**** BEGIN LOGGING AT Thu Apr 01 02:59:56 2021 Apr 01 03:15:13 zmatt: I probably swapped it with a signal to solve some other function. one of the many examples of "if you don't test it, it probably won't work" Apr 01 03:36:52 jkridner: to be fair, figuring out pinmux is usually already bad enough for an application-specific device, let alone for something like the BBAI where you want as much as possible accessible while also meeting a bunch of backwards compatibility requirements Apr 01 03:36:57 not an enviable task Apr 01 03:38:55 still, at least there's one full-featured video input interface for 8/16/24-bit formats (even if 24-bit RGB/YUV will lose the least significant bit of the green/Cb channel) Apr 01 03:39:10 and two additional 8-bit interfaces Apr 01 03:40:18 (and the two 8-bit interfaces are actually proper iosets, except for vin2b vsync) Apr 01 05:52:06 jkridner: I updated my BBAI video input note ( https://pastebin.com/Lm1KnHXr ) with some more details Apr 01 06:16:32 There is so much available! I am being ripped off...'puters! Apr 01 06:17:38 GenTooMan: When you awake from your slumber, build the kernel w/ ftdi_sio support for linux for it to work w/ the BBB/BBBW! Apr 01 06:19:09 Damn WIndozz keeps telling me I have nothing installed. It is their damn fault. You have to alter everything in the kernel to have anything work these days. Blah. Apr 01 06:26:09 I plugged in the FTDI cable to early...argh. Apr 01 07:37:59 Hello Guys, Greetings! If you have any update for the mac issue that I asked for, please fell free to pass. Apr 01 07:46:51 Raj: you overestimate our ability to remember people and what issues they had :P Apr 01 07:47:53 if it's about usb networking: as far as I know it should just work plug and play (no driver installation), as long as the beaglebone is running a sufficiently recent image Apr 01 07:53:17 I thought, not to ask the same question again and againšŸ™‚ Apr 01 07:53:54 in general yeah Apr 01 07:54:01 ah, I see your question in scrollback now Apr 01 07:54:24 Yes, I read it like that everywhere but it don't work like that in the mac OS version 11.3 Apr 01 07:54:39 Thank you for that scroll Apr 01 07:54:53 so, first of all, you don't need to install any drivers Apr 01 07:55:17 Yeah but it didn't recognise Apr 01 07:55:30 have you reflashed it with the latest firmware yet? Apr 01 07:55:51 I see drive but ping to either 6.2 or 7.2 doesn't get reply Apr 01 07:56:11 Yes, debian 10.3 Apr 01 07:57:10 I know Apple somehow broke cdc-ecm in Catalina, but the latest image implemented a workaround for that Apr 01 07:57:53 ehm, hmm Apr 01 07:58:23 so, just to confirm, it doesn't show up as a network interface in settings? Apr 01 07:58:33 Yes, correct Apr 01 07:59:11 what steps did you do to reflash the beaglebone exactly? Apr 01 08:00:05 oh shit Apr 01 08:00:15 maybe the workaround isn't in the latest released image yet Apr 01 08:00:46 can you try reflashing it with the latest Debian 10 IoT testing image: https://elinux.org/Beagleboard:Latest-images-testing#Debian_10_.28Buster.29_IoT Apr 01 08:01:33 Sure, thank you so much. I will try and come back with the updates. Thank you again šŸ™‚ Apr 01 08:12:52 zmatt, if that is working then can you please tel me the patch to apply in the existing version?šŸ™‚ Apr 01 08:13:55 I mean if you have any link with you similar to the one above. Apr 01 08:14:17 https://github.com/RobertCNelson/boot-scripts/commit/b61125c1485bee929340cacc06c85c6fcfd678bc Apr 01 08:14:52 Thank youšŸ‘ Apr 01 08:15:05 basically just replace all occurrences of the word "ecm" by "ncm" in /opt/scripts/boot/am335x_evm.sh Apr 01 08:15:18 (the other files modified in that commit are for other boards) Apr 01 08:18:20 Oh ok thank you for the additional updates. Apr 01 08:18:48 do let me know if the testing image works Apr 01 08:19:52 Sure, doing it now. I will update you shortly. Apr 01 08:23:49 Hello! I have a problem with the black Beaglebone. I tested the two cans and they both work, but there is a problem. For example, if I use can0 and then I use can1 then neither can0 nor can1 work anymore. To make can1 work I have to restart the beaglebone Apr 01 10:07:37 zmatt, it is working ok in 10.5. thank you. I am going to do the patch update in 4.14 that I am using it now. I hope it will work ok. Apr 01 10:08:00 Pls ignore if sent already. Have network issue on my end Apr 01 11:38:56 jkridner: hmm, so the current images on https://beagleboard.org/latest-images have not had working usb networking for Mac users for basically a year now? I don't know what the status is of the next released image, but if it's not soon it might perhaps be worth considering hot-fixing the current release-images... the patch is pretty trivial: ... Apr 01 11:39:01 ...https://github.com/RobertCNelson/boot-scripts/commit/b61125c1485bee929340cacc06c85c6fcfd678bc Apr 01 11:44:21 pretty confident that works? at one point, I thought windows also had NCM. When I was trying to debug, I found order mattered greatly to Windows as it would stop once it found an interface it couldnā€™t load. As I recall, I got hung trying to figure out how to get Windows to load its NCM driver. Apr 01 11:58:28 jkridner: windows uses RNDIS, not NCM Apr 01 11:59:21 and as the commit explains, Apple managed to break CDC-ECM (specifically in composite devices iirc) in catalina, and that commit is the fix for it (but the released image does not include it) Apr 01 12:01:02 and just an hour ago we had a mac user who indeed had no working usb networking when the BBB was running the release image, while it worked fine with the latest testing image Apr 01 12:02:23 oof Apr 01 12:02:38 why can't OS's behave slightly more consistently LOL Apr 01 12:19:30 I'm really curious wtf Apple managed to do Apr 01 12:20:01 to break CDC-ECM Apr 01 12:20:56 there's also not much about it on the internet, so I'm guessing it doesn't actually affect many people Apr 01 12:21:27 so maybe a better question is, what does rcn's gadget do that causes ECM to break on Mac since Catalina Apr 01 12:27:16 My concern is the patch breaks windows, so if it is Windows tested, I'll encourage the merge and ASAP release. Apr 01 12:27:37 I think there are so many pending changes that it has made releases nearly impossible with the regressions. :-( Apr 01 12:27:50 oof Apr 01 12:27:53 not fun Apr 01 12:27:55 Perhaps rcn-ee can roll-back the clock and just apply a few incremental upgrades. Apr 01 12:28:02 Instead of doing Debian point releases, can we have a rolling Debian release? Apr 01 12:28:13 lol @ rolling release Apr 01 12:28:15 these days, we've been in the mire of trying to replace Cloud9-IDE with Coder (VS Code). Apr 01 12:28:25 Oh Apr 01 12:28:48 Yup, VSCode is good Apr 01 12:28:52 I don't see how it could affect windows, it supports neither CDC-ECM nor CDC-NCM, which is why we also have an RNDIS gadget Apr 01 12:28:52 As far as I know, rcn-ee was still hung-up with all the merges coming from LAST year's GSoC. Apr 01 12:29:14 What can be done to unblock him? Apr 01 12:29:46 zmatt: my experience is that Windows barfed as soon as it saw an interface it didn't know and would then not load the RNDIS gadget. As last I left it, whichever one was indexed first would cause either Windows or Mac to work. Apr 01 12:30:16 jkridner: but, it knows neither ecm nor ncm, so I don't see how replacing one by the other could have any impact Apr 01 12:30:31 Abhishek[m]_: I think if people give more attention to the weekly builds he does and search-for/fix regressions, one of those could become the release. Apr 01 12:31:05 zmatt: yeah, I don't know. If it is a simple swap-out and it makes Mac work, I agree, it seems it shouldn't break Windows. I'm just looking for some verification. Apr 01 12:32:01 jkridner: I mean, if you have a windows system around it's easy enough to test Apr 01 12:32:11 When someone did a HoRNDIS hack for Catalina, I installed it and haven't backed it out again yet. Apr 01 12:32:22 ew Apr 01 12:32:35 HoRNDIS should not be needed on mac, it hasn't been needed for ages Apr 01 12:32:48 I've got a drive around here somewhere with Windows installed. Apr 01 12:33:47 zmatt: indeed. The ECM breakage caused me to put it back in. At workshops, I've done pretty much exactly the patch you linked to get people's Macs working. Man, workshops, guess *THAT* was a long time ago. Apr 01 12:34:01 jkridner: One thing is that to fix issues, you have to replace the whole image vs. just some files here and there. Itā€™d be really nice to be just able to patch Apr 01 12:34:15 I think thereā€™s already git pull for opt scripts Apr 01 12:34:50 sure. I've been trying to think of better ways to create a rolling test image, but I don't think it is an easy problem. Apr 01 12:34:54 But swapping out whole images is wasteful Apr 01 12:35:09 many things can simply be done with live images pulling in Robert's latest packages along with his latest kernel.... Apr 01 12:35:15 that tests about 90% of things. Apr 01 12:36:30 Lately, I've been playing with containers, but -- Apr 01 12:36:39 Iā€™ve been thinking about it from 2 angles - one of them is to make Beagles boot ā€œdisklessā€ with TFTP and rootfs via NFS Apr 01 12:36:46 * jkridner listens to the sound of worms flying out of a can Apr 01 12:37:17 Abhishek[m]_: yeah, rcn-ee tries to support that with his u-boot image. Apr 01 12:37:27 he even releases a rootfs tarball. Apr 01 12:38:04 Yup, I doubt a 1-click system has been put together for that? Apr 01 12:38:08 would be good to create a clean flow that could be done in a test farm. Apr 01 12:38:28 Abhishek[m]_: don't think so. Apr 01 12:38:57 I day-dream sometimes about BeagleBoot providing all the features. Apr 01 12:39:32 at some point I've had a working bootp/tftp setup that allowed bootrom to netboot from scratch, i.e. bootrom fetches SPL, SPL fetches u-boot, u-boot fetches FIT and boots it... all via tftp Apr 01 12:39:39 having a fake network to run all the protocols on such that you can have a single-app provide all the interfaces cross-platform could be pretty handy. Apr 01 12:39:56 you don't really need a fake network Apr 01 12:40:28 all these steps perform quite specific bootp requests that normal dhcp servers are not going to answer, so there's no problem doing with on your main network Apr 01 12:40:34 the advantage of using the USB connection rather than a real network is removing the OS-specific networking integration in the test app. Apr 01 12:41:03 yeah usb makes things easier since you can just use libusb and do everything in unprivileged userspace Apr 01 12:41:07 which is what BBBlfs did Apr 01 12:41:15 (albeit poorly) Apr 01 12:41:18 exactly. Apr 01 12:41:23 And then BeagleBoot Apr 01 12:41:34 Whatā€™s the current status of that? Apr 01 12:41:56 An improved BeagleBoot/BBlfs that goes all the way to running an NFS image would be amazing for testing. Apr 01 12:42:06 well Apr 01 12:42:18 well, I tried to get parthi to push the code "upstream" into BalenaEtcher. Apr 01 12:42:18 BBBlfs/beagleboot has no reason to do anything with NFS Apr 01 12:42:26 * jkridner looks around for anuj Apr 01 12:42:35 it boots into an initramfs that makes the eMMC show up as usb mass storage Apr 01 12:42:48 so that you can then flash it just like an sd card Apr 01 12:43:12 right, but it could boot a u-boot that asks for a kernel and NFS rootfs. Apr 01 12:43:57 anyway, parthi made https://github.com/balena-io-modules/etcher-sdk/pull/204 Apr 01 12:44:04 wait, it got merged?!? Apr 01 12:44:24 if you're going for nfsroot then you'll probably want it on ethernet rather than usb, and it requires enough network infrastructure that setting up the bootp/tftp server (dnsmasq can do both) is not a big deal Apr 01 12:45:40 zmatt: again, the usage of libusb means that the NFS implementation would be OS independent. Sure, writing it in nodejs will probably make it pretty slow. Apr 01 12:46:04 what would the point of this be? Apr 01 12:46:16 like, why would you want to use nfsroot like that? Apr 01 12:46:45 even trying to get random people to pick a single tool like dnsmasq on Linux is about impossible. Apr 01 12:46:55 just for Abhishek[m]_ 's use-case... Apr 01 12:47:12 though, if reflashing is made easy enough, maybe doing the NFSROOT is not so necessary? Apr 01 12:47:13 well he was talking about proper nfsroot / diskless booting Apr 01 12:47:35 you don't want a crappy nfs reimplementation in javascript for that :P Apr 01 12:47:55 while for beagleboot/reflashing, there's absolutely no reason to use nfsroot at all Apr 01 12:48:08 sure. and maybe you just boot enough to get a diskless boot working. I was just going down the road of answering Abhishek[m]_ 's question "Whatā€™s the current status of that?" Apr 01 12:48:38 BBBlfs/BeagleBoot has no use for nfsroot Apr 01 12:49:14 it just sticks all of userspace into initramfs Apr 01 12:50:11 that's what it does for small distros, but for rcn-ee big debian images, something like NFS could be reasonable. Apr 01 12:50:53 * jkridner really wanted BeagleBoot to use Buildroot to build it's u-boot/kernel images. I don't recall if that was changed rather than dropping in random pre-built images. Apr 01 12:51:29 wait, back up, are we still talking about flashing beaglebones via usb, or do you just want to boot an image onto a beaglebone (without reflashing eMMC) ? Apr 01 12:53:42 for flashing the initramfs just contains a script that sets up g_mass_storage and that's it Apr 01 12:54:10 usb-booting a system would be a very different thing indeed Apr 01 13:05:19 jkridner: so, I just looked at bit at what my usb-flashing setup looked like (which we used for a while years ago), which was such that if you connected a beaglebone to its usb with S2 held (and no sd card), it would automatically boot the beaglebone into BBBlfs to make its eMMC show up as usb mass storage Apr 01 13:06:25 this was my dnsmasq.conf: https://pastebin.com/raw/U73FzAdi Apr 01 13:07:39 zmatt: yeah, I'd gone onto talking about not flashing the eMMC, but rather just serving up the image live over USB. Apr 01 13:07:58 * jkridner has to go AFK for breakfast. Apr 01 13:08:00 yeah, that's a different story Apr 01 13:08:12 though I still wouldn't use nfsroot probably, I'd be more inclined to nbd Apr 01 13:11:11 jkridner: is there a demand for that though? I don't think I've ever seen anyone ask for it Apr 01 13:12:25 Getting BalenaEtcher to natively flash is a huge deal. Apr 01 13:12:37 like, easy and reliable usb-flashing would be amazing to have, and would probably render sd-card based flashing largely obsolete. Apr 01 13:12:43 yeah exactly Apr 01 13:13:09 and as a cherry on top, once you have the beaglebone in UMS mode you can also use that to do make an image of the eMMC as backup Apr 01 13:13:15 Yeah! Apr 01 13:13:24 UMS is sloow though Apr 01 13:13:27 I think the problem @Abhishek_ and I are talking about in terms of lowering barriers to testing new images is half-solved there. Apr 01 13:13:32 When I tried Apr 01 13:13:44 UMS in u-boot is too slow. Apr 01 13:13:45 Abhishek[m]_: that sounds like an implementation issue, it shouldn't have to Apr 01 13:13:47 oh Apr 01 13:13:53 ums in u-boot is slow as fuck yeah Apr 01 13:14:14 that's why BBBlfs didn't use that, it booted a tiny linux system and used the g_mass_storage driver Apr 01 13:14:35 switching to a nice Buildroot image to do UMS with the kernel should be fine. Apr 01 13:15:21 BBBlfs and others didnā€™t do much to address the u-boot/kernel/rootfs rebuilding. Apr 01 13:15:25 I just extracted the initramfs... it's just the firmware/modules belonging to the kernel, busybox, half a million symlinks to busybox, and /init which is a shell script that loads g_mass_storage Apr 01 13:15:30 that's it Apr 01 13:15:37 (the BBBlfs initramfs) Apr 01 13:16:09 I recently saw a tweet that buildroot now works with precompiled toolchains so itā€™s faster to build images Apr 01 13:16:11 I personally packed that initramfs together with the kernel and dtb into a FIT Apr 01 13:16:57 jkridner: buildroot sounds like overkill Apr 01 13:17:06 there is no rootfs even Apr 01 13:17:29 it's just kernel, busybox, shell script Apr 01 13:18:00 and busybox + shell script could probably be replaced by a small C executable :P Apr 01 13:18:11 well Apr 01 13:18:24 maybe not Apr 01 13:18:47 I forgot for a moment that it still needs to load the kernel module, and that process is somewhat involved P Apr 01 13:18:51 Also USB MS in Beagle is slow as well but still faster than UMS Apr 01 13:19:13 than u-boot ums you mean Apr 01 13:19:16 "UMS" = usb mass storage Apr 01 13:19:23 Yup Apr 01 13:19:45 G_MS vs U-boot MS I mean Apr 01 13:20:20 I wonder if BBBlfs had USB DMA enabled in its kernel config Apr 01 13:21:09 Also RNDIS/CDC. I barely ever got 70Mbps on the link and I donā€™t know if thereā€™s enough financial incentive to improve the drivers to achieve full potential Apr 01 13:21:45 yeah the musb driver doesn't seem great Apr 01 13:22:26 but you could try recompiling the kernel with usb dma enabled (disable CONFIG_MUSB_PIO_ONLY) Apr 01 13:22:46 Didnā€™t change things much last time I tried Apr 01 13:22:52 apparently usb dma has been disabled due to "stability" issues, but it's possible those are specific to certain use cases only Apr 01 13:22:57 ah, bummer Apr 01 13:23:39 Thereā€™s cheap hardware all abound but I guess thereā€™s scarcity of HW tuned optimally for performance Apr 01 13:23:52 I don't think the HW is the problem Apr 01 13:24:13 I meant software stack and drivers Apr 01 13:24:31 s/HW/SW Apr 01 13:24:44 yeah, linux also tends to add tons of overhead everywhere Apr 01 13:25:05 that's the cost of versatility Apr 01 13:25:30 Iā€™ve been reading about io_urings and gift page kernel mmaping something Apr 01 13:25:39 Zero copy stuff Apr 01 13:25:58 well in this case the data never passes through userspace Apr 01 13:27:04 this case - ? Apr 01 13:27:10 I've been toying with the thought of making a little baremetal test to see how well the am335x usb can *actually* perform, but I don't actually care enough to justify making time for it Apr 01 13:27:14 Easiest way to build u-boot, kernel, rootfs with simple ā€˜makeā€™. Maybe setup to use host tools. Apr 01 13:27:20 in the case of UMS / beagleboot Apr 01 13:27:46 zmatt: How much time do you estimate it could take? Apr 01 13:28:07 I'm so bad at estimating such things that I'm not even going to try. Apr 01 13:28:28 note that I already have done some baremetal tests on the BBB a while ago Apr 01 13:29:40 the last status of my baremetal code to access eMMC was that I had completed initialization and read the eCSD :P Apr 01 13:29:48 and then I got distracted by more important things Apr 01 13:31:24 jkridner: I've never used it but I'll take your word for it... Apr 01 13:31:40 or maybe I have used it, hmm, but so long ago I don't really remember Apr 01 13:34:16 I think u-boot and kernel are universal solutions but if you really want blink-in-eye boot youā€™d have to fine tune whole stack Apr 01 13:34:46 Abhishek[m]_: my baremetal tests were always directly loaded by bootrom yes, no u-boot Apr 01 13:35:17 I used tftp boot since then I could just do "make" and have the bbb download the new binary via tftp Apr 01 13:35:53 the whole process was basically instant Apr 01 13:36:54 I'd press "r" on the serial console to reset the bbb, and the new application would be running before I'd lifted my finger from the keyboard Apr 01 13:37:21 of course it also didn't *do* much, but still fun Apr 01 13:45:15 hmm, another way to test the usb subsystem's actual capabilities would be to use linux but instead of using the kernel driver for the usb subsystem, use it directly from userspace Apr 01 13:46:05 (and if it needs a lot of constant attention, PRU could lend a hand) Apr 01 13:53:31 Hi, Good-M! ifconfig up/down modifies /etc/resolv.conf file by many apps. Any way to fix it? or name resolution not to use the file at all? Apr 01 13:54:11 you shouldn't use ifconfig to mess with interfaces when you're using connman Apr 01 13:54:30 you shouldn't use ifconfig at all Apr 01 13:54:36 it's old and obsolete Apr 01 13:54:49 use connman connect/disconnect? Apr 01 13:54:49 connman manages /etc/resolv.conf based on the configuration and status of network interfaces Apr 01 13:54:53 yeah that too Apr 01 13:55:24 connman can be told not to mess with resolv.conf Apr 01 13:55:33 mru to the rescue Apr 01 13:55:35 what does connman use to resolve such as 8.8.8.8 ? Apr 01 13:55:48 whatever dhcp says or you configure manually Apr 01 13:56:02 tiger: well normally DHCP informs you which nameservers to use Apr 01 13:58:20 In cell modem case, I know the server, but it needs a DNS for 'ping google.com' to resolve the name. Wifi has different DNS, so I try to fix it as well-known DNS 8.8.8.8 Apr 01 14:00:12 for both wifi and cell modem it normally should be auto-configured based on what the network is dictating Apr 01 14:00:17 via dhcp or via ppp Apr 01 14:00:51 but if both are available, connman and pppd will fight over resolv.conf Apr 01 14:00:56 yeaap Apr 01 14:01:22 we solved that by telling them to use different files Apr 01 14:01:31 like, the real solution is using a network manager that natively supports both ppp and wifi Apr 01 14:01:37 or, well, both ppp and dhcp Apr 01 14:01:40 and then a script watches them for changes and makes a symlink that points to the best one Apr 01 14:01:48 It is true that it auto-configured, but they add/remove nameserver into resolv.conf. Not just that, it also removes whatever was in the file Apr 01 14:01:50 mru: ew Apr 01 14:02:06 I'd love to get rid of connman Apr 01 14:02:08 tiger: yes, that's expected behaviour Apr 01 14:02:28 but it's not my choice Apr 01 14:02:58 tiger: if your modem was connected via usb instead of uart (so you could use cdc-ecm and dhcp instead of having to use PPP) you wouldn't be having this problem :/ Apr 01 14:03:03 so, the resolv.con is only be used by dhcp or dhcpd, right? Apr 01 14:03:24 connman is the dhcp client I'm pretty sure Apr 01 14:03:42 PPP through UART is FIXed. Apr 01 14:04:14 I know :) Apr 01 14:04:24 There are so many apps, tools, and configurations, very messy. Apr 01 14:04:34 and none of them quite work Apr 01 14:05:27 that's right. And one has some advantages over the other Apr 01 14:06:29 hmm, systemd-resolved supports changing the dns associated with an interface at runtime (using resolvectl or via dbus) ... I'd probably be inclined to see if I could rig ppp up write the dns settings it acquired there Apr 01 14:06:51 I had metric added into routing table, so both wifi and cell can co-exist with wifi higher priority Apr 01 14:06:53 but that would be what I'd try because I'm already using systemd-networkd and systemd-resolved Apr 01 14:07:04 yes, we know Apr 01 14:09:05 but, now if wifi turns on/off or cell on/off can still be a problem because of resolv.conf.Ā  Currently I have to manually add an entry such as 8.8.8.8 into the file Apr 01 14:09:20 why are you repeating this? Apr 01 14:10:20 we know that's the problem you're having, we've been discussing that problem for the past 15 minutes Apr 01 14:10:45 because if wifi off, cell stays on, but nameserver 8.8.8.8 is removed by wifi. So even cell is still on, it can not resovle 'ping google.com' the google.com name Apr 01 14:11:21 I feel like we're going in circles here Apr 01 14:11:21 sorry for repeating Apr 01 14:12:07 you guys typing too fast, faster than I can read Apr 01 14:12:44 ehm, you're free to take your time to read what's said Apr 01 14:13:13 :] Apr 01 14:14:02 mru: it could be nice if pppd didn't configure the interface at all (but just bring the link up) but intercepted dhcp requests and answered them with the information acquired from the ppp negoation... then you could trivially combine it with any network manager Apr 01 14:14:13 *would be nice Apr 01 14:14:47 feel free to make a patch Apr 01 14:16:04 ah it seems pppd at least doesn't rudely overwrite resolv.conf, it creates a separate file.. and also passes the nameservers via env vars to the /etc/ppp/ip-up script Apr 01 14:16:23 that would make the systemd-resolved solution potentially very easy Apr 01 14:17:27 ditch connman, use systemd-networkd + systemd-resolved and (manually configured) wpa_supplicant for wifi Apr 01 14:17:54 and have the /etc/ppp/ip-up script tell systemd-resolved what the nameserver(s) are for the ppp interface Apr 01 14:18:10 and let it sort what to do with that information Apr 01 14:18:34 where is the location of the script for wifi up/down? Apr 01 14:19:06 why would there be one? Apr 01 14:19:21 I don't see any reason connman would invoke a script when wifi goes up or down Apr 01 14:22:33 I saw a link for fixing resolv.conf using resolvconf.service Apr 01 14:23:11 sounds like it has a decent chance of making an even bigger mess Apr 01 14:24:56 mess patched by another mess? Apr 01 14:25:10 I don't think resolvconf supports connman Apr 01 14:30:38 yeah I'd lean towards the solution I outlined earlier Apr 01 14:33:53 I'm pretty sure systemd-networkd doesn't mind pppd, and like I said pppd passes the dns settings to the /etc/ppp/ip-up which would tell systemd-resolved about it (which can either manage /etc/resolv.conf or more commonly is used as a dns proxy that handles all the details internally) Apr 01 14:34:13 so none of the daemons should be fighting each other Apr 01 14:40:21 I found a solution:Ā  move etc/resolv.conf symlink into /run/systemd/resolve/resolv.conf, add a fixed nameserver into that resolv.conf Apr 01 14:40:32 tiger: what? Apr 01 14:41:19 why on earth would you move the symlink connman created to *that* location? Apr 01 14:42:08 /run/systemd/resolve/resolv.conf is the location of the resolv.conf file managed by systemd-resolved (which you're currently not using so I have no idea why you're doing anything in that directory at all) Apr 01 14:43:05 and yes you can remove the /etc/resolv.conf symlink and create your own settings file, but I know connman will still replace it again with the symlink (though I'm not quite sure when it does so) Apr 01 14:43:06 In my BBBw, /run/systemd/resolve/resolv.conf was not there. I just created Apr 01 14:43:10 WHY Apr 01 14:43:22 I don't know Apr 01 14:43:28 your actions make no sense whatsoever Apr 01 14:43:34 you're just blending random things together Apr 01 14:43:37 https://xkcd.com/763/ Apr 01 14:43:59 tiger: like I just said, that path belongs to systemd-resolved Apr 01 14:44:05 which you're not using (currently) Apr 01 14:44:16 hemce, that path is unused on your system Apr 01 14:44:33 if it _were_ used, you should never move things there or otherwise mess with its content Apr 01 14:44:43 I see Apr 01 14:44:52 /run/systemd/ is private state for systemd and its associated daemons Apr 01 14:45:13 just like /run/connman/ is the private state for connman Apr 01 14:46:20 Then, If I create ~/mydevice/resolv.conf, and /etc/resolv.conf - > that file, will that be ok? Apr 01 14:46:40 if you want to manually manage resolve.conf yourself that's fine, but you'll still need to ensure connman doesn't assume control of /etc/resolv.conf Apr 01 14:46:53 one of the first things mru said was: 14:55 < mru> connman can be told not to mess with resolv.conf Apr 01 14:47:06 you should probably have asked him "how do I tell it that?" Apr 01 14:47:29 tiger: there's no point to create a symlink like you just described Apr 01 14:47:40 a custom /etc/resolv.conf should just be a normal file located at /etc/resolv.conf Apr 01 14:48:12 things were so much easier before someone came up with the concept of network managers Apr 01 14:48:23 symlinks are used to make /etc/resolv.conf point to resolv.conf files created by connman, systemd-networkd, or pppd Apr 01 14:48:29 Sorry, I might miss mru's point. So How do I tell connman not to mess the file Apr 01 14:50:21 --nodnsproxy Apr 01 14:50:39 then it leaves /etc/resolv.conf alone Apr 01 14:50:54 it still puts whatever nameservers it knows about in /run/connman/resolv.conf or something similar Apr 01 14:51:02 you sure about that? it doesn't sound like that's what that option does Apr 01 14:51:12 dns proxying has nothing to do with maintaining /etc/resolv.conf Apr 01 14:51:13 well, it also disables the local dns proxy Apr 01 14:51:30 when the proxy is active, it puts localhost in /etc/resolv.conf Apr 01 14:51:52 it doesn't seem to be active by default then Apr 01 14:51:59 if you want to use nameservers connman doesn't know about, you obviously can't use its proxy Apr 01 14:51:59 since I don't have localhost in /etc/resolv.conf Apr 01 14:52:45 it has the dns caches of my isp Apr 01 14:53:02 it's a while since I fought this fight Apr 01 14:53:28 ExecStart=/usr/sbin/connmand -n --nodnsproxy Apr 01 14:53:31 it's already disabled Apr 01 14:53:41 on your system, yes Apr 01 14:53:53 it's the default in connmand itself Apr 01 14:53:53 on the latest IoT image Apr 01 14:54:28 right, but it's default disabled on debian evidently Apr 01 14:55:04 I can't be expected to know how every system has been configured Apr 01 14:55:17 doesn't matter anyhow since it clearly doesn't fix the problem :P Apr 01 14:55:37 I lost track of what the actual problem was long ago Apr 01 14:56:04 it doesn't look like connman has any option to make it not assume control over /etc/resolv.conf ... which seems to fit its theme of just assuming control of everyting Apr 01 14:57:22 my device has a read-only rootfs with an overlayfs Apr 01 14:57:34 and the overlay has nothing for /etc/resolv.conf Apr 01 14:57:40 so nothing has touched it Apr 01 14:57:45 it's easy enough to make /etc/ read-only for connman.service Apr 01 14:57:54 I wonder how it would react to that Apr 01 14:58:01 I expect badly Apr 01 14:58:31 evil solution: LD_PRELOAD hijack open() and replace /etc/resolv.conf with /dev/null Apr 01 14:58:45 @Abhishek_ can we get students and mentors to give https://rcn-ee.net/rootfs/bb.org/testing/2021-03-29/ a test? Apr 01 14:58:51 please nobody actually do that Apr 01 14:58:52 well I found a post of someone who used chattr +i as, apparently successful, workaround Apr 01 14:59:02 so maybe it just complains but moves on Apr 01 14:59:37 easy enough to test Apr 01 15:01:04 uhh Apr 01 15:01:08 huh Apr 01 15:01:24 it... already has ProtectSystem=full Apr 01 15:02:36 (i.e. /etc/ should be read-only to connman) Apr 01 15:04:22 yeah, connman doesn't create the symlink even if /etc/resolv.conf doesn't exist Apr 01 15:04:27 but I know something does Apr 01 15:04:41 found it! Apr 01 15:05:01 /usr/lib/tmpfiles.d/connman_resolvconf.conf Apr 01 15:06:23 let's see if this workaround works Apr 01 15:07:26 yep, it does Apr 01 15:07:57 tiger: ok, if you want to manually configure your /etc/resolv.conf, here's what you do: Apr 01 15:08:31 OK, steps? Apr 01 15:08:37 1. run command: sudo touch /etc/tmpfiles.d/connman_resolvconf.conf Apr 01 15:09:08 next? Apr 01 15:09:15 2. remove /etc/resolv.conf (the symlink), e.g. just "sudo rm /etc/resolv.conf" Apr 01 15:09:29 3. create /etc/resolv.conf as normal file with whatever settings you want Apr 01 15:11:17 of course this still isn't really a good way to handle it, since you're supposed to use the dns servers that are provided via dhcp (for wifi) or ppp (for cellular), but I guess just using 8.8.8.8 might work fine in practice Apr 01 15:11:20 I created a static file/etc/resolv.conf and added nameserver 8.8.8.8 Apr 01 15:12:11 using 8.8.8.8 as nameserver, will it have any side-effects? Apr 01 15:13:42 it might be slower, it's possible there are networks (especially corporate ones) that will block it, and you won't be able to resolve hostname of local devices (except using mDNS) Apr 01 15:14:33 and you're depending on google keeping the 8.8.8.8 up :P Apr 01 15:15:15 can resolv.conf include multiple nameserver entries? I guess yes, in order of the lines in the file? Apr 01 15:15:17 oh crap Apr 01 15:15:18 hold on Apr 01 15:15:40 so, step 1 earlier created an empty file at /etc/tmpfiles.d/connman_resolvconf.conf Apr 01 15:16:07 but I noticed it should not be empty, it needs to contain this line: Apr 01 15:16:10 If so, I can add manually a namesver entry into the file, and usually that is fixed. Apr 01 15:16:11 d /run/connman - - - - Apr 01 15:17:51 add "dĀ  /run/connman ----" into the file? Apr 01 15:18:30 the spaces between the dashes matter Apr 01 15:18:37 ok Apr 01 15:18:39 actually, just run this command as root: Apr 01 15:18:42 grep -v resolv /usr/lib/tmpfiles.d/connman_resolvconf.conf >/etc/tmpfiles.d/connman_resolvconf.conf Apr 01 15:18:52 that will create the file with correct contents Apr 01 15:20:52 added using your grep command, and cat the file, it has the line above Apr 01 15:21:48 Ok. It works great! Thank you guys! Apr 01 15:42:55 https://github.com/balena-io/etcher/issues/3477 Apr 01 15:43:14 beagleboot is now part of etcher-sdk, but not yet part of the Etcher app. Apr 01 15:43:43 nice Apr 01 15:48:17 at least it actually checks the contents of the packets it receives, unlike the original BBBlfs which just assumed it would receive specific packets at specific times and blindly replied based on those assumptions XD Apr 01 15:49:21 oh, this uses u-boot ums? Apr 01 15:49:48 so probably gonna be slow Apr 01 16:28:29 interesting that the new AM64xx already has 5 variants with product page and price listed for the non-experimental part number (though not orderable yet), suggesting it's preparing to go into production, while for the older AM65xx (from which the AM64xx clearly derives) only the top variant (AM6548) has a product page, and it only lists an experimental part... which is out of stock Apr 01 16:34:19 then again I suppose it's not that strange, given how much more stuff there's on the AM65xx... Apr 01 17:10:36 jkridner: I managed to make the LCD cape from Element 14 work anyway with the latest mainline kernel, U-Boot and Buildroot anyway (https://tinyurl.com/u94k5u7). I hope we can find the schematics anyway. This would help. Apr 01 19:46:22 https://www.newark.com/element14/6100310/beaglebone-ai-fan-cape/dp/50AH3704?COM=rew-BeagleBoard-102_94942-BEAGLEBOARD%20CMPNULL Apr 01 19:47:12 cheap Fan Capes, stand-alone fans and serial cables. Apr 01 19:47:22 er, AI serial adapter cables. Apr 01 19:51:00 $7 fan capes, $0.12 serial cables and $1.65 fans. Apr 01 20:05:36 Wow, cool! Apr 01 20:10:59 they still selling accessories for AI? thought that is becoming the next panda :D Apr 01 20:12:30 I got two! Apr 01 20:13:00 So, why are they only charging $7.00? Apr 01 20:13:55 Anyway, hopefully I can start using it more w/ the EVEs and DSPs onboard once this Maker Faire is done/kaput. Apr 01 20:14:19 set_: will there be a remote/streamed walk through? Apr 01 20:14:32 Of what? Apr 01 20:14:42 What do you mean walk through? Apr 01 20:14:48 Oh...of the maker faire? Apr 01 20:14:58 Maybe. Apr 01 20:15:01 the MF Apr 01 20:15:16 I'd be curious to see what bio related hacks are there Apr 01 20:15:25 I will have to see. I am going to video some of it (or try to). Apr 01 20:15:26 Oh. Apr 01 20:15:45 We are sharing the venue grounds w/ an Art Walk. Apr 01 20:16:05 how big is it? i am mainly used to MFBA which is huge Apr 01 20:16:20 So, there will be tons of people looking for art and us, the Maker Faire participants, will be a few. Apr 01 20:16:23 Tiny. Apr 01 20:16:27 One street. Apr 01 20:16:43 Our town is sort of a hub city where people just pass through. Apr 01 20:16:50 Not a large place. Apr 01 20:17:00 BA? Apr 01 20:17:02 oh Apr 01 20:17:07 Makerfaire Bay Area Apr 01 20:17:10 Oh! Apr 01 20:17:12 Yea. Apr 01 20:17:25 you can spend 2 days there and not get through it all Apr 01 20:17:25 We are in no way comparable to that size. Apr 01 20:17:30 No way. Apr 01 20:17:47 You can probably walk the street in six blocks. Apr 01 20:17:50 a. Apr 01 20:17:52 Ha. Apr 01 20:18:12 that's not bad... BN had a micro-MF and that was small Apr 01 20:18:39 I mean...last time, in 2019, we had it all indoors except for the wood/metal workers showing off stuff. Apr 01 20:19:30 This time, it is outdoors due to COVID. We are all supposed to wear masks. Apr 01 20:19:37 I will be one of those all. Apr 01 20:20:46 For some reason, they are loosening restrictions in most states. Louisiana is the place that is 50th on the list of places to visit, travel to, and move away from! Apr 01 20:21:01 Ha. We are dead last. Apr 01 20:21:22 Think of the poor children and their education. Argh! Apr 01 20:22:04 Teachers probably talk about what they did last weekend instead of studying to teach actual subject matter. Apr 01 20:22:13 is MF still mostly people showing off amateurish recreations of TV show props? Apr 01 20:22:24 SOmetimes. Apr 01 20:22:32 Oh! Apr 01 20:22:34 with a raspberry pi, obviously Apr 01 20:22:43 and some blinkenlights Apr 01 20:23:00 I think we have a dude or two that will do cosplay and another that is into star wars. Apr 01 20:23:38 He made a R2D2 in 2019. Apr 01 20:23:45 I think he was from TX, though. Apr 01 20:24:06 "The show must go on!" Apr 01 20:24:49 The children were fascinated by his R2D2 and lights/sounds. Apr 01 20:25:07 GenTooMan! Apr 01 20:25:42 Forget it, GenTooMan: I have to go. Second vaccine time. Blah. I will wake up in three days and be back! Apr 01 20:26:31 side effects that bad? Apr 01 22:00:55 mru, do you know how the C19 vaccines work? they basically give you a virus without genetic material in it so it can't take over your cells. However it will instigate an immune response. Apr 01 22:01:22 which vaccine? Apr 01 22:01:27 there are several types Apr 01 22:01:41 mru, the majority do that. Apr 01 22:02:05 there are three kinds of vaccine in use or being developed Apr 01 22:02:28 the simplest is the pure mrna method used by pfizer and moderna Apr 01 22:02:46 the latest is more like the traditional flu inoculation one shot. Apr 01 22:03:27 the side affects mostly depend on how the persons immune system response Apr 01 22:03:31 this consists of mrna strands coding the sars-cov-2 spike protein encased in a lipid shell Apr 01 22:03:59 when this enters the human cells, they start producing the spikes but nothing else Apr 01 22:04:34 the immune system recognises the spikes as hostile and develops antibodies, hopefully remembering how to do this quickly shold the real virus show up Apr 01 22:04:49 the loose spikes are actually totally harmless Apr 01 22:04:52 Aw! Apr 01 22:05:09 after a few days, the mrna disintegrates and the spike production stops Apr 01 22:05:46 the second dose acts as a reminder to the immune system that this is important, like revising before an exam Apr 01 22:06:19 see set_ was asleep already. Apr 01 22:06:28 * GenTooMan grins. Apr 01 22:07:15 SLeepy time dose is something I am born to fight. Buildroot! Apr 01 22:07:21 ha. Apr 01 22:07:23 SOrry. Apr 01 22:07:39 GenTooMan! Apr 01 22:07:54 hmm I'm wondering what the Johnson & Johnson one does to only need one injection. Apr 01 22:08:02 the AZ, Janssen (J&S), and Sputnik vaccines is both adenovirus vector based Apr 01 22:08:14 I had the serial tty to USB working. Apr 01 22:08:27 Now, I cannot sign in. It was a like a one time sign in. Apr 01 22:09:01 this means they've taken an already pretty benign adenovirus, removed the genes that allow it to replicate, and added the gene for the spike protein Apr 01 22:09:38 when injected, the vector (carrier) virus enters the human cells which start making the spike protein Apr 01 22:09:41 etc Apr 01 22:11:12 I am in! Apr 01 22:12:10 sinovac and a few others use a deactivated sars-cov-2 virus like traditional vaccines Apr 01 22:13:02 Also Covishield Apr 01 22:13:03 I don't know what's different between the single-dose and double-dose variants Apr 01 22:14:05 I think this is it Apr 01 22:14:16 maybe they've added some adjuvants to provoke a stronger immune response the first time Apr 01 22:14:27 you can probably find out if you dig around Apr 01 22:14:41 It probably also depends on how adversely body reacts, mild ones need more doses Apr 01 22:15:27 https://xkcd.com/2425/ Apr 01 22:15:37 it is an interesting subject. Cell division and replication is neat stuff... Apr 01 22:17:17 But that rna stuff is not something I study largely any longer. I think it is a nice subject. Apr 01 22:17:32 to be clear, I am not a biotech person Apr 01 22:17:42 Still, you can learn about it Apr 01 22:17:46 Like me! Apr 01 22:18:01 I am not biotech person either. Apr 01 22:18:36 I have no clue if my SD Card is sdb or sdc. Apr 01 22:19:15 Hmm. Apr 01 22:19:21 this is pretty interesting: https://berthub.eu/articles/posts/reverse-engineering-source-code-of-the-biontech-pfizer-vaccine/ Apr 01 22:19:59 Wild! Did someone steal the vaccine to reverse engineer it? Apr 01 22:19:59 my workstation has nvme for all fixed storage Apr 01 22:20:12 so removable things are sd-whatever-exists Apr 01 22:20:22 no, nothing is stolen Apr 01 22:20:25 OH. Apr 01 22:20:27 Okay. Apr 01 22:20:30 that's an analysis of published data Apr 01 22:20:35 Aw. Apr 01 22:20:47 the rna sequence is public Apr 01 22:20:53 Dang! Apr 01 22:20:56 exactly how they arrived at it is not Apr 01 22:21:01 Ha. Apr 01 22:21:18 People eating fungus that is uncooked from caves? Apr 01 22:21:33 That may just do it! Apr 01 22:21:37 it's obviously based on the sequence from the actual virus Apr 01 22:21:42 Oh. Apr 01 22:21:43 but there are some interesting tweaks Apr 01 22:21:48 * GenTooMan gives set_ a soap stone lamp so he can relax. Apr 01 22:22:03 I am fightin' the good fight over trying to not pass out! Apr 01 22:22:09 ha. Apr 01 22:24:39 hmm the vaccine isn't available to mere mortals yet in my state. They are still stuck in phase 1a (IE only vital people get it). Apr 01 22:25:09 still doing 50y and up here Apr 01 22:26:28 I smoke. I am a liability! Apr 01 22:26:32 well it's a good thing more vaccines are available. Apr 01 22:27:10 I just read that the Phizer doses got mfg. incorrectly, 15 million doses. Apr 01 22:27:18 Oops. Apr 01 22:27:34 A little lag in distribution now. Apr 02 01:32:39 whelp @zmatt, you are right. I rendered Ubuntu useless for cross-compilation (again). It is me! Apr 02 01:33:29 a reinstallation is needed. Aw and argh. Apr 02 01:34:21 Updating the WSL2 Ubuntu kernel did it. Blah, I should have never listened to the older script in which I was applying to get 'goodies.' Apr 02 01:35:09 I was ready too. I was about to apply the image to a SD Card. Apr 02 01:35:11 Argh. Apr 02 01:36:35 or...I dd'd the wrong thing. Apr 02 01:36:42 Probably the later. Apr 02 01:38:05 what is cfdisk? Apr 02 01:43:05 Forget it. It works now. This is too odd. Apr 02 01:47:55 I forgot to umount the dang sd card first. I am so mad at myself. aw... Apr 02 02:41:09 Anyway, there is no way to connect to the BBB imageless via six pin header and the FTDI cable...this is what buildroot states. Blah. Apr 02 02:41:22 Oh. On WSL2. Apr 02 02:41:34 WSL2 has no way to use usb devices... Apr 02 02:43:38 But, it saves it to the 'puter at some location. So, I can use the FTDI cable to port the image to SD Card and then to the BBBW (I have not figured out how complicated it is yet). Apr 02 02:52:06 two and a half hours to figure it out! Apr 02 02:59:32 If anyone is interested, the buildroot ideas work better than yocto on WSL2 on Win 10 w/ ubuntu 20.04 LTS and Debian 10. **** ENDING LOGGING AT Fri Apr 02 02:59:57 2021