**** BEGIN LOGGING AT Mon Feb 24 02:59:57 2020 Feb 24 09:19:39 karlp: is your mbusd published somewhere? maybe we can push something upstream Feb 24 09:20:06 I'm using it for the right part of this screenshot: https://www.linux-ipv6.be/Screenshot_20200224_105724.png Feb 24 09:20:33 * stintel had a productive weekend Feb 24 09:27:32 stintel: :-) Feb 24 09:28:21 blogic: that's what happens when the gf is sick at home :) Feb 24 09:28:42 try getting divorced, productivity goes through the roof :-D Feb 24 09:28:47 hahaha Feb 24 09:37:31 is xback ok? Feb 24 09:37:43 ldir: huh ? Feb 24 09:38:15 I've not seen any kernel bumps for a while Feb 24 09:39:15 I think he's focusing on 5.4? Feb 24 09:39:29 I was wondering if it was worth putting effort in bumping 4.19 still Feb 24 09:39:51 as I flashed some new images this week and ended up with a non-latest kernel :P Feb 24 09:44:00 at least 4.19.105 didn't really need any changes for the targets I've built (ipq806x, ipq40xx, ath79, lantiq) Feb 24 09:54:50 106 is out Feb 24 09:55:58 104 & 105 required no patch refreshing at all(!) Feb 24 09:58:01 There's 'always' 1 patch that requires a slight line number change so to get 2 no change bumps in a row was a shock :-) Feb 24 09:58:41 probably means 106 will be refresh hell Feb 24 10:18:47 is anyone interested in or working on support for an Ubiquiti EdgeRouter 10X? Feb 24 10:19:12 the device is a MT7621 (ramips) and _very_ similar to EdgeRouter X which is already supported Feb 24 10:20:05 oh, that looks interesting. I've just ordered a 10gigE PoE switch Feb 24 10:20:24 ah bah. it's passive PoE Feb 24 10:20:48 why they keep doing this is beyond me Feb 24 10:21:51 stintel: and the 10X only has a single PoE passthrough port Feb 24 10:23:28 stintel: and the speed and hardware design of the 10X is really a router platform not a switch... Feb 24 10:23:45 if it only has a single 10gigE port I'll probably only use that one. don't have much use for the other ports. but looks cheap and might be an alternative for the expensive macchiatobin I've been looking at Feb 24 10:23:55 stintel: the upper 5 GBit ports are all on a single 1GBit link to the CPU ... Feb 24 10:24:22 yeah weird designs are weird Feb 24 10:25:17 and from what I have read about that realtek chip it does not have the offloading features that are available on the first 5 ports Feb 24 10:26:40 wut, they're using realtek now too? Feb 24 10:26:41 so it really is SoHo and not able to saturate more than 1 GBit interface - but it consumes 2.1W idle Feb 24 10:26:44 * stintel lost interest Feb 24 10:27:21 stintel: the 10X use an MT7621 with an integrated 5port switch within the SoC as it seems Feb 24 10:28:16 stintel: and the 5port switch has an rgmii link with an 1GBit capacity (which is used for a SFP module on the PoE model) and there a RT8367 is connected with another 5 ports Feb 24 10:29:37 don't understand why they decided to design it that way... the MT7621 also has a unused PCIe interface Feb 24 10:30:27 feels a bit mikrotik'y Feb 24 10:31:53 do you have any info on the 10gig chip? Feb 24 10:32:16 stintel: there is no 10gig chip - which one do you mean? Feb 24 10:33:04 it's called 10X because it has 10x 1GBit ports (5 first class MediaTek, 5 snd class RealTek) Feb 24 10:33:31 HEH Feb 24 10:34:04 > Ten Gigabit RJ45 ports offer copper connectivity with PoE input on port 1 Feb 24 10:34:08 ok\ Feb 24 10:34:17 * stintel goes to make coffee Feb 24 10:34:34 I guess I read port instead of ports Feb 24 10:35:11 stintel: but 10GBit would be cool - i'd buy one Feb 24 10:38:45 yeah, so I guess macchiatobin is still on the table :) Feb 24 10:39:26 although the switch I ordered is full layer 3 (I just realised that after ordering it, ordered it for the 10gigE and PoE++ specs) Feb 24 10:39:48 so I don't even need a 10gigE router for now, as my WAN link is only 200/120 and inter-vlan will be handled by the switch Feb 24 11:14:23 ldir, the bumps are staged ;-) Feb 24 11:15:37 currently very busy at work, and focusing on 5.4 as stintel indicated. Feb 24 11:15:40 welcome (x)back :p Feb 24 11:24:38 stintel: I think it is still useful to bump 4.19 as it make transitioning to 5.4 "easier" Feb 24 11:27:58 xback: do you have any 4.19 bumps in your staging tree as well ? Feb 24 11:28:39 stintel: https://git.openwrt.org/?p=openwrt/staging/xback.git;a=commit;h=742124970c5fe6781cac05238f7ad1bffdaa70e0 Feb 24 11:29:36 xback: have time to bump to .106 ? Feb 24 11:29:46 if not I can have a go at it tonight after work Feb 24 11:29:53 and push it to master Feb 24 11:32:04 stintel: it's scheduled for today Feb 24 11:32:27 xback: awesome, thanks! Feb 24 12:35:39 106 bump if it helps https://git.openwrt.org/?p=openwrt/staging/ldir.git;a=commitdiff;h=8294c750efc6e2a6a63daae2f13f7f699f225632 Feb 24 13:16:52 argh .. so annoying that someone refreshed all patches while my patches were already staged .. git am error .. Feb 24 13:17:46 ldir: thanks! Feb 24 13:30:15 Hauke: pong Feb 24 13:36:21 Hauke: nm, see mail please. Feb 24 14:11:54 Hauke: x86_64 builds fine now. Testing it now .. Feb 24 14:29:31 Hauke: and it runs nicely :-) Great! Feb 24 17:34:22 hey guys, take a look at this https://bugs.openwrt.org/index.php?do=details&task_id=2848 Feb 24 17:34:28 do you identify the commit that introduce this problem? Feb 24 18:12:00 stintel: kernel bump pushed to master (4.19.106) Feb 24 19:24:29 xback: thanks again! Feb 24 19:42:20 Hauke: so ath79 is broken after all 🙂 http://downloads.openwrt.org/snapshots/faillogs/mips_24kc/base/iw/full/compile.txt Feb 24 19:45:46 mips24kc buildbot seems broken :( maybe disk full? see http://downloads.openwrt.org/snapshots/faillogs/mips_24kc/base/iw/tiny/compile.txt Feb 24 19:58:48 dangole: hey Feb 24 19:58:56 dangole: i fixed the wifi reload bad channel thing Feb 24 19:59:04 will push the next couple of days Feb 24 19:59:16 we basically now also add a radio_config_id Feb 24 19:59:30 and force a full restart of the radio when things change Feb 24 19:59:36 blogic: nice! i thought that before and had it implemented, then dropped it because i thought it's unneeded... Feb 24 19:59:55 turns out that a dynamic channel change, even when using CSA would not zwizzle chanwidth Feb 24 20:01:14 blogic: ouf... stations needing to follow a chan-switch which is not DFS/radar related very fast seems like a very rare corner case anyway Feb 24 20:01:44 dangole: i will next add a patch that will trigger csa if only the channel is switched Feb 24 20:02:12 blogic: perfectionst... :p Feb 24 20:02:48 blogic: have you seen that report from Stijn? i just arrived back to europe and can look into it, but for now i haven't found out why things are different to what they were before Feb 24 20:03:09 did, not looked Feb 24 20:03:15 wont have time till early next week Feb 24 20:03:27 dangole: sorry for the delay. holidays and frequent travel often get in the way of stuff :( Feb 24 20:03:33 i'll be wrapping up u80211d the next couple of days Feb 24 20:03:37 then move on to MBSSID Feb 24 20:04:06 but I wrote it on my whiteboard to not forget about it entirely Feb 24 20:04:14 which seems to work somehow Feb 24 20:05:22 I've worked around it for rpi0w by enabling rngd with /dev/hwrng Feb 24 20:05:44 but it's still a problem when after a power outage an openwrt client comes up before the AP Feb 24 20:06:06 stintel: i also did not see you comment on how to reproduce this on arbitrary hardware by simply power-on the AP a but later Feb 24 20:06:22 yeah that's currently my main issue Feb 24 20:06:52 stintel: netifd should wait for IFUP event before starting udhcpcd, but maybe this was not the case before either Feb 24 20:07:10 blogic: MBSSID? Feb 24 20:07:13 dangole: it doesn't start udhcpc at all afaik Feb 24 20:07:42 stintel: ? just starts it too early i thought? Feb 24 20:07:59 stintel, blogic: brb, food is ready ;) Feb 24 20:08:15 if that were the case it would still get an IP once the interface comes up I'd expect? Feb 24 20:08:16 dangole: multi bssid Feb 24 20:08:26 let me verify again Feb 24 20:08:28 dangole: basically 11ax sta requires this Feb 24 20:08:41 AP is optional and as of 11be/eht its required Feb 24 20:08:52 so the radio broadcasts a single beacon for all vaps Feb 24 20:09:06 and they are daisy chanied in that frame Feb 24 20:09:19 and then 6ghz is data only with nothin and the beacons are sent out as mbssib ones on 5g while 6g does qam4096 on 320mhz channels Feb 24 20:09:45 :-D Feb 24 20:10:06 that will allow above 10g throughput Feb 24 20:10:32 basically in high density dwellings you can have a pile of folks stream 4k over wifi Feb 24 20:10:48 i dont see any other use case Feb 24 20:10:57 mbssid as in, the ability to configure multi APs in a client device ? Feb 24 20:11:09 no Feb 24 20:11:23 multiple beacons inside a dingle mgmt frame Feb 24 20:11:36 they blundered the naming and introduced a namespace collission Feb 24 20:11:59 on high density areas you will find that on 2,4g 1/4 of the band is spewed with beacons Feb 24 20:12:02 and probes Feb 24 20:12:10 as they are sent out in 6mbit Feb 24 20:12:44 so in EHT bss's donot broadcast beacons anymore on their channel Feb 24 20:12:52 but use 5g to sent aggregates of beacons Feb 24 20:25:55 10g throughput with a >=15km link margin and I'll buy it :p Feb 24 20:32:24 :D Feb 24 20:34:22 * mamarley drives up in a forklift with a large spool of fiber. Feb 24 20:34:22 Someone said they needed 10Gbps over 15km? Feb 24 20:42:47 stintel: no, mine is still closed source. It's someth8ing I've been considering opensourcing for a while, but it still has some unrelated stuff in it, and it's work I can't justify right now, but it's probably the best target of stuff for me to move from the private side to the public side. Feb 24 20:47:27 I see, in that case I would consider throwing what I have to openwrt-packages Feb 24 20:47:34 if I figure out that ::data instance Feb 24 20:52:00 it's really, like select() and then modbus_tcp_pi_accept() and modbus_receive_confirmation stuff if you use libmodbus Feb 24 20:52:08 that mbusd hthing looks supppppper complicated. Feb 24 20:52:42 I think there's even an implementation linked as a PR on libmodbus asking for it be standard functionality available :) Feb 24 20:53:54 I honestly did not look at the code Feb 24 20:56:23 karlp: https://github.com/stephane/libmodbus/pull/341 ? Feb 24 20:56:40 you do like select on some sockets, modbus_receive(), on the tcp side, check or open the rtu side, modbus_send_raw_request on the rtu side, modbus_receive_confirmation, then a little bit of write() trickery to unwrap/rewrap the mbap header Feb 24 20:56:56 that's the one I remember seeing Feb 24 20:57:03 haven't tested it, but it looked similar to mine Feb 24 20:57:03 the fact that libmodbus author didn't even bother commenting on that PR is not confidence inspiring :( Feb 24 20:57:15 libmodbus is a veryyyyyyy slow moving project Feb 24 20:57:33 it has activity every now and again, but it's mostly "done" Feb 24 20:57:42 feature creep is ... feature creep? Feb 24 20:58:45 that's how forks happen Feb 24 20:59:17 no, that's not the right pr, that' sjust the tcp server. Feb 24 20:59:24 you're looking for the tcp/rtu relay somewhere Feb 24 20:59:38 https://github.com/stephane/libmodbus/pull/260 ? Feb 24 20:59:48 well, forks have to provide value beyond, "I merged my own half baked pr, so this fork is better now" Feb 24 21:00:24 yeah, #260 and the linked 445. Feb 24 21:00:27 * karlp shrugs Feb 24 21:00:35 or just package and use mbusd :) no library deps Feb 24 21:00:42 oh well, I'll stick with that mbusd thing for now Feb 24 21:00:50 you could probably even use lua-libmodbus and make it even simpler :) Feb 24 21:01:32 I mean, you've got something that works, no need to mess with it :) Feb 24 21:01:46 what are you using client side? Feb 24 21:01:51 home-assistant Feb 24 21:03:49 I should probably get on with migrating my Z-wave network from domoticz to home-assistant and then ditch domoticz. I much prefer the tiny C++ thing over the mastodon that is home-assistant, but maintaining domoticz package is a huge pita, and home-assistant supports so much more Feb 24 21:04:37 heh, I thought domoticz was a "huge" c++ thing, but a tiny one, but I gbuess that says more about home-assistant :) Feb 24 21:05:01 karlp: I was able to run domoticz on an rspro with 16MB flash Feb 24 21:05:25 homeassistant/home-assistant stable 035fb01da530 6 days ago 1.16GB Feb 24 21:05:37 1.2G /home/docker/home-assistant/ Feb 24 21:05:58 so that's 2.4GB in total and the home-assistant sqlite db .. it grows at an insane rate Feb 24 21:06:32 * karlp laughs Feb 24 21:06:33 ok. Feb 24 21:07:43 and domoticz has in my opinion much prettier graphs and reports for power consumption etc Feb 24 21:07:59 with home-assistant I am now also running influxdb and grafana to have a fancy dashboard Feb 24 21:13:56 what does domoticz store in? on 16meg falsh? Feb 24 21:14:04 also sqlite Feb 24 21:14:18 -rwxrwx--- 1 domoticz domoticz 2723840 Feb 24 23:10 /var/lib/domoticz/domoticz.db Feb 24 21:14:26 this is after 4 years of usage or so Feb 24 21:15:17 well since 09/2017 Feb 24 21:20:36 what's it storing then? hourly or something? Feb 24 21:21:39 every five minutes iirc Feb 24 21:23:40 there is another table where it keeps daily usage which is used for historical reports Feb 24 21:23:47 it's quite efficient Feb 24 21:38:52 ugh, after changing most configs home-assistant needs to be restarted, no reload for most things Feb 24 21:42:51 oh, you're only dealing with a single three phase mains point aren't you... Feb 24 21:56:31 karlp: yeah Feb 24 21:56:59 karlp: why you ask? or you mean I'm only measuring that? Feb 24 22:24:41 always thinking about things to do on device. Feb 24 23:36:59 hi. builds of 19.07 branch currently fail for me when trying to include hostapd+hostapd-utils , maybe also wpa-cli. build host is current arch linux, maybe the tools are too new? maybe this is the relevant part of the buildlog, i can only continue debugging tomorrow night maybe. https://paste.tecff.de/?3c490ea6627c3d8d#PzjBmbOkuWwqHXTXiwpDJc0pJ0cak40XFZnsMiJaD10= Feb 25 00:19:09 seems someone broke the snapshot buildbotsd **** ENDING LOGGING AT Tue Feb 25 02:59:57 2020