 I either remember wrong or it's been fixed Sep 12 14:30:16 and the hwrng generates 64 bits per 650ns with the config used in linux Sep 12 14:31:46 http://lxr.free-electrons.com/source/drivers/char/hw_random/omap-rng.c#L154 Sep 12 14:31:58 Anyway, it point to the boot script, that boot script exists, it's executable. Wading around inside to see what it may be missing. Sep 12 14:33:33 I wonder what the hell the kernel is doing to get such sucky speeds as mentioned in that comment (2 Mbit/s) Sep 12 14:33:46 zmatt: no clue. sounds wrong? Sep 12 14:33:51 it should be able to get 100 Mbit/s Sep 12 14:33:58 Not that I'm doing much with the rng Sep 12 14:34:12 can't honestly say that I trust it ;P Sep 12 14:34:27 of course whether extracting at such a rate is a good idea I don't know Sep 12 14:34:30 but it's probably better than the openssl built in one. Sep 12 14:34:46 Because that one is fucking _stupid_ in what it litters on your drive. Sep 12 14:35:11 I thought openssl uses /dev/urandom normally? Sep 12 14:35:23 "ha ha ha" Sep 12 14:35:33 or getrandom() Sep 12 14:35:37 No, openssl seeds it's own rng with something.. .sometimes.. random Sep 12 14:35:41 And tosses that seed in your home dir Sep 12 14:35:46 Or somewhere else where it might be Sep 12 14:35:51 And then starts fucking around Sep 12 14:36:01 Most distributions these days patch it to not be a piece of shit Sep 12 14:36:20 I'm not sure if OpenSSL 1.1 released with that code removed Sep 12 14:36:54 The rand command outputs num pseudo-random bytes after seeding the random number generator once. As in other openssl command line tools, PRNG seeding uses the file $HOME/.rnd or .rnd in addition to Sep 12 14:36:58 the files given in the -rand option. A new $HOME/.rnd or .rnd file will be written back if enough seeding was obtained from these sources. Sep 12 14:37:10 "Blubb-blubb-blubb-I'm-fucking-drunk-and-I-should-lay-off-the-coding" Sep 12 14:38:49 on all calls to openssl rand I see it reads from /dev/urandom Sep 12 14:39:10 in addition to fucking around with its own .rnd file, which I'll now want to disable if I can :/ Sep 12 14:39:17 Thanks for the clues, all. Sep 12 14:39:32 zmatt: there's an environment variable... lemme see Sep 12 14:40:27 zmatt: $RANDFILE Sep 12 14:40:37 yeah I see it in openssl.cnf Sep 12 14:40:54 if neither RANDFILE nor HOME is set, it will use .rnd in current dir, or none at all.... Sep 12 14:41:06 It's like a collection of stupid ideas all at once Sep 12 14:42:22 it seems to ignore it though Sep 12 14:42:48 Clearly Sep 12 14:42:55 ah it works if provided via env, though not in /usr/lib/ssl/openssl.cnf Sep 12 14:43:01