**** BEGIN LOGGING AT Thu Feb 18 02:59:56 2021 Feb 18 04:17:42 ynezz: ping? Feb 18 04:47:02 For those using malta for dev, this enables 5.10: https://github.com/openwrt/openwrt/pull/3881. As always, very handy for word-size/endianness testing. Feb 18 05:19:04 build #2 of ipq40xx/generic is complete: Failure [failed targetupload] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/ipq40xx%2Fgeneric/builds/2 blamelist: Adrian Schmutzler , Felix Fietkau , Eneas U de Queiroz Feb 18 06:05:22 rmilecki: just pinging you about a follow-up rewiew request for an ath79 migration PR: https://github.com/openwrt/openwrt/pull/3062 Feb 18 06:06:13 rmilecki: appreciate if you have a chance to look, we're trying to get it done for 21.xx Feb 18 06:06:17 guidosarducci: oh, some hacky mtd code Feb 18 06:06:19 i'll try to check tha Feb 18 06:07:18 rmilecki: originally, but it's been updated the "right way" since then. Thanks! Feb 18 08:35:41 nbd: ping - I've got that build problem on another mac - https://pastebin.com/wAmFqM4K - I'm confused to say the least! Feb 18 08:56:27 mangix: ping Feb 18 09:11:46 rsalvaterra, ping Feb 18 09:11:51 morning guys Feb 18 09:12:02 Feb 18 09:12:11 nitroshift: pong o/ Feb 18 09:12:23 pm please? Feb 18 09:13:14 morning Feb 18 09:14:27 Sure, go ahead. :) Feb 18 09:27:37 I'm bleary eyed and tired, but can anyone think of a reason the build system would refuse to use the .xz package it created in ./dl/ on download but instead goes out every step and downloads trhe git source over and over? Feb 18 09:28:08 It's 550mb each time, and I'm not sure why Feb 18 09:29:07 wrong PKG_MIRROR_HASH? Feb 18 09:29:19 the Makefile has the HASH which I did manually from the pack it did during a make package/xxx/download, because I wondered if it was freakout because of the skip Feb 18 09:29:26 PKG_MIRROR_HASH isn't even defined at this point Feb 18 09:29:32 PKG_HASH is however Feb 18 09:30:18 The package is only avail as source, so it had to go out and get it to tarball it, but it's doing it each step.. prepare, configure, compile, etc Feb 18 09:30:52 see https://openwrt.org/docs/guide-developer/packages#use_source_repository Feb 18 09:30:54 and I think I'd have noticed it prior if it was doing it, but I may have been dense enough not to Feb 18 09:31:18 Yes zorun, but this is NEW behavior Feb 18 09:32:19 I physically can see the build-system created .xz file in dl/ Feb 18 09:32:28 and it refuses to use it Feb 18 09:35:05 Which is why I'm confused and wondering if something changed.. I updated my working tree with the ABI changes that happened a few days ago Feb 18 09:36:07 no idea why something changed, but if you define PKG_MIRROR_HASH with the sha256 of the file it should work Feb 18 09:36:34 Ok.. Then what is PKG_HASH used for? Is that for pulling a tarball whole? Feb 18 09:36:56 I mean, I commented it out so it isn't a big deal, but then PKG_HASH and PKG_MIRROR_HASH will be the sameA? Feb 18 09:38:38 PKG_HASH is for tarballs, it doesn't make sense for a git repo Feb 18 09:38:40 Sorry if I sound petulant :) I've just been staring at this for a while and not understand :) Feb 18 09:39:12 I thought PKG_HASH was used against the dl/xxxx.xz file... I'll set it and see what happens Feb 18 09:39:16 Thank you :) Feb 18 09:40:13 (as in, I'll set PKG_MIRROR_HASH) Feb 18 09:56:44 looks like linux-mips.org is off the air Feb 18 10:03:39 quick question: 19.07.7 ath79 target is missing most of the packages e.g. luci-app-aria2 missing which makes it imossible to use imagebuilder with this target. Is this a known problem? Feb 18 10:05:19 russell-- Not only gone, but really gone.. I can't even do a historical IP pull on it Feb 18 10:05:51 nast: this would be a problem, but the packages seem to be there. what is your error? Feb 18 10:06:45 sorry, you're right. Feb 18 10:06:53 I was wrong about it. Feb 18 10:07:42 it was couple of hours ago I was building mt7621, x86 and ath79 images, and only last one failed. but it should be ok now. Feb 18 10:08:19 ok Feb 18 10:13:35 Grommish: seems like a dns problem, maybe their server is in texas Feb 18 10:14:31 My domain can be tracked by historical IP back to 2011.. linux-mips.org I can't even find a single historical one Feb 18 10:14:38 and it is registared out of india Feb 18 10:15:04 It was also renewed in 2020 Feb 18 10:15:15 I wonder if someone stole it on expiration Feb 18 10:16:47 mangix: ping Feb 18 10:17:07 rsalvaterra: pong Feb 18 10:17:20 my docker setup blew up Feb 18 10:17:36 i have no idea how to debug this Feb 18 10:17:59 mangix: I'm now chasing the IRQ problem on the Omnia… you told me you have a couple of Omnias, right? Feb 18 10:18:08 just one Feb 18 10:18:54 Ok, one is enough. :) Feb 18 10:19:00 Grommish: i only noticed because i set up a git remote to their tree some time ago and a fetch wasn't connecting Feb 18 10:19:11 What version of OpenWrt is it running? Feb 18 10:19:30 it isnt Feb 18 10:20:24 Oh… /o\ Feb 18 10:20:26 russell-- Yeah, I tried to find some patches there a while back and couldn't ocnnect, but didn't think about it.. viewdns.info is a tool I use that can give interestring info.. https://viewdns.info/iphistory/?domain=openwrt.org Feb 18 10:20:52 Don't do google.com :D Takes forever Feb 18 10:21:51 I want to go to bed but I want to see if zorun has saved me from my own idiocy first Feb 18 10:23:16 docker...how amazing. the only error it's giving me is that something went wrong Feb 18 10:24:20 Next it'll be all about the kubers :) Feb 18 10:47:58 sigh. looks like i'll have to rebuild my NAS Feb 18 10:48:23 * mangix hates debian Feb 18 10:52:26 * Borromini loves debian Feb 18 10:52:37 * rsalvaterra loves Debian Feb 18 10:52:39 mangix: i thought you liked arch? why are you using debian on your NAS? Feb 18 10:54:36 Borromini: It doesn't necessarily have to be pure Debian. OpenMediaVault is Debian-based, for example. Feb 18 10:54:58 true. Feb 18 10:59:07 TrueNAS is FBSD.. Besides the OS, truenas itself is actually awesome thing =) Feb 18 11:01:11 it is Feb 18 11:08:56 * Borromini doesn't mind plain debian with zfs on his x86_64 Feb 18 11:09:10 have armbian on my odroid xu4 though. Feb 18 11:09:47 I use ZFS on selected use computers w/ debian, but on NAS specifically, TrueNAS does rule (: Feb 18 11:10:45 Borromini: what are you using the xu4 for? Feb 18 11:13:20 anyone knows why samba-libs fails to build due to libncursesw6 dependency not found? Feb 18 11:13:52 it isn't available in buildbot packages either Feb 18 11:14:00 nitroshift: jow was discussing that yesterday with nbd. Seems to be realated to ABI dependency changes Feb 18 11:14:14 f00b4r0, thanks Feb 18 11:15:31 Borromini: my NAS is ARM based Feb 18 11:15:53 kinda have to use Armbian Feb 18 11:16:40 mangix: It's the Helios, right? Feb 18 11:19:40 yeah Feb 18 11:20:02 mangix: what's the state of arch linux arm? Feb 18 11:20:12 none. lol Feb 18 11:20:15 you got the helios or the helios64? Feb 18 11:20:22 4 Feb 18 11:21:05 meh it requires custom patches to work eight Feb 18 11:21:13 right Feb 18 11:21:17 with 'none' are you talking in general or for your device in particular Feb 18 11:21:29 device Feb 18 11:21:44 ALARM supports mvebu Feb 18 11:22:02 I installed ir on the Omnia at one point actually Feb 18 11:22:38 everything was broken ao I didn't go further Feb 18 11:27:25 rsalvaterra: hrrm I was considering putting some sort of mini-arm thing in an old scsi hdd cage ;p Feb 18 11:28:21 rsalvaterra: make small nas box... already has small very nice quality 5/12v psu in itwith plenty spare 5v to run a PI/smmilar Feb 18 11:28:25 beware of SATA performance Feb 18 11:28:38 mangix: sure, and SATA breakout fun.... Feb 18 11:29:02 mangix: some but not all mini-ARMs can have a pci-e/nvme ssd attached.... I wonder! Feb 18 11:29:25 yeah the hwlios64 has one Feb 18 11:30:05 seems like owrt 21.02 forked already -- but not to stage of test images with new mac80211 yet? Feb 18 11:30:46 probably too warly Feb 18 11:31:11 I'm really wondering where to "watch" per-se Feb 18 11:31:14 Could someone offer suggestions as to proceed with https://pastebin.com/TgAaPL0T Feb 18 11:32:28 enyc: the 5.10 mac80211 is in there if that's what you're talking about Feb 18 11:32:50 haha the first 'backport' to 21.02 snuck in =D Feb 18 11:33:03 Borromini: wassat? Feb 18 11:34:16 sorry? Feb 18 11:34:45 https://git.openwrt.org/?p=openwrt/openwrt.git;a=log;h=refs/heads/openwrt-21.02 Feb 18 11:34:56 what's your question Feb 18 11:35:35 Borromini: nm, badly worded "whats that" -- I think I can see on openwrt git... Feb 18 11:36:13 enyc: i gathered that, my reply was aabout which statement you were asking about Feb 18 11:36:25 and yes, the UniFI 6 LR snuck in Feb 18 11:36:33 Borromini: ok, what is the snuck_in backport -- just new device Feb 18 11:37:14 ldir, i had the same question but for a different package, the answer i got was that they are related to ABI dependency changes Feb 18 11:38:17 enyc: yeah Feb 18 11:39:00 hrrm upstream 5.10.17 has ath9k: fix build error with LEDS_CLASS=m .... When CONFIG_ATH9K is built-in but LED support is in a loadable module, both ath9k drivers fails to link .... doubt this is openwrt issue Feb 18 11:39:07 nitroshift: it's all very confusing - I had this problem on one build system and it magically went away. The other build system is throwing this error however the ipkg files etc produced are identical. Feb 18 11:39:37 enyc: mac80211 is at 5.10.16 atm Feb 18 11:39:49 ldir, same here Feb 18 11:39:54 ldir: it looks very random every time. i have had it with dmesg every single time Feb 18 11:40:02 sometimes cleaning tmp/ and bin/ helps Feb 18 11:40:35 Borromini: yes i mean... this build error probably doesn't affect openwrt strongly expect led support built in default kernel Feb 18 11:46:08 mangix: mt7621 has no hardware crypto, does it? Feb 18 11:46:22 Borromini: I've done that a couple of times - I'm thinking it's a bug in opkg Feb 18 11:47:04 ldir: yeah it always pops up at the image generation stage Feb 18 11:48:14 blocktrron: ping Feb 18 11:48:25 blocktrron: did you get one of the mt7622 unifis ? Feb 18 11:48:33 my pre-order got cancled Feb 18 11:49:54 Borromini: It has hw crypto, This is a very recent WIP driver https://github.com/vschagen/mtk-eip93 Feb 18 11:51:58 Rene__: oh, ok. Feb 18 11:52:12 thanks :) Feb 18 11:52:25 stripping my openssl config a bit so i can squeeze it into multiple devices Feb 18 11:52:32 bye bye wolfssl where i can help it. Feb 18 12:11:00 build #856 of sunxi/cortexa53 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/sunxi%2Fcortexa53/builds/856 Feb 18 12:31:08 ynezz: Would it be possible to add to your staging for buildbot this patch https://lists.openwrt.org/pipermail/openwrt-devel/2021-January/033245.html ? Feb 18 12:46:52 blogic: yes, its here Feb 18 12:47:03 Pre orderes beginning of dec 20 Feb 18 12:49:48 Their shipping estimations are a big lie. "In stock" can mean "8 weeks" Feb 18 12:51:52 blocktrron: i recently got a second hand mt7610 device to play with, is the "mt76: mt76x02: set bssi_idx in mt76x02_mac_wcid_set_key" in your mt7610-fix branch something one would want? Feb 18 12:54:06 Borromini: yes it doea Feb 18 12:54:41 mangix: yeah, thanks Feb 18 12:55:22 Borromini: this patch is upstream and backported to 19.07 Feb 18 12:55:31 supposwdly an XOR engine too. no driver though Feb 18 12:56:10 blocktrron: ok thanks, will comb through mt76 commits, in pre-21.02 a well then probably Feb 18 13:01:34 Well, ope Feb 18 13:01:54 mt76 in 19.07 is pretty much 2 years old at this point. Not sure if i'd waste time on that Feb 18 13:02:30 OTOH, there were not big changes for mt76x{023} i'm aware of Feb 18 13:03:29 build #725 of lantiq/xway is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/lantiq%2Fxway/builds/725 Feb 18 13:03:39 blocktrron: i cannot be bothered with 19.07, have been on master for a while and will be switching to 21.02 builds :) Feb 18 13:03:52 test driving this EAP235 with MT7613 on master as well since weeks Feb 18 13:04:27 also the whole ethernet driver & DSA migration thing made me switch to master earlier than i would have otherwise Feb 18 13:08:51 I'm awaiting my first realtek-based switch :P Feb 18 13:08:58 should be here tomorrow Feb 18 13:09:42 =D Feb 18 13:09:45 what did you get? Feb 18 13:09:55 XGS1210? :P Feb 18 13:12:07 GS108T v3 Feb 18 13:12:41 will use it for my amp/tv/xbox/vim3/vim3l Feb 18 13:12:43 :) Feb 18 13:12:50 nice want to get that too Feb 18 13:13:07 should save one cable at least (and free up another socket) Feb 18 13:13:29 i have a primal forest of extension blocks for all the stuff hiding behind the tv :-/ Feb 18 13:13:44 I've got myself two Zyxel GS1900-8HP for 69 from Amazon WHD on that last sale. Feb 18 13:13:50 s/primal/primeval/ Feb 18 13:14:03 blocktrron: neat! i have a 10HP here and an 8HP v1 Feb 18 13:14:12 Actually, pretty nice with the accesible console and even the stock firmware looked pretty solid for that price point Feb 18 13:14:20 yeah, agreed Feb 18 13:14:23 compared with the TP-Link JetStream accidents I've been using Feb 18 13:14:33 and the angled pins are a nice and thoughtful plus Feb 18 13:15:19 btw is realtek gonna get disabled on 21.02 or is the jury still out on that? Feb 18 13:15:27 since there are images https://downloads.openwrt.org/releases/21.02-SNAPSHOT/targets/realtek/generic/ Feb 18 13:15:50 (i remember there being talk of it being too young to include in the release) Feb 18 13:15:52 I wanted something PoE-PD Feb 18 13:15:58 stintel: same here. Feb 18 13:16:00 can't wait for a PoE-PD with PT Feb 18 13:16:06 PT? Feb 18 13:16:15 and preferably NBaseT PoE-PD uplink :P Feb 18 13:16:20 PassThrough Feb 18 13:16:24 ah ok. Feb 18 13:16:40 hehe Feb 18 14:10:27 blocktrron: so you have one and it works ? Feb 18 14:25:10 blogic: yes Feb 18 14:25:17 writing this message with it Feb 18 14:26:38 * olmari is glad that in very recent years the "active" PoE stuff entry-level has came well cheap Feb 18 14:27:03 I hate the active word there, but can't fight windmills there =) Feb 18 14:48:00 blocktrron: ok, found 1 on ebay Feb 18 14:48:03 ordered it Feb 18 14:49:48 :) Feb 18 14:53:44 blocktrron: what radio is used for 2,4g ? Feb 18 14:53:49 and does the pcb have jtag / Feb 18 14:53:50 ? Feb 18 14:55:32 it's the mt7615 of the mt7622 Feb 18 14:55:45 hmmmz Feb 18 14:55:52 I haven't seen jtag on the PCB, i can do some pictures if you'd like Feb 18 14:56:21 although i'm not able to get the cans of, as they are either glued together or require different tools than the ghetto screwdriver i have here Feb 18 14:58:01 blogic: https://i.imgur.com/Y7HtD8E.jpg Feb 18 14:58:09 blunt end? O-) Feb 18 15:04:01 blogic: what's wrong with that 2,4 radio? Feb 18 15:04:08 Borromini: its 11n Feb 18 15:04:13 no ax Feb 18 15:04:34 is there really a lot of difference in 2,4 Ghz AX? Feb 18 15:04:44 yes Feb 18 15:04:52 HE vs HT rates Feb 18 15:04:56 like 670 vs 300 Feb 18 15:05:01 oh. Feb 18 15:05:13 I'm most interested in figuring out if the 2.5GE port works. the WiFi is currently capped with my 1GE equipment Feb 18 15:05:30 over those few 2,4 GHz channels still huh Feb 18 15:05:38 blocktrron: hehe. the bottleneck always moves ;) Feb 18 15:08:56 round pc and copper around looks interesting Feb 18 15:09:17 10 "pcb" antennas ? Feb 18 15:10:27 blocktrron, "SA" on antenna connector - is there a "spectral analyze" feature ? Feb 18 15:10:56 also bluetooth antenna ? Feb 18 15:12:18 BT ist from the 7915 --> No firmware atm Feb 18 15:12:37 But yeah, it seems to have a spectral monitor. presumably for zero wait DFS Feb 18 15:12:53 But all this is not listed on Ubiquitis spec sheet. Feb 18 15:14:05 The other side looks like this: https://i.imgur.com/DFBX5bC.jpg Feb 18 15:14:47 The casing is really bad. It's glued together and reinforced with plastic clips. Feb 18 15:14:50 Not fun opening Feb 18 15:15:14 ahhh finally took the time to call my ISP to complain about the speed, was getting 310/310 on a 1000/600 connection Feb 18 15:15:35 https://www.speedtest.net/result/10952095330 Feb 18 15:15:44 * stintel happy camper Feb 18 15:15:59 =) Feb 18 15:16:10 blocktrron: 4x4 huh that's pretty neat Feb 18 15:17:14 maybe I should have another go at getting OpenWrt on my 8x8 ax AP with 10GbE PoE-PD uplink :P Feb 18 15:18:05 that EAP245 is disappointingly slow still Feb 18 15:19:08 ha Feb 18 15:19:10 bah, your speedtest result reminded me that I need to replace a gig switch,I'm on 100Monly at themoment. Feb 18 15:19:22 stintel: which one is that 8x8? Feb 18 15:19:31 i have the "fdisk" incompatible with the architectures configured bug Feb 18 15:19:42 lucky me -.- Feb 18 15:19:55 plntyk: queue at the back :P Feb 18 15:20:07 stintel: are you renting out your pipe to the rest of the building? Feb 18 15:20:11 stintel: 1000/600…? o_O Feb 18 15:20:15 could make some money :P Feb 18 15:20:20 * rsalvaterra cries in 200/20… Feb 18 15:20:29 * Borromini hides his 100/40 Feb 18 15:20:37 100/50 altough i pay for 250/100 Feb 18 15:20:49 * plntyk hits his 32/4 Feb 18 15:20:57 Borromini: I'd happily trade my 200/20 for your 100/40. :P Feb 18 15:20:59 wait till you hear the price :P Feb 18 15:21:03 i was on 6/2 until 4 months ago ;) Feb 18 15:21:19 blogic: I remember something like that, must be a huge improvement nonetheless Feb 18 15:21:35 blogic: ouch, ADSL? Feb 18 15:21:37 blogic: Uhh…? What, ADSL? Feb 18 15:23:29 rsalvaterra: you need the upload? :) Feb 18 15:23:51 paying 35lv/month Feb 18 15:23:53 stintel: that's why i am saying you should re-sell :P Feb 18 15:23:54 Who doesn't need the upload too? Feb 18 15:24:21 Borromini: so ~EUR18/mo Feb 18 15:24:25 stintel: oh. you said it because it was effing low :( Feb 18 15:24:28 yes. lv is BGN Feb 18 15:24:38 ddg doesn't like two letter currencies :P Feb 18 15:24:55 ddg ? Feb 18 15:25:01 Borromini: Who doesn't need the upload? :) Feb 18 15:25:07 stintel: duckduckgo has a currency converter Feb 18 15:25:11 rsalvaterra: true. Feb 18 15:25:12 ahhh Feb 18 15:25:18 especially with the asynchronous part Feb 18 15:25:20 I gave up on duckduckgo Feb 18 15:25:30 i fall back to google if it doesn't work Feb 18 15:25:31 I was type !g foo all the time Feb 18 15:25:33 but most of the time it does Feb 18 15:25:35 hehehe Feb 18 15:25:39 Connection asymmetry is a cancer. Feb 18 15:26:06 well I can live with 1000/600, but in Belgium they now offer 1000/40 Feb 18 15:26:07 😂 Feb 18 15:26:13 Even dial-up was symmetric up until V.90. :P Feb 18 15:26:39 s/asynchronous/asymetric/ Feb 18 15:26:57 so so two m's anyway, lovely Feb 18 15:27:02 rsalvaterra: well, it started off at 1200/75, then it was symmetric for a while until HST 9600/300 etc Feb 18 15:27:10 i need me an irssi spellchecker :P Feb 18 15:27:19 1000/40…? That needs SQM with aggressive ack filtering. :P Feb 18 15:27:39 rsalvaterra: it's full retard :( Feb 18 15:27:46 I actually de-bloat my 1000/1000 down to 900/900 with FQ_CODEL Feb 18 15:28:33 right, there HAS to be a word in German for this - what is it when you're both pleased but also quite cross at the same thing? Feb 18 15:28:33 SwedeMike: Well, your 1000/1000 is probably already about 950/950 accounting for the overhead… :) Feb 18 15:28:50 rsalvaterra: all depends on how one counts the overhead, indeed. Feb 18 15:29:01 and now I'm really going to have to replace the APU2. it just isn't strong enough Feb 18 15:29:21 stintel: my APU2 can't do 900/900 with CAKE, but it seems fairly ok at FQ_CODEL Feb 18 15:29:38 SwedeMike: I usually count the goodput. :) Feb 18 15:29:53 I have an odroid h2+ I'm waiting for openwrt to gain native support for Feb 18 15:29:54 SwedeMike: you have a command handy for that fq_codel thing ? Feb 18 15:29:55 'macbot' has ridded itself of the 'can't find fdisk incompatible arch' problem - but I have no definitive reason as to how/why - grrrr Feb 18 15:30:09 stintel: I just use the built in luci-sqm-apps or whatever it's called Feb 18 15:30:12 SwedeMike: I was looking into it but meh those realtek NICs :( Feb 18 15:30:26 SwedeMike: ah, ok. I'll have a look Feb 18 15:30:36 stintel: they even have an expansion board with i think 4 extra 2,5 Gbit NICs Feb 18 15:30:41 yeah Feb 18 15:30:46 it looks super fancy Feb 18 15:30:55 but maybe I should try again to get a Macchiatobin Feb 18 15:31:42 with layer_cake my speed drops below 200Mbps :/ Feb 18 15:32:02 stintel: sqm-scripts-extra and luci-app-sqm Feb 18 15:32:36 yeah, CAKE doesn't work well at high speeds, I had to stop doing that. I reported it to the people who made it and they started talking about irq load etc Feb 18 15:32:54 I see, so it's not just me Feb 18 15:33:04 good to know at least Feb 18 15:33:16 stintel: https://www.mail-archive.com/bloat@lists.bufferbloat.net/msg06117.html for a thread on the topic Feb 18 15:33:35 what do I need to do to get luci-app-sqm in luci after installing it? it's not in the menuys Feb 18 15:34:19 under admin > network? Feb 18 15:34:34 well just under network Feb 18 15:34:41 * Borromini was looking at the URL Feb 18 15:34:42 it's in network->SQM QoS for me Feb 18 15:34:55 and I didn't have to do anything Feb 18 15:35:27 stintel: I install luci-app-sqm sqm-scripts and sqm-scripts-extra , did it do all of those for you based on dependencies? Feb 18 15:37:33 SwedeMike: I already had the latter 2 Feb 18 15:38:33 stintel: maybe a forced reload of your web interface? Feb 18 15:38:55 nope. also remove /tmp/luci* Feb 18 15:38:59 still not visible Feb 18 15:39:12 there was also some purge thing to append at the end if it kept showing stale data but i forgot the exact wording Feb 18 15:39:46 or ?nocache Feb 18 15:48:04 cake is a bit of a cpu beast but then it's potentially doing quite a bit - shaping the link to a rate (unless in unlimited mode) - host fairness (by default) - nat lookups (not default for the host fairness) - diffserv fairness and finally an fq-codel/blue hybrid queue management thing. Feb 18 15:50:23 ldir: so cake even works if i do not know the correct rate, right ? Feb 18 15:51:21 it can use the 'back pressure' from the network interface driver...which hopefully implements byte queue limits. Feb 18 15:52:15 if the back-pressure doesn't exist as is the case for adsl/vdsl drivers then your only option is to shape what you send into that driver. Feb 18 15:56:59 ok Feb 18 15:57:09 i need to look into the cfg Feb 18 15:57:19 do you have a working uci cfg at hand ? Feb 18 15:58:42 the wealth of options always confused me Feb 18 16:03:34 no, not really - I have to shape the link to my vdsl modem 'cos it has a bufferbloat problem... my 100Mbit eth link to modem will overwhelm the 80/20 vdsl link Feb 18 16:04:51 I'll have a think/test with how to select 'unlimited' rather than 'bandwidth 'n'' mode within the sqm-scripts paradigm, 'cos it isn't obvious to me either! Feb 18 16:20:34 does anyone here have a feel for how fast the apple m1 macbooks are for building openwrt with respect to say a 2017 intel based box? Feb 18 16:23:18 build #866 of pistachio/generic is complete: Failure [failed updatefeeds] Build details are at http://buildbot.openwrt.org/master/images/builders/pistachio%2Fgeneric/builds/866 blamelist: Mathias Kresin Feb 18 16:23:20 ldir: It should be fast, especially with Linux, but that will take some time. Feb 18 17:16:47 yeah my ssh connection definitely starts lagging without SQM Feb 18 17:28:14 I know this is probably the wrong channel to ask this, but… how do you guys debug multicast connectivity problems? :/ Feb 18 17:32:29 stintel: can you send me your sqm uci ? Feb 18 17:32:45 blogic: sure Feb 18 17:39:40 rsalvaterra: what kind of multicast? I've debugged some multicast in my life... Feb 18 17:40:09 SwedeMike: basically, a triple-play service. IPTV. Feb 18 17:41:19 rsalvaterra: so this is for an HGW point of view? How much do you know multicast? Feb 18 17:41:21 I still can't get (surprise!) the TV box streaming without stopping after about 5 minutes. Feb 18 17:42:13 rsalvaterra: start by tcpdumping IGMP/MLD on wan and see if it's sending periodic updates Feb 18 17:43:16 correlate with IGMP/MLD on LAN and see what's going on there, then if that looks fine, check if stream still comes in on wan and not out to lan (indicating the actual routing is the problem) Feb 18 17:43:31 SwedeMike: I'm running igmpproxy at verbosity level 2 and it seems to be working. Receives requests from the lan, repeats them upstream in the wan. Feb 18 17:44:08 The network is IPv4 only, so it's only IGMP, not MLD. Feb 18 17:45:26 rsalvaterra: https://unix.stackexchange.com/questions/132604/br0-causes-to-drop-multicast-connection-after-5-minutes might be of use Feb 18 17:46:23 rsalvaterra: I have less experience fault-finding linux bridges etc, I mostly deployed multicast in ISP network so it was more PIM/MLD/IGMP/MSDP etc Feb 18 17:46:29 The lan side is a single interface, not a bridge, so I suppose IGMP snooping won't buy my anything. Is my reasoning correct? Feb 18 17:46:46 yeah, if it's not a bridge then I guess that's out of the question Feb 18 17:47:18 but again, I'd verify if the ISP stops sending you the stream (tcpdump on wan) and that's the problem, or if it's the actual forwarding on your device that is the problem Feb 18 17:47:26 I hate these new-fangled boxes which require both IPTV and internet access. :P Feb 18 17:48:44 Yeah, I've done a fair bit of tcpdump and I see the stream stopping, of course. But that's the symptom, I'm trying to find the cause… Feb 18 17:48:54 stream stopping where? Feb 18 17:49:27 does it stop coming in on wan or does it stop being forwarded? Feb 18 17:50:18 Good question, I'll have to recheck that. Feb 18 17:51:45 I need forwading from the lantv to both wantv and wan (these are the actual zone names). Feb 18 17:52:08 Both wantv and wan are masqueraded. Feb 18 17:53:07 Both wantv and wan input is dropped by default, but from what I seen, igmpproxy does the right thing with the firewall rules. Feb 18 17:54:01 good fault isolation is to see if your ISP is still sending you the stream or not, before you know that there is little reason to start poking at other places Feb 18 17:55:14 Ok, tcpdump on the wantv side it is. :P Feb 18 17:56:57 Thanks for the hints! :) Feb 18 18:41:58 Any recommendations on an 8+ port SFP+ managed switch? Feb 18 18:48:18 Hmm, kernel patch refreshing is so slow. Or the standard script has to do it in a slow way at least. For every target the generic kernel patches are refreshed. Ideally the generic stuff would be refreshed once and then the target specific patches refreshed on top of that. Feb 18 18:53:14 It's not a task that lends itself to parallelism either - hmmm. Feb 18 18:55:16 ldir: what's the current order of patch application? Feb 18 18:56:11 thought out loud - download new kernel archive from upstream, apply/refresh patches to it and produce a local 'patched generic kernel' tar archive. Then use that 'openwrt localised generic patched kernel' archive for everything else, refreshing the target patches on it. Feb 18 18:56:55 Sounds like a good job for overlayfs/docker Feb 18 18:57:17 hurricos: get vanilla kernel archive. Apply openwrt generic patches to it. Apply openwrt target specific patches on top of that. Feb 18 18:57:37 but if I had to guess, a good reason against changing it is that the patching process already leaks internal logic Feb 18 18:58:32 thus a refresh of all 'n' targets also refreshes the generic patches 'n' times, where 'n-1' runs will (or should) result in no changes to the generic patches. Feb 18 18:58:58 Equally, s#docker/overlayfs#git plumbing# Feb 18 19:03:22 ldir: I also asked myself that question a few times, but I did have an idea how to improve that Feb 18 19:03:39 at least applying the generic patches is necessary on every target ... Feb 18 19:04:17 so, one could only save updating the patch files afterwards, i.e. half of the job Feb 18 19:04:31 and making that possible would make the whole thing more ugly and error-prone Feb 18 19:05:57 another potential thought - instead of a full patched kernel archive, just have the generic differences in an archive that is restored over the kernel Feb 18 19:07:12 yes, but the problem is when and how this kernel+generic would be created Feb 18 19:07:32 unless we make refreshing a two-step process and hand this over to the user Feb 18 19:08:37 but given the saved time, one should actually consider ugly options if they are sufficiently "safe" Feb 18 19:09:49 it should be something like 'make target/generic/refresh' - I don't know really but the current situation is very slow and seems to be wasting a lot of time. Feb 18 19:11:04 yes, no argument about the latter Feb 18 19:11:13 unless you can parallel at target level it's a lot of sequential 'quilt foo' Feb 18 19:24:34 https://www.youtube.com/watch?v=gm0b_ijaYMQ Feb 18 19:24:42 rover v2 landing in 2 hrs Feb 18 19:24:57 Its a must watch imho Feb 18 19:25:29 openwrt on board? Feb 18 19:25:38 i doubt Feb 18 19:25:39 oh it's got a couple of microphones on it so you can hear it scream! Feb 18 19:25:51 ldir: ;) Feb 18 19:26:09 we might know that there is life outside our small bubble in a few weeks Feb 18 19:26:31 and they have a drone on board o first flight on an other planet Feb 18 19:26:45 probably vxworks Feb 18 19:26:54 and there is a device on the unit that might convert the mars atmosphere to oxygen Feb 18 19:26:58 hurricos: no Feb 18 19:27:03 no? Feb 18 19:27:12 hurricos: its most likely xilinx fpga Feb 18 19:27:21 .... running vxworks Feb 18 19:27:21 they have a special line of product Feb 18 19:27:24 :^) Feb 18 19:27:38 so the bkd radiation is pretty heavy Feb 18 19:28:02 so xilinx has this thing where all syntezised block come in pairs of 3 Feb 18 19:28:07 and it take the best 2 Feb 18 19:28:27 so if bkd radiation flips a bit then the other 3 will be sane Feb 18 19:28:36 similar to how nand eec works Feb 18 19:28:44 it's vxworks Feb 18 19:28:46 https://ai.jpl.nasa.gov/public/documents/papers/rabideau_iwpss2017_prototyping.pdf Feb 18 19:28:52 wow Feb 18 19:29:02 I'm sure you have some kind of low-level architecture to verify bkd radiation isn't messing up your hardware Feb 18 19:29:21 but vxworks *software* also plays a part in that stack Feb 18 19:29:32 that suprises me Feb 18 19:29:52