**** BEGIN LOGGING AT Mon Nov 30 03:00:00 2020 Nov 30 04:47:49 Hauke: so, i built my config on a bigger memory device (tplink wdr3600, which has 128MB of RAM) and it seems usage (reported by /usr/bin/free) climbs to about 20-21 MB and stabilizes, so it looks like not a leak as such. also the netifd process itself doesn't grow, which implies there is some kernel data structure or buffer that is growing. usage above about 19MB becomes problematic on the 32MB Nov 30 04:47:55 ubnt bullet. Nov 30 04:50:12 mangix: mind becoming the maintainer of whois? Nov 30 06:27:45 why? Nov 30 07:27:17 mangix: thanks for fixing the broken packages Nov 30 09:00:13 what a patch... Nov 30 09:00:27 do people actually tether their iphones to their routers? Nov 30 09:01:32 they do apparently. there's been people complaining in here Nov 30 09:02:03 One of my gripes with iOS is that you can't tether your wifi connection Nov 30 09:02:20 unless you jailbreak and buy a $5 tweak Nov 30 09:04:42 you don't own your apple hardware, apple owns you :) Nov 30 09:04:55 how do you figure? Nov 30 09:05:45 you said it yourself. to get functionality you have to jailbreak, otherwise your stuck to whatever apps Apple sanctions. Nov 30 09:06:39 it's actually worse than that, but better than selling yourself to advertisers Nov 30 09:07:14 mangix: I don't remember the last time I saw an ad on (my) Android… Nov 30 09:07:18 I'm not suggesting Google are the good guys here, don't get me wrong. Nov 30 09:08:35 mangix: it's a difficult tradeoff... i'd rather get and use an iphone, but android at least you can tailor to your liking. or use stuff like lineage. Nov 30 09:12:07 Do you know if there is a way to roll back iOS 13? not much is showing up online Nov 30 09:12:10 Apple doesn't sign iOS 13 anymore, so no. Nov 30 09:12:14 hahaha Nov 30 09:12:24 apple dropped wired tethering in ios 14. it still works in ios 13 Nov 30 09:13:02 what do you mean they dropped it? Nov 30 09:13:10 no longer support Nov 30 09:13:18 dropped aka nuked. :) Nov 30 09:13:22 in ios 14, it's wifi tethering only Nov 30 09:13:42 so what's this about? https://github.com/openwrt/openwrt/commit/f64496f30f2ef97124dc4e13a48ee0de9d51832e Nov 30 09:14:11 https://support.apple.com/en-gb/guide/iphone/iph45447ca6/ios seems to indicate that USB based tethering still works Nov 30 09:15:38 and it works just fine with iOS 14 and MacOS 11 Nov 30 09:15:48 just went through the libimobiledevice issue. no idea how they figured it out Nov 30 09:15:56 I just turned off wifi on my macbook, hooked it up to my iphone via USB cable, and it just worked Nov 30 09:16:37 SwedeMike: interesting. someone knowledgeable said it was wifi only now... i guess either they were mistaken or i remember incorrectly Nov 30 09:16:43 good to hear, though! Nov 30 09:16:52 "ne thing I noticed is that the Windows and macOS drivers seem to issue multiple simultaneous receiving URBs, whereas the Linux driver issues only one at a time and waits for it to complete before receiving again." Nov 30 09:16:56 interesting... Nov 30 09:24:28 oh great. once people found the fix the thread turned into kernel tech support. Nov 30 09:25:13 :P :P Nov 30 09:41:14 I really need to start using SDKs instead of building my own toolchains Nov 30 09:43:09 and why don't you just use directly github actions? Nov 30 09:43:44 well, I indirectly use that Nov 30 09:44:07 I basically mean for testing locally Nov 30 09:44:53 I mean, anyone can fork the repo and have the testing for free on github infra Nov 30 09:45:16 I think, it's just that people don't realize this fact Nov 30 09:45:45 in this case it's quite pointless as I don't know what the issue is :) Nov 30 09:45:55 need to test locally Nov 30 09:46:38 I find it interesting that GitHub Actions is free Nov 30 09:47:39 on GitLab I usually add my own runner, then put `sleep 6000` in the CI YAML when I need to debug the state and then simply exec myself into that container Nov 30 09:47:58 AFAIK GH allows custom runners as well, but I didn't tried that yet Nov 30 09:48:53 interesting... Nov 30 10:19:16 phew, I managed to recover the router - the internet is working again...and I get to keep my testicles :-) Nov 30 10:20:34 ldir-: What have you done? :P Nov 30 10:21:24 tethered his ipad :) Nov 30 10:22:33 I managed to flash an image that didn't include 'procd', which is all a bit fundamental and hilariously stupid. Nov 30 10:22:57 Is that even possible? O_o Nov 30 10:23:48 yes, make something idiot proof and they'll build a better idiot...ie me! :-) Nov 30 10:24:09 Come to think of it, anything is possible when tinkering with the source… Nov 30 10:24:25 ldir-: did you try to install systemd? Nov 30 10:24:52 It's possible to select 'procd-selinux' and disable 'procd' - and then disable 'procd-selinux' and forget to re-enable 'procd' Nov 30 10:25:38 ldir-: Sounds like that should be fixed. At least one variant should always be enabled, no? Nov 30 10:25:38 flash the image and errr, well it boots to a shell but your options are limited! Nov 30 10:27:04 yes - at least that's what I think I managed to do. I'll investigate once I've recovered myself... it's been an interesting 'learning' experience, not helped by being interrupted by paid work. Nov 30 10:27:28 it's all working now and I didn't lose my config. phew Nov 30 10:28:56 ldir-: Nowadays I always backup my config. No more yoloflashing. :P Nov 30 10:30:25 Yeah I've learned a valuable lesson - I thought I had a backup... turns out it was 6 months old Nov 30 10:31:06 that's why I really wanted to maintain the config if at all possible. Nov 30 10:32:02 and since it's an apu2 with a 16GB disc I wanted to maintain the partitioning etc. Nov 30 10:32:38 Heh… I believe we, as humans, always learn our lessons the hard way. In my case, I lost a config with three WireGuard endpoints. Not exactly easy to recreate the keys. :P Nov 30 10:33:59 :P Nov 30 10:34:41 ha ha, yes, urgh, unpleasant lesson :-) Nov 30 10:37:42 And speaking of configs, it seems I still haven't managed to get enforced airtime policy to work… :/ Nov 30 10:40:01 I'm pleasantly surprised to find my ISP has provided me with the same ipv6 prefix despite being offline for nearly 24hours Nov 30 10:43:14 I've learned that vodafone in Germany kind of keeps it linked to the MAC address of the modem virtually forever ... Nov 30 10:43:50 IPv4 may change, but the prefix stays for at least a week (haven't tested longer ...) Nov 30 10:44:12 adrianschmutzler: Here in Portugal they like to change the prefixed often… because greed (selling fixed IPs). Nov 30 10:44:19 *prefix Nov 30 10:44:37 yes, makes sense somehow Nov 30 10:45:06 ldir-: how did the usb booting go? Nov 30 10:45:16 By "often" I mean about once a month. Nov 30 10:45:24 the problem in Germany is that you only get a prefix if you have your own router, so this is effectively only affected a subgroup anyway Nov 30 10:45:58 if you use the provided device, you just get one IPv4/IPv6 and no ipv6-pd at all ... Nov 30 10:46:28 so, maybe it's just broken and they don't know ;-) Nov 30 10:46:42 * rsalvaterra is still waiting for NOS to provide him with IPv6… ¬_¬ Nov 30 10:48:09 russell--: I wrote an openwrt image to a usb stick, booted that, then mounted the msata partitions, copied the 'sysupgrade.tgz' file on sda1 to another stick, then overwrote vmlinuz & sda2 with rootfs image, sacrificed a goat, rebooted :-) Nov 30 10:48:24 interesting Nov 30 10:48:41 package/x/refresh cannot handle file renames properly Nov 30 10:49:01 fwiw, i got a msata adapter so i could plug it directly into my computer for fiddling, but then switched to SD cards because i wasn't using the msata's immense space for anything. it's kind of embarassing to have all that space when openwrt fits into <11MB. Nov 30 10:49:22 well, vodafone is really fun ... Nov 30 10:49:46 i think my first apu was before it could boot from SD card Nov 30 10:49:48 russell--: turris omnia router has 8GB eMMC Nov 30 10:50:17 with their device in standard mode you get DS lite, with bridge-mode (their device working as a modem) you get full IPv4 but no IPv6, and with your own device you get full stack with pd but no support :P Nov 30 10:50:48 and no updates for the modem firmware ... Nov 30 10:50:49 i actually tried to find some small SD cards, thinking people would be throwing them away and couldn't. ended up buying 16GB uSD's for $5/each. Nov 30 10:51:53 Yes I've lots of unused space but I really don't care...am just grateful to be working again Nov 30 10:52:27 the other interesting discovery is that the apu4 (at least as they arrived here) doesn't like the 4GB sd cards that pcengines sells for use in apu's. apu2 is okay with them, but not the apu4. Nov 30 10:54:12 https://gist.github.com/neheb/64e8203d27e0be7c6c20d513d0cc442a Nov 30 10:54:18 go home quilt, you're drunk Nov 30 10:56:27 https://fivethirtyeight.com/wp-content/uploads/2016/10/gettyimages-517387760.jpg Nov 30 10:56:36 oh, different kind of poll Nov 30 11:31:51 ldir-: instead of sacrificing a goat, you could have dd'd a copy of what was on your msata before the copying, and then you could get back to the pre-goat state any time you wanted. Nov 30 11:33:00 for your sake, if not the goats Nov 30 11:33:29 lol - no actual goats were involved in any way during the incident :-) Nov 30 11:34:30 the dog stuck its nose in a couple of times...'oooh that's an interesting silver box...is it edible?' Nov 30 11:34:53 * rsalvaterra prefers waving dead chickens. Nov 30 11:38:24 russell--: I was happy enough just to get the config archive... and retaining the partitioning is another bonus. I've learned a few things about the upgrade process too :-) Nov 30 11:39:03 but yes I appreciate I could have dd'd the entire drive to my usb stick. Nov 30 12:04:21 it had only just occurred to me ;-) Nov 30 13:14:47 * russell-- is debugging a problem with an ath9k radio and it's reported tx power's. vendor says max power is 600mW (about 27 dBm), the reg domain says (in US, max power on channel 165 is 30 dBm), but iw phy0 info says 19 dBm, and /sys/kernel/debug/ieee80211/phy0/power says 19. wtf? Nov 30 13:16:39 Hm… maybe the vendor is lying…? Nov 30 13:20:00 pg 6 of https://dl.ubnt.com/datasheets/bulletm/bm_ds_web.pdf Nov 30 13:26:47 My AirGrid's ath9k also reports 19 dBm, IIRC. Are you sure those values aren't already accounting for the directional gain? Nov 30 13:30:42 in the US, you can do 30dBm with a 23 dBi antenna on that channel Nov 30 13:31:12 Also page 6… https://dl.ubnt.com/datasheets/airgridm/airGrid_HP.pdf Nov 30 13:31:53 russell--: You can do anything anywhere with the right regdb, but I didn't write this. :P Nov 30 13:33:04 i'm just curious where the 19 is coming from Nov 30 13:33:08 * enyc meows Nov 30 13:38:26 russell--: have you tried decoding the "art" contents with https://github.com/pepe2k/ar9300_eeprom ? Probably it has different limits for different channels or something like that? More realistically, probably there's an additional PA installed and when you set 19 you actually get 27 on the antenna? Nov 30 13:39:37 russell--: the tool prints power in half-dBm Nov 30 13:40:13 And there can be different limits for different channels and link parameters. Nov 30 13:40:47 i have just got a few of mikrotik hap ac2 and it seem to work well. is there some explaination about how ipq40xx work ? i see switch0 in system that is supposed to have all 5 ports connected but in system we get eth0 for lan/switch and eth1 for wan. Nov 30 13:41:28 The AirGrid M5 has an external power amplifier. I'd assume the M2 also has one. Nov 30 13:43:08 The Bullet M2 also has a PA: https://macpod.net/misc/ubiquiti_bullet_m2_teardown/ubiquiti_bullet_m2_teardown.php Nov 30 13:44:35 fwiw, the origin of my inquiry is that (if i believe the readings, which i naturally don't) i seem to be seeing about 30 dB less at the receiver than i should Nov 30 13:45:42 when it's day light i might go over an try to measure with a objective (if uncalibrated) receiver Nov 30 13:46:37 i believe the relative measurements, mostly, but not absolute Nov 30 13:48:52 russell--: that 19 comes from how the kernel decodes the art partition most probably. Nov 30 13:54:55 russell--: high antenna gain encoded in eeprom? Nov 30 13:57:25 the ar9300_eeprom either doesn't decode my art partition or i'm using it wrong Nov 30 14:05:25 * russell-- goes Zzzzzzz Nov 30 14:13:04 interesting LuCI bug. it seems when there are more than a few thousand connections, the connections graph won't display and then LuCI dies, complaining it can't read main section in config. Nov 30 14:13:32 that sounds more like rcpd dying Nov 30 14:13:35 *rpcd Nov 30 14:13:50 and considering your earlier remark about adguard, you might be running into an OOM situation Nov 30 14:14:17 check dmesg for oomkiller traces Nov 30 14:14:38 211/500MB used as per htop Nov 30 14:15:41 no OOM related errors from dmesg Nov 30 14:17:47 ok Nov 30 14:18:16 looks like the is firewall also misbehaving. no longer doing port forwards, just dropping traffic :| Nov 30 14:20:31 restarting rpcd gives me LuCI back. Nov 30 14:21:30 that is, until I try and view the connections graph again.. Nov 30 14:21:48 "No informatino available" and then LuCI dies. Nov 30 14:24:02 probably `ubus call luci getConntrackList` exceeding the bus capacity Nov 30 14:25:13 that one might be a candidate for the new eventsource machanism Nov 30 14:25:22 ah, yeah, you might be onto something there. I've pushed the max connection limit to 65535. Nov 30 14:26:25 there's really not much that can be fixed about it, apart from limiting the output or paging the data as ubus has a fixed message size limit which might be hit here Nov 30 14:26:52 what could be fixed is the rpcd crash, but unclear if the fault lies there or within the libubus client library Nov 30 14:28:33 currently hosting ~9.5K connections. Nov 30 14:29:36 at least, that's what it reports on the overview page. Nov 30 14:31:43 answering my own question above - seems like few ipq4018 devices (zyxel NBG6617, mikrotik hap ac2 and some others) share same 'quirky' wan port - it is hardwired to one of the switch ports with vid 2. Nov 30 14:40:54 damex: indeed, it's documented https://openwrt.org/toh/zyxel/zyxel_nbg6617#switch_ports_for_vlans Nov 30 14:41:54 ynezz: yeah, i just asked it to be documented for mikrotik hap ac2 in its github merge request. Nov 30 14:42:14 looking forward to DSA Nov 30 14:42:33 that would be nice Nov 30 15:29:04 damex: ipq4018 is pesky. don't know about 4019. Nov 30 15:33:14 hello, the IRC log advertised for this channel does not seem to work :( Nov 30 15:33:34 502 Bad Gateway from nginx Nov 30 15:40:21 13/topic Nov 30 15:58:40 Borromini: well, i couldn't find anything better. Nov 30 16:14:58 damex: wasn't 'hating'. i have an EA6350 here as well. as long as you don't need the VLAN stuff, it should be fine i think. Nov 30 16:16:29 Borromini: i am using vlans one ipq4019 just fine. i have untagged for users and dedicted tags for management and services. sff homelab so i don't have dedicated physical ports for management. Nov 30 16:16:35 s/one/on/ Nov 30 16:18:09 so technically three tags passed over trunk between routers ;) Nov 30 16:18:33 is there a problem with that on ea6350? Nov 30 16:19:49 erm nbg6617 sorry. it broke here but i wanted to use vlan 2 Nov 30 16:20:37 Borromini: i think you can just use any other vid? Nov 30 16:21:46 oh, it is using ipq4018 Nov 30 16:21:50 not 4019 ;p Nov 30 16:22:42 <10 tags and >4090 should not normally be used. they're generally have some platform quirks even on lots of enterprise/datacenter hardware Nov 30 16:22:56 yeah it's 4018 :) Nov 30 16:23:30 had been fighting with it for hours before i realised there was a known issue with vlans etc Nov 30 16:23:38 glad you got it working. Nov 30 16:23:40 gotta run. Nov 30 16:23:56 i had hoped that er-4 would get a black friday discount over here but no dice :( Nov 30 16:24:10 haha Nov 30 16:24:39 keeping an eye on it though, but we just moved, so lots of other expenses as well :P Nov 30 16:24:41 apparently should have spent your money on btc anyway ;) Nov 30 16:24:47 we got https://mikrotik.com/product/rb4011igs_5hacq2hnd_in this bad boy discounted :D Nov 30 16:24:55 stintel: yes, my financial advisor told me so ;) Nov 30 16:25:00 anyway ttyl! Nov 30 16:25:14 Borromini: did you :) Nov 30 16:27:20 Borromini: you could get used er4 :) Nov 30 18:24:13 russell--: ok debugging which data structure in the kernel needs too much RAM is much harder ;-) Nov 30 19:21:52 anyone online right now good at setting up Strongswan? Nov 30 20:09:00 mangix: i just ldd'ed /usr/lib/lua/luci/template/parser.so on my master build - same symbols not found Nov 30 20:12:38 got a UCI scripting question…. I see “config_foreach func section” being called in places like /etc/init.d/snmp … even for sections that are unique (i.e. “agent” of which there should only be one instance)… or the “config system” section in /etc/config/system… Why? Why use a “foreach” when only one is expected? And then what happens when there’s more than one but shouldn’t be? Nov 30 20:13:48 philipp64: I think because technically there could be more than one of the same sections. Nov 30 20:14:09 and that’s not a configuration error? Nov 30 20:14:14 It's common to have only one "config system", but I don't think it's mandatory Nov 30 20:14:52 I don't have any idea about snmp, so just doing arguments from a general POV here Nov 30 20:15:05 (because there's no better way to write them, but yes, they could set global vars and abort if there's a secnd, I've donen that for things where it mattered) Nov 30 20:15:54 karlp: where? so I can copy? Nov 30 20:16:37 hrm, let me see if it's public Nov 30 20:17:14 also… how come we have “config_get_bool” but that’s the only typing? “uci_load_validate” supports “uinteger” for instance... Nov 30 20:17:29 shells don't have types... Nov 30 20:17:49 bool just just does magic convertion from yes/no/true/false/1/0 to "the right thing" Nov 30 20:18:26 to restrict to one section, you just have a global var called "DONE" or osmething, and in your foreach shell handler you just return if it was already set, or set it and continue Nov 30 20:29:07 karlp: sure, but it wouldn’t be that hard to match against digit strings for integers, do range checking, etc. Nov 30 20:33:21 the uci shell api is poorly designed Nov 30 20:33:39 the get_bool helper was once added because it is very frequently needed Nov 30 20:34:42 uci_load_validate was then eventually introduced and provides somewhat overlapping functionality Nov 30 20:36:38 imho the solution for robust uci file parsing in shell is not adding further config_get_$type methods but providing the ability to declare a spec according to which the file is parsed and resulting values being put into variables Nov 30 20:37:49 uci_load_validate provides that to some extend, but it uses ancient C code for validating datatypes, which in turn was forked from an ancient version of luci Nov 30 20:40:10 most uci converting init scripts do not really perform any type checking at all, they rely on the uci config to be syntactically and semantically sane Nov 30 20:40:40 so the problem is essentially deferred elsewhere, e.g. to luci's form validation, or to whatever is writing the configs (e.g. the user using vi) Nov 30 20:42:57 here's some code (not shell) which illustrates the approach: https://git.openwrt.org/?p=openwrt/staging/jow.git;a=blob;f=package/network/config/firewall4/files/usr/share/ucode/fw4.uc;h=ffc7e4132064d5adf44d56a4e6227ba071da51e6;hb=97da695775afbb835b81da59c7dbc7004301256a#l1814 Nov 30 20:43:10 first we parse an entire section according to a spec which declares types and constraints Nov 30 20:43:31 due to that we'll also be able to warn users about unknown (e.g. misspelled) options found in the config Nov 30 20:44:34 and after the generic parsing, parser code can be reduced to high level semantic checking while leaving all the gruntwork of testing whether an int option contains an int is left to generic parsing code Nov 30 20:56:59 i like the idea of using schemas… embedded logic maybe less so. Nov 30 21:02:04 jow: do you plan to use this for uci? Nov 30 21:02:29 or do you plan to implement such a feature to uci Nov 30 21:05:41 karlp: still around? Nov 30 21:06:06 can you have a look at https://github.com/openwrt/packages/pull/14028 ? Nov 30 21:46:51 sorry, was putting the girls to bed. Nov 30 21:55:58 looks fine to me, personally, I'm a fan of ${} all the time, but I certainly don't mind if people have decided, "this particular case doesn't need it, this time" Nov 30 22:01:05 hahahahaha Nov 30 22:01:08 https://www.phoronix.com/scan.php?page=news_item&px=Intel-IWD-1.10-Released Nov 30 22:02:24 i still wonder if iwd will ever come to OpenWrt. Dec 01 00:02:33 mangix: have you ever tried using it? Dec 01 00:06:09 karlp: I’m a fan of keeping things simple and consistent. I like ${} only where it’s necessary… it’s ugly to look at. But if it had been used 100% of the time already, I would have gone with that. Dec 01 00:56:40 russell--: on my desktop yes Dec 01 00:56:58 iwd's AP mode is very barebone Dec 01 01:26:56 mangix: it sounds interesting, but i never managed to wade my way to seeing it in action Dec 01 01:39:22 it would certainly be nice to take a look at (give it a sping for testing), but I wouldn't quite trust it for daily operations Dec 01 02:31:58 iwd is basically the systemd of wifi Dec 01 02:32:13 meaning it handles stuff that is normally handles by other stuff Dec 01 02:32:20 *handled Dec 01 02:37:11 Any one know what's rong with this PR? Dec 01 02:37:13 https://github.com/openwrt/openwrt/pull/2522#issuecomment-736132984 Dec 01 02:37:42 mac80211/hostapd: add support for 256-QAM on 802.11n Dec 01 02:38:00 Will it brake things? Dec 01 02:46:51 How can disable the file hash check during a build? Dec 01 02:50:36 I modified a tar file in ./dl and the build is complaining that the file hashes don't match. How can I disable the file hash check? **** ENDING LOGGING AT Tue Dec 01 03:00:01 2020