**** BEGIN LOGGING AT Mon Sep 17 02:59:59 2018 Sep 17 10:47:26 * russell-- just built 8065-ge51aa699f7 for oxnas-pogoplug-v3, and like 7581-g7880a6f7fe, i'm not getting a shell on the serial console (i see boot messages, no shell, can ssh in) Sep 17 10:48:25 [ 0.590000] console [ttyS0] enabled Sep 17 10:48:40 wrong speed? Sep 17 10:49:31 i'd expect garbage Sep 17 10:49:42 pidof ash shows only the ssh session Sep 17 10:50:08 oh so no getty Sep 17 10:50:22 ? Sep 17 10:50:31 russell--: is that DT target? Sep 17 10:50:59 russell--: what kernel? Sep 17 10:51:09 actually, i think i found the problem Sep 17 10:51:34 system.@system[0].ttylogin='0' Sep 17 10:51:39 ok, just in case: i was considering to point target/linux/bcm53xx/patches-4.14/910-Revert-earlycon-initialise-baud-field-of-earlycon-de.patch Sep 17 10:53:54 mmm, I can just depend on USE_GLIBC right? Sep 17 10:53:59 or am I doing it wrong Sep 17 10:54:04 for a package, that is Sep 17 10:54:29 your package might be... ;) Sep 17 10:55:00 yeah well Sep 17 10:55:10 too much just doesn't work with musl without serious maintaining Sep 17 10:55:48 if they'd just implement even fake ucontexts, that would solve most Sep 17 10:57:13 hmm, the uci fix didn't do the trick Sep 17 11:03:45 question regarding /etc/config/wireless: was there ever an ifname option for wifi-iface? Sep 17 11:12:23 nemesis-ninux: there is still, isn't there? Sep 17 11:13:04 mine do Sep 17 11:13:46 wireless.@wifi-iface[1].ifname='wlan1' Sep 17 11:14:30 yes, OpenWISP uses this a lot Sep 17 11:14:31 russell--: does you kernel cmdline have a "console=" argument? last time I checked procd relies on that and doesn't support parsing the device-tree stdout property Sep 17 11:14:51 rmilecki: Linux pogoplug 4.14.68 #0 SMP Sat Sep 15 17:18:42 2018 armv6l GNU/Linux Sep 17 11:14:53 but a contributor is working on something and told me `ifname` is not documented in `wifi-iface`, only in `wifi-device` Sep 17 11:15:05 so I think the wiki now is missing this info Sep 17 11:15:16 KanjiMonster: looking... Sep 17 11:15:25 okraits: I was talking about you :-) Sep 17 11:16:02 russell--: if you see boot messages past "console [ttyS0] enabled" and it's not that issue Sep 17 11:16:04 KanjiMonster: /proc/cmdline just says earlyprintk Sep 17 11:16:19 rmilecki: i do Sep 17 11:16:23 just no shell Sep 17 11:17:11 tmp/.config-package.in:12517:error: recursive dependency detected! Sep 17 11:17:11 uh h Sep 17 11:17:13 oh Sep 17 11:17:15 what did I do Sep 17 11:17:29 from the ssh session, i can echo "HELLO IN THERE" to /dev/console and that appears Sep 17 11:17:56 russell--: is it just missing from inittab? Sep 17 11:18:15 inittab looks normal Sep 17 11:18:19 russell--: do you have console=... in bootards? Sep 17 11:18:21 *bootargs Sep 17 11:18:44 russell--: it's the last idea I got: target/linux/bcm53xx/patches-4.14/320-ARM-dts-BCM5301X-Add-serial-to-the-bootargs.patch Sep 17 11:18:45 russell--: then you need to add "console=" to the kernel cmdline (I guess in the dts?) Sep 17 11:19:17 that sounds promising Sep 17 11:20:58 russell--: the alternative is to enhance procd to parse the stdout-path property from device-tree, but I assume that's a bit out of your comfort zone ;) Sep 17 11:21:10 hm, I'm not selecting any of the deps mentioned, is it just broken depends in the packages, rather than my config? Sep 17 11:23:02 jwh: yes, that's an issue in package dependencies, regardless of the config Sep 17 11:23:11 aaah ok Sep 17 11:23:15 who done it? :D Sep 17 11:23:36 PACKAGE_kmod-openvswitch && PACKAGE_kmod-openvswitch-geneve Sep 17 11:24:09 it means there is essentially a "option A depends on B" ... "option B depends on A" (or something to that effect) Sep 17 11:24:14 yeah Sep 17 11:24:32 I thought it was my config but I realised I don't have that enabled anyway Sep 17 11:25:08 KanjiMonster: i'd be starting from scratch on procd, for sure ;-) Sep 17 11:26:10 russell--: I stumbled upon that when I added the uart nodes to brcm63xx, thinking I can now replace the serial=foo stuff from cmdline with stdout-path properties Sep 17 11:27:02 i'm building with the dts fix Sep 17 11:41:39 procd support for stdout-path would be great Sep 17 11:41:54 it's getting annoying to carry on that bcm53xx patch Sep 17 11:41:58 and it's getting bigger Sep 17 11:45:03 i'm also getting this during sysupgrade: [ 666.450000] Reboot failed -- System halted Sep 17 11:45:24 upside, i'm getting excercise going down in the basement to power cycle it Sep 17 12:01:19 hey are you guys all going to the summit? Sep 17 12:03:51 I won't :P Sep 17 12:04:03 just looking at flights, pretty dumb Sep 17 12:04:25 pas moi, /me is a-bunch-a-km's away Sep 17 12:04:51 heh Sep 17 12:05:15 if I go... my return trip is via dublin and ireland, seems like a farce Sep 17 12:05:27 err Sep 17 12:05:29 dublin and london Sep 17 13:28:19 hm, whatever happened to that pppd bcp patch Sep 17 13:29:39 speaking of which, not sure it made it into the kernel either Sep 17 13:32:45 nope, guess not Sep 17 13:32:46 awesome Sep 17 14:00:48 KanjiMonster, rmilecki: patching bootargs in the dts fixed the console, as you both suggested Sep 17 14:01:28 jwh: i'm going, flight booked Sep 17 14:01:35 russell--: hooray! :) Sep 17 14:01:53 i managed to hose it up (accidentally pasted ttyS1 instead of ttyS0, took a while to remember how to recover from that, lol) Sep 17 14:02:43 sysupgrade is claiming the firmware doesn't match, so i have to force and also reboot is broken, but those can wait until after sleepy time Sep 17 14:06:52 rmilecki: o/ Sep 17 14:07:28 https://lwn.net/Articles/376761/ Sep 17 14:07:31 well this is gonna be fun Sep 17 14:32:39 hm, forgotten how to prepare the kernel, there is a patch target or something I can specify right? Sep 17 14:34:38 jwh: www.openwrt.org : serach for "working with patches" -> adding/editing kernel patches Sep 17 14:35:47 thanks <3 Sep 17 14:37:37 google openwrt quilt works too ;) Sep 17 14:37:43 gonna give updating that bcp patch a whirl, my last one was for uh 3.x Sep 17 14:37:54 but you'd have to know the "quilt" keyword then of course Sep 17 14:37:57 stintel: yeah well I didn't know quilt was the keyword :P Sep 17 14:38:04 ;) Sep 17 14:39:06 lucky ppp hasn't really changed since then, so it shouldn't be that terrible Sep 17 14:39:14 BCP would be nice Sep 17 14:40:36 stintel: pretty sure googling openwrt quilt will still take you to the old page... Sep 17 15:08:23 meh that wasn't so hard, only 2 bits of the original patch that didn't apply, now to see if they've changed all the underlying stuff in the meantime :D Sep 17 15:12:07 I'm going to pretend I know what the wiki page is talking about Sep 17 15:15:47 either I can't read or its just broken english Sep 17 15:16:31 oh, I see Sep 17 15:20:17 right arse about face just to drop an existing diff in, no? Sep 17 15:29:11 hm Sep 17 15:29:38 question is now, do I alter config-4.14 (as an example), or hint some other way Sep 17 17:11:04 if I package an sdk from my own buildroot, adn then untar it and try and use it, what causes it to try and use key signing by default, and how can I turn it off? https://zerobin.net/?34b6bf53e29f416b#XBORpeCwPRAo9wTvgUvsPrL/fxyfErTRIyUzmWrfK2U= Sep 17 17:11:42 I can turn off CONFIG_SIGN in menuconfig within the sdk, but can I set that properly in my own buildroot that _makes_ the sdk?Ð Sep 17 17:13:23 I guess it's carrying "SIGNED_PACKAGES" frm my own tree? Sep 17 17:13:52 this doesn't feel very end user friendly, what's the intended use case? Sep 17 18:22:13 might be a stupid question, but I am unable to find it in documentation: can I access LuCi webinterface from LAN if I assign a different VLAN id to LAN? Sep 17 18:38:36 hm Sep 17 18:39:14 so if I want to set a kernel symbol as a bool, in .config, how can I add that? abusing kmod is obviously dumb and won't work since it assumes tristate Sep 17 18:39:37 so I wanna set a kernel config symbol if kmod-ppp is enabled Sep 17 18:43:14 I mean I guess I could enable it explicitly Sep 17 18:43:23 err, implicitly Sep 17 18:57:21 lynxis: so, about that simple netifd change we were discussing :-) Sep 17 18:57:39 lynxis: can I send a pr somewhere to make that glob less greedy? Sep 17 19:13:29 lynxis: hey, did you read my mail reagrding reproducible squashfs? I never received any reaction :( Sep 17 19:41:03 Pensacola: luci is jsut listening liek any other webserver. you can see what it's listening on with "cat /proc/$(pidof uhttpd)/cmdline | xargs -0" Sep 17 19:41:18 whether your network or firewall allows access to it is nothing to do with luci per se Sep 17 19:51:03 since I suck at taking notes and I've forgotten already, where are the netifd event hook thingers documented? Sep 17 19:51:28 wik isearch for hotplug or netifd Sep 17 19:51:50 mmm, I got the hotplug stuff Sep 17 19:52:56 guess I could just ubus calls to extract more info Sep 17 19:53:48 starting to think I should just write it in C rather than event triggered shell scripts Sep 17 19:58:59 what are you doing? Sep 17 19:59:52 mmm, NTE/CPE monitoring - but rather than meh snmp support (and the associated bloat that is net snmp), I started writing a shell script that procd and netifd can exec Sep 17 19:59:56 post load, interface info, etc Sep 17 20:00:45 I could literally just exec ubus call | curl every 60 seconds but that seems silly, especially if its got broken connectivity and needs to get a message out Sep 17 20:02:01 with a shell script its gonna be execing stuff, hoping to not totally break the state of the device in the process Sep 17 20:02:51 for example uh, reading forwarding tables without doing things like `ip ro` parsing Sep 17 20:04:10 I mean I guess I could use that openwisp thing but it doesn't look that complete Sep 17 20:05:05 and its kinda overkill for just reporting stats Sep 17 20:05:24 use snmp and call it a day? Sep 17 20:06:01 fine if theres flash to spare and I can usefully hook everything up as oids Sep 17 20:07:10 not really that practical though, and uh, may have unidirectional comms Sep 17 20:07:30 wonder if I should just shove json into udp packets and be done with it Sep 17 20:09:06 did you look at minisnmpd? Sep 17 20:09:17 i'm using it on an ipcam, it's lightweight & stable Sep 17 20:09:20 no v3 though. Sep 17 20:14:35 karlp thank you, so I think it should work Sep 17 20:14:40 I will try tomorrow Sep 18 00:05:24 Zero_Chaos: you can send a patch via the ML. maybe jow have time to setup a netifd mirror, so you can also send PR. Sep 18 00:06:20 lynxis: any chance I can just send you the patch? I'm on enough mailing lists :-) **** ENDING LOGGING AT Tue Sep 18 03:00:00 2018